blob: d0a941a4e62007034a2367a5fbff27bc5571e851 [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070040#include <asm/uaccess.h>
41
42
Linus Torvalds1da177e2005-04-16 15:20:36 -070043LIST_HEAD(super_blocks);
44DEFINE_SPINLOCK(sb_lock);
45
46/**
47 * alloc_super - create new superblock
Henrik Kretzschmarfe2bbc42006-09-06 00:03:41 -070048 * @type: filesystem type superblock should belong to
Linus Torvalds1da177e2005-04-16 15:20:36 -070049 *
50 * Allocates and initializes a new &struct super_block. alloc_super()
51 * returns a pointer new superblock or %NULL if allocation had failed.
52 */
Ingo Molnarcf516242006-07-03 00:25:27 -070053static struct super_block *alloc_super(struct file_system_type *type)
Linus Torvalds1da177e2005-04-16 15:20:36 -070054{
Oliver Neukum11b0b5a2006-03-25 03:08:13 -080055 struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER);
Linus Torvalds1da177e2005-04-16 15:20:36 -070056 static struct super_operations default_op;
57
58 if (s) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070059 if (security_sb_alloc(s)) {
60 kfree(s);
61 s = NULL;
62 goto out;
63 }
64 INIT_LIST_HEAD(&s->s_dirty);
65 INIT_LIST_HEAD(&s->s_io);
Ken Chen0e0f4fc2007-10-16 23:30:38 -070066 INIT_LIST_HEAD(&s->s_more_io);
Linus Torvalds1da177e2005-04-16 15:20:36 -070067 INIT_LIST_HEAD(&s->s_files);
68 INIT_LIST_HEAD(&s->s_instances);
69 INIT_HLIST_HEAD(&s->s_anon);
70 INIT_LIST_HEAD(&s->s_inodes);
71 init_rwsem(&s->s_umount);
Ingo Molnar7892f2f2006-01-09 15:59:25 -080072 mutex_init(&s->s_lock);
Arjan van de Ven897c6ff2006-07-03 00:25:28 -070073 lockdep_set_class(&s->s_umount, &type->s_umount_key);
Ingo Molnarcf516242006-07-03 00:25:27 -070074 /*
75 * The locking rules for s_lock are up to the
76 * filesystem. For example ext3fs has different
77 * lock ordering than usbfs:
78 */
79 lockdep_set_class(&s->s_lock, &type->s_lock_key);
Linus Torvalds1da177e2005-04-16 15:20:36 -070080 down_write(&s->s_umount);
81 s->s_count = S_BIAS;
82 atomic_set(&s->s_active, 1);
Arjan van de Vena11f3a02006-03-23 03:00:33 -080083 mutex_init(&s->s_vfs_rename_mutex);
Ingo Molnard3be9152006-03-23 03:00:29 -080084 mutex_init(&s->s_dquot.dqio_mutex);
85 mutex_init(&s->s_dquot.dqonoff_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -070086 init_rwsem(&s->s_dquot.dqptr_sem);
87 init_waitqueue_head(&s->s_wait_unfrozen);
88 s->s_maxbytes = MAX_NON_LFS;
89 s->dq_op = sb_dquot_ops;
90 s->s_qcop = sb_quotactl_ops;
91 s->s_op = &default_op;
92 s->s_time_gran = 1000000000;
93 }
94out:
95 return s;
96}
97
98/**
99 * destroy_super - frees a superblock
100 * @s: superblock to free
101 *
102 * Frees a superblock.
103 */
104static inline void destroy_super(struct super_block *s)
105{
106 security_sb_free(s);
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700107 kfree(s->s_subtype);
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800108 kfree(s->s_options);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109 kfree(s);
110}
111
112/* Superblock refcounting */
113
114/*
115 * Drop a superblock's refcount. Returns non-zero if the superblock was
116 * destroyed. The caller must hold sb_lock.
