blob: 6987824d0dce0ec94713d3f394e8d2c8535c04aa [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
43void get_filesystem(struct file_system_type *fs);
44void put_filesystem(struct file_system_type *fs);
45struct file_system_type *get_fs_type(const char *name);
46
47LIST_HEAD(super_blocks);
48DEFINE_SPINLOCK(sb_lock);
49
50/**
51 * alloc_super - create new superblock
Henrik Kretzschmarfe2bbc42006-09-06 00:03:41 -070052 * @type: filesystem type superblock should belong to
Linus Torvalds1da177e2005-04-16 15:20:36 -070053 *
54 * Allocates and initializes a new &struct super_block. alloc_super()
55 * returns a pointer new superblock or %NULL if allocation had failed.
56 */
Ingo Molnarcf516242006-07-03 00:25:27 -070057static struct super_block *alloc_super(struct file_system_type *type)
Linus Torvalds1da177e2005-04-16 15:20:36 -070058{
Oliver Neukum11b0b5a2006-03-25 03:08:13 -080059 struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER);
Linus Torvalds1da177e2005-04-16 15:20:36 -070060 static struct super_operations default_op;
61
62 if (s) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070063 if (security_sb_alloc(s)) {
64 kfree(s);
65 s = NULL;
66 goto out;
67 }
68 INIT_LIST_HEAD(&s->s_dirty);
69 INIT_LIST_HEAD(&s->s_io);
70 INIT_LIST_HEAD(&s->s_files);
71 INIT_LIST_HEAD(&s->s_instances);
72 INIT_HLIST_HEAD(&s->s_anon);
73 INIT_LIST_HEAD(&s->s_inodes);
74 init_rwsem(&s->s_umount);
Ingo Molnar7892f2f2006-01-09 15:59:25 -080075 mutex_init(&s->s_lock);
Arjan van de Ven897c6ff2006-07-03 00:25:28 -070076 lockdep_set_class(&s->s_umount, &type->s_umount_key);
Ingo Molnarcf516242006-07-03 00:25:27 -070077 /*
78 * The locking rules for s_lock are up to the
79 * filesystem. For example ext3fs has different
80 * lock ordering than usbfs:
81 */
82 lockdep_set_class(&s->s_lock, &type->s_lock_key);
Linus Torvalds1da177e2005-04-16 15:20:36 -070083 down_write(&s->s_umount);
84 s->s_count = S_BIAS;
85 atomic_set(&s->s_active, 1);
Arjan van de Vena11f3a02006-03-23 03:00:33 -080086 mutex_init(&s->s_vfs_rename_mutex);
Ingo Molnard3be9152006-03-23 03:00:29 -080087 mutex_init(&s->s_dquot.dqio_mutex);
88 mutex_init(&s->s_dquot.dqonoff_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -070089 init_rwsem(&s->s_dquot.dqptr_sem);
90 init_waitqueue_head(&s->s_wait_unfrozen);
91 s->s_maxbytes = MAX_NON_LFS;
92 s->dq_op = sb_dquot_ops;
93 s->s_qcop = sb_quotactl_ops;
94 s->s_op = &default_op;
95 s->s_time_gran = 1000000000;
96 }
97out:
98 return s;
99}
100
101/**
102 * destroy_super - frees a superblock
103 * @s: superblock to free
104 *
105 * Frees a superblock.
106 */
107static inline void destroy_super(struct super_block *s)
108{
109 security_sb_free(s);
110 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
223/**
224 * generic_shutdown_super - common helper for ->kill_sb()
225 * @sb: superblock to kill
226 *
227 * generic_shutdown_super() does all fs-independent work on superblock
228 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
229 * that need destruction out of superblock, call generic_shutdown_super()
230 * and release aforementioned objects. Note: dentries and inodes _are_
231 * taken care of and do not need specific handling.
