blob: 01d5c40e9119341f71c632a1f6052248b9860808 [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>
42
43
Linus Torvalds1da177e2005-04-16 15:20:36 -070044LIST_HEAD(super_blocks);
45DEFINE_SPINLOCK(sb_lock);
46
47/**
48 * alloc_super - create new superblock
Henrik Kretzschmarfe2bbc42006-09-06 00:03:41 -070049 * @type: filesystem type superblock should belong to
Linus Torvalds1da177e2005-04-16 15:20:36 -070050 *
51 * Allocates and initializes a new &struct super_block. alloc_super()
52 * returns a pointer new superblock or %NULL if allocation had failed.
53 */
Ingo Molnarcf516242006-07-03 00:25:27 -070054static struct super_block *alloc_super(struct file_system_type *type)
Linus Torvalds1da177e2005-04-16 15:20:36 -070055{
Oliver Neukum11b0b5a2006-03-25 03:08:13 -080056 struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER);
Linus Torvalds1da177e2005-04-16 15:20:36 -070057 static struct super_operations default_op;
58
59 if (s) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070060 if (security_sb_alloc(s)) {
61 kfree(s);
62 s = NULL;
63 goto out;
64 }
65 INIT_LIST_HEAD(&s->s_dirty);
66 INIT_LIST_HEAD(&s->s_io);
Ken Chen0e0f4fc2007-10-16 23:30:38 -070067 INIT_LIST_HEAD(&s->s_more_io);
Linus Torvalds1da177e2005-04-16 15:20:36 -070068 INIT_LIST_HEAD(&s->s_files);
69 INIT_LIST_HEAD(&s->s_instances);
70 INIT_HLIST_HEAD(&s->s_anon);
71 INIT_LIST_HEAD(&s->s_inodes);
72 init_rwsem(&s->s_umount);
Ingo Molnar7892f2f2006-01-09 15:59:25 -080073 mutex_init(&s->s_lock);
Arjan van de Ven897c6ff2006-07-03 00:25:28 -070074 lockdep_set_class(&s->s_umount, &type->s_umount_key);
Ingo Molnarcf516242006-07-03 00:25:27 -070075 /*
76 * The locking rules for s_lock are up to the
77 * filesystem. For example ext3fs has different
78 * lock ordering than usbfs:
79 */
80 lockdep_set_class(&s->s_lock, &type->s_lock_key);
Linus Torvalds1da177e2005-04-16 15:20:36 -070081 down_write(&s->s_umount);
82 s->s_count = S_BIAS;
83 atomic_set(&s->s_active, 1);
Arjan van de Vena11f3a02006-03-23 03:00:33 -080084 mutex_init(&s->s_vfs_rename_mutex);
Ingo Molnard3be9152006-03-23 03:00:29 -080085 mutex_init(&s->s_dquot.dqio_mutex);
86 mutex_init(&s->s_dquot.dqonoff_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -070087 init_rwsem(&s->s_dquot.dqptr_sem);
88 init_waitqueue_head(&s->s_wait_unfrozen);
89 s->s_maxbytes = MAX_NON_LFS;
90 s->dq_op = sb_dquot_ops;
91 s->s_qcop = sb_quotactl_ops;
92 s->s_op = &default_op;
93 s->s_time_gran = 1000000000;
94 }
95out:
96 return s;
97}
98
99/**
100 * destroy_super - frees a superblock
101 * @s: superblock to free
102 *
103 * Frees a superblock.
104 */
105static inline void destroy_super(struct super_block *s)
106{
107 security_sb_free(s);
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700108 kfree(s->s_subtype);
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800109 kfree(s->s_options);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110 kfree(s);
111}
112
113/* Superblock refcounting */
114
115/*
116 * Drop a superblock's refcount. Returns non-zero if the superblock was
117 * destroyed. The caller must hold sb_lock.
118 */
119int __put_super(struct super_block *sb)
120{
121 int ret = 0;
122
123 if (!--sb->s_count) {
124 destroy_super(sb);
125 ret = 1;
126 }
127 return ret;
128}
129
130/*
131 * Drop a superblock's refcount.
