]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - fs/super.c
fs: move mark_files_ro into file_table.c
[mirror_ubuntu-bionic-kernel.git] / fs / super.c
CommitLineData
1da177e4
LT
1/*
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:
96de0e25 18 * Torbjörn Lindh (torbjorn.lindh@gopta.se), April 14, 1996.
1da177e4
LT
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
1da177e4
LT
23#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>
353ab6e9 39#include <linux/mutex.h>
49e0d02c 40#include <linux/file.h>
efaee192 41#include <linux/async.h>
1da177e4 42#include <asm/uaccess.h>
6d59e7f5 43#include "internal.h"
1da177e4
LT
44
45
1da177e4
LT
46LIST_HEAD(super_blocks);
47DEFINE_SPINLOCK(sb_lock);
48
49/**
50 * alloc_super - create new superblock
fe2bbc48 51 * @type: filesystem type superblock should belong to
1da177e4
LT
52 *
53 * Allocates and initializes a new &struct super_block. alloc_super()
54 * returns a pointer new superblock or %NULL if allocation had failed.
55 */
cf516249 56static struct super_block *alloc_super(struct file_system_type *type)
1da177e4 57{
11b0b5ab 58 struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER);
1da177e4
LT
59 static struct super_operations default_op;
60
61 if (s) {
1da177e4
LT
62 if (security_sb_alloc(s)) {
63 kfree(s);
64 s = NULL;
65 goto out;
66 }
67 INIT_LIST_HEAD(&s->s_dirty);
68 INIT_LIST_HEAD(&s->s_io);
0e0f4fc2 69 INIT_LIST_HEAD(&s->s_more_io);
1da177e4
LT
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);
da3bbdd4 74 INIT_LIST_HEAD(&s->s_dentry_lru);
efaee192 75 INIT_LIST_HEAD(&s->s_async_list);
1da177e4 76 init_rwsem(&s->s_umount);
7892f2f4 77 mutex_init(&s->s_lock);
897c6ff9 78 lockdep_set_class(&s->s_umount, &type->s_umount_key);
cf516249
IM
79 /*
80 * The locking rules for s_lock are up to the
81 * filesystem. For example ext3fs has different
82 * lock ordering than usbfs:
83 */
84 lockdep_set_class(&s->s_lock, &type->s_lock_key);
ada723dc
PZ
85 /*
86 * sget() can have s_umount recursion.
87 *
88 * When it cannot find a suitable sb, it allocates a new
89 * one (this one), and tries again to find a suitable old
90 * one.
91 *
92 * In case that succeeds, it will acquire the s_umount
93 * lock of the old one. Since these are clearly distrinct
94 * locks, and this object isn't exposed yet, there's no
95 * risk of deadlocks.
96 *
97 * Annotate this by putting this lock in a different
98 * subclass.
99 */
100 down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING);
1da177e4
LT
101 s->s_count = S_BIAS;
102 atomic_set(&s->s_active, 1);
a11f3a05 103 mutex_init(&s->s_vfs_rename_mutex);
d3be915f
IM
104 mutex_init(&s->s_dquot.dqio_mutex);
105 mutex_init(&s->s_dquot.dqonoff_mutex);
1da177e4
LT
106 init_rwsem(&s->s_dquot.dqptr_sem);
107 init_waitqueue_head(&s->s_wait_unfrozen);
108 s->s_maxbytes = MAX_NON_LFS;
109 s->dq_op = sb_dquot_ops;
110 s->s_qcop = sb_quotactl_ops;
111 s->s_op = &default_op;
112 s->s_time_gran = 1000000000;
113 }
114out:
115 return s;
116}
117
118/**
119 * destroy_super - frees a superblock
120 * @s: superblock to free
121 *
122 * Frees a superblock.
123 */
124static inline void destroy_super(struct super_block *s)
125{
126 security_sb_free(s);
79c0b2df 127 kfree(s->s_subtype);
b3b304a2 128 kfree(s->s_options);
1da177e4
LT
129 kfree(s);
130}
131
132/* Superblock refcounting */
133
134/*
135 * Drop a superblock's refcount. Returns non-zero if the superblock was
136 * destroyed. The caller must hold sb_lock.
137 */
6b09ae66 138static int __put_super(struct super_block *sb)
1da177e4
LT
139{
140 int ret = 0;
141
142 if (!--sb->s_count) {
143 destroy_super(sb);
144 ret = 1;
145 }
146 return ret;
147}
148
149/*
150 * Drop a superblock's refcount.
