blob: feaae7eeaffdfb2eaba44b17c802f24a0bc3fd4b [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:
18 * Torbjörn Lindh (torbjorn.lindh@gopta.se), April 14, 1996.
19 * 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 Molnar7892f2f42006-01-09 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);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108 kfree(s);
109}
110
111/* Superblock refcounting */
112
113/*
114 * Drop a superblock's refcount. Returns non-zero if the superblock was
115 * destroyed. The caller must hold sb_lock.
116 */
117int __put_super(struct super_block *sb)
118{
119 int ret = 0;
120
121 if (!--sb->s_count) {
122 destroy_super(sb);
123 ret = 1;
124 }
125 return ret;
126}
127
128/*
129 * Drop a superblock's refcount.
130 * Returns non-zero if the superblock is about to be destroyed and
131 * at least is already removed from super_blocks list, so if we are
132 * making a loop through super blocks then we need to restart.
133 * The caller must hold sb_lock.
134 */
135int __put_super_and_need_restart(struct super_block *sb)
136{
137 /* check for race with generic_shutdown_super() */
138 if (list_empty(&sb->s_list)) {
139 /* super block is removed, need to restart... */
140 __put_super(sb);
141 return 1;
142 }
143 /* can't be the last, since s_list is still in use */
144 sb->s_count--;
145 BUG_ON(sb->s_count == 0);
146 return 0;
147}
148
149/**
150 * put_super - drop a temporary reference to superblock
151 * @sb: superblock in question
152 *
153 * Drops a temporary reference, frees superblock if there's no
154 * references left.
155 */
156static void put_super(struct super_block *sb)
157{
158 spin_lock(&sb_lock);
159 __put_super(sb);
160 spin_unlock(&sb_lock);
161}
162
163
164/**
165 * deactivate_super - drop an active reference to superblock
166 * @s: superblock to deactivate
167 *
168 * Drops an active reference to superblock, acquiring a temprory one if
169 * there is no active references left. In that case we lock superblock,
170 * tell fs driver to shut it down and drop the temporary reference we
171 * had just acquired.
172 */
173void deactivate_super(struct super_block *s)
174{
175 struct file_system_type *fs = s->s_type;
176 if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
177 s->s_count -= S_BIAS-1;
178 spin_unlock(&sb_lock);
Al Viro7b7b1ac2005-11-07 17:13:39 -0500179 DQUOT_OFF(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180 down_write(&s->s_umount);
181 fs->kill_sb(s);
182 put_filesystem(fs);
183 put_super(s);
184 }
185}
186
187EXPORT_SYMBOL(deactivate_super);
188
189/**
190 * grab_super - acquire an active reference
191 * @s: reference we are trying to make active
192 *
193 * Tries to acquire an active reference. grab_super() is used when we
194 * had just found a superblock in super_blocks or fs_type->fs_supers
195 * and want to turn it into a full-blown active reference. grab_super()
196 * is called with sb_lock held and drops it. Returns 1 in case of
197 * success, 0 if we had failed (superblock contents was already dead or
198 * dying when grab_super() had been called).
199 */
Josh Triplett9c4dbee2006-09-29 01:59:29 -0700200static int grab_super(struct super_block *s) __releases(sb_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201{
202 s->s_count++;
203 spin_unlock(&sb_lock);
204 down_write(&s->s_umount);
205 if (s->s_root) {
206 spin_lock(&sb_lock);
207 if (s->s_count > S_BIAS) {
208 atomic_inc(&s->s_active);
209 s->s_count--;
210 spin_unlock(&sb_lock);
211 return 1;
212 }
213 spin_unlock(&sb_lock);
214 }
215 up_write(&s->s_umount);
216 put_super(s);
217 yield();
218 return 0;
219}
220
David Howellscf9a2ae2006-08-29 19:05:54 +0100221/*
Al Viro914e2632006-10-18 13:55:46 -0400222 * Superblock locking. We really ought to get rid of these two.
