blob: 453877c5697be1b927bc8b9654f98f9cad71644b [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>
25#include <linux/init.h>
26#include <linux/smp_lock.h>
27#include <linux/acct.h>
28#include <linux/blkdev.h>
29#include <linux/quotaops.h>
30#include <linux/namei.h>
31#include <linux/buffer_head.h> /* for fsync_super() */
32#include <linux/mount.h>
33#include <linux/security.h>
34#include <linux/syscalls.h>
35#include <linux/vfs.h>
36#include <linux/writeback.h> /* for the emergency remount stuff */
37#include <linux/idr.h>
38#include <linux/kobject.h>
Ingo Molnar353ab6e2006-03-26 01:37:12 -080039#include <linux/mutex.h>
Dave Hansen49e0d022008-02-15 14:37:32 -080040#include <linux/file.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070041#include <asm/uaccess.h>
Al Viro6d59e7f2008-03-22 15:48:17 -040042#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070043
44
Linus Torvalds1da177e2005-04-16 15:20:36 -070045LIST_HEAD(super_blocks);
46DEFINE_SPINLOCK(sb_lock);
47
48/**
49 * alloc_super - create new superblock
Henrik Kretzschmarfe2bbc42006-09-06 00:03:41 -070050 * @type: filesystem type superblock should belong to
Linus Torvalds1da177e2005-04-16 15:20:36 -070051 *
52 * Allocates and initializes a new &struct super_block. alloc_super()
53 * returns a pointer new superblock or %NULL if allocation had failed.
54 */
Ingo Molnarcf516242006-07-03 00:25:27 -070055static struct super_block *alloc_super(struct file_system_type *type)
Linus Torvalds1da177e2005-04-16 15:20:36 -070056{
Oliver Neukum11b0b5a2006-03-25 03:08:13 -080057 struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER);
Linus Torvalds1da177e2005-04-16 15:20:36 -070058 static struct super_operations default_op;
59
60 if (s) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070061 if (security_sb_alloc(s)) {
62 kfree(s);
63 s = NULL;
64 goto out;
65 }
66 INIT_LIST_HEAD(&s->s_dirty);
67 INIT_LIST_HEAD(&s->s_io);
Ken Chen0e0f4fc2007-10-16 23:30:38 -070068 INIT_LIST_HEAD(&s->s_more_io);
Linus Torvalds1da177e2005-04-16 15:20:36 -070069 INIT_LIST_HEAD(&s->s_files);
70 INIT_LIST_HEAD(&s->s_instances);
71 INIT_HLIST_HEAD(&s->s_anon);
72 INIT_LIST_HEAD(&s->s_inodes);
73 init_rwsem(&s->s_umount);
Ingo Molnar7892f2f2006-01-09 15:59:25 -080074 mutex_init(&s->s_lock);
Arjan van de Ven897c6ff2006-07-03 00:25:28 -070075 lockdep_set_class(&s->s_umount, &type->s_umount_key);
Ingo Molnarcf516242006-07-03 00:25:27 -070076 /*
77 * The locking rules for s_lock are up to the
78 * filesystem. For example ext3fs has different
79 * lock ordering than usbfs:
80 */
81 lockdep_set_class(&s->s_lock, &type->s_lock_key);
Linus Torvalds1da177e2005-04-16 15:20:36 -070082 down_write(&s->s_umount);
83 s->s_count = S_BIAS;
84 atomic_set(&s->s_active, 1);
Arjan van de Vena11f3a02006-03-23 03:00:33 -080085 mutex_init(&s->s_vfs_rename_mutex);
Ingo Molnard3be9152006-03-23 03:00:29 -080086 mutex_init(&s->s_dquot.dqio_mutex);
87 mutex_init(&s->s_dquot.dqonoff_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -070088 init_rwsem(&s->s_dquot.dqptr_sem);
89 init_waitqueue_head(&s->s_wait_unfrozen);
90 s->s_maxbytes = MAX_NON_LFS;
91 s->dq_op = sb_dquot_ops;
92 s->s_qcop = sb_quotactl_ops;
93 s->s_op = &default_op;
94 s->s_time_gran = 1000000000;
95 }
96out:
97 return s;
98}
99
100/**
101 * destroy_super - frees a superblock
102 * @s: superblock to free
103 *
104 * Frees a superblock.