151 * Returns non-zero if the superblock is about to be destroyed and
152 * at least is already removed from super_blocks list, so if we are
153 * making a loop through super blocks then we need to restart.
154 * The caller must hold sb_lock.
155 */
156int __put_super_and_need_restart(struct super_block *sb)
157{
158 /* check for race with generic_shutdown_super() */
159 if (list_empty(&sb->s_list)) {
160 /* super block is removed, need to restart... */
161 __put_super(sb);
162 return 1;
163 }
164 /* can't be the last, since s_list is still in use */
165 sb->s_count--;
166 BUG_ON(sb->s_count == 0);
167 return 0;
168}
169
170/**
171 * put_super - drop a temporary reference to superblock
172 * @sb: superblock in question
173 *
174 * Drops a temporary reference, frees superblock if there's no
175 * references left.
176 */
177static void put_super(struct super_block *sb)
178{
179 spin_lock(&sb_lock);
180 __put_super(sb);
181 spin_unlock(&sb_lock);
182}
183
184
185/**
186 * deactivate_super - drop an active reference to superblock
187 * @s: superblock to deactivate
188 *
189 * Drops an active reference to superblock, acquiring a temprory one if
190 * there is no active references left. In that case we lock superblock,
191 * tell fs driver to shut it down and drop the temporary reference we
192 * had just acquired.
193 */
194void deactivate_super(struct super_block *s)
195{
196 struct file_system_type *fs = s->s_type;
197 if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
198 s->s_count -= S_BIAS-1;
199 spin_unlock(&sb_lock);
9e3509e2 200 vfs_dq_off(s, 0);
1da177e4
LT
201 down_write(&s->s_umount);
202 fs->kill_sb(s);
203 put_filesystem(fs);
204 put_super(s);
205 }
206}
207
208EXPORT_SYMBOL(deactivate_super);
209
74dbbdd7
AV
210/**
211 * deactivate_locked_super - drop an active reference to superblock
212 * @s: superblock to deactivate
213 *
214 * Equivalent of up_write(&s->s_umount); deactivate_super(s);, except that
215 * it does not unlock it until it's all over. As the result, it's safe to
216 * use to dispose of new superblock on ->get_sb() failure exits - nobody
217 * will see the sucker until it's all over. Equivalent using up_write +
218 * deactivate_super is safe for that purpose only if superblock is either
219 * safe to use or has NULL ->s_root when we unlock.
220 */
221void deactivate_locked_super(struct super_block *s)
222{
223 struct file_system_type *fs = s->s_type;
224 if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
225 s->s_count -= S_BIAS-1;
226 spin_unlock(&sb_lock);
227 vfs_dq_off(s, 0);
228 fs->kill_sb(s);
229 put_filesystem(fs);
230 put_super(s);
231 } else {
232 up_write(&s->s_umount);
233 }
234}
235
236EXPORT_SYMBOL(deactivate_locked_super);
237
1da177e4
LT
238/**
239 * grab_super - acquire an active reference
240 * @s: reference we are trying to make active
241 *
242 * Tries to acquire an active reference. grab_super() is used when we
243 * had just found a superblock in super_blocks or fs_type->fs_supers
244 * and want to turn it into a full-blown active reference. grab_super()
245 * is called with sb_lock held and drops it. Returns 1 in case of
246 * success, 0 if we had failed (superblock contents was already dead or
247 * dying when grab_super() had been called).
248 */
9c4dbee7 249static int grab_super(struct super_block *s) __releases(sb_lock)
1da177e4
LT
250{
251 s->s_count++;
252 spin_unlock(&sb_lock);
253 down_write(&s->s_umount);
254 if (s->s_root) {
255 spin_lock(&sb_lock);
256 if (s->s_count > S_BIAS) {
257 atomic_inc(&s->s_active);
258 s->s_count--;
259 spin_unlock(&sb_lock);
260 return 1;
261 }
262 spin_unlock(&sb_lock);
263 }
264 up_write(&s->s_umount);
265 put_super(s);
266 yield();
267 return 0;
268}
269
914e2637
AV
270/*
271 * Superblock locking. We really ought to get rid of these two.
272 */
273void lock_super(struct super_block * sb)
274{
275 get_fs_excl();
276 mutex_lock(&sb->s_lock);
277}
278
279void unlock_super(struct super_block * sb)
280{
281 put_fs_excl();
282 mutex_unlock(&sb->s_lock);
283}
284
285EXPORT_SYMBOL(lock_super);
286EXPORT_SYMBOL(unlock_super);
287
cf9a2ae8
DH
288/*
289 * Write out and wait upon all dirty data associated with this
290 * superblock. Filesystem data as well as the underlying block
291 * device. Takes the superblock lock. Requires a second blkdev
292 * flush by the caller to complete the operation.