117 */
118int __put_super(struct super_block *sb)
119{
120 int ret = 0;
121
122 if (!--sb->s_count) {
123 destroy_super(sb);
124 ret = 1;
125 }
126 return ret;
127}
128
129/*
130 * Drop a superblock's refcount.
131 * Returns non-zero if the superblock is about to be destroyed and
132 * at least is already removed from super_blocks list, so if we are
133 * making a loop through super blocks then we need to restart.
134 * The caller must hold sb_lock.
135 */
136int __put_super_and_need_restart(struct super_block *sb)
137{
138 /* check for race with generic_shutdown_super() */
139 if (list_empty(&sb->s_list)) {
140 /* super block is removed, need to restart... */
141 __put_super(sb);
142 return 1;
143 }
144 /* can't be the last, since s_list is still in use */
145 sb->s_count--;
146 BUG_ON(sb->s_count == 0);
147 return 0;
148}
149
150/**
151 * put_super - drop a temporary reference to superblock
152 * @sb: superblock in question
153 *
154 * Drops a temporary reference, frees superblock if there's no
155 * references left.
156 */
157static void put_super(struct super_block *sb)
158{
159 spin_lock(&sb_lock);
160 __put_super(sb);
161 spin_unlock(&sb_lock);
162}
163
164
165/**
166 * deactivate_super - drop an active reference to superblock
167 * @s: superblock to deactivate
168 *
169 * Drops an active reference to superblock, acquiring a temprory one if
170 * there is no active references left. In that case we lock superblock,
171 * tell fs driver to shut it down and drop the temporary reference we
172 * had just acquired.
173 */
174void deactivate_super(struct super_block *s)
175{
176 struct file_system_type *fs = s->s_type;
177 if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
178 s->s_count -= S_BIAS-1;
179 spin_unlock(&sb_lock);
Al Viro7b7b1ac2005-11-07 17:13:39 -0500180 DQUOT_OFF(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181 down_write(&s->s_umount);
182 fs->kill_sb(s);
183 put_filesystem(fs);
184 put_super(s);
185 }
186}
187
188EXPORT_SYMBOL(deactivate_super);
189
190/**
191 * grab_super - acquire an active reference
192 * @s: reference we are trying to make active
193 *
194 * Tries to acquire an active reference. grab_super() is used when we
195 * had just found a superblock in super_blocks or fs_type->fs_supers
196 * and want to turn it into a full-blown active reference. grab_super()
197 * is called with sb_lock held and drops it. Returns 1 in case of
198 * success, 0 if we had failed (superblock contents was already dead or
199 * dying when grab_super() had been called).
200 */
Josh Triplett9c4dbee2006-09-29 01:59:29 -0700201static int grab_super(struct super_block *s) __releases(sb_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202{
203 s->s_count++;
204 spin_unlock(&sb_lock);
205 down_write(&s->s_umount);
206 if (s->s_root) {
207 spin_lock(&sb_lock);
208 if (s->s_count > S_BIAS) {
209 atomic_inc(&s->s_active);
210 s->s_count--;
211 spin_unlock(&sb_lock);
212 return 1;
213 }
214 spin_unlock(&sb_lock);
215 }
216 up_write(&s->s_umount);
217 put_super(s);
218 yield();
219 return 0;
220}
221
David Howellscf9a2ae2006-08-29 19:05:54 +0100222/*
Al Viro914e2632006-10-18 13:55:46 -0400223 * Superblock locking. We really ought to get rid of these two.
224 */
225void lock_super(struct super_block * sb)
226{
227 get_fs_excl();
228 mutex_lock(&sb->s_lock);
229}
230
231void unlock_super(struct super_block * sb)
232{
233 put_fs_excl();
234 mutex_unlock(&sb->s_lock);
235}
236
237EXPORT_SYMBOL(lock_super);
238EXPORT_SYMBOL(unlock_super);
239
240/*
David Howellscf9a2ae2006-08-29 19:05:54 +0100241 * Write out and wait upon all dirty data associated with this
242 * superblock. Filesystem data as well as the underlying block
243 * device. Takes the superblock lock. Requires a second blkdev
244 * flush by the caller to complete the operation.