223 */
224void lock_super(struct super_block * sb)
225{
226 get_fs_excl();
227 mutex_lock(&sb->s_lock);
228}
229
230void unlock_super(struct super_block * sb)
231{
232 put_fs_excl();
233 mutex_unlock(&sb->s_lock);
234}
235
236EXPORT_SYMBOL(lock_super);
237EXPORT_SYMBOL(unlock_super);
238
239/*
David Howellscf9a2ae2006-08-29 19:05:54 +0100240 * Write out and wait upon all dirty data associated with this
241 * superblock. Filesystem data as well as the underlying block
242 * device. Takes the superblock lock. Requires a second blkdev
243 * flush by the caller to complete the operation.
244 */
245void __fsync_super(struct super_block *sb)
246{
247 sync_inodes_sb(sb, 0);
248 DQUOT_SYNC(sb);
249 lock_super(sb);
250 if (sb->s_dirt && sb->s_op->write_super)
251 sb->s_op->write_super(sb);
252 unlock_super(sb);
253 if (sb->s_op->sync_fs)
254 sb->s_op->sync_fs(sb, 1);
255 sync_blockdev(sb->s_bdev);
256 sync_inodes_sb(sb, 1);
257}
258
259/*
260 * Write out and wait upon all dirty data associated with this
261 * superblock. Filesystem data as well as the underlying block
262 * device. Takes the superblock lock.
263 */
264int fsync_super(struct super_block *sb)
265{
266 __fsync_super(sb);
267 return sync_blockdev(sb->s_bdev);
268}
269
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270/**
271 * generic_shutdown_super - common helper for ->kill_sb()
272 * @sb: superblock to kill
273 *
274 * generic_shutdown_super() does all fs-independent work on superblock
275 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
276 * that need destruction out of superblock, call generic_shutdown_super()
277 * and release aforementioned objects. Note: dentries and inodes _are_
278 * taken care of and do not need specific handling.
David Howellsc636ebd2006-10-11 01:22:19 -0700279 *
280 * Upon calling this function, the filesystem may no longer alter or
281 * rearrange the set of dentries belonging to this super_block, nor may it
282 * change the attachments of dentries to inodes.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283 */
284void generic_shutdown_super(struct super_block *sb)
285{
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -0800286 const struct super_operations *sop = sb->s_op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287
David Howellsc636ebd2006-10-11 01:22:19 -0700288 if (sb->s_root) {
289 shrink_dcache_for_umount(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290 fsync_super(sb);
291 lock_super(sb);
292 sb->s_flags &= ~MS_ACTIVE;
293 /* bad name - it should be evict_inodes() */
294 invalidate_inodes(sb);
295 lock_kernel();
296
297 if (sop->write_super && sb->s_dirt)
298 sop->write_super(sb);
299 if (sop->put_super)
300 sop->put_super(sb);
301
302 /* Forget any remaining inodes */
303 if (invalidate_inodes(sb)) {
Dave Jones7b4fe292006-02-07 12:58:48 -0800304 printk("VFS: Busy inodes after unmount of %s. "
305 "Self-destruct in 5 seconds. Have a nice day...\n",
306 sb->s_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307 }
308
309 unlock_kernel();
310 unlock_super(sb);
311 }
312 spin_lock(&sb_lock);
313 /* should be initialized for __put_super_and_need_restart() */
314 list_del_init(&sb->s_list);
315 list_del(&sb->s_instances);
316 spin_unlock(&sb_lock);
317 up_write(&sb->s_umount);
318}
319
320EXPORT_SYMBOL(generic_shutdown_super);
321
322/**
323 * sget - find or create a superblock
324 * @type: filesystem type superblock should belong to
325 * @test: comparison callback
326 * @set: setup callback
327 * @data: argument to each of them
328 */
329struct super_block *sget(struct file_system_type *type,
330 int (*test)(struct super_block *,void *),
331 int (*set)(struct super_block *,void *),
332 void *data)
333{
334 struct super_block *s = NULL;
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700335 struct super_block *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336 int err;
337
338retry:
339 spin_lock(&sb_lock);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700340 if (test) {
341 list_for_each_entry(old, &type->fs_supers, s_instances) {
342 if (!test(old, data))
343 continue;
344 if (!grab_super(old))
345 goto retry;
346 if (s)
347 destroy_super(s);
348 return old;
349 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350 }
351 if (!s) {
352 spin_unlock(&sb_lock);