293 */
294void __fsync_super(struct super_block *sb)
295{
296 sync_inodes_sb(sb, 0);
9e3509e2 297 vfs_dq_sync(sb);
cf9a2ae8
DH
298 lock_super(sb);
299 if (sb->s_dirt && sb->s_op->write_super)
300 sb->s_op->write_super(sb);
301 unlock_super(sb);
302 if (sb->s_op->sync_fs)
303 sb->s_op->sync_fs(sb, 1);
304 sync_blockdev(sb->s_bdev);
305 sync_inodes_sb(sb, 1);
306}
307
308/*
309 * Write out and wait upon all dirty data associated with this
310 * superblock. Filesystem data as well as the underlying block
311 * device. Takes the superblock lock.
312 */
313int fsync_super(struct super_block *sb)
314{
315 __fsync_super(sb);
316 return sync_blockdev(sb->s_bdev);
317}
800a9647 318EXPORT_SYMBOL_GPL(fsync_super);
cf9a2ae8 319
1da177e4
LT
320/**
321 * generic_shutdown_super - common helper for ->kill_sb()
322 * @sb: superblock to kill
323 *
324 * generic_shutdown_super() does all fs-independent work on superblock
325 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
326 * that need destruction out of superblock, call generic_shutdown_super()
327 * and release aforementioned objects. Note: dentries and inodes _are_
328 * taken care of and do not need specific handling.
c636ebdb
DH
329 *
330 * Upon calling this function, the filesystem may no longer alter or
331 * rearrange the set of dentries belonging to this super_block, nor may it
332 * change the attachments of dentries to inodes.
1da177e4
LT
333 */
334void generic_shutdown_super(struct super_block *sb)
335{
ee9b6d61 336 const struct super_operations *sop = sb->s_op;
1da177e4 337
efaee192 338
c636ebdb
DH
339 if (sb->s_root) {
340 shrink_dcache_for_umount(sb);
1da177e4
LT
341 fsync_super(sb);
342 lock_super(sb);
343 sb->s_flags &= ~MS_ACTIVE;
efaee192
AV
344
345 /*
346 * wait for asynchronous fs operations to finish before going further
347 */
766ccb9e 348 async_synchronize_full_domain(&sb->s_async_list);
efaee192 349
1da177e4
LT
350 /* bad name - it should be evict_inodes() */
351 invalidate_inodes(sb);
352 lock_kernel();
353
354 if (sop->write_super && sb->s_dirt)
355 sop->write_super(sb);
356 if (sop->put_super)
357 sop->put_super(sb);
358
359 /* Forget any remaining inodes */
360 if (invalidate_inodes(sb)) {
7b4fe29e
DJ
361 printk("VFS: Busy inodes after unmount of %s. "
362 "Self-destruct in 5 seconds. Have a nice day...\n",
363 sb->s_id);
1da177e4
LT
364 }
365
366 unlock_kernel();
367 unlock_super(sb);
368 }
369 spin_lock(&sb_lock);
370 /* should be initialized for __put_super_and_need_restart() */
371 list_del_init(&sb->s_list);
372 list_del(&sb->s_instances);
373 spin_unlock(&sb_lock);
374 up_write(&sb->s_umount);
375}
376
377EXPORT_SYMBOL(generic_shutdown_super);
378
379/**
380 * sget - find or create a superblock
381 * @type: filesystem type superblock should belong to
382 * @test: comparison callback
383 * @set: setup callback
384 * @data: argument to each of them
385 */
386struct super_block *sget(struct file_system_type *type,
387 int (*test)(struct super_block *,void *),
388 int (*set)(struct super_block *,void *),
389 void *data)
390{
391 struct super_block *s = NULL;
d4730127 392 struct super_block *old;
1da177e4
LT
393 int err;
394
395retry:
396 spin_lock(&sb_lock);
d4730127
MK
397 if (test) {
398 list_for_each_entry(old, &type->fs_supers, s_instances) {
399 if (!test(old, data))
400 continue;
401 if (!grab_super(old))
402 goto retry;
a3cfbb53
LZ
403 if (s) {
404 up_write(&s->s_umount);
d4730127 405 destroy_super(s);
a3cfbb53 406 }
d4730127
MK
407 return old;
408 }
1da177e4
LT
409 }
410 if (!s) {
411 spin_unlock(&sb_lock);
cf516249 412 s = alloc_super(type);
1da177e4
LT
413 if (!s)
414 return ERR_PTR(-ENOMEM);
415 goto retry;
416 }
417
418 err = set(s, data);
419 if (err) {
420 spin_unlock(&sb_lock);
a3cfbb53 421 up_write(&s->s_umount);
1da177e4
LT
422 destroy_super(s);
423 return ERR_PTR(err);
424 }
425 s->s_type = type;
426 strlcpy(s->s_id, type->name, sizeof(s->s_id));
427 list_add_tail(&s->s_list, &super_blocks);
428 list_add(&s->s_instances, &type->fs_supers);
429 spin_unlock(&sb_lock);
430 get_filesystem(type);
431 return s;
432}
433
434EXPORT_SYMBOL(sget);
435
436void drop_super(struct super_block *sb)
437{
438 up_read(&sb->s_umount);
439 put_super(sb);
440}
441
442EXPORT_SYMBOL(drop_super);
443
444static inline void write_super(struct super_block *sb)
445{
446 lock_super(sb);
447 if (sb->s_root && sb->s_dirt)
448 if (sb->s_op->write_super)
449 sb->s_op->write_super(sb);
450 unlock_super(sb);
451}
452
453/*
454 * Note: check the dirty flag before waiting, so we don't
455 * hold up the sync while mounting a device. (The newly
456 * mounted device won't need syncing.)