245 */
246void __fsync_super(struct super_block *sb)
247{
248 sync_inodes_sb(sb, 0);
249 DQUOT_SYNC(sb);
250 lock_super(sb);
251 if (sb->s_dirt && sb->s_op->write_super)
252 sb->s_op->write_super(sb);
253 unlock_super(sb);
254 if (sb->s_op->sync_fs)
255 sb->s_op->sync_fs(sb, 1);
256 sync_blockdev(sb->s_bdev);
257 sync_inodes_sb(sb, 1);
258}
259
260/*
261 * Write out and wait upon all dirty data associated with this
262 * superblock. Filesystem data as well as the underlying block
263 * device. Takes the superblock lock.
264 */
265int fsync_super(struct super_block *sb)
266{
267 __fsync_super(sb);
268 return sync_blockdev(sb->s_bdev);
269}
270
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271/**
272 * generic_shutdown_super - common helper for ->kill_sb()
273 * @sb: superblock to kill
274 *
275 * generic_shutdown_super() does all fs-independent work on superblock
276 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
277 * that need destruction out of superblock, call generic_shutdown_super()
278 * and release aforementioned objects. Note: dentries and inodes _are_
279 * taken care of and do not need specific handling.
David Howellsc636ebd2006-10-11 01:22:19 -0700280 *
281 * Upon calling this function, the filesystem may no longer alter or
282 * rearrange the set of dentries belonging to this super_block, nor may it
283 * change the attachments of dentries to inodes.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284 */
285void generic_shutdown_super(struct super_block *sb)
286{
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -0800287 const struct super_operations *sop = sb->s_op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288
David Howellsc636ebd2006-10-11 01:22:19 -0700289 if (sb->s_root) {
290 shrink_dcache_for_umount(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291 fsync_super(sb);
292 lock_super(sb);
293 sb->s_flags &= ~MS_ACTIVE;
294 /* bad name - it should be evict_inodes() */
295 invalidate_inodes(sb);
296 lock_kernel();
297
298 if (sop->write_super && sb->s_dirt)
299 sop->write_super(sb);
300 if (sop->put_super)
301 sop->put_super(sb);
302
303 /* Forget any remaining inodes */
304 if (invalidate_inodes(sb)) {
Dave Jones7b4fe292006-02-07 12:58:48 -0800305 printk("VFS: Busy inodes after unmount of %s. "
306 "Self-destruct in 5 seconds. Have a nice day...\n",
307 sb->s_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308 }
309
310 unlock_kernel();
311 unlock_super(sb);
312 }
313 spin_lock(&sb_lock);
314 /* should be initialized for __put_super_and_need_restart() */
315 list_del_init(&sb->s_list);
316 list_del(&sb->s_instances);
317 spin_unlock(&sb_lock);
318 up_write(&sb->s_umount);
319}
320
321EXPORT_SYMBOL(generic_shutdown_super);
322
323/**
324 * sget - find or create a superblock
325 * @type: filesystem type superblock should belong to
326 * @test: comparison callback
327 * @set: setup callback
328 * @data: argument to each of them
329 */
330struct super_block *sget(struct file_system_type *type,
331 int (*test)(struct super_block *,void *),
332 int (*set)(struct super_block *,void *),
333 void *data)
334{
335 struct super_block *s = NULL;
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700336 struct super_block *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337 int err;
338
339retry:
340 spin_lock(&sb_lock);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700341 if (test) {
342 list_for_each_entry(old, &type->fs_supers, s_instances) {
343 if (!test(old, data))
344 continue;
345 if (!grab_super(old))
346 goto retry;
347 if (s)
348 destroy_super(s);
349 return old;
350 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 }
352 if (!s) {
353 spin_unlock(&sb_lock);
Ingo Molnarcf516242006-07-03 00:25:27 -0700354 s = alloc_super(type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355 if (!s)
356 return ERR_PTR(-ENOMEM);
357 goto retry;
358 }
359
360 err = set(s, data);
361 if (err) {
362 spin_unlock(&sb_lock);
363 destroy_super(s);
364 return ERR_PTR(err);
365 }
366 s->s_type = type;
367 strlcpy(s->s_id, type->name, sizeof(s->s_id));
368 list_add_tail(&s->s_list, &super_blocks);
369 list_add(&s->s_instances, &type->fs_supers);
370 spin_unlock(&sb_lock);
371 get_filesystem(type);
372 return s;
373}
374
375EXPORT_SYMBOL(sget);
376
377void drop_super(struct super_block *sb)
378{
379 up_read(&sb->s_umount);
380 put_super(sb);
381}
382
383EXPORT_SYMBOL(drop_super);
384
385static inline void write_super(struct super_block *sb)
386{
387 lock_super(sb);
388 if (sb->s_root && sb->s_dirt)
389 if (sb->s_op->write_super)
390 sb->s_op->write_super(sb);
391 unlock_super(sb);
392}
393
394/*
395 * Note: check the dirty flag before waiting, so we don't
396 * hold up the sync while mounting a device. (The newly
397 * mounted device won't need syncing.)