105 */
106static inline void destroy_super(struct super_block *s)
107{
108 security_sb_free(s);
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700109 kfree(s->s_subtype);
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800110 kfree(s->s_options);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111 kfree(s);
112}
113
114/* Superblock refcounting */
115
116/*
117 * Drop a superblock's refcount. Returns non-zero if the superblock was
118 * destroyed. The caller must hold sb_lock.
119 */
Adrian Bunk6b09ae62008-04-29 00:58:54 -0700120static int __put_super(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121{
122 int ret = 0;
123
124 if (!--sb->s_count) {
125 destroy_super(sb);
126 ret = 1;
127 }
128 return ret;
129}
130
131/*
132 * Drop a superblock's refcount.
133 * Returns non-zero if the superblock is about to be destroyed and
134 * at least is already removed from super_blocks list, so if we are
135 * making a loop through super blocks then we need to restart.
136 * The caller must hold sb_lock.
137 */
138int __put_super_and_need_restart(struct super_block *sb)
139{
140 /* check for race with generic_shutdown_super() */
141 if (list_empty(&sb->s_list)) {
142 /* super block is removed, need to restart... */
143 __put_super(sb);
144 return 1;
145 }
146 /* can't be the last, since s_list is still in use */
147 sb->s_count--;
148 BUG_ON(sb->s_count == 0);
149 return 0;
150}
151
152/**
153 * put_super - drop a temporary reference to superblock
154 * @sb: superblock in question
155 *
156 * Drops a temporary reference, frees superblock if there's no
157 * references left.
158 */
159static void put_super(struct super_block *sb)
160{
161 spin_lock(&sb_lock);
162 __put_super(sb);
163 spin_unlock(&sb_lock);
164}
165
166
167/**
168 * deactivate_super - drop an active reference to superblock
169 * @s: superblock to deactivate
170 *
171 * Drops an active reference to superblock, acquiring a temprory one if
172 * there is no active references left. In that case we lock superblock,
173 * tell fs driver to shut it down and drop the temporary reference we
174 * had just acquired.
175 */
176void deactivate_super(struct super_block *s)
177{
178 struct file_system_type *fs = s->s_type;
179 if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
180 s->s_count -= S_BIAS-1;
181 spin_unlock(&sb_lock);
Jan Kara0ff5af82008-04-28 02:14:33 -0700182 DQUOT_OFF(s, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183 down_write(&s->s_umount);
184 fs->kill_sb(s);
185 put_filesystem(fs);
186 put_super(s);
187 }
188}
189
190EXPORT_SYMBOL(deactivate_super);
191
192/**
193 * grab_super - acquire an active reference
194 * @s: reference we are trying to make active
195 *
196 * Tries to acquire an active reference. grab_super() is used when we
197 * had just found a superblock in super_blocks or fs_type->fs_supers
198 * and want to turn it into a full-blown active reference. grab_super()
199 * is called with sb_lock held and drops it. Returns 1 in case of
200 * success, 0 if we had failed (superblock contents was already dead or
201 * dying when grab_super() had been called).
202 */
Josh Triplett9c4dbee2006-09-29 01:59:29 -0700203static int grab_super(struct super_block *s) __releases(sb_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204{
205 s->s_count++;
206 spin_unlock(&sb_lock);
207 down_write(&s->s_umount);
208 if (s->s_root) {
209 spin_lock(&sb_lock);
210 if (s->s_count > S_BIAS) {
211 atomic_inc(&s->s_active);
212 s->s_count--;
213 spin_unlock(&sb_lock);
214 return 1;
215 }
216 spin_unlock(&sb_lock);
217 }
218 up_write(&s->s_umount);
219 put_super(s);
220 yield();
221 return 0;
222}
223
David Howellscf9a2ae2006-08-29 19:05:54 +0100224/*
Al Viro914e2632006-10-18 13:55:46 -0400225 * Superblock locking. We really ought to get rid of these two.