457 */
458void sync_supers(void)
459{
618f0636
KK
460 struct super_block *sb;
461
1da177e4 462 spin_lock(&sb_lock);
618f0636
KK
463restart:
464 list_for_each_entry(sb, &super_blocks, s_list) {
1da177e4
LT
465 if (sb->s_dirt) {
466 sb->s_count++;
467 spin_unlock(&sb_lock);
468 down_read(&sb->s_umount);
469 write_super(sb);
618f0636
KK
470 up_read(&sb->s_umount);
471 spin_lock(&sb_lock);
472 if (__put_super_and_need_restart(sb))
473 goto restart;
474 }
475 }
1da177e4
LT
476 spin_unlock(&sb_lock);
477}
478
479/*
3a4fa0a2 480 * Call the ->sync_fs super_op against all filesystems which are r/w and
1da177e4
LT
481 * which implement it.
482 *
483 * This operation is careful to avoid the livelock which could easily happen
484 * if two or more filesystems are being continuously dirtied. s_need_sync_fs
485 * is used only here. We set it against all filesystems and then clear it as
486 * we sync them. So redirtied filesystems are skipped.
487 *
3a4fa0a2 488 * But if process A is currently running sync_filesystems and then process B
1da177e4
LT
489 * calls sync_filesystems as well, process B will set all the s_need_sync_fs
490 * flags again, which will cause process A to resync everything. Fix that with
491 * a local mutex.
492 *
493 * (Fabian) Avoid sync_fs with clean fs & wait mode 0
494 */
495void sync_filesystems(int wait)
496{
497 struct super_block *sb;
353ab6e9 498 static DEFINE_MUTEX(mutex);
1da177e4 499
353ab6e9 500 mutex_lock(&mutex); /* Could be down_interruptible */
1da177e4 501 spin_lock(&sb_lock);
618f0636 502 list_for_each_entry(sb, &super_blocks, s_list) {
1da177e4
LT
503 if (!sb->s_op->sync_fs)
504 continue;
505 if (sb->s_flags & MS_RDONLY)
506 continue;
507 sb->s_need_sync_fs = 1;
508 }
1da177e4
LT
509
510restart:
618f0636 511 list_for_each_entry(sb, &super_blocks, s_list) {
1da177e4
LT
512 if (!sb->s_need_sync_fs)
513 continue;
514 sb->s_need_sync_fs = 0;
515 if (sb->s_flags & MS_RDONLY)
516 continue; /* hm. Was remounted r/o meanwhile */
517 sb->s_count++;
518 spin_unlock(&sb_lock);
519 down_read(&sb->s_umount);
766ccb9e 520 async_synchronize_full_domain(&sb->s_async_list);
1da177e4
LT
521 if (sb->s_root && (wait || sb->s_dirt))
522 sb->s_op->sync_fs(sb, wait);
618f0636
KK
523 up_read(&sb->s_umount);
524 /* restart only when sb is no longer on the list */
525 spin_lock(&sb_lock);
526 if (__put_super_and_need_restart(sb))
527 goto restart;
1da177e4
LT
528 }
529 spin_unlock(&sb_lock);
353ab6e9 530 mutex_unlock(&mutex);
1da177e4
LT
531}
532
533/**
534 * get_super - get the superblock of a device
535 * @bdev: device to get the superblock for
536 *
537 * Scans the superblock list and finds the superblock of the file system
538 * mounted on the device given. %NULL is returned if no match is found.