398 */
399void sync_supers(void)
400{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700401 struct super_block *sb;
402
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700404restart:
405 list_for_each_entry(sb, &super_blocks, s_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406 if (sb->s_dirt) {
407 sb->s_count++;
408 spin_unlock(&sb_lock);
409 down_read(&sb->s_umount);
410 write_super(sb);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700411 up_read(&sb->s_umount);
412 spin_lock(&sb_lock);
413 if (__put_super_and_need_restart(sb))
414 goto restart;
415 }
416 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417 spin_unlock(&sb_lock);
418}
419
420/*
Robert P. J. Day3a4fa0a2007-10-19 23:10:43 +0200421 * Call the ->sync_fs super_op against all filesystems which are r/w and
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422 * which implement it.
423 *
424 * This operation is careful to avoid the livelock which could easily happen
425 * if two or more filesystems are being continuously dirtied. s_need_sync_fs
426 * is used only here. We set it against all filesystems and then clear it as
427 * we sync them. So redirtied filesystems are skipped.
428 *
Robert P. J. Day3a4fa0a2007-10-19 23:10:43 +0200429 * But if process A is currently running sync_filesystems and then process B
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430 * calls sync_filesystems as well, process B will set all the s_need_sync_fs
431 * flags again, which will cause process A to resync everything. Fix that with
432 * a local mutex.
433 *
434 * (Fabian) Avoid sync_fs with clean fs & wait mode 0
435 */
436void sync_filesystems(int wait)
437{
438 struct super_block *sb;
Ingo Molnar353ab6e2006-03-26 01:37:12 -0800439 static DEFINE_MUTEX(mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440
Ingo Molnar353ab6e2006-03-26 01:37:12 -0800441 mutex_lock(&mutex); /* Could be down_interruptible */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700443 list_for_each_entry(sb, &super_blocks, s_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 if (!sb->s_op->sync_fs)
445 continue;
446 if (sb->s_flags & MS_RDONLY)
447 continue;
448 sb->s_need_sync_fs = 1;
449 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450
451restart:
Kirill Korotaev618f0632005-06-23 00:09:54 -0700452 list_for_each_entry(sb, &super_blocks, s_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 if (!sb->s_need_sync_fs)
454 continue;
455 sb->s_need_sync_fs = 0;
456 if (sb->s_flags & MS_RDONLY)
457 continue; /* hm. Was remounted r/o meanwhile */
458 sb->s_count++;
459 spin_unlock(&sb_lock);
460 down_read(&sb->s_umount);
461 if (sb->s_root && (wait || sb->s_dirt))
462 sb->s_op->sync_fs(sb, wait);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700463 up_read(&sb->s_umount);
464 /* restart only when sb is no longer on the list */
465 spin_lock(&sb_lock);
466 if (__put_super_and_need_restart(sb))
467 goto restart;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468 }
469 spin_unlock(&sb_lock);
Ingo Molnar353ab6e2006-03-26 01:37:12 -0800470 mutex_unlock(&mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471}
472
473/**
474 * get_super - get the superblock of a device
475 * @bdev: device to get the superblock for
476 *
477 * Scans the superblock list and finds the superblock of the file system
478 * mounted on the device given. %NULL is returned if no match is found.
