]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/super.c
Make allocation of anon devices cheaper
[net-next-2.6.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>
1da177e4
LT
31#include <linux/mount.h>
32#include <linux/security.h>
33#include <linux/syscalls.h>
34#include <linux/vfs.h>
35#include <linux/writeback.h> /* for the emergency remount stuff */
36#include <linux/idr.h>
37#include <linux/kobject.h>
353ab6e9 38#include <linux/mutex.h>
49e0d02c 39#include <linux/file.h>
1da177e4 40#include <asm/uaccess.h>
6d59e7f5 41#include "internal.h"
1da177e4
LT
42
43
1da177e4
LT
44LIST_HEAD(super_blocks);
45DEFINE_SPINLOCK(sb_lock);
46
47/**
48 * alloc_super - create new superblock
fe2bbc48 49 * @type: filesystem type superblock should belong to
1da177e4
LT
50 *
51 * Allocates and initializes a new &struct super_block. alloc_super()
52 * returns a pointer new superblock or %NULL if allocation had failed.
53 */
cf516249 54static struct super_block *alloc_super(struct file_system_type *type)
1da177e4 55{
11b0b5ab 56 struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER);
1da177e4
LT
57 static struct super_operations default_op;
58
59 if (s) {
1da177e4
LT
60 if (security_sb_alloc(s)) {
61 kfree(s);
62 s = NULL;
63 goto out;
64 }
65 INIT_LIST_HEAD(&s->s_dirty);
66 INIT_LIST_HEAD(&s->s_io);
0e0f4fc2 67 INIT_LIST_HEAD(&s->s_more_io);
1da177e4
LT
68 INIT_LIST_HEAD(&s->s_files);
69 INIT_LIST_HEAD(&s->s_instances);
70 INIT_HLIST_HEAD(&s->s_anon);
71 INIT_LIST_HEAD(&s->s_inodes);
da3bbdd4 72 INIT_LIST_HEAD(&s->s_dentry_lru);
1da177e4 73 init_rwsem(&s->s_umount);
7892f2f4 74 mutex_init(&s->s_lock);
897c6ff9 75 lockdep_set_class(&s->s_umount, &type->s_umount_key);
cf516249
IM
76 /*
77 * The locking rules for s_lock are up to the
78 * filesystem. For example ext3fs has different
79 * lock ordering than usbfs:
80 */
81 lockdep_set_class(&s->s_lock, &type->s_lock_key);
ada723dc
PZ
82 /*
83 * sget() can have s_umount recursion.
84 *
85 * When it cannot find a suitable sb, it allocates a new
86 * one (this one), and tries again to find a suitable old
87 * one.
88 *
89 * In case that succeeds, it will acquire the s_umount
90 * lock of the old one. Since these are clearly distrinct
91 * locks, and this object isn't exposed yet, there's no
92 * risk of deadlocks.
93 *
94 * Annotate this by putting this lock in a different
95 * subclass.
96 */
97 down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING);
1da177e4
LT
98 s->s_count = S_BIAS;
99 atomic_set(&s->s_active, 1);
a11f3a05 100 mutex_init(&s->s_vfs_rename_mutex);
d3be915f
IM
101 mutex_init(&s->s_dquot.dqio_mutex);
102 mutex_init(&s->s_dquot.dqonoff_mutex);
1da177e4
LT
103 init_rwsem(&s->s_dquot.dqptr_sem);
104 init_waitqueue_head(&s->s_wait_unfrozen);
105 s->s_maxbytes = MAX_NON_LFS;
106 s->dq_op = sb_dquot_ops;
107 s->s_qcop = sb_quotactl_ops;
108 s->s_op = &default_op;
109 s->s_time_gran = 1000000000;
110 }
111out:
112 return s;
113}
114
115/**
116 * destroy_super - frees a superblock
117 * @s: superblock to free
118 *
119 * Frees a superblock.
120 */
121static inline void destroy_super(struct super_block *s)
122{
123 security_sb_free(s);
79c0b2df 124 kfree(s->s_subtype);
b3b304a2 125 kfree(s->s_options);
1da177e4
LT
126 kfree(s);
127}
128
129/* Superblock refcounting */
130
131/*
132 * Drop a superblock's refcount. Returns non-zero if the superblock was
133 * destroyed. The caller must hold sb_lock.
134 */
6b09ae66 135static int __put_super(struct super_block *sb)
1da177e4
LT
136{
137 int ret = 0;
138
139 if (!--sb->s_count) {
140 destroy_super(sb);
141 ret = 1;
142 }
143 return ret;
144}
145
146/*
147 * Drop a superblock's refcount.
148 * Returns non-zero if the superblock is about to be destroyed and
149 * at least is already removed from super_blocks list, so if we are
150 * making a loop through super blocks then we need to restart.
151 * The caller must hold sb_lock.
152 */
153int __put_super_and_need_restart(struct super_block *sb)
154{
155 /* check for race with generic_shutdown_super() */
156 if (list_empty(&sb->s_list)) {
157 /* super block is removed, need to restart... */
158 __put_super(sb);
159 return 1;
160 }
161 /* can't be the last, since s_list is still in use */
162 sb->s_count--;
163 BUG_ON(sb->s_count == 0);
164 return 0;
165}
166
167/**
168 * put_super - drop a temporary reference to superblock
169 * @sb: superblock in question
170 *
171 * Drops a temporary reference, frees superblock if there's no
172 * references left.
173 */
174static void put_super(struct super_block *sb)
175{
176 spin_lock(&sb_lock);
177 __put_super(sb);
178 spin_unlock(&sb_lock);
179}
180
181
182/**
183 * deactivate_super - drop an active reference to superblock
184 * @s: superblock to deactivate
185 *
186 * Drops an active reference to superblock, acquiring a temprory one if
187 * there is no active references left. In that case we lock superblock,
188 * tell fs driver to shut it down and drop the temporary reference we
189 * had just acquired.
190 */
191void deactivate_super(struct super_block *s)
192{
193 struct file_system_type *fs = s->s_type;
194 if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
195 s->s_count -= S_BIAS-1;
196 spin_unlock(&sb_lock);
9e3509e2 197 vfs_dq_off(s, 0);
1da177e4
LT
198 down_write(&s->s_umount);
199 fs->kill_sb(s);
200 put_filesystem(fs);
201 put_super(s);
202 }
203}
204
205EXPORT_SYMBOL(deactivate_super);