132 * Returns non-zero if the superblock is about to be destroyed and
133 * at least is already removed from super_blocks list, so if we are
134 * making a loop through super blocks then we need to restart.
135 * The caller must hold sb_lock.
136 */
137int __put_super_and_need_restart(struct super_block *sb)
138{
139 /* check for race with generic_shutdown_super() */
140 if (list_empty(&sb->s_list)) {
141 /* super block is removed, need to restart... */
142 __put_super(sb);
143 return 1;
144 }
145 /* can't be the last, since s_list is still in use */
146 sb->s_count--;
147 BUG_ON(sb->s_count == 0);
148 return 0;
149}
150
151/**
152 * put_super - drop a temporary reference to superblock
153 * @sb: superblock in question
154 *
155 * Drops a temporary reference, frees superblock if there's no
156 * references left.
157 */
158static void put_super(struct super_block *sb)
159{
160 spin_lock(&sb_lock);
161 __put_super(sb);
162 spin_unlock(&sb_lock);
163}
164
165
166/**
167 * deactivate_super - drop an active reference to superblock
168 * @s: superblock to deactivate
169 *
170 * Drops an active reference to superblock, acquiring a temprory one if
171 * there is no active references left. In that case we lock superblock,
172 * tell fs driver to shut it down and drop the temporary reference we
173 * had just acquired.
174 */
175void deactivate_super(struct super_block *s)
176{
177 struct file_system_type *fs = s->s_type;
178 if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
179 s->s_count -= S_BIAS-1;
180 spin_unlock(&sb_lock);
Al Viro7b7b1ac2005-11-07 17:13:39 -0500181 DQUOT_OFF(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 down_write(&s->s_umount);
183 fs->kill_sb(s);
184 put_filesystem(fs);
185 put_super(s);
186 }
187}
188
189EXPORT_SYMBOL(deactivate_super);
190
191/**
192 * grab_super - acquire an active reference
193 * @s: reference we are trying to make active
194 *
195 * Tries to acquire an active reference. grab_super() is used when we
196 * had just found a superblock in super_blocks or fs_type->fs_supers
197 * and want to turn it into a full-blown active reference. grab_super()
198 * is called with sb_lock held and drops it. Returns 1 in case of
199 * success, 0 if we had failed (superblock contents was already dead or
200 * dying when grab_super() had been called).
201 */
Josh Triplett9c4dbee2006-09-29 01:59:29 -0700202static int grab_super(struct super_block *s) __releases(sb_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203{
204 s->s_count++;
205 spin_unlock(&sb_lock);
206 down_write(&s->s_umount);
207 if (s->s_root) {
208 spin_lock(&sb_lock);
209 if (s->s_count > S_BIAS) {
210 atomic_inc(&s->s_active);
211 s->s_count--;
212 spin_unlock(&sb_lock);
213 return 1;
214 }
215 spin_unlock(&sb_lock);
216 }
217 up_write(&s->s_umount);
218 put_super(s);
219 yield();
220 return 0;
221}
222
David Howellscf9a2ae2006-08-29 19:05:54 +0100223/*
Al Viro914e2632006-10-18 13:55:46 -0400224 * Superblock locking. We really ought to get rid of these two.
225 */
226void lock_super(struct super_block * sb)
227{
228 get_fs_excl();
229 mutex_lock(&sb->s_lock);
230}
231
232void unlock_super(struct super_block * sb)
233{
234 put_fs_excl();
235 mutex_unlock(&sb->s_lock);
236}
237
238EXPORT_SYMBOL(lock_super);
239EXPORT_SYMBOL(unlock_super);
240
241/*
David Howellscf9a2ae2006-08-29 19:05:54 +0100242 * Write out and wait upon all dirty data associated with this
243 * superblock. Filesystem data as well as the underlying block
244 * device. Takes the superblock lock. Requires a second blkdev
245 * flush by the caller to complete the operation.