Ingo Molnarcf516242006-07-03 00:25:27 -0700353 s = alloc_super(type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354 if (!s)
355 return ERR_PTR(-ENOMEM);
356 goto retry;
357 }
358
359 err = set(s, data);
360 if (err) {
361 spin_unlock(&sb_lock);
362 destroy_super(s);
363 return ERR_PTR(err);
364 }
365 s->s_type = type;
366 strlcpy(s->s_id, type->name, sizeof(s->s_id));
367 list_add_tail(&s->s_list, &super_blocks);
368 list_add(&s->s_instances, &type->fs_supers);
369 spin_unlock(&sb_lock);
370 get_filesystem(type);
371 return s;
372}
373
374EXPORT_SYMBOL(sget);
375
376void drop_super(struct super_block *sb)
377{
378 up_read(&sb->s_umount);
379 put_super(sb);
380}
381
382EXPORT_SYMBOL(drop_super);
383
384static inline void write_super(struct super_block *sb)
385{
386 lock_super(sb);
387 if (sb->s_root && sb->s_dirt)
388 if (sb->s_op->write_super)
389 sb->s_op->write_super(sb);
390 unlock_super(sb);
391}
392
393/*
394 * Note: check the dirty flag before waiting, so we don't
395 * hold up the sync while mounting a device. (The newly
396 * mounted device won't need syncing.)
397 */
398void sync_supers(void)
399{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700400 struct super_block *sb;
401
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700403restart:
404 list_for_each_entry(sb, &super_blocks, s_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 if (sb->s_dirt) {
406 sb->s_count++;
407 spin_unlock(&sb_lock);
408 down_read(&sb->s_umount);
409 write_super(sb);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700410 up_read(&sb->s_umount);
411 spin_lock(&sb_lock);
412 if (__put_super_and_need_restart(sb))
413 goto restart;
414 }
415 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 spin_unlock(&sb_lock);
417}
418
419/*
420 * Call the ->sync_fs super_op against all filesytems which are r/w and
421 * which implement it.
422 *
423 * This operation is careful to avoid the livelock which could easily happen
424 * if two or more filesystems are being continuously dirtied. s_need_sync_fs
425 * is used only here. We set it against all filesystems and then clear it as
426 * we sync them. So redirtied filesystems are skipped.
427 *
428 * But if process A is currently running sync_filesytems and then process B
429 * calls sync_filesystems as well, process B will set all the s_need_sync_fs
430 * flags again, which will cause process A to resync everything. Fix that with
431 * a local mutex.
432 *
433 * (Fabian) Avoid sync_fs with clean fs & wait mode 0
434 */
435void sync_filesystems(int wait)
436{
437 struct super_block *sb;
Ingo Molnar353ab6e2006-03-26 01:37:12 -0800438 static DEFINE_MUTEX(mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439
Ingo Molnar353ab6e2006-03-26 01:37:12 -0800440 mutex_lock(&mutex); /* Could be down_interruptible */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700442 list_for_each_entry(sb, &super_blocks, s_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 if (!sb->s_op->sync_fs)
444 continue;
445 if (sb->s_flags & MS_RDONLY)
446 continue;
447 sb->s_need_sync_fs = 1;
448 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449
450restart:
Kirill Korotaev618f0632005-06-23 00:09:54 -0700451 list_for_each_entry(sb, &super_blocks, s_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 if (!sb->s_need_sync_fs)
453 continue;
454 sb->s_need_sync_fs = 0;
455 if (sb->s_flags & MS_RDONLY)
456 continue; /* hm. Was remounted r/o meanwhile */
457 sb->s_count++;
458 spin_unlock(&sb_lock);
459 down_read(&sb->s_umount);
460 if (sb->s_root && (wait || sb->s_dirt))
461 sb->s_op->sync_fs(sb, wait);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700462 up_read(&sb->s_umount);
463 /* restart only when sb is no longer on the list */
464 spin_lock(&sb_lock);
465 if (__put_super_and_need_restart(sb))
466 goto restart;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467 }
468 spin_unlock(&sb_lock);
Ingo Molnar353ab6e2006-03-26 01:37:12 -0800469 mutex_unlock(&mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470}
471
472/**
473 * get_super - get the superblock of a device
474 * @bdev: device to get the superblock for
475 *
476 * Scans the superblock list and finds the superblock of the file system
477 * mounted on the device given. %NULL is returned if no match is found.