226 */
227void lock_super(struct super_block * sb)
228{
229 get_fs_excl();
230 mutex_lock(&sb->s_lock);
231}
232
233void unlock_super(struct super_block * sb)
234{
235 put_fs_excl();
236 mutex_unlock(&sb->s_lock);
237}
238
239EXPORT_SYMBOL(lock_super);
240EXPORT_SYMBOL(unlock_super);
241
242/*
David Howellscf9a2ae2006-08-29 19:05:54 +0100243 * Write out and wait upon all dirty data associated with this
244 * superblock. Filesystem data as well as the underlying block
245 * device. Takes the superblock lock. Requires a second blkdev
246 * flush by the caller to complete the operation.
247 */
248void __fsync_super(struct super_block *sb)
249{
250 sync_inodes_sb(sb, 0);
251 DQUOT_SYNC(sb);
252 lock_super(sb);
253 if (sb->s_dirt && sb->s_op->write_super)
254 sb->s_op->write_super(sb);
255 unlock_super(sb);
256 if (sb->s_op->sync_fs)
257 sb->s_op->sync_fs(sb, 1);
258 sync_blockdev(sb->s_bdev);
259 sync_inodes_sb(sb, 1);
260}
261
262/*
263 * Write out and wait upon all dirty data associated with this
264 * superblock. Filesystem data as well as the underlying block
265 * device. Takes the superblock lock.
266 */
267int fsync_super(struct super_block *sb)
268{
269 __fsync_super(sb);
270 return sync_blockdev(sb->s_bdev);
271}
272
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273/**
274 * generic_shutdown_super - common helper for ->kill_sb()
275 * @sb: superblock to kill
276 *
277 * generic_shutdown_super() does all fs-independent work on superblock
278 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
279 * that need destruction out of superblock, call generic_shutdown_super()
280 * and release aforementioned objects. Note: dentries and inodes _are_
281 * taken care of and do not need specific handling.
David Howellsc636ebd2006-10-11 01:22:19 -0700282 *
283 * Upon calling this function, the filesystem may no longer alter or
284 * rearrange the set of dentries belonging to this super_block, nor may it
285 * change the attachments of dentries to inodes.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286 */
287void generic_shutdown_super(struct super_block *sb)
288{
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -0800289 const struct super_operations *sop = sb->s_op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290
David Howellsc636ebd2006-10-11 01:22:19 -0700291 if (sb->s_root) {
292 shrink_dcache_for_umount(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 fsync_super(sb);
294 lock_super(sb);
295 sb->s_flags &= ~MS_ACTIVE;
296 /* bad name - it should be evict_inodes() */
297 invalidate_inodes(sb);
298 lock_kernel();
299
300 if (sop->write_super && sb->s_dirt)
301 sop->write_super(sb);
302 if (sop->put_super)
303 sop->put_super(sb);
304
305 /* Forget any remaining inodes */
306 if (invalidate_inodes(sb)) {
Dave Jones7b4fe292006-02-07 12:58:48 -0800307 printk("VFS: Busy inodes after unmount of %s. "
308 "Self-destruct in 5 seconds. Have a nice day...\n",
309 sb->s_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310 }
311
312 unlock_kernel();
313 unlock_super(sb);
314 }
315 spin_lock(&sb_lock);
316 /* should be initialized for __put_super_and_need_restart() */
317 list_del_init(&sb->s_list);
318 list_del(&sb->s_instances);
319 spin_unlock(&sb_lock);
320 up_write(&sb->s_umount);
321}
322
323EXPORT_SYMBOL(generic_shutdown_super);
324
325/**
326 * sget - find or create a superblock
327 * @type: filesystem type superblock should belong to
328 * @test: comparison callback
329 * @set: setup callback
330 * @data: argument to each of them
331 */
332struct super_block *sget(struct file_system_type *type,
333 int (*test)(struct super_block *,void *),
334 int (*set)(struct super_block *,void *),
335 void *data)
336{
337 struct super_block *s = NULL;
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700338 struct super_block *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339 int err;
340
341retry:
342 spin_lock(&sb_lock);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700343 if (test) {
344 list_for_each_entry(old, &type->fs_supers, s_instances) {
345 if (!test(old, data))
346 continue;
347 if (!grab_super(old))
348 goto retry;
349 if (s)
350 destroy_super(s);
351 return old;
352 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353 }
354 if (!s) {
355 spin_unlock(&sb_lock);
Ingo Molnarcf516242006-07-03 00:25:27 -0700356 s = alloc_super(type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357 if (!s)
358 return ERR_PTR(-ENOMEM);
359 goto retry;
360 }
361
362 err = set(s, data);
363 if (err) {
364 spin_unlock(&sb_lock);
365 destroy_super(s);
366 return ERR_PTR(err);
367 }
368 s->s_type = type;
369 strlcpy(s->s_id, type->name, sizeof(s->s_id));
370 list_add_tail(&s->s_list, &super_blocks);
371 list_add(&s->s_instances, &type->fs_supers);
372 spin_unlock(&sb_lock);
373 get_filesystem(type);
374 return s;
375}
376
377EXPORT_SYMBOL(sget);
378
379void drop_super(struct super_block *sb)
380{
381 up_read(&sb->s_umount);
382 put_super(sb);
383}
384
385EXPORT_SYMBOL(drop_super);
386
387static inline void write_super(struct super_block *sb)
388{
389 lock_super(sb);
390 if (sb->s_root && sb->s_dirt)
391 if (sb->s_op->write_super)
392 sb->s_op->write_super(sb);
393 unlock_super(sb);
394}
395
396/*
397 * Note: check the dirty flag before waiting, so we don't
398 * hold up the sync while mounting a device. (The newly
399 * mounted device won't need syncing.)