232 */
233void generic_shutdown_super(struct super_block *sb)
234{
235 struct dentry *root = sb->s_root;
236 struct super_operations *sop = sb->s_op;
237
238 if (root) {
239 sb->s_root = NULL;
240 shrink_dcache_parent(root);
David Howells454e2392006-06-23 02:02:57 -0700241 shrink_dcache_sb(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242 dput(root);
243 fsync_super(sb);
244 lock_super(sb);
245 sb->s_flags &= ~MS_ACTIVE;
246 /* bad name - it should be evict_inodes() */
247 invalidate_inodes(sb);
248 lock_kernel();
249
250 if (sop->write_super && sb->s_dirt)
251 sop->write_super(sb);
252 if (sop->put_super)
253 sop->put_super(sb);
254
255 /* Forget any remaining inodes */
256 if (invalidate_inodes(sb)) {
Dave Jones7b4fe292006-02-07 12:58:48 -0800257 printk("VFS: Busy inodes after unmount of %s. "
258 "Self-destruct in 5 seconds. Have a nice day...\n",
259 sb->s_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260 }
261
262 unlock_kernel();
263 unlock_super(sb);
264 }
265 spin_lock(&sb_lock);
266 /* should be initialized for __put_super_and_need_restart() */
267 list_del_init(&sb->s_list);
268 list_del(&sb->s_instances);
269 spin_unlock(&sb_lock);
270 up_write(&sb->s_umount);
271}
272
273EXPORT_SYMBOL(generic_shutdown_super);
274
275/**
276 * sget - find or create a superblock
277 * @type: filesystem type superblock should belong to
278 * @test: comparison callback
279 * @set: setup callback
280 * @data: argument to each of them
281 */
282struct super_block *sget(struct file_system_type *type,
283 int (*test)(struct super_block *,void *),
284 int (*set)(struct super_block *,void *),
285 void *data)
286{
287 struct super_block *s = NULL;
288 struct list_head *p;
289 int err;
290
291retry:
292 spin_lock(&sb_lock);
293 if (test) list_for_each(p, &type->fs_supers) {
294 struct super_block *old;
295 old = list_entry(p, struct super_block, s_instances);
296 if (!test(old, data))
297 continue;
298 if (!grab_super(old))
299 goto retry;
300 if (s)
301 destroy_super(s);
302 return old;
303 }
304 if (!s) {
305 spin_unlock(&sb_lock);
Ingo Molnarcf516242006-07-03 00:25:27 -0700306 s = alloc_super(type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307 if (!s)
308 return ERR_PTR(-ENOMEM);
309 goto retry;
310 }
311
312 err = set(s, data);
313 if (err) {
314 spin_unlock(&sb_lock);
315 destroy_super(s);
316 return ERR_PTR(err);
317 }
318 s->s_type = type;
319 strlcpy(s->s_id, type->name, sizeof(s->s_id));
320 list_add_tail(&s->s_list, &super_blocks);
321 list_add(&s->s_instances, &type->fs_supers);
322 spin_unlock(&sb_lock);
323 get_filesystem(type);
324 return s;
325}
326
327EXPORT_SYMBOL(sget);
328
329void drop_super(struct super_block *sb)
330{
331 up_read(&sb->s_umount);
332 put_super(sb);
333}
334
335EXPORT_SYMBOL(drop_super);
336
337static inline void write_super(struct super_block *sb)
338{
339 lock_super(sb);
340 if (sb->s_root && sb->s_dirt)
341 if (sb->s_op->write_super)
342 sb->s_op->write_super(sb);
343 unlock_super(sb);
344}
345
346/*
347 * Note: check the dirty flag before waiting, so we don't
348 * hold up the sync while mounting a device. (The newly
349 * mounted device won't need syncing.)
350 */
351void sync_supers(void)
352{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700353 struct super_block *sb;
354
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700356restart:
357 list_for_each_entry(sb, &super_blocks, s_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358 if (sb->s_dirt) {
359 sb->s_count++;
360 spin_unlock(&sb_lock);
361 down_read(&sb->s_umount);
362 write_super(sb);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700363 up_read(&sb->s_umount);
364 spin_lock(&sb_lock);
365 if (__put_super_and_need_restart(sb))
366 goto restart;
367 }
368 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369 spin_unlock(&sb_lock);
370}
371
372/*
373 * Call the ->sync_fs super_op against all filesytems which are r/w and
374 * which implement it.
375 *
376 * This operation is careful to avoid the livelock which could easily happen
377 * if two or more filesystems are being continuously dirtied. s_need_sync_fs
378 * is used only here. We set it against all filesystems and then clear it as
379 * we sync them. So redirtied filesystems are skipped.