478 */
479
480struct super_block * get_super(struct block_device *bdev)
481{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700482 struct super_block *sb;
483
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 if (!bdev)
485 return NULL;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700486
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700488rescan:
489 list_for_each_entry(sb, &super_blocks, s_list) {
490 if (sb->s_bdev == bdev) {
491 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700493 down_read(&sb->s_umount);
494 if (sb->s_root)
495 return sb;
496 up_read(&sb->s_umount);
497 /* restart only when sb is no longer on the list */
498 spin_lock(&sb_lock);
499 if (__put_super_and_need_restart(sb))
500 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501 }
502 }
503 spin_unlock(&sb_lock);
504 return NULL;
505}
506
507EXPORT_SYMBOL(get_super);
508
509struct super_block * user_get_super(dev_t dev)
510{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700511 struct super_block *sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700514rescan:
515 list_for_each_entry(sb, &super_blocks, s_list) {
516 if (sb->s_dev == dev) {
517 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700519 down_read(&sb->s_umount);
520 if (sb->s_root)
521 return sb;
522 up_read(&sb->s_umount);
523 /* restart only when sb is no longer on the list */
524 spin_lock(&sb_lock);
525 if (__put_super_and_need_restart(sb))
526 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 }
528 }
529 spin_unlock(&sb_lock);
530 return NULL;
531}
532
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533asmlinkage long sys_ustat(unsigned dev, struct ustat __user * ubuf)
534{
535 struct super_block *s;
536 struct ustat tmp;
537 struct kstatfs sbuf;
538 int err = -EINVAL;
539
540 s = user_get_super(new_decode_dev(dev));
541 if (s == NULL)
542 goto out;
David Howells726c3342006-06-23 02:02:58 -0700543 err = vfs_statfs(s->s_root, &sbuf);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544 drop_super(s);
545 if (err)
546 goto out;
547
548 memset(&tmp,0,sizeof(struct ustat));
549 tmp.f_tfree = sbuf.f_bfree;
550 tmp.f_tinode = sbuf.f_ffree;
551
552 err = copy_to_user(ubuf,&tmp,sizeof(struct ustat)) ? -EFAULT : 0;
553out:
554 return err;
555}
556
557/**
558 * mark_files_ro
559 * @sb: superblock in question
560 *
561 * All files are marked read/only. We don't care about pending
562 * delete files so this should be used in 'force' mode only
563 */
564
565static void mark_files_ro(struct super_block *sb)
566{
567 struct file *f;
568
569 file_list_lock();
Eric Dumazet2f512012005-10-30 15:02:16 -0800570 list_for_each_entry(f, &sb->s_files, f_u.fu_list) {
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800571 if (S_ISREG(f->f_path.dentry->d_inode->i_mode) && file_count(f))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572 f->f_mode &= ~FMODE_WRITE;
573 }
574 file_list_unlock();
575}
576
577/**
578 * do_remount_sb - asks filesystem to change mount options.
579 * @sb: superblock in question
580 * @flags: numeric part of options
581 * @data: the rest of options
582 * @force: whether or not to force the change
583 *
584 * Alters the mount options of a mounted file system.