246 */
247void __fsync_super(struct super_block *sb)
248{
249 sync_inodes_sb(sb, 0);
250 DQUOT_SYNC(sb);
251 lock_super(sb);
252 if (sb->s_dirt && sb->s_op->write_super)
253 sb->s_op->write_super(sb);
254 unlock_super(sb);
255 if (sb->s_op->sync_fs)
256 sb->s_op->sync_fs(sb, 1);
257 sync_blockdev(sb->s_bdev);
258 sync_inodes_sb(sb, 1);
259}
260
261/*
262 * Write out and wait upon all dirty data associated with this
263 * superblock. Filesystem data as well as the underlying block
264 * device. Takes the superblock lock.
265 */
266int fsync_super(struct super_block *sb)
267{
268 __fsync_super(sb);
269 return sync_blockdev(sb->s_bdev);
270}
271
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272/**
273 * generic_shutdown_super - common helper for ->kill_sb()
274 * @sb: superblock to kill
275 *
276 * generic_shutdown_super() does all fs-independent work on superblock
277 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
278 * that need destruction out of superblock, call generic_shutdown_super()
279 * and release aforementioned objects. Note: dentries and inodes _are_
280 * taken care of and do not need specific handling.
David Howellsc636ebd2006-10-11 01:22:19 -0700281 *
282 * Upon calling this function, the filesystem may no longer alter or
283 * rearrange the set of dentries belonging to this super_block, nor may it
284 * change the attachments of dentries to inodes.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285 */
286void generic_shutdown_super(struct super_block *sb)
287{
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -0800288 const struct super_operations *sop = sb->s_op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289
David Howellsc636ebd2006-10-11 01:22:19 -0700290 if (sb->s_root) {
291 shrink_dcache_for_umount(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292 fsync_super(sb);
293 lock_super(sb);
294 sb->s_flags &= ~MS_ACTIVE;
295 /* bad name - it should be evict_inodes() */
296 invalidate_inodes(sb);
297 lock_kernel();
298
299 if (sop->write_super && sb->s_dirt)
300 sop->write_super(sb);
301 if (sop->put_super)
302 sop->put_super(sb);
303
304 /* Forget any remaining inodes */
305 if (invalidate_inodes(sb)) {
Dave Jones7b4fe292006-02-07 12:58:48 -0800306 printk("VFS: Busy inodes after unmount of %s. "
307 "Self-destruct in 5 seconds. Have a nice day...\n",
308 sb->s_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309 }
310
311 unlock_kernel();
312 unlock_super(sb);
313 }
314 spin_lock(&sb_lock);
315 /* should be initialized for __put_super_and_need_restart() */
316 list_del_init(&sb->s_list);
317 list_del(&sb->s_instances);
318 spin_unlock(&sb_lock);
319 up_write(&sb->s_umount);
320}
321
322EXPORT_SYMBOL(generic_shutdown_super);
323
324/**
325 * sget - find or create a superblock
326 * @type: filesystem type superblock should belong to
327 * @test: comparison callback
328 * @set: setup callback
329 * @data: argument to each of them
330 */
331struct super_block *sget(struct file_system_type *type,
332 int (*test)(struct super_block *,void *),
333 int (*set)(struct super_block *,void *),
334 void *data)
335{
336 struct super_block *s = NULL;
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700337 struct super_block *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338 int err;
339
340retry:
341 spin_lock(&sb_lock);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700342 if (test) {
343 list_for_each_entry(old, &type->fs_supers, s_instances) {
344 if (!test(old, data))
345 continue;
346 if (!grab_super(old))
347 goto retry;
348 if (s)
349 destroy_super(s);
350 return old;
351 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352 }
353 if (!s) {
354 spin_unlock(&sb_lock);
Ingo Molnarcf516242006-07-03 00:25:27 -0700355 s = alloc_super(type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356 if (!s)
357 return ERR_PTR(-ENOMEM);
358 goto retry;
359 }
360
361 err = set(s, data);
362 if (err) {
363 spin_unlock(&sb_lock);
364 destroy_super(s);
365 return ERR_PTR(err);
366 }
367 s->s_type = type;
368 strlcpy(s->s_id, type->name, sizeof(s->s_id));
369 list_add_tail(&s->s_list, &super_blocks);
370 list_add(&s->s_instances, &type->fs_supers);
371 spin_unlock(&sb_lock);
372 get_filesystem(type);
373 return s;
374}
375
376EXPORT_SYMBOL(sget);
377
378void drop_super(struct super_block *sb)
379{
380 up_read(&sb->s_umount);
381 put_super(sb);
382}
383
384EXPORT_SYMBOL(drop_super);
385
386static inline void write_super(struct super_block *sb)
387{
388 lock_super(sb);
389 if (sb->s_root && sb->s_dirt)
390 if (sb->s_op->write_super)
391 sb->s_op->write_super(sb);
392 unlock_super(sb);
393}
394
395/*
396 * Note: check the dirty flag before waiting, so we don't
397 * hold up the sync while mounting a device. (The newly
398 * mounted device won't need syncing.)