479 */
480
481struct super_block * get_super(struct block_device *bdev)
482{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700483 struct super_block *sb;
484
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485 if (!bdev)
486 return NULL;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700487
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700489rescan:
490 list_for_each_entry(sb, &super_blocks, s_list) {
491 if (sb->s_bdev == bdev) {
492 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700494 down_read(&sb->s_umount);
495 if (sb->s_root)
496 return sb;
497 up_read(&sb->s_umount);
498 /* restart only when sb is no longer on the list */
499 spin_lock(&sb_lock);
500 if (__put_super_and_need_restart(sb))
501 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502 }
503 }
504 spin_unlock(&sb_lock);
505 return NULL;
506}
507
508EXPORT_SYMBOL(get_super);
509
510struct super_block * user_get_super(dev_t dev)
511{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700512 struct super_block *sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700515rescan:
516 list_for_each_entry(sb, &super_blocks, s_list) {
517 if (sb->s_dev == dev) {
518 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700520 down_read(&sb->s_umount);
521 if (sb->s_root)
522 return sb;
523 up_read(&sb->s_umount);
524 /* restart only when sb is no longer on the list */
525 spin_lock(&sb_lock);
526 if (__put_super_and_need_restart(sb))
527 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528 }
529 }
530 spin_unlock(&sb_lock);
531 return NULL;
532}
533
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534asmlinkage long sys_ustat(unsigned dev, struct ustat __user * ubuf)
535{
536 struct super_block *s;
537 struct ustat tmp;
538 struct kstatfs sbuf;
539 int err = -EINVAL;
540
541 s = user_get_super(new_decode_dev(dev));
542 if (s == NULL)
543 goto out;
David Howells726c3342006-06-23 02:02:58 -0700544 err = vfs_statfs(s->s_root, &sbuf);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545 drop_super(s);
546 if (err)
547 goto out;
548
549 memset(&tmp,0,sizeof(struct ustat));
550 tmp.f_tfree = sbuf.f_bfree;
551 tmp.f_tinode = sbuf.f_ffree;
552
553 err = copy_to_user(ubuf,&tmp,sizeof(struct ustat)) ? -EFAULT : 0;
554out:
555 return err;
556}
557
558/**
Randy Dunlapa6b91912008-03-19 17:01:00 -0700559 * mark_files_ro - mark all files read-only
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560 * @sb: superblock in question
561 *
Randy Dunlapa6b91912008-03-19 17:01:00 -0700562 * All files are marked read-only. We don't care about pending
563 * delete files so this should be used in 'force' mode only.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 */
565
566static void mark_files_ro(struct super_block *sb)
567{
568 struct file *f;
569
570 file_list_lock();
Eric Dumazet2f512012005-10-30 15:02:16 -0800571 list_for_each_entry(f, &sb->s_files, f_u.fu_list) {
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800572 if (S_ISREG(f->f_path.dentry->d_inode->i_mode) && file_count(f))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 f->f_mode &= ~FMODE_WRITE;
574 }
575 file_list_unlock();
576}
577
578/**
579 * do_remount_sb - asks filesystem to change mount options.
580 * @sb: superblock in question
581 * @flags: numeric part of options
582 * @data: the rest of options
583 * @force: whether or not to force the change
584 *
585 * Alters the mount options of a mounted file system.