585 */
586int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
587{
588 int retval;
589
David Howells93614012006-09-30 20:45:40 +0200590#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
592 return -EACCES;
David Howells93614012006-09-30 20:45:40 +0200593#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594 if (flags & MS_RDONLY)
595 acct_auto_close(sb);
596 shrink_dcache_sb(sb);
597 fsync_super(sb);
598
599 /* If we are remounting RDONLY and current sb is read/write,
600 make sure there are no rw files opened */
601 if ((flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY)) {
602 if (force)
603 mark_files_ro(sb);
604 else if (!fs_may_remount_ro(sb))
605 return -EBUSY;
606 }
607
608 if (sb->s_op->remount_fs) {
609 lock_super(sb);
610 retval = sb->s_op->remount_fs(sb, &flags, data);
611 unlock_super(sb);
612 if (retval)
613 return retval;
614 }
615 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
616 return 0;
617}
618
619static void do_emergency_remount(unsigned long foo)
620{
621 struct super_block *sb;
622
623 spin_lock(&sb_lock);
624 list_for_each_entry(sb, &super_blocks, s_list) {
625 sb->s_count++;
626 spin_unlock(&sb_lock);
627 down_read(&sb->s_umount);
628 if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) {
629 /*
630 * ->remount_fs needs lock_kernel().
631 *
632 * What lock protects sb->s_flags??
633 */
634 lock_kernel();
635 do_remount_sb(sb, MS_RDONLY, NULL, 1);
636 unlock_kernel();
637 }
638 drop_super(sb);
639 spin_lock(&sb_lock);
640 }
641 spin_unlock(&sb_lock);
642 printk("Emergency Remount complete\n");
643}
644
645void emergency_remount(void)
646{
647 pdflush_operation(do_emergency_remount, 0);
648}
649
650/*
651 * Unnamed block devices are dummy devices used by virtual
652 * filesystems which don't use real block-devices. -- jrs
653 */
654
655static struct idr unnamed_dev_idr;
656static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
657
658int set_anon_super(struct super_block *s, void *data)
659{
660 int dev;
661 int error;
662
663 retry:
664 if (idr_pre_get(&unnamed_dev_idr, GFP_ATOMIC) == 0)
665 return -ENOMEM;
666 spin_lock(&unnamed_dev_lock);
667 error = idr_get_new(&unnamed_dev_idr, NULL, &dev);
668 spin_unlock(&unnamed_dev_lock);
669 if (error == -EAGAIN)
670 /* We raced and lost with another CPU. */
671 goto retry;
672 else if (error)
673 return -EAGAIN;
674
675 if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
676 spin_lock(&unnamed_dev_lock);
677 idr_remove(&unnamed_dev_idr, dev);
678 spin_unlock(&unnamed_dev_lock);
679 return -EMFILE;
680 }
681 s->s_dev = MKDEV(0, dev & MINORMASK);
682 return 0;
683}
684
685EXPORT_SYMBOL(set_anon_super);
686
687void kill_anon_super(struct super_block *sb)
688{
689 int slot = MINOR(sb->s_dev);
690
691 generic_shutdown_super(sb);
692 spin_lock(&unnamed_dev_lock);
693 idr_remove(&unnamed_dev_idr, slot);
694 spin_unlock(&unnamed_dev_lock);
695}
696
697EXPORT_SYMBOL(kill_anon_super);
698
699void __init unnamed_dev_init(void)
700{
701 idr_init(&unnamed_dev_idr);
702}
703
704void kill_litter_super(struct super_block *sb)
705{
706 if (sb->s_root)
707 d_genocide(sb->s_root);
708 kill_anon_super(sb);
709}
710
711EXPORT_SYMBOL(kill_litter_super);
712
David Howells93614012006-09-30 20:45:40 +0200713#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714static int set_bdev_super(struct super_block *s, void *data)
715{
716 s->s_bdev = data;
717 s->s_dev = s->s_bdev->bd_dev;
718 return 0;
719}
720
721static int test_bdev_super(struct super_block *s, void *data)
722{
723 return (void *)s->s_bdev == data;
724}
725
David Howells454e2392006-06-23 02:02:57 -0700726int get_sb_bdev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727 int flags, const char *dev_name, void *data,
David Howells454e2392006-06-23 02:02:57 -0700728 int (*fill_super)(struct super_block *, void *, int),
729 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730{
731 struct block_device *bdev;
732 struct super_block *s;
733 int error = 0;
734
735 bdev = open_bdev_excl(dev_name, flags, fs_type);
736 if (IS_ERR(bdev))
David Howells454e2392006-06-23 02:02:57 -0700737 return PTR_ERR(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738
739 /*
740 * once the super is inserted into the list by sget, s_umount