539 */
540
541struct super_block * get_super(struct block_device *bdev)
542{
618f0636
KK
543 struct super_block *sb;
544
1da177e4
LT
545 if (!bdev)
546 return NULL;
618f0636 547
1da177e4 548 spin_lock(&sb_lock);
618f0636
KK
549rescan:
550 list_for_each_entry(sb, &super_blocks, s_list) {
551 if (sb->s_bdev == bdev) {
552 sb->s_count++;
1da177e4 553 spin_unlock(&sb_lock);
618f0636
KK
554 down_read(&sb->s_umount);
555 if (sb->s_root)
556 return sb;
557 up_read(&sb->s_umount);
558 /* restart only when sb is no longer on the list */
559 spin_lock(&sb_lock);
560 if (__put_super_and_need_restart(sb))
561 goto rescan;
1da177e4
LT
562 }
563 }
564 spin_unlock(&sb_lock);
565 return NULL;
566}
567
568EXPORT_SYMBOL(get_super);
569
570struct super_block * user_get_super(dev_t dev)
571{
618f0636 572 struct super_block *sb;
1da177e4 573
1da177e4 574 spin_lock(&sb_lock);
618f0636
KK
575rescan:
576 list_for_each_entry(sb, &super_blocks, s_list) {
577 if (sb->s_dev == dev) {
578 sb->s_count++;
1da177e4 579 spin_unlock(&sb_lock);
618f0636
KK
580 down_read(&sb->s_umount);
581 if (sb->s_root)
582 return sb;
583 up_read(&sb->s_umount);
584 /* restart only when sb is no longer on the list */
585 spin_lock(&sb_lock);
586 if (__put_super_and_need_restart(sb))
587 goto rescan;
1da177e4
LT
588 }
589 }
590 spin_unlock(&sb_lock);
591 return NULL;
592}
593
257ac264 594SYSCALL_DEFINE2(ustat, unsigned, dev, struct ustat __user *, ubuf)
1da177e4
LT
595{
596 struct super_block *s;
597 struct ustat tmp;
598 struct kstatfs sbuf;
599 int err = -EINVAL;
600
601 s = user_get_super(new_decode_dev(dev));
602 if (s == NULL)
603 goto out;
726c3342 604 err = vfs_statfs(s->s_root, &sbuf);
1da177e4
LT
605 drop_super(s);
606 if (err)
607 goto out;
608
609 memset(&tmp,0,sizeof(struct ustat));
610 tmp.f_tfree = sbuf.f_bfree;
611 tmp.f_tinode = sbuf.f_ffree;
612
613 err = copy_to_user(ubuf,&tmp,sizeof(struct ustat)) ? -EFAULT : 0;
614out:
615 return err;
616}
617
1da177e4
LT
618/**
619 * do_remount_sb - asks filesystem to change mount options.
620 * @sb: superblock in question
621 * @flags: numeric part of options
622 * @data: the rest of options
623 * @force: whether or not to force the change
624 *
625 * Alters the mount options of a mounted file system.
626 */
627int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
628{
629 int retval;
0ff5af83 630 int remount_rw;
1da177e4 631
9361401e 632#ifdef CONFIG_BLOCK
1da177e4
LT
633 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
634 return -EACCES;
9361401e 635#endif
1da177e4
LT
636 if (flags & MS_RDONLY)
637 acct_auto_close(sb);
638 shrink_dcache_sb(sb);
639 fsync_super(sb);
640
641 /* If we are remounting RDONLY and current sb is read/write,
642 make sure there are no rw files opened */
643 if ((flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY)) {
644 if (force)
645 mark_files_ro(sb);
646 else if (!fs_may_remount_ro(sb))
647 return -EBUSY;
9e3509e2 648 retval = vfs_dq_off(sb, 1);
0ff5af83
JK
649 if (retval < 0 && retval != -ENOSYS)
650 return -EBUSY;
1da177e4 651 }
0ff5af83 652 remount_rw = !(flags & MS_RDONLY) && (sb->s_flags & MS_RDONLY);
1da177e4
LT
653
654 if (sb->s_op->remount_fs) {
655 lock_super(sb);
656 retval = sb->s_op->remount_fs(sb, &flags, data);
657 unlock_super(sb);
658 if (retval)
659 return retval;
660 }
661 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
0ff5af83 662 if (remount_rw)
9e3509e2 663 vfs_dq_quota_on_remount(sb);
1da177e4
LT
664 return 0;
665}
666
a2a9537a 667static void do_emergency_remount(struct work_struct *work)
1da177e4
LT
668{
669 struct super_block *sb;
670
671 spin_lock(&sb_lock);
672 list_for_each_entry(sb, &super_blocks, s_list) {
673 sb->s_count++;
674 spin_unlock(&sb_lock);
675 down_read(&sb->s_umount);
676 if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) {
677 /*
678 * ->remount_fs needs lock_kernel().