399 */
400void sync_supers(void)
401{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700402 struct super_block *sb;
403
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700405restart:
406 list_for_each_entry(sb, &super_blocks, s_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 if (sb->s_dirt) {
408 sb->s_count++;
409 spin_unlock(&sb_lock);
410 down_read(&sb->s_umount);
411 write_super(sb);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700412 up_read(&sb->s_umount);
413 spin_lock(&sb_lock);
414 if (__put_super_and_need_restart(sb))
415 goto restart;
416 }
417 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418 spin_unlock(&sb_lock);
419}
420
421/*
Robert P. J. Day3a4fa0a2007-10-19 23:10:43 +0200422 * Call the ->sync_fs super_op against all filesystems which are r/w and
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 * which implement it.
424 *
425 * This operation is careful to avoid the livelock which could easily happen
426 * if two or more filesystems are being continuously dirtied. s_need_sync_fs
427 * is used only here. We set it against all filesystems and then clear it as
428 * we sync them. So redirtied filesystems are skipped.
429 *
Robert P. J. Day3a4fa0a2007-10-19 23:10:43 +0200430 * But if process A is currently running sync_filesystems and then process B
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431 * calls sync_filesystems as well, process B will set all the s_need_sync_fs
432 * flags again, which will cause process A to resync everything. Fix that with
433 * a local mutex.
434 *
435 * (Fabian) Avoid sync_fs with clean fs & wait mode 0
436 */
437void sync_filesystems(int wait)
438{
439 struct super_block *sb;
Ingo Molnar353ab6e2006-03-26 01:37:12 -0800440 static DEFINE_MUTEX(mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441
Ingo Molnar353ab6e2006-03-26 01:37:12 -0800442 mutex_lock(&mutex); /* Could be down_interruptible */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700444 list_for_each_entry(sb, &super_blocks, s_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 if (!sb->s_op->sync_fs)
446 continue;
447 if (sb->s_flags & MS_RDONLY)
448 continue;
449 sb->s_need_sync_fs = 1;
450 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451
452restart:
Kirill Korotaev618f0632005-06-23 00:09:54 -0700453 list_for_each_entry(sb, &super_blocks, s_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454 if (!sb->s_need_sync_fs)
455 continue;
456 sb->s_need_sync_fs = 0;
457 if (sb->s_flags & MS_RDONLY)
458 continue; /* hm. Was remounted r/o meanwhile */
459 sb->s_count++;
460 spin_unlock(&sb_lock);
461 down_read(&sb->s_umount);
462 if (sb->s_root && (wait || sb->s_dirt))
463 sb->s_op->sync_fs(sb, wait);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700464 up_read(&sb->s_umount);
465 /* restart only when sb is no longer on the list */
466 spin_lock(&sb_lock);
467 if (__put_super_and_need_restart(sb))
468 goto restart;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469 }
470 spin_unlock(&sb_lock);
Ingo Molnar353ab6e2006-03-26 01:37:12 -0800471 mutex_unlock(&mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472}
473
474/**
475 * get_super - get the superblock of a device
476 * @bdev: device to get the superblock for
477 *
478 * Scans the superblock list and finds the superblock of the file system
479 * mounted on the device given. %NULL is returned if no match is found.