400 */
401void sync_supers(void)
402{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700403 struct super_block *sb;
404
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700406restart:
407 list_for_each_entry(sb, &super_blocks, s_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408 if (sb->s_dirt) {
409 sb->s_count++;
410 spin_unlock(&sb_lock);
411 down_read(&sb->s_umount);
412 write_super(sb);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700413 up_read(&sb->s_umount);
414 spin_lock(&sb_lock);
415 if (__put_super_and_need_restart(sb))
416 goto restart;
417 }
418 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419 spin_unlock(&sb_lock);
420}
421
422/*
Robert P. J. Day3a4fa0a2007-10-19 23:10:43 +0200423 * Call the ->sync_fs super_op against all filesystems which are r/w and
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 * which implement it.
425 *
426 * This operation is careful to avoid the livelock which could easily happen
427 * if two or more filesystems are being continuously dirtied. s_need_sync_fs
428 * is used only here. We set it against all filesystems and then clear it as
429 * we sync them. So redirtied filesystems are skipped.
430 *
Robert P. J. Day3a4fa0a2007-10-19 23:10:43 +0200431 * But if process A is currently running sync_filesystems and then process B
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 * calls sync_filesystems as well, process B will set all the s_need_sync_fs
433 * flags again, which will cause process A to resync everything. Fix that with
434 * a local mutex.
435 *
436 * (Fabian) Avoid sync_fs with clean fs & wait mode 0
437 */
438void sync_filesystems(int wait)
439{
440 struct super_block *sb;
Ingo Molnar353ab6e2006-03-26 01:37:12 -0800441 static DEFINE_MUTEX(mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442
Ingo Molnar353ab6e2006-03-26 01:37:12 -0800443 mutex_lock(&mutex); /* Could be down_interruptible */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700445 list_for_each_entry(sb, &super_blocks, s_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446 if (!sb->s_op->sync_fs)
447 continue;
448 if (sb->s_flags & MS_RDONLY)
449 continue;
450 sb->s_need_sync_fs = 1;
451 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452
453restart:
Kirill Korotaev618f0632005-06-23 00:09:54 -0700454 list_for_each_entry(sb, &super_blocks, s_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455 if (!sb->s_need_sync_fs)
456 continue;
457 sb->s_need_sync_fs = 0;
458 if (sb->s_flags & MS_RDONLY)
459 continue; /* hm. Was remounted r/o meanwhile */
460 sb->s_count++;
461 spin_unlock(&sb_lock);
462 down_read(&sb->s_umount);
463 if (sb->s_root && (wait || sb->s_dirt))
464 sb->s_op->sync_fs(sb, wait);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700465 up_read(&sb->s_umount);
466 /* restart only when sb is no longer on the list */
467 spin_lock(&sb_lock);
468 if (__put_super_and_need_restart(sb))
469 goto restart;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470 }
471 spin_unlock(&sb_lock);
Ingo Molnar353ab6e2006-03-26 01:37:12 -0800472 mutex_unlock(&mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473}
474
475/**
476 * get_super - get the superblock of a device
477 * @bdev: device to get the superblock for
478 *
479 * Scans the superblock list and finds the superblock of the file system
480 * mounted on the device given. %NULL is returned if no match is found.