741 * will protect the lockfs code from trying to start a snapshot
742 * while we are mounting
743 */
David Chinnerf73ca1b2007-01-10 23:15:41 -0800744 down(&bdev->bd_mount_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745 s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
David Chinnerf73ca1b2007-01-10 23:15:41 -0800746 up(&bdev->bd_mount_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700748 goto error_s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749
750 if (s->s_root) {
751 if ((flags ^ s->s_flags) & MS_RDONLY) {
752 up_write(&s->s_umount);
753 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700754 error = -EBUSY;
755 goto error_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756 }
David Howells454e2392006-06-23 02:02:57 -0700757
758 close_bdev_excl(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759 } else {
760 char b[BDEVNAME_SIZE];
761
762 s->s_flags = flags;
763 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
Pekka Enberge78c9a02006-01-08 01:03:39 -0800764 sb_set_blocksize(s, block_size(bdev));
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800765 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766 if (error) {
767 up_write(&s->s_umount);
768 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700769 goto error;
Greg Kroah-Hartmanfa675762006-02-22 09:39:02 -0800770 }
David Howells454e2392006-06-23 02:02:57 -0700771
772 s->s_flags |= MS_ACTIVE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773 }
774
David Howells454e2392006-06-23 02:02:57 -0700775 return simple_set_mnt(mnt, s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776
David Howells454e2392006-06-23 02:02:57 -0700777error_s:
778 error = PTR_ERR(s);
779error_bdev:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780 close_bdev_excl(bdev);
David Howells454e2392006-06-23 02:02:57 -0700781error:
782 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783}
784
785EXPORT_SYMBOL(get_sb_bdev);
786
787void kill_block_super(struct super_block *sb)
788{
789 struct block_device *bdev = sb->s_bdev;
790
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791 generic_shutdown_super(sb);
792 sync_blockdev(bdev);
793 close_bdev_excl(bdev);
794}
795
796EXPORT_SYMBOL(kill_block_super);
David Howells93614012006-09-30 20:45:40 +0200797#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798
David Howells454e2392006-06-23 02:02:57 -0700799int get_sb_nodev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800 int flags, void *data,
David Howells454e2392006-06-23 02:02:57 -0700801 int (*fill_super)(struct super_block *, void *, int),
802 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803{
804 int error;
805 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
806
807 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700808 return PTR_ERR(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809
810 s->s_flags = flags;
811
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800812 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813 if (error) {
814 up_write(&s->s_umount);
815 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700816 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817 }
818 s->s_flags |= MS_ACTIVE;
David Howells454e2392006-06-23 02:02:57 -0700819 return simple_set_mnt(mnt, s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820}
821
822EXPORT_SYMBOL(get_sb_nodev);
823
824static int compare_single(struct super_block *s, void *p)
825{
826 return 1;
827}
828
David Howells454e2392006-06-23 02:02:57 -0700829int get_sb_single(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830 int flags, void *data,
David Howells454e2392006-06-23 02:02:57 -0700831 int (*fill_super)(struct super_block *, void *, int),
832 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833{
834 struct super_block *s;
835 int error;
836
837 s = sget(fs_type, compare_single, set_anon_super, NULL);
838 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700839 return PTR_ERR(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840 if (!s->s_root) {
841 s->s_flags = flags;
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800842 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843 if (error) {