586 */
587int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
588{
589 int retval;
590
David Howells93614012006-09-30 20:45:40 +0200591#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
593 return -EACCES;
David Howells93614012006-09-30 20:45:40 +0200594#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 if (flags & MS_RDONLY)
596 acct_auto_close(sb);
597 shrink_dcache_sb(sb);
598 fsync_super(sb);
599
600 /* If we are remounting RDONLY and current sb is read/write,
601 make sure there are no rw files opened */
602 if ((flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY)) {
603 if (force)
604 mark_files_ro(sb);
605 else if (!fs_may_remount_ro(sb))
606 return -EBUSY;
Jan Kara66191dc2008-02-08 04:22:13 -0800607 DQUOT_OFF(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 }
609
610 if (sb->s_op->remount_fs) {
611 lock_super(sb);
612 retval = sb->s_op->remount_fs(sb, &flags, data);
613 unlock_super(sb);
614 if (retval)
615 return retval;
616 }
617 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
618 return 0;
619}
620
621static void do_emergency_remount(unsigned long foo)
622{
623 struct super_block *sb;
624
625 spin_lock(&sb_lock);
626 list_for_each_entry(sb, &super_blocks, s_list) {
627 sb->s_count++;
628 spin_unlock(&sb_lock);
629 down_read(&sb->s_umount);
630 if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) {
631 /*
632 * ->remount_fs needs lock_kernel().
633 *
634 * What lock protects sb->s_flags??
635 */
636 lock_kernel();
637 do_remount_sb(sb, MS_RDONLY, NULL, 1);
638 unlock_kernel();
639 }
640 drop_super(sb);
641 spin_lock(&sb_lock);
642 }
643 spin_unlock(&sb_lock);
644 printk("Emergency Remount complete\n");
645}
646
647void emergency_remount(void)
648{
649 pdflush_operation(do_emergency_remount, 0);
650}
651
652/*
653 * Unnamed block devices are dummy devices used by virtual
654 * filesystems which don't use real block-devices. -- jrs
655 */
656
657static struct idr unnamed_dev_idr;
658static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
659
660int set_anon_super(struct super_block *s, void *data)
661{
662 int dev;
663 int error;
664
665 retry:
666 if (idr_pre_get(&unnamed_dev_idr, GFP_ATOMIC) == 0)
667 return -ENOMEM;
668 spin_lock(&unnamed_dev_lock);
669 error = idr_get_new(&unnamed_dev_idr, NULL, &dev);
670 spin_unlock(&unnamed_dev_lock);
671 if (error == -EAGAIN)
672 /* We raced and lost with another CPU. */
673 goto retry;
674 else if (error)
675 return -EAGAIN;
676
677 if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
678 spin_lock(&unnamed_dev_lock);
679 idr_remove(&unnamed_dev_idr, dev);
680 spin_unlock(&unnamed_dev_lock);
681 return -EMFILE;
682 }
683 s->s_dev = MKDEV(0, dev & MINORMASK);
684 return 0;
685}
686
687EXPORT_SYMBOL(set_anon_super);
688
689void kill_anon_super(struct super_block *sb)
690{
691 int slot = MINOR(sb->s_dev);
692
693 generic_shutdown_super(sb);
694 spin_lock(&unnamed_dev_lock);
695 idr_remove(&unnamed_dev_idr, slot);
696 spin_unlock(&unnamed_dev_lock);
697}
698
699EXPORT_SYMBOL(kill_anon_super);
700
701void __init unnamed_dev_init(void)
702{
703 idr_init(&unnamed_dev_idr);
704}
705
706void kill_litter_super(struct super_block *sb)
707{
708 if (sb->s_root)
709 d_genocide(sb->s_root);
710 kill_anon_super(sb);
711}
712
713EXPORT_SYMBOL(kill_litter_super);
714
David Howells93614012006-09-30 20:45:40 +0200715#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716static int set_bdev_super(struct super_block *s, void *data)
717{
718 s->s_bdev = data;
719 s->s_dev = s->s_bdev->bd_dev;
720 return 0;
721}
722
723static int test_bdev_super(struct super_block *s, void *data)
724{
725 return (void *)s->s_bdev == data;
726}
727
David Howells454e2392006-06-23 02:02:57 -0700728int get_sb_bdev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 int flags, const char *dev_name, void *data,
David Howells454e2392006-06-23 02:02:57 -0700730 int (*fill_super)(struct super_block *, void *, int),
731 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732{
733 struct block_device *bdev;
734 struct super_block *s;
735 int error = 0;
736
737 bdev = open_bdev_excl(dev_name, flags, fs_type);
738 if (IS_ERR(bdev))
David Howells454e2392006-06-23 02:02:57 -0700739 return PTR_ERR(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740
741 /*
742 * once the super is inserted into the list by sget, s_umount