380 *
381 * But if process A is currently running sync_filesytems and then process B
382 * calls sync_filesystems as well, process B will set all the s_need_sync_fs
383 * flags again, which will cause process A to resync everything. Fix that with
384 * a local mutex.
385 *
386 * (Fabian) Avoid sync_fs with clean fs & wait mode 0
387 */
388void sync_filesystems(int wait)
389{
390 struct super_block *sb;
Ingo Molnar353ab6e2006-03-26 01:37:12 -0800391 static DEFINE_MUTEX(mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392
Ingo Molnar353ab6e2006-03-26 01:37:12 -0800393 mutex_lock(&mutex); /* Could be down_interruptible */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700395 list_for_each_entry(sb, &super_blocks, s_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 if (!sb->s_op->sync_fs)
397 continue;
398 if (sb->s_flags & MS_RDONLY)
399 continue;
400 sb->s_need_sync_fs = 1;
401 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402
403restart:
Kirill Korotaev618f0632005-06-23 00:09:54 -0700404 list_for_each_entry(sb, &super_blocks, s_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 if (!sb->s_need_sync_fs)
406 continue;
407 sb->s_need_sync_fs = 0;
408 if (sb->s_flags & MS_RDONLY)
409 continue; /* hm. Was remounted r/o meanwhile */
410 sb->s_count++;
411 spin_unlock(&sb_lock);
412 down_read(&sb->s_umount);
413 if (sb->s_root && (wait || sb->s_dirt))
414 sb->s_op->sync_fs(sb, wait);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700415 up_read(&sb->s_umount);
416 /* restart only when sb is no longer on the list */
417 spin_lock(&sb_lock);
418 if (__put_super_and_need_restart(sb))
419 goto restart;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420 }
421 spin_unlock(&sb_lock);
Ingo Molnar353ab6e2006-03-26 01:37:12 -0800422 mutex_unlock(&mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423}
424
425/**
426 * get_super - get the superblock of a device
427 * @bdev: device to get the superblock for
428 *
429 * Scans the superblock list and finds the superblock of the file system
430 * mounted on the device given. %NULL is returned if no match is found.
431 */
432
433struct super_block * get_super(struct block_device *bdev)
434{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700435 struct super_block *sb;
436
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 if (!bdev)
438 return NULL;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700439
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700441rescan:
442 list_for_each_entry(sb, &super_blocks, s_list) {
443 if (sb->s_bdev == bdev) {
444 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700446 down_read(&sb->s_umount);
447 if (sb->s_root)
448 return sb;
449 up_read(&sb->s_umount);
450 /* restart only when sb is no longer on the list */
451 spin_lock(&sb_lock);
452 if (__put_super_and_need_restart(sb))
453 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454 }
455 }
456 spin_unlock(&sb_lock);
457 return NULL;
458}
459
460EXPORT_SYMBOL(get_super);
461
462struct super_block * user_get_super(dev_t dev)
463{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700464 struct super_block *sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700467rescan:
468 list_for_each_entry(sb, &super_blocks, s_list) {
469 if (sb->s_dev == dev) {
470 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700472 down_read(&sb->s_umount);
473 if (sb->s_root)
474 return sb;
475 up_read(&sb->s_umount);
476 /* restart only when sb is no longer on the list */
477 spin_lock(&sb_lock);
478 if (__put_super_and_need_restart(sb))
479 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480 }
481 }
482 spin_unlock(&sb_lock);
483 return NULL;
484}
485
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486asmlinkage long sys_ustat(unsigned dev, struct ustat __user * ubuf)
487{
488 struct super_block *s;
489 struct ustat tmp;
490 struct kstatfs sbuf;
491 int err = -EINVAL;
492
493 s = user_get_super(new_decode_dev(dev));
494 if (s == NULL)
495 goto out;
David Howells726c3342006-06-23 02:02:58 -0700496 err = vfs_statfs(s->s_root, &sbuf);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 drop_super(s);
498 if (err)
499 goto out;
500
501 memset(&tmp,0,sizeof(struct ustat));
502 tmp.f_tfree = sbuf.f_bfree;
503 tmp.f_tinode = sbuf.f_ffree;
504
505 err = copy_to_user(ubuf,&tmp,sizeof(struct ustat)) ? -EFAULT : 0;
506out:
507 return err;
508}
509
510/**
511 * mark_files_ro
512 * @sb: superblock in question
513 *
514 * All files are marked read/only. We don't care about pending
515 * delete files so this should be used in 'force' mode only
516 */
517
518static void mark_files_ro(struct super_block *sb)
519{
520 struct file *f;
521
522 file_list_lock();
Eric Dumazet2f512012005-10-30 15:02:16 -0800523 list_for_each_entry(f, &sb->s_files, f_u.fu_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524 if (S_ISREG(f->f_dentry->d_inode->i_mode) && file_count(f))
525 f->f_mode &= ~FMODE_WRITE;
526 }
527 file_list_unlock();
528}
529
530/**
531 * do_remount_sb - asks filesystem to change mount options.