679 *
680 * What lock protects sb->s_flags??
681 */
682 lock_kernel();
683 do_remount_sb(sb, MS_RDONLY, NULL, 1);
684 unlock_kernel();
685 }
686 drop_super(sb);
687 spin_lock(&sb_lock);
688 }
689 spin_unlock(&sb_lock);
a2a9537a 690 kfree(work);
1da177e4
LT
691 printk("Emergency Remount complete\n");
692}
693
694void emergency_remount(void)
695{
a2a9537a
JA
696 struct work_struct *work;
697
698 work = kmalloc(sizeof(*work), GFP_ATOMIC);
699 if (work) {
700 INIT_WORK(work, do_emergency_remount);
701 schedule_work(work);
702 }
1da177e4
LT
703}
704
705/*
706 * Unnamed block devices are dummy devices used by virtual
707 * filesystems which don't use real block-devices. -- jrs
708 */
709
ad76cbc6 710static DEFINE_IDA(unnamed_dev_ida);
1da177e4
LT
711static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
712
713int set_anon_super(struct super_block *s, void *data)
714{
715 int dev;
716 int error;
717
718 retry:
ad76cbc6 719 if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0)
1da177e4
LT
720 return -ENOMEM;
721 spin_lock(&unnamed_dev_lock);
ad76cbc6 722 error = ida_get_new(&unnamed_dev_ida, &dev);
1da177e4
LT
723 spin_unlock(&unnamed_dev_lock);
724 if (error == -EAGAIN)
725 /* We raced and lost with another CPU. */
726 goto retry;
727 else if (error)
728 return -EAGAIN;
729
730 if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
731 spin_lock(&unnamed_dev_lock);
ad76cbc6 732 ida_remove(&unnamed_dev_ida, dev);
1da177e4
LT
733 spin_unlock(&unnamed_dev_lock);
734 return -EMFILE;
735 }
736 s->s_dev = MKDEV(0, dev & MINORMASK);
737 return 0;
738}
739
740EXPORT_SYMBOL(set_anon_super);
741
742void kill_anon_super(struct super_block *sb)
743{
744 int slot = MINOR(sb->s_dev);
745
746 generic_shutdown_super(sb);
747 spin_lock(&unnamed_dev_lock);
ad76cbc6 748 ida_remove(&unnamed_dev_ida, slot);
1da177e4
LT
749 spin_unlock(&unnamed_dev_lock);
750}
751
752EXPORT_SYMBOL(kill_anon_super);
753
1da177e4
LT
754void kill_litter_super(struct super_block *sb)
755{
756 if (sb->s_root)
757 d_genocide(sb->s_root);
758 kill_anon_super(sb);
759}
760
761EXPORT_SYMBOL(kill_litter_super);
762
909e6d94
SH
763static int ns_test_super(struct super_block *sb, void *data)
764{
765 return sb->s_fs_info == data;
766}
767
768static int ns_set_super(struct super_block *sb, void *data)
769{
770 sb->s_fs_info = data;
771 return set_anon_super(sb, NULL);
772}
773
774int get_sb_ns(struct file_system_type *fs_type, int flags, void *data,
775 int (*fill_super)(struct super_block *, void *, int),
776 struct vfsmount *mnt)
777{
778 struct super_block *sb;
779
780 sb = sget(fs_type, ns_test_super, ns_set_super, data);
781 if (IS_ERR(sb))
782 return PTR_ERR(sb);
783
784 if (!sb->s_root) {
785 int err;
786 sb->s_flags = flags;
787 err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
788 if (err) {
74dbbdd7 789 deactivate_locked_super(sb);
909e6d94
SH
790 return err;
791 }
792
793 sb->s_flags |= MS_ACTIVE;
794 }
795
796 simple_set_mnt(mnt, sb);
797 return 0;
798}
799
800EXPORT_SYMBOL(get_sb_ns);
801
9361401e 802#ifdef CONFIG_BLOCK
1da177e4
LT
803static int set_bdev_super(struct super_block *s, void *data)
804{
805 s->s_bdev = data;
806 s->s_dev = s->s_bdev->bd_dev;
807 return 0;
808}
809
810static int test_bdev_super(struct super_block *s, void *data)
811{
812 return (void *)s->s_bdev == data;
813}
814
454e2398 815int get_sb_bdev(struct file_system_type *fs_type,