481 */
482
483struct super_block * get_super(struct block_device *bdev)
484{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700485 struct super_block *sb;
486
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 if (!bdev)
488 return NULL;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700489
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700491rescan:
492 list_for_each_entry(sb, &super_blocks, s_list) {
493 if (sb->s_bdev == bdev) {
494 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700496 down_read(&sb->s_umount);
497 if (sb->s_root)
498 return sb;
499 up_read(&sb->s_umount);
500 /* restart only when sb is no longer on the list */
501 spin_lock(&sb_lock);
502 if (__put_super_and_need_restart(sb))
503 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504 }
505 }
506 spin_unlock(&sb_lock);
507 return NULL;
508}
509
510EXPORT_SYMBOL(get_super);
511
512struct super_block * user_get_super(dev_t dev)
513{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700514 struct super_block *sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700517rescan:
518 list_for_each_entry(sb, &super_blocks, s_list) {
519 if (sb->s_dev == dev) {
520 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700522 down_read(&sb->s_umount);
523 if (sb->s_root)
524 return sb;
525 up_read(&sb->s_umount);
526 /* restart only when sb is no longer on the list */
527 spin_lock(&sb_lock);
528 if (__put_super_and_need_restart(sb))
529 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530 }
531 }
532 spin_unlock(&sb_lock);
533 return NULL;
534}
535
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536asmlinkage long sys_ustat(unsigned dev, struct ustat __user * ubuf)
537{
538 struct super_block *s;
539 struct ustat tmp;
540 struct kstatfs sbuf;
541 int err = -EINVAL;
542
543 s = user_get_super(new_decode_dev(dev));
544 if (s == NULL)
545 goto out;
David Howells726c3342006-06-23 02:02:58 -0700546 err = vfs_statfs(s->s_root, &sbuf);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547 drop_super(s);
548 if (err)
549 goto out;
550
551 memset(&tmp,0,sizeof(struct ustat));
552 tmp.f_tfree = sbuf.f_bfree;
553 tmp.f_tinode = sbuf.f_ffree;
554
555 err = copy_to_user(ubuf,&tmp,sizeof(struct ustat)) ? -EFAULT : 0;
556out:
557 return err;
558}
559
560/**
Randy Dunlapa6b91912008-03-19 17:01:00 -0700561 * mark_files_ro - mark all files read-only
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562 * @sb: superblock in question
563 *
Randy Dunlapa6b91912008-03-19 17:01:00 -0700564 * All files are marked read-only. We don't care about pending
565 * delete files so this should be used in 'force' mode only.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 */
567
568static void mark_files_ro(struct super_block *sb)
569{
570 struct file *f;
571
Dave Hansen49e0d022008-02-15 14:37:32 -0800572retry:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 file_list_lock();
Eric Dumazet2f512012005-10-30 15:02:16 -0800574 list_for_each_entry(f, &sb->s_files, f_u.fu_list) {
Dave Hansen49e0d022008-02-15 14:37:32 -0800575 struct vfsmount *mnt;
576 if (!S_ISREG(f->f_path.dentry->d_inode->i_mode))
577 continue;
578 if (!file_count(f))
579 continue;
580 if (!(f->f_mode & FMODE_WRITE))
581 continue;
582 f->f_mode &= ~FMODE_WRITE;
Dave Hansenad775f52008-02-15 14:38:01 -0800583 if (file_check_writeable(f) != 0)
584 continue;
585 file_release_write(f);
Dave Hansen49e0d022008-02-15 14:37:32 -0800586 mnt = mntget(f->f_path.mnt);
587 file_list_unlock();
588 /*
589 * This can sleep, so we can't hold
590 * the file_list_lock() spinlock.
591 */
592 mnt_drop_write(mnt);
593 mntput(mnt);
594 goto retry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 }
596 file_list_unlock();
597}
598
599/**
600 * do_remount_sb - asks filesystem to change mount options.
601 * @sb: superblock in question
602 * @flags: numeric part of options
603 * @data: the rest of options
604 * @force: whether or not to force the change
605 *
606 * Alters the mount options of a mounted file system.