480 */
481
482struct super_block * get_super(struct block_device *bdev)
483{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700484 struct super_block *sb;
485
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486 if (!bdev)
487 return NULL;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700488
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700490rescan:
491 list_for_each_entry(sb, &super_blocks, s_list) {
492 if (sb->s_bdev == bdev) {
493 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700495 down_read(&sb->s_umount);
496 if (sb->s_root)
497 return sb;
498 up_read(&sb->s_umount);
499 /* restart only when sb is no longer on the list */
500 spin_lock(&sb_lock);
501 if (__put_super_and_need_restart(sb))
502 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503 }
504 }
505 spin_unlock(&sb_lock);
506 return NULL;
507}
508
509EXPORT_SYMBOL(get_super);
510
511struct super_block * user_get_super(dev_t dev)
512{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700513 struct super_block *sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700516rescan:
517 list_for_each_entry(sb, &super_blocks, s_list) {
518 if (sb->s_dev == dev) {
519 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700521 down_read(&sb->s_umount);
522 if (sb->s_root)
523 return sb;
524 up_read(&sb->s_umount);
525 /* restart only when sb is no longer on the list */
526 spin_lock(&sb_lock);
527 if (__put_super_and_need_restart(sb))
528 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529 }
530 }
531 spin_unlock(&sb_lock);
532 return NULL;
533}
534
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535asmlinkage long sys_ustat(unsigned dev, struct ustat __user * ubuf)
536{
537 struct super_block *s;
538 struct ustat tmp;
539 struct kstatfs sbuf;
540 int err = -EINVAL;
541
542 s = user_get_super(new_decode_dev(dev));
543 if (s == NULL)
544 goto out;
David Howells726c3342006-06-23 02:02:58 -0700545 err = vfs_statfs(s->s_root, &sbuf);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546 drop_super(s);
547 if (err)
548 goto out;
549
550 memset(&tmp,0,sizeof(struct ustat));
551 tmp.f_tfree = sbuf.f_bfree;
552 tmp.f_tinode = sbuf.f_ffree;
553
554 err = copy_to_user(ubuf,&tmp,sizeof(struct ustat)) ? -EFAULT : 0;
555out:
556 return err;
557}
558
559/**
Randy Dunlapa6b91912008-03-19 17:01:00 -0700560 * mark_files_ro - mark all files read-only
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561 * @sb: superblock in question
562 *
Randy Dunlapa6b91912008-03-19 17:01:00 -0700563 * All files are marked read-only. We don't care about pending
564 * delete files so this should be used in 'force' mode only.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565 */
566
567static void mark_files_ro(struct super_block *sb)
568{
569 struct file *f;
570
Dave Hansen49e0d022008-02-15 14:37:32 -0800571retry:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572 file_list_lock();
Eric Dumazet2f512012005-10-30 15:02:16 -0800573 list_for_each_entry(f, &sb->s_files, f_u.fu_list) {
Dave Hansen49e0d022008-02-15 14:37:32 -0800574 struct vfsmount *mnt;
575 if (!S_ISREG(f->f_path.dentry->d_inode->i_mode))
576 continue;
577 if (!file_count(f))
578 continue;
579 if (!(f->f_mode & FMODE_WRITE))
580 continue;
581 f->f_mode &= ~FMODE_WRITE;
582 mnt = mntget(f->f_path.mnt);
583 file_list_unlock();
584 /*
585 * This can sleep, so we can't hold
586 * the file_list_lock() spinlock.
587 */
588 mnt_drop_write(mnt);
589 mntput(mnt);
590 goto retry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 }
592 file_list_unlock();
593}
594
595/**
596 * do_remount_sb - asks filesystem to change mount options.
597 * @sb: superblock in question
598 * @flags: numeric part of options
599 * @data: the rest of options
600 * @force: whether or not to force the change
601 *
602 * Alters the mount options of a mounted file system.