1da177e4 816 int flags, const char *dev_name, void *data,
454e2398
DH
817 int (*fill_super)(struct super_block *, void *, int),
818 struct vfsmount *mnt)
1da177e4
LT
819{
820 struct block_device *bdev;
821 struct super_block *s;
30c40d2c 822 fmode_t mode = FMODE_READ;
1da177e4
LT
823 int error = 0;
824
30c40d2c
AV
825 if (!(flags & MS_RDONLY))
826 mode |= FMODE_WRITE;
827
828 bdev = open_bdev_exclusive(dev_name, mode, fs_type);
1da177e4 829 if (IS_ERR(bdev))
454e2398 830 return PTR_ERR(bdev);
1da177e4
LT
831
832 /*
833 * once the super is inserted into the list by sget, s_umount
834 * will protect the lockfs code from trying to start a snapshot
835 * while we are mounting
836 */
f73ca1b7 837 down(&bdev->bd_mount_sem);
1da177e4 838 s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
f73ca1b7 839 up(&bdev->bd_mount_sem);
1da177e4 840 if (IS_ERR(s))
454e2398 841 goto error_s;
1da177e4
LT
842
843 if (s->s_root) {
844 if ((flags ^ s->s_flags) & MS_RDONLY) {
74dbbdd7 845 deactivate_locked_super(s);
454e2398
DH
846 error = -EBUSY;
847 goto error_bdev;
1da177e4 848 }
454e2398 849
30c40d2c 850 close_bdev_exclusive(bdev, mode);
1da177e4
LT
851 } else {
852 char b[BDEVNAME_SIZE];
853
854 s->s_flags = flags;
30c40d2c 855 s->s_mode = mode;
1da177e4 856 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
e78c9a00 857 sb_set_blocksize(s, block_size(bdev));
9b04c997 858 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
1da177e4 859 if (error) {
74dbbdd7 860 deactivate_locked_super(s);
454e2398 861 goto error;
fa675765 862 }
454e2398
DH
863
864 s->s_flags |= MS_ACTIVE;
87d8fe1e 865 bdev->bd_super = s;
1da177e4
LT
866 }
867
a3ec947c
SB
868 simple_set_mnt(mnt, s);
869 return 0;
1da177e4 870
454e2398
DH
871error_s:
872 error = PTR_ERR(s);
873error_bdev:
30c40d2c 874 close_bdev_exclusive(bdev, mode);
454e2398
DH
875error:
876 return error;
1da177e4
LT
877}
878
879EXPORT_SYMBOL(get_sb_bdev);
880
881void kill_block_super(struct super_block *sb)
882{
883 struct block_device *bdev = sb->s_bdev;
30c40d2c 884 fmode_t mode = sb->s_mode;
1da177e4 885
ddbaaf30 886 bdev->bd_super = NULL;
1da177e4
LT
887 generic_shutdown_super(sb);
888 sync_blockdev(bdev);
30c40d2c 889 close_bdev_exclusive(bdev, mode);
1da177e4
LT
890}
891
892EXPORT_SYMBOL(kill_block_super);
9361401e 893#endif
1da177e4 894
454e2398 895int get_sb_nodev(struct file_system_type *fs_type,
1da177e4 896 int flags, void *data,
454e2398
DH
897 int (*fill_super)(struct super_block *, void *, int),
898 struct vfsmount *mnt)
1da177e4
LT
899{
900 int error;
901 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
902
903 if (IS_ERR(s))
454e2398 904 return PTR_ERR(s);
1da177e4
LT
905
906 s->s_flags = flags;
907
9b04c997 908 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
1da177e4 909 if (error) {
74dbbdd7 910 deactivate_locked_super(s);
454e2398 911 return error;
1da177e4
LT
912 }
913 s->s_flags |= MS_ACTIVE;
a3ec947c
SB
914 simple_set_mnt(mnt, s);
915 return 0;
1da177e4
LT
916}
917
918EXPORT_SYMBOL(get_sb_nodev);
919
920static int compare_single(struct super_block *s, void *p)
921{
922 return 1;
923}
924
454e2398 925int get_sb_single(struct file_system_type *fs_type,
1da177e4 926 int flags, void *data,
454e2398
DH
927 int (*fill_super)(struct super_block *, void *, int),
928 struct vfsmount *mnt)
1da177e4
LT
929{
930 struct super_block *s;
931 int error;