607 */
608int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
609{
610 int retval;
Jan Kara0ff5af82008-04-28 02:14:33 -0700611 int remount_rw;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612
David Howells93614012006-09-30 20:45:40 +0200613#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
615 return -EACCES;
David Howells93614012006-09-30 20:45:40 +0200616#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617 if (flags & MS_RDONLY)
618 acct_auto_close(sb);
619 shrink_dcache_sb(sb);
620 fsync_super(sb);
621
622 /* If we are remounting RDONLY and current sb is read/write,
623 make sure there are no rw files opened */
624 if ((flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY)) {
625 if (force)
626 mark_files_ro(sb);
627 else if (!fs_may_remount_ro(sb))
628 return -EBUSY;
Jan Kara0ff5af82008-04-28 02:14:33 -0700629 retval = DQUOT_OFF(sb, 1);
630 if (retval < 0 && retval != -ENOSYS)
631 return -EBUSY;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632 }
Jan Kara0ff5af82008-04-28 02:14:33 -0700633 remount_rw = !(flags & MS_RDONLY) && (sb->s_flags & MS_RDONLY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634
635 if (sb->s_op->remount_fs) {
636 lock_super(sb);
637 retval = sb->s_op->remount_fs(sb, &flags, data);
638 unlock_super(sb);
639 if (retval)
640 return retval;
641 }
642 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
Jan Kara0ff5af82008-04-28 02:14:33 -0700643 if (remount_rw)
644 DQUOT_ON_REMOUNT(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645 return 0;
646}
647
648static void do_emergency_remount(unsigned long foo)
649{
650 struct super_block *sb;
651
652 spin_lock(&sb_lock);
653 list_for_each_entry(sb, &super_blocks, s_list) {
654 sb->s_count++;
655 spin_unlock(&sb_lock);
656 down_read(&sb->s_umount);
657 if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) {
658 /*
659 * ->remount_fs needs lock_kernel().
660 *
661 * What lock protects sb->s_flags??
662 */
663 lock_kernel();
664 do_remount_sb(sb, MS_RDONLY, NULL, 1);
665 unlock_kernel();
666 }
667 drop_super(sb);
668 spin_lock(&sb_lock);
669 }
670 spin_unlock(&sb_lock);
671 printk("Emergency Remount complete\n");
672}
673
674void emergency_remount(void)
675{
676 pdflush_operation(do_emergency_remount, 0);
677}
678
679/*
680 * Unnamed block devices are dummy devices used by virtual
681 * filesystems which don't use real block-devices. -- jrs
682 */
683
684static struct idr unnamed_dev_idr;
685static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
686
687int set_anon_super(struct super_block *s, void *data)
688{
689 int dev;
690 int error;
691
692 retry:
693 if (idr_pre_get(&unnamed_dev_idr, GFP_ATOMIC) == 0)
694 return -ENOMEM;
695 spin_lock(&unnamed_dev_lock);
696 error = idr_get_new(&unnamed_dev_idr, NULL, &dev);
697 spin_unlock(&unnamed_dev_lock);
698 if (error == -EAGAIN)
699 /* We raced and lost with another CPU. */
700 goto retry;
701 else if (error)
702 return -EAGAIN;
703
704 if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
705 spin_lock(&unnamed_dev_lock);
706 idr_remove(&unnamed_dev_idr, dev);
707 spin_unlock(&unnamed_dev_lock);
708 return -EMFILE;
709 }
710 s->s_dev = MKDEV(0, dev & MINORMASK);
711 return 0;
712}
713
714EXPORT_SYMBOL(set_anon_super);
715
716void kill_anon_super(struct super_block *sb)
717{
718 int slot = MINOR(sb->s_dev);
719
720 generic_shutdown_super(sb);
721 spin_lock(&unnamed_dev_lock);
722 idr_remove(&unnamed_dev_idr, slot);
723 spin_unlock(&unnamed_dev_lock);
724}
725
726EXPORT_SYMBOL(kill_anon_super);
727
728void __init unnamed_dev_init(void)
729{
730 idr_init(&unnamed_dev_idr);
731}
732
733void kill_litter_super(struct super_block *sb)
734{
735 if (sb->s_root)
736 d_genocide(sb->s_root);
737 kill_anon_super(sb);
738}
739
740EXPORT_SYMBOL(kill_litter_super);
741
David Howells93614012006-09-30 20:45:40 +0200742#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743static int set_bdev_super(struct super_block *s, void *data)
744{
745 s->s_bdev = data;
746 s->s_dev = s->s_bdev->bd_dev;
747 return 0;
748}
749
750static int test_bdev_super(struct super_block *s, void *data)
751{
752 return (void *)s->s_bdev == data;
753}
754
David Howells454e2392006-06-23 02:02:57 -0700755int get_sb_bdev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756 int flags, const char *dev_name, void *data,
David Howells454e2392006-06-23 02:02:57 -0700757 int (*fill_super)(struct super_block *, void *, int),
758 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759{
760 struct block_device *bdev;
761 struct super_block *s;