603 */
604int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
605{
606 int retval;
607
David Howells93614012006-09-30 20:45:40 +0200608#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
610 return -EACCES;
David Howells93614012006-09-30 20:45:40 +0200611#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612 if (flags & MS_RDONLY)
613 acct_auto_close(sb);
614 shrink_dcache_sb(sb);
615 fsync_super(sb);
616
617 /* If we are remounting RDONLY and current sb is read/write,
618 make sure there are no rw files opened */
619 if ((flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY)) {
620 if (force)
621 mark_files_ro(sb);
622 else if (!fs_may_remount_ro(sb))
623 return -EBUSY;
Jan Kara66191dc2008-02-08 04:22:13 -0800624 DQUOT_OFF(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625 }
626
627 if (sb->s_op->remount_fs) {
628 lock_super(sb);
629 retval = sb->s_op->remount_fs(sb, &flags, data);
630 unlock_super(sb);
631 if (retval)
632 return retval;
633 }
634 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
635 return 0;
636}
637
638static void do_emergency_remount(unsigned long foo)
639{
640 struct super_block *sb;
641
642 spin_lock(&sb_lock);
643 list_for_each_entry(sb, &super_blocks, s_list) {
644 sb->s_count++;
645 spin_unlock(&sb_lock);
646 down_read(&sb->s_umount);
647 if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) {
648 /*
649 * ->remount_fs needs lock_kernel().
650 *
651 * What lock protects sb->s_flags??
652 */
653 lock_kernel();
654 do_remount_sb(sb, MS_RDONLY, NULL, 1);
655 unlock_kernel();
656 }
657 drop_super(sb);
658 spin_lock(&sb_lock);
659 }
660 spin_unlock(&sb_lock);
661 printk("Emergency Remount complete\n");
662}
663
664void emergency_remount(void)
665{
666 pdflush_operation(do_emergency_remount, 0);
667}
668
669/*
670 * Unnamed block devices are dummy devices used by virtual
671 * filesystems which don't use real block-devices. -- jrs
672 */
673
674static struct idr unnamed_dev_idr;
675static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
676
677int set_anon_super(struct super_block *s, void *data)
678{
679 int dev;
680 int error;
681
682 retry:
683 if (idr_pre_get(&unnamed_dev_idr, GFP_ATOMIC) == 0)
684 return -ENOMEM;
685 spin_lock(&unnamed_dev_lock);
686 error = idr_get_new(&unnamed_dev_idr, NULL, &dev);
687 spin_unlock(&unnamed_dev_lock);
688 if (error == -EAGAIN)
689 /* We raced and lost with another CPU. */
690 goto retry;
691 else if (error)
692 return -EAGAIN;
693
694 if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
695 spin_lock(&unnamed_dev_lock);
696 idr_remove(&unnamed_dev_idr, dev);
697 spin_unlock(&unnamed_dev_lock);
698 return -EMFILE;
699 }
700 s->s_dev = MKDEV(0, dev & MINORMASK);
701 return 0;
702}
703
704EXPORT_SYMBOL(set_anon_super);
705
706void kill_anon_super(struct super_block *sb)
707{
708 int slot = MINOR(sb->s_dev);
709
710 generic_shutdown_super(sb);
711 spin_lock(&unnamed_dev_lock);
712 idr_remove(&unnamed_dev_idr, slot);
713 spin_unlock(&unnamed_dev_lock);
714}
715
716EXPORT_SYMBOL(kill_anon_super);
717
718void __init unnamed_dev_init(void)
719{
720 idr_init(&unnamed_dev_idr);
721}
722
723void kill_litter_super(struct super_block *sb)
724{
725 if (sb->s_root)
726 d_genocide(sb->s_root);
727 kill_anon_super(sb);
728}
729
730EXPORT_SYMBOL(kill_litter_super);
731
David Howells93614012006-09-30 20:45:40 +0200732#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733static int set_bdev_super(struct super_block *s, void *data)
734{
735 s->s_bdev = data;
736 s->s_dev = s->s_bdev->bd_dev;
737 return 0;
738}
739
740static int test_bdev_super(struct super_block *s, void *data)
741{
742 return (void *)s->s_bdev == data;
743}
744
David Howells454e2392006-06-23 02:02:57 -0700745int get_sb_bdev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746 int flags, const char *dev_name, void *data,
David Howells454e2392006-06-23 02:02:57 -0700747 int (*fill_super)(struct super_block *, void *, int),
748 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749{
750 struct block_device *bdev;
751 struct super_block *s;
752 int error = 0;
753
754 bdev = open_bdev_excl(dev_name, flags, fs_type);
755 if (IS_ERR(bdev))
David Howells454e2392006-06-23 02:02:57 -0700756 return PTR_ERR(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757
758 /*
759 * once the super is inserted into the list by sget, s_umount