532 * @sb: superblock in question
533 * @flags: numeric part of options
534 * @data: the rest of options
535 * @force: whether or not to force the change
536 *
537 * Alters the mount options of a mounted file system.
538 */
539int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
540{
541 int retval;
542
543 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
544 return -EACCES;
545 if (flags & MS_RDONLY)
546 acct_auto_close(sb);
547 shrink_dcache_sb(sb);
548 fsync_super(sb);
549
550 /* If we are remounting RDONLY and current sb is read/write,
551 make sure there are no rw files opened */
552 if ((flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY)) {
553 if (force)
554 mark_files_ro(sb);
555 else if (!fs_may_remount_ro(sb))
556 return -EBUSY;
557 }
558
559 if (sb->s_op->remount_fs) {
560 lock_super(sb);
561 retval = sb->s_op->remount_fs(sb, &flags, data);
562 unlock_super(sb);
563 if (retval)
564 return retval;
565 }
566 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
567 return 0;
568}
569
570static void do_emergency_remount(unsigned long foo)
571{
572 struct super_block *sb;
573
574 spin_lock(&sb_lock);
575 list_for_each_entry(sb, &super_blocks, s_list) {
576 sb->s_count++;
577 spin_unlock(&sb_lock);
578 down_read(&sb->s_umount);
579 if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) {
580 /*
581 * ->remount_fs needs lock_kernel().
582 *
583 * What lock protects sb->s_flags??
584 */
585 lock_kernel();
586 do_remount_sb(sb, MS_RDONLY, NULL, 1);
587 unlock_kernel();
588 }
589 drop_super(sb);
590 spin_lock(&sb_lock);
591 }
592 spin_unlock(&sb_lock);
593 printk("Emergency Remount complete\n");
594}
595
596void emergency_remount(void)
597{
598 pdflush_operation(do_emergency_remount, 0);
599}
600
601/*
602 * Unnamed block devices are dummy devices used by virtual
603 * filesystems which don't use real block-devices. -- jrs
604 */
605
606static struct idr unnamed_dev_idr;
607static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
608
609int set_anon_super(struct super_block *s, void *data)
610{
611 int dev;
612 int error;
613
614 retry:
615 if (idr_pre_get(&unnamed_dev_idr, GFP_ATOMIC) == 0)
616 return -ENOMEM;
617 spin_lock(&unnamed_dev_lock);
618 error = idr_get_new(&unnamed_dev_idr, NULL, &dev);
619 spin_unlock(&unnamed_dev_lock);
620 if (error == -EAGAIN)
621 /* We raced and lost with another CPU. */
622 goto retry;
623 else if (error)
624 return -EAGAIN;
625
626 if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
627 spin_lock(&unnamed_dev_lock);
628 idr_remove(&unnamed_dev_idr, dev);
629 spin_unlock(&unnamed_dev_lock);
630 return -EMFILE;
631 }
632 s->s_dev = MKDEV(0, dev & MINORMASK);
633 return 0;
634}
635
636EXPORT_SYMBOL(set_anon_super);
637
638void kill_anon_super(struct super_block *sb)
639{
640 int slot = MINOR(sb->s_dev);
641
642 generic_shutdown_super(sb);
643 spin_lock(&unnamed_dev_lock);
644 idr_remove(&unnamed_dev_idr, slot);
645 spin_unlock(&unnamed_dev_lock);
646}
647
648EXPORT_SYMBOL(kill_anon_super);
649
650void __init unnamed_dev_init(void)
651{
652 idr_init(&unnamed_dev_idr);
653}
654
655void kill_litter_super(struct super_block *sb)
656{
657 if (sb->s_root)
658 d_genocide(sb->s_root);
659 kill_anon_super(sb);
660}
661
662EXPORT_SYMBOL(kill_litter_super);
663
664static int set_bdev_super(struct super_block *s, void *data)
665{
666 s->s_bdev = data;
667 s->s_dev = s->s_bdev->bd_dev;
668 return 0;
669}
670
671static int test_bdev_super(struct super_block *s, void *data)
672{
673 return (void *)s->s_bdev == data;
674}
675
Greg Kroah-Hartmanfa675762006-02-22 09:39:02 -0800676static void bdev_uevent(struct block_device *bdev, enum kobject_action action)
677{
678 if (bdev->bd_disk) {
679 if (bdev->bd_part)
680 kobject_uevent(&bdev->bd_part->kobj, action);
681 else
682 kobject_uevent(&bdev->bd_disk->kobj, action);
683 }
684}
685
David Howells454e2392006-06-23 02:02:57 -0700686int get_sb_bdev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687 int flags, const char *dev_name, void *data,