762 int error = 0;
763
764 bdev = open_bdev_excl(dev_name, flags, fs_type);
765 if (IS_ERR(bdev))
David Howells454e2392006-06-23 02:02:57 -0700766 return PTR_ERR(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767
768 /*
769 * once the super is inserted into the list by sget, s_umount
770 * will protect the lockfs code from trying to start a snapshot
771 * while we are mounting
772 */
David Chinnerf73ca1b2007-01-10 23:15:41 -0800773 down(&bdev->bd_mount_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774 s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
David Chinnerf73ca1b2007-01-10 23:15:41 -0800775 up(&bdev->bd_mount_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700777 goto error_s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778
779 if (s->s_root) {
780 if ((flags ^ s->s_flags) & MS_RDONLY) {
781 up_write(&s->s_umount);
782 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700783 error = -EBUSY;
784 goto error_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785 }
David Howells454e2392006-06-23 02:02:57 -0700786
787 close_bdev_excl(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788 } else {
789 char b[BDEVNAME_SIZE];
790
791 s->s_flags = flags;
792 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
Pekka Enberge78c9a02006-01-08 01:03:39 -0800793 sb_set_blocksize(s, block_size(bdev));
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800794 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795 if (error) {
796 up_write(&s->s_umount);
797 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700798 goto error;
Greg Kroah-Hartmanfa675762006-02-22 09:39:02 -0800799 }
David Howells454e2392006-06-23 02:02:57 -0700800
801 s->s_flags |= MS_ACTIVE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802 }
803
David Howells454e2392006-06-23 02:02:57 -0700804 return simple_set_mnt(mnt, s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805
David Howells454e2392006-06-23 02:02:57 -0700806error_s:
807 error = PTR_ERR(s);
808error_bdev:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809 close_bdev_excl(bdev);
David Howells454e2392006-06-23 02:02:57 -0700810error:
811 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812}
813
814EXPORT_SYMBOL(get_sb_bdev);
815
816void kill_block_super(struct super_block *sb)
817{
818 struct block_device *bdev = sb->s_bdev;
819
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820 generic_shutdown_super(sb);
821 sync_blockdev(bdev);
822 close_bdev_excl(bdev);
823}
824
825EXPORT_SYMBOL(kill_block_super);
David Howells93614012006-09-30 20:45:40 +0200826#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827
David Howells454e2392006-06-23 02:02:57 -0700828int get_sb_nodev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829 int flags, void *data,
David Howells454e2392006-06-23 02:02:57 -0700830 int (*fill_super)(struct super_block *, void *, int),
831 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832{
833 int error;
834 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
835
836 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700837 return PTR_ERR(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838
839 s->s_flags = flags;
840
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800841 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842 if (error) {
843 up_write(&s->s_umount);
844 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700845 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846 }
847 s->s_flags |= MS_ACTIVE;
David Howells454e2392006-06-23 02:02:57 -0700848 return simple_set_mnt(mnt, s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849}
850
851EXPORT_SYMBOL(get_sb_nodev);
852
853static int compare_single(struct super_block *s, void *p)
854{
855 return 1;
856}
857
David Howells454e2392006-06-23 02:02:57 -0700858int get_sb_single(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 int flags, void *data,
David Howells454e2392006-06-23 02:02:57 -0700860 int (*fill_super)(struct super_block *, void *, int),
861 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862{
863 struct super_block *s;
864 int error;
865
866 s = sget(fs_type, compare_single, set_anon_super, NULL);
867 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700868 return PTR_ERR(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 if (!s->s_root) {
870 s->s_flags = flags;
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800871 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872 if (error) {
873 up_write(&s->s_umount);