932
933 s = sget(fs_type, compare_single, set_anon_super, NULL);
934 if (IS_ERR(s))
454e2398 935 return PTR_ERR(s);
1da177e4
LT
936 if (!s->s_root) {
937 s->s_flags = flags;
9b04c997 938 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
1da177e4 939 if (error) {
74dbbdd7 940 deactivate_locked_super(s);
454e2398 941 return error;
1da177e4
LT
942 }
943 s->s_flags |= MS_ACTIVE;
944 }
945 do_remount_sb(s, flags, data, 0);
a3ec947c
SB
946 simple_set_mnt(mnt, s);
947 return 0;
1da177e4
LT
948}
949
950EXPORT_SYMBOL(get_sb_single);
951
952struct vfsmount *
bb4a58bf 953vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data)
1da177e4 954{
1da177e4 955 struct vfsmount *mnt;
1da177e4 956 char *secdata = NULL;
454e2398 957 int error;
1da177e4
LT
958
959 if (!type)
960 return ERR_PTR(-ENODEV);
1da177e4 961
454e2398 962 error = -ENOMEM;
1da177e4
LT
963 mnt = alloc_vfsmnt(name);
964 if (!mnt)
965 goto out;
966
e0007529 967 if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) {
1da177e4 968 secdata = alloc_secdata();
454e2398 969 if (!secdata)
1da177e4 970 goto out_mnt;
1da177e4 971
e0007529 972 error = security_sb_copy_data(data, secdata);
454e2398 973 if (error)
1da177e4 974 goto out_free_secdata;
1da177e4
LT
975 }
976
454e2398
DH
977 error = type->get_sb(type, flags, name, data, mnt);
978 if (error < 0)
1da177e4 979 goto out_free_secdata;
b4c07bce 980 BUG_ON(!mnt->mnt_sb);
454e2398 981
12204e24 982 error = security_sb_kern_mount(mnt->mnt_sb, flags, secdata);
1da177e4
LT
983 if (error)
984 goto out_sb;
454e2398
DH
985
986 mnt->mnt_mountpoint = mnt->mnt_root;
1da177e4 987 mnt->mnt_parent = mnt;
454e2398 988 up_write(&mnt->mnt_sb->s_umount);
8680e22f 989 free_secdata(secdata);
1da177e4
LT
990 return mnt;
991out_sb:
454e2398 992 dput(mnt->mnt_root);
74dbbdd7 993 deactivate_locked_super(mnt->mnt_sb);
1da177e4
LT
994out_free_secdata:
995 free_secdata(secdata);
996out_mnt:
997 free_vfsmnt(mnt);
998out:
454e2398 999 return ERR_PTR(error);
1da177e4
LT
1000}
1001
bb4a58bf
TM
1002EXPORT_SYMBOL_GPL(vfs_kern_mount);
1003
79c0b2df
MS
1004static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype)
1005{
1006 int err;
1007 const char *subtype = strchr(fstype, '.');
1008 if (subtype) {
1009 subtype++;
1010 err = -EINVAL;
1011 if (!subtype[0])
1012 goto err;
1013 } else
1014 subtype = "";
1015
1016 mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL);
1017 err = -ENOMEM;
1018 if (!mnt->mnt_sb->s_subtype)
1019 goto err;
1020 return mnt;
1021
1022 err:
1023 mntput(mnt);
1024 return ERR_PTR(err);
1025}
1026
bb4a58bf
TM
1027struct vfsmount *
1028do_kern_mount(const char *fstype, int flags, const char *name, void *data)
1029{
1030 struct file_system_type *type = get_fs_type(fstype);
1031 struct vfsmount *mnt;
1032 if (!type)
1033 return ERR_PTR(-ENODEV);
1034 mnt = vfs_kern_mount(type, flags, name, data);
79c0b2df
MS
1035 if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) &&
1036 !mnt->mnt_sb->s_subtype)
1037 mnt = fs_set_subtype(mnt, fstype);
bb4a58bf
TM
1038 put_filesystem(type);
1039 return mnt;
1040}
8a4e98d9 1041EXPORT_SYMBOL_GPL(do_kern_mount);
bb4a58bf 1042
8bf9725c 1043struct vfsmount *kern_mount_data(struct file_system_type *type, void *data)
1da177e4 1044{
8bf9725c 1045 return vfs_kern_mount(type, MS_KERNMOUNT, type->name, data);
1da177e4
LT
1046}
1047
8bf9725c 1048EXPORT_SYMBOL_GPL(kern_mount_data);