David Howells454e2392006-06-23 02:02:57 -0700688 int (*fill_super)(struct super_block *, void *, int),
689 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690{
691 struct block_device *bdev;
692 struct super_block *s;
693 int error = 0;
694
695 bdev = open_bdev_excl(dev_name, flags, fs_type);
696 if (IS_ERR(bdev))
David Howells454e2392006-06-23 02:02:57 -0700697 return PTR_ERR(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698
699 /*
700 * once the super is inserted into the list by sget, s_umount
701 * will protect the lockfs code from trying to start a snapshot
702 * while we are mounting
703 */
Arjan van de Venc039e312006-03-23 03:00:28 -0800704 mutex_lock(&bdev->bd_mount_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705 s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
Arjan van de Venc039e312006-03-23 03:00:28 -0800706 mutex_unlock(&bdev->bd_mount_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700707 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700708 goto error_s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709
710 if (s->s_root) {
711 if ((flags ^ s->s_flags) & MS_RDONLY) {
712 up_write(&s->s_umount);
713 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700714 error = -EBUSY;
715 goto error_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 }
David Howells454e2392006-06-23 02:02:57 -0700717
718 close_bdev_excl(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719 } else {
720 char b[BDEVNAME_SIZE];
721
722 s->s_flags = flags;
723 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
Pekka Enberge78c9a02006-01-08 01:03:39 -0800724 sb_set_blocksize(s, block_size(bdev));
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800725 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726 if (error) {
727 up_write(&s->s_umount);
728 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700729 goto error;
Greg Kroah-Hartmanfa675762006-02-22 09:39:02 -0800730 }
David Howells454e2392006-06-23 02:02:57 -0700731
732 s->s_flags |= MS_ACTIVE;
733 bdev_uevent(bdev, KOBJ_MOUNT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 }
735
David Howells454e2392006-06-23 02:02:57 -0700736 return simple_set_mnt(mnt, s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737
David Howells454e2392006-06-23 02:02:57 -0700738error_s:
739 error = PTR_ERR(s);
740error_bdev:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741 close_bdev_excl(bdev);
David Howells454e2392006-06-23 02:02:57 -0700742error:
743 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744}
745
746EXPORT_SYMBOL(get_sb_bdev);
747
748void kill_block_super(struct super_block *sb)
749{
750 struct block_device *bdev = sb->s_bdev;
751
Greg Kroah-Hartmanfa675762006-02-22 09:39:02 -0800752 bdev_uevent(bdev, KOBJ_UMOUNT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753 generic_shutdown_super(sb);
754 sync_blockdev(bdev);
755 close_bdev_excl(bdev);
756}
757
758EXPORT_SYMBOL(kill_block_super);
759
David Howells454e2392006-06-23 02:02:57 -0700760int get_sb_nodev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761 int flags, void *data,
David Howells454e2392006-06-23 02:02:57 -0700762 int (*fill_super)(struct super_block *, void *, int),
763 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764{
765 int error;
766 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
767
768 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700769 return PTR_ERR(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770
771 s->s_flags = flags;
772
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800773 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774 if (error) {
775 up_write(&s->s_umount);
776 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700777 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778 }
779 s->s_flags |= MS_ACTIVE;
David Howells454e2392006-06-23 02:02:57 -0700780 return simple_set_mnt(mnt, s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781}
782
783EXPORT_SYMBOL(get_sb_nodev);
784
785static int compare_single(struct super_block *s, void *p)
786{
787 return 1;
788}
789