874 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700875 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876 }
877 s->s_flags |= MS_ACTIVE;
878 }
879 do_remount_sb(s, flags, data, 0);
David Howells454e2392006-06-23 02:02:57 -0700880 return simple_set_mnt(mnt, s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881}
882
883EXPORT_SYMBOL(get_sb_single);
884
885struct vfsmount *
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400886vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888 struct vfsmount *mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889 char *secdata = NULL;
David Howells454e2392006-06-23 02:02:57 -0700890 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891
892 if (!type)
893 return ERR_PTR(-ENODEV);
894
David Howells454e2392006-06-23 02:02:57 -0700895 error = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896 mnt = alloc_vfsmnt(name);
897 if (!mnt)
898 goto out;
899
Eric Parise0007522008-03-05 10:31:54 -0500900 if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901 secdata = alloc_secdata();
David Howells454e2392006-06-23 02:02:57 -0700902 if (!secdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903 goto out_mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904
Eric Parise0007522008-03-05 10:31:54 -0500905 error = security_sb_copy_data(data, secdata);
David Howells454e2392006-06-23 02:02:57 -0700906 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907 goto out_free_secdata;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908 }
909
David Howells454e2392006-06-23 02:02:57 -0700910 error = type->get_sb(type, flags, name, data, mnt);
911 if (error < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912 goto out_free_secdata;
Lee Schermerhornb4c07bc2007-07-15 23:40:54 -0700913 BUG_ON(!mnt->mnt_sb);
David Howells454e2392006-06-23 02:02:57 -0700914
915 error = security_sb_kern_mount(mnt->mnt_sb, secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916 if (error)
917 goto out_sb;
David Howells454e2392006-06-23 02:02:57 -0700918
919 mnt->mnt_mountpoint = mnt->mnt_root;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920 mnt->mnt_parent = mnt;
David Howells454e2392006-06-23 02:02:57 -0700921 up_write(&mnt->mnt_sb->s_umount);
Gerald Schaefer8680e222005-06-21 17:15:16 -0700922 free_secdata(secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923 return mnt;
924out_sb:
David Howells454e2392006-06-23 02:02:57 -0700925 dput(mnt->mnt_root);
926 up_write(&mnt->mnt_sb->s_umount);
927 deactivate_super(mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928out_free_secdata:
929 free_secdata(secdata);
930out_mnt:
931 free_vfsmnt(mnt);
932out:
David Howells454e2392006-06-23 02:02:57 -0700933 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934}
935
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400936EXPORT_SYMBOL_GPL(vfs_kern_mount);
937
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700938static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype)
939{
940 int err;
941 const char *subtype = strchr(fstype, '.');
942 if (subtype) {
943 subtype++;
944 err = -EINVAL;
945 if (!subtype[0])
946 goto err;
947 } else
948 subtype = "";
949
950 mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL);
951 err = -ENOMEM;
952 if (!mnt->mnt_sb->s_subtype)
953 goto err;
954 return mnt;
955
956 err:
957 mntput(mnt);
958 return ERR_PTR(err);
959}
960
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400961struct vfsmount *
962do_kern_mount(const char *fstype, int flags, const char *name, void *data)
963{
964 struct file_system_type *type = get_fs_type(fstype);
965 struct vfsmount *mnt;
966 if (!type)
967 return ERR_PTR(-ENODEV);
968 mnt = vfs_kern_mount(type, flags, name, data);
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700969 if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) &&
970 !mnt->mnt_sb->s_subtype)
971 mnt = fs_set_subtype(mnt, fstype);
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400972 put_filesystem(type);
973 return mnt;
974}
Al Viro8a4e98d2008-02-24 01:43:03 -0500975EXPORT_SYMBOL_GPL(do_kern_mount);
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400976
Pavel Emelyanov8bf97252007-10-18 23:40:02 -0700977struct vfsmount *kern_mount_data(struct file_system_type *type, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978{
Pavel Emelyanov8bf97252007-10-18 23:40:02 -0700979 return vfs_kern_mount(type, MS_KERNMOUNT, type->name, data);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980}
981
Pavel Emelyanov8bf97252007-10-18 23:40:02 -0700982EXPORT_SYMBOL_GPL(kern_mount_data);