743 * will protect the lockfs code from trying to start a snapshot
744 * while we are mounting
745 */
David Chinnerf73ca1b2007-01-10 23:15:41 -0800746 down(&bdev->bd_mount_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
David Chinnerf73ca1b2007-01-10 23:15:41 -0800748 up(&bdev->bd_mount_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700750 goto error_s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751
752 if (s->s_root) {
753 if ((flags ^ s->s_flags) & MS_RDONLY) {
754 up_write(&s->s_umount);
755 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700756 error = -EBUSY;
757 goto error_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758 }
David Howells454e2392006-06-23 02:02:57 -0700759
760 close_bdev_excl(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761 } else {
762 char b[BDEVNAME_SIZE];
763
764 s->s_flags = flags;
765 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
Pekka Enberge78c9a02006-01-08 01:03:39 -0800766 sb_set_blocksize(s, block_size(bdev));
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800767 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768 if (error) {
769 up_write(&s->s_umount);
770 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700771 goto error;
Greg Kroah-Hartmanfa675762006-02-22 09:39:02 -0800772 }
David Howells454e2392006-06-23 02:02:57 -0700773
774 s->s_flags |= MS_ACTIVE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 }
776
David Howells454e2392006-06-23 02:02:57 -0700777 return simple_set_mnt(mnt, s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778
David Howells454e2392006-06-23 02:02:57 -0700779error_s:
780 error = PTR_ERR(s);
781error_bdev:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782 close_bdev_excl(bdev);
David Howells454e2392006-06-23 02:02:57 -0700783error:
784 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785}
786
787EXPORT_SYMBOL(get_sb_bdev);
788
789void kill_block_super(struct super_block *sb)
790{
791 struct block_device *bdev = sb->s_bdev;
792
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793 generic_shutdown_super(sb);
794 sync_blockdev(bdev);
795 close_bdev_excl(bdev);
796}
797
798EXPORT_SYMBOL(kill_block_super);
David Howells93614012006-09-30 20:45:40 +0200799#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800
David Howells454e2392006-06-23 02:02:57 -0700801int get_sb_nodev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802 int flags, void *data,
David Howells454e2392006-06-23 02:02:57 -0700803 int (*fill_super)(struct super_block *, void *, int),
804 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805{
806 int error;
807 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
808
809 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700810 return PTR_ERR(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811
812 s->s_flags = flags;
813
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800814 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815 if (error) {
816 up_write(&s->s_umount);
817 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700818 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819 }
820 s->s_flags |= MS_ACTIVE;
David Howells454e2392006-06-23 02:02:57 -0700821 return simple_set_mnt(mnt, s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822}
823
824EXPORT_SYMBOL(get_sb_nodev);
825
826static int compare_single(struct super_block *s, void *p)
827{
828 return 1;
829}
830
David Howells454e2392006-06-23 02:02:57 -0700831int get_sb_single(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832 int flags, void *data,
David Howells454e2392006-06-23 02:02:57 -0700833 int (*fill_super)(struct super_block *, void *, int),
834 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835{
836 struct super_block *s;
837 int error;
838
839 s = sget(fs_type, compare_single, set_anon_super, NULL);
840 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700841 return PTR_ERR(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842 if (!s->s_root) {
843 s->s_flags = flags;
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800844 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845 if (error) {
846 up_write(&s->s_umount);
847 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700848 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849 }