David Howells454e2392006-06-23 02:02:57 -0700790int get_sb_single(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791 int flags, void *data,
David Howells454e2392006-06-23 02:02:57 -0700792 int (*fill_super)(struct super_block *, void *, int),
793 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794{
795 struct super_block *s;
796 int error;
797
798 s = sget(fs_type, compare_single, set_anon_super, NULL);
799 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700800 return PTR_ERR(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801 if (!s->s_root) {
802 s->s_flags = flags;
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800803 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804 if (error) {
805 up_write(&s->s_umount);
806 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700807 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808 }
809 s->s_flags |= MS_ACTIVE;
810 }
811 do_remount_sb(s, flags, data, 0);
David Howells454e2392006-06-23 02:02:57 -0700812 return simple_set_mnt(mnt, s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813}
814
815EXPORT_SYMBOL(get_sb_single);
816
817struct vfsmount *
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400818vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820 struct vfsmount *mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821 char *secdata = NULL;
David Howells454e2392006-06-23 02:02:57 -0700822 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823
824 if (!type)
825 return ERR_PTR(-ENODEV);
826
David Howells454e2392006-06-23 02:02:57 -0700827 error = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828 mnt = alloc_vfsmnt(name);
829 if (!mnt)
830 goto out;
831
832 if (data) {
833 secdata = alloc_secdata();
David Howells454e2392006-06-23 02:02:57 -0700834 if (!secdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835 goto out_mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836
837 error = security_sb_copy_data(type, data, secdata);
David Howells454e2392006-06-23 02:02:57 -0700838 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839 goto out_free_secdata;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840 }
841
David Howells454e2392006-06-23 02:02:57 -0700842 error = type->get_sb(type, flags, name, data, mnt);
843 if (error < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844 goto out_free_secdata;
David Howells454e2392006-06-23 02:02:57 -0700845
846 error = security_sb_kern_mount(mnt->mnt_sb, secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847 if (error)
848 goto out_sb;
David Howells454e2392006-06-23 02:02:57 -0700849
850 mnt->mnt_mountpoint = mnt->mnt_root;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851 mnt->mnt_parent = mnt;
David Howells454e2392006-06-23 02:02:57 -0700852 up_write(&mnt->mnt_sb->s_umount);
Gerald Schaefer8680e222005-06-21 17:15:16 -0700853 free_secdata(secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854 return mnt;
855out_sb:
David Howells454e2392006-06-23 02:02:57 -0700856 dput(mnt->mnt_root);
857 up_write(&mnt->mnt_sb->s_umount);
858 deactivate_super(mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859out_free_secdata:
860 free_secdata(secdata);
861out_mnt:
862 free_vfsmnt(mnt);
863out:
David Howells454e2392006-06-23 02:02:57 -0700864 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865}
866
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400867EXPORT_SYMBOL_GPL(vfs_kern_mount);
868
869struct vfsmount *
870do_kern_mount(const char *fstype, int flags, const char *name, void *data)
871{
872 struct file_system_type *type = get_fs_type(fstype);
873 struct vfsmount *mnt;
874 if (!type)
875 return ERR_PTR(-ENODEV);
876 mnt = vfs_kern_mount(type, flags, name, data);
877 put_filesystem(type);
878 return mnt;
879}
880
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881struct vfsmount *kern_mount(struct file_system_type *type)
882{
Trond Myklebust1f5ce9e2006-06-09 09:34:16 -0400883 return vfs_kern_mount(type, 0, type->name, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884}
885
886EXPORT_SYMBOL(kern_mount);