760 * will protect the lockfs code from trying to start a snapshot
761 * while we are mounting
762 */
David Chinnerf73ca1b2007-01-10 23:15:41 -0800763 down(&bdev->bd_mount_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764 s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
David Chinnerf73ca1b2007-01-10 23:15:41 -0800765 up(&bdev->bd_mount_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700767 goto error_s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768
769 if (s->s_root) {
770 if ((flags ^ s->s_flags) & MS_RDONLY) {
771 up_write(&s->s_umount);
772 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700773 error = -EBUSY;
774 goto error_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 }
David Howells454e2392006-06-23 02:02:57 -0700776
777 close_bdev_excl(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778 } else {
779 char b[BDEVNAME_SIZE];
780
781 s->s_flags = flags;
782 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
Pekka Enberge78c9a02006-01-08 01:03:39 -0800783 sb_set_blocksize(s, block_size(bdev));
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800784 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785 if (error) {
786 up_write(&s->s_umount);
787 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700788 goto error;
Greg Kroah-Hartmanfa675762006-02-22 09:39:02 -0800789 }
David Howells454e2392006-06-23 02:02:57 -0700790
791 s->s_flags |= MS_ACTIVE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792 }
793
David Howells454e2392006-06-23 02:02:57 -0700794 return simple_set_mnt(mnt, s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795
David Howells454e2392006-06-23 02:02:57 -0700796error_s:
797 error = PTR_ERR(s);
798error_bdev:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799 close_bdev_excl(bdev);
David Howells454e2392006-06-23 02:02:57 -0700800error:
801 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802}
803
804EXPORT_SYMBOL(get_sb_bdev);
805
806void kill_block_super(struct super_block *sb)
807{
808 struct block_device *bdev = sb->s_bdev;
809
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810 generic_shutdown_super(sb);
811 sync_blockdev(bdev);
812 close_bdev_excl(bdev);
813}
814
815EXPORT_SYMBOL(kill_block_super);
David Howells93614012006-09-30 20:45:40 +0200816#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817
David Howells454e2392006-06-23 02:02:57 -0700818int get_sb_nodev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819 int flags, void *data,
David Howells454e2392006-06-23 02:02:57 -0700820 int (*fill_super)(struct super_block *, void *, int),
821 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822{
823 int error;
824 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
825
826 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700827 return PTR_ERR(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828
829 s->s_flags = flags;
830
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800831 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832 if (error) {
833 up_write(&s->s_umount);
834 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700835 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836 }
837 s->s_flags |= MS_ACTIVE;
David Howells454e2392006-06-23 02:02:57 -0700838 return simple_set_mnt(mnt, s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839}
840
841EXPORT_SYMBOL(get_sb_nodev);
842
843static int compare_single(struct super_block *s, void *p)
844{
845 return 1;
846}
847
David Howells454e2392006-06-23 02:02:57 -0700848int get_sb_single(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849 int flags, void *data,
David Howells454e2392006-06-23 02:02:57 -0700850 int (*fill_super)(struct super_block *, void *, int),
851 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852{
853 struct super_block *s;
854 int error;
855
856 s = sget(fs_type, compare_single, set_anon_super, NULL);
857 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700858 return PTR_ERR(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 if (!s->s_root) {
860 s->s_flags = flags;
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800861 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862 if (error) {
863 up_write(&s->s_umount);
864 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700865 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866 }
867 s->s_flags |= MS_ACTIVE;
868 }
869 do_remount_sb(s, flags, data, 0);