844 up_write(&s->s_umount);
845 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700846 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847 }
848 s->s_flags |= MS_ACTIVE;
849 }
850 do_remount_sb(s, flags, data, 0);
David Howells454e2392006-06-23 02:02:57 -0700851 return simple_set_mnt(mnt, s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852}
853
854EXPORT_SYMBOL(get_sb_single);
855
856struct vfsmount *
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400857vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 struct vfsmount *mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 char *secdata = NULL;
David Howells454e2392006-06-23 02:02:57 -0700861 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862
863 if (!type)
864 return ERR_PTR(-ENODEV);
865
David Howells454e2392006-06-23 02:02:57 -0700866 error = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867 mnt = alloc_vfsmnt(name);
868 if (!mnt)
869 goto out;
870
871 if (data) {
872 secdata = alloc_secdata();
David Howells454e2392006-06-23 02:02:57 -0700873 if (!secdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874 goto out_mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875
876 error = security_sb_copy_data(type, data, secdata);
David Howells454e2392006-06-23 02:02:57 -0700877 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878 goto out_free_secdata;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879 }
880
David Howells454e2392006-06-23 02:02:57 -0700881 error = type->get_sb(type, flags, name, data, mnt);
882 if (error < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883 goto out_free_secdata;
Lee Schermerhornb4c07bc2007-07-15 23:40:54 -0700884 BUG_ON(!mnt->mnt_sb);
David Howells454e2392006-06-23 02:02:57 -0700885
886 error = security_sb_kern_mount(mnt->mnt_sb, secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887 if (error)
888 goto out_sb;
David Howells454e2392006-06-23 02:02:57 -0700889
890 mnt->mnt_mountpoint = mnt->mnt_root;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 mnt->mnt_parent = mnt;
David Howells454e2392006-06-23 02:02:57 -0700892 up_write(&mnt->mnt_sb->s_umount);
Gerald Schaefer8680e222005-06-21 17:15:16 -0700893 free_secdata(secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894 return mnt;
895out_sb:
David Howells454e2392006-06-23 02:02:57 -0700896 dput(mnt->mnt_root);
897 up_write(&mnt->mnt_sb->s_umount);
898 deactivate_super(mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899out_free_secdata:
900 free_secdata(secdata);
901out_mnt:
902 free_vfsmnt(mnt);
903out:
David Howells454e2392006-06-23 02:02:57 -0700904 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905}
906
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400907EXPORT_SYMBOL_GPL(vfs_kern_mount);
908
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700909static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype)
910{
911 int err;
912 const char *subtype = strchr(fstype, '.');
913 if (subtype) {
914 subtype++;
915 err = -EINVAL;
916 if (!subtype[0])
917 goto err;
918 } else
919 subtype = "";
920
921 mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL);
922 err = -ENOMEM;
923 if (!mnt->mnt_sb->s_subtype)
924 goto err;
925 return mnt;
926
927 err:
928 mntput(mnt);
929 return ERR_PTR(err);
930}
931
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400932struct vfsmount *
933do_kern_mount(const char *fstype, int flags, const char *name, void *data)
934{
935 struct file_system_type *type = get_fs_type(fstype);
936 struct vfsmount *mnt;
937 if (!type)
938 return ERR_PTR(-ENODEV);
939 mnt = vfs_kern_mount(type, flags, name, data);
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700940 if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) &&
941 !mnt->mnt_sb->s_subtype)
942 mnt = fs_set_subtype(mnt, fstype);
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400943 put_filesystem(type);
944 return mnt;
945}
946
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947struct vfsmount *kern_mount(struct file_system_type *type)
948{
Trond Myklebust1f5ce9e2006-06-09 09:34:16 -0400949 return vfs_kern_mount(type, 0, type->name, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950}
951
952EXPORT_SYMBOL(kern_mount);