850 s->s_flags |= MS_ACTIVE;
851 }
852 do_remount_sb(s, flags, data, 0);
David Howells454e2392006-06-23 02:02:57 -0700853 return simple_set_mnt(mnt, s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854}
855
856EXPORT_SYMBOL(get_sb_single);
857
858struct vfsmount *
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400859vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861 struct vfsmount *mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862 char *secdata = NULL;
David Howells454e2392006-06-23 02:02:57 -0700863 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864
865 if (!type)
866 return ERR_PTR(-ENODEV);
867
David Howells454e2392006-06-23 02:02:57 -0700868 error = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 mnt = alloc_vfsmnt(name);
870 if (!mnt)
871 goto out;
872
Eric Parise0007522008-03-05 10:31:54 -0500873 if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874 secdata = alloc_secdata();
David Howells454e2392006-06-23 02:02:57 -0700875 if (!secdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876 goto out_mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877
Eric Parise0007522008-03-05 10:31:54 -0500878 error = security_sb_copy_data(data, secdata);
David Howells454e2392006-06-23 02:02:57 -0700879 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880 goto out_free_secdata;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881 }
882
David Howells454e2392006-06-23 02:02:57 -0700883 error = type->get_sb(type, flags, name, data, mnt);
884 if (error < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885 goto out_free_secdata;
Lee Schermerhornb4c07bc2007-07-15 23:40:54 -0700886 BUG_ON(!mnt->mnt_sb);
David Howells454e2392006-06-23 02:02:57 -0700887
888 error = security_sb_kern_mount(mnt->mnt_sb, secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889 if (error)
890 goto out_sb;
David Howells454e2392006-06-23 02:02:57 -0700891
892 mnt->mnt_mountpoint = mnt->mnt_root;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893 mnt->mnt_parent = mnt;
David Howells454e2392006-06-23 02:02:57 -0700894 up_write(&mnt->mnt_sb->s_umount);
Gerald Schaefer8680e222005-06-21 17:15:16 -0700895 free_secdata(secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896 return mnt;
897out_sb:
David Howells454e2392006-06-23 02:02:57 -0700898 dput(mnt->mnt_root);
899 up_write(&mnt->mnt_sb->s_umount);
900 deactivate_super(mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901out_free_secdata:
902 free_secdata(secdata);
903out_mnt:
904 free_vfsmnt(mnt);
905out:
David Howells454e2392006-06-23 02:02:57 -0700906 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907}
908
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400909EXPORT_SYMBOL_GPL(vfs_kern_mount);
910
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700911static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype)
912{
913 int err;
914 const char *subtype = strchr(fstype, '.');
915 if (subtype) {
916 subtype++;
917 err = -EINVAL;
918 if (!subtype[0])
919 goto err;
920 } else
921 subtype = "";
922
923 mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL);
924 err = -ENOMEM;
925 if (!mnt->mnt_sb->s_subtype)
926 goto err;
927 return mnt;
928
929 err:
930 mntput(mnt);
931 return ERR_PTR(err);
932}
933
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400934struct vfsmount *
935do_kern_mount(const char *fstype, int flags, const char *name, void *data)
936{
937 struct file_system_type *type = get_fs_type(fstype);
938 struct vfsmount *mnt;
939 if (!type)
940 return ERR_PTR(-ENODEV);
941 mnt = vfs_kern_mount(type, flags, name, data);
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700942 if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) &&
943 !mnt->mnt_sb->s_subtype)
944 mnt = fs_set_subtype(mnt, fstype);
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400945 put_filesystem(type);
946 return mnt;
947}
948
Pavel Emelyanov8bf97252007-10-18 23:40:02 -0700949struct vfsmount *kern_mount_data(struct file_system_type *type, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950{
Pavel Emelyanov8bf97252007-10-18 23:40:02 -0700951 return vfs_kern_mount(type, MS_KERNMOUNT, type->name, data);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952}
953
Pavel Emelyanov8bf97252007-10-18 23:40:02 -0700954EXPORT_SYMBOL_GPL(kern_mount_data);