David Howells454e2392006-06-23 02:02:57 -0700870 return simple_set_mnt(mnt, s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871}
872
873EXPORT_SYMBOL(get_sb_single);
874
875struct vfsmount *
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400876vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878 struct vfsmount *mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879 char *secdata = NULL;
David Howells454e2392006-06-23 02:02:57 -0700880 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881
882 if (!type)
883 return ERR_PTR(-ENODEV);
884
David Howells454e2392006-06-23 02:02:57 -0700885 error = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 mnt = alloc_vfsmnt(name);
887 if (!mnt)
888 goto out;
889
Eric Parise0007522008-03-05 10:31:54 -0500890 if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 secdata = alloc_secdata();
David Howells454e2392006-06-23 02:02:57 -0700892 if (!secdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893 goto out_mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894
Eric Parise0007522008-03-05 10:31:54 -0500895 error = security_sb_copy_data(data, secdata);
David Howells454e2392006-06-23 02:02:57 -0700896 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897 goto out_free_secdata;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 }
899
David Howells454e2392006-06-23 02:02:57 -0700900 error = type->get_sb(type, flags, name, data, mnt);
901 if (error < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 goto out_free_secdata;
Lee Schermerhornb4c07bc2007-07-15 23:40:54 -0700903 BUG_ON(!mnt->mnt_sb);
David Howells454e2392006-06-23 02:02:57 -0700904
905 error = security_sb_kern_mount(mnt->mnt_sb, secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906 if (error)
907 goto out_sb;
David Howells454e2392006-06-23 02:02:57 -0700908
909 mnt->mnt_mountpoint = mnt->mnt_root;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910 mnt->mnt_parent = mnt;
David Howells454e2392006-06-23 02:02:57 -0700911 up_write(&mnt->mnt_sb->s_umount);
Gerald Schaefer8680e222005-06-21 17:15:16 -0700912 free_secdata(secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913 return mnt;
914out_sb:
David Howells454e2392006-06-23 02:02:57 -0700915 dput(mnt->mnt_root);
916 up_write(&mnt->mnt_sb->s_umount);
917 deactivate_super(mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918out_free_secdata:
919 free_secdata(secdata);
920out_mnt:
921 free_vfsmnt(mnt);
922out:
David Howells454e2392006-06-23 02:02:57 -0700923 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924}
925
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400926EXPORT_SYMBOL_GPL(vfs_kern_mount);
927
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700928static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype)
929{
930 int err;
931 const char *subtype = strchr(fstype, '.');
932 if (subtype) {
933 subtype++;
934 err = -EINVAL;
935 if (!subtype[0])
936 goto err;
937 } else
938 subtype = "";
939
940 mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL);
941 err = -ENOMEM;
942 if (!mnt->mnt_sb->s_subtype)
943 goto err;
944 return mnt;
945
946 err:
947 mntput(mnt);
948 return ERR_PTR(err);
949}
950
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400951struct vfsmount *
952do_kern_mount(const char *fstype, int flags, const char *name, void *data)
953{
954 struct file_system_type *type = get_fs_type(fstype);
955 struct vfsmount *mnt;
956 if (!type)
957 return ERR_PTR(-ENODEV);
958 mnt = vfs_kern_mount(type, flags, name, data);
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700959 if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) &&
960 !mnt->mnt_sb->s_subtype)
961 mnt = fs_set_subtype(mnt, fstype);
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400962 put_filesystem(type);
963 return mnt;
964}
Al Viro8a4e98d2008-02-24 01:43:03 -0500965EXPORT_SYMBOL_GPL(do_kern_mount);
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400966
Pavel Emelyanov8bf97252007-10-18 23:40:02 -0700967struct vfsmount *kern_mount_data(struct file_system_type *type, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968{
Pavel Emelyanov8bf97252007-10-18 23:40:02 -0700969 return vfs_kern_mount(type, MS_KERNMOUNT, type->name, data);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970}
971
Pavel Emelyanov8bf97252007-10-18 23:40:02 -0700972EXPORT_SYMBOL_GPL(kern_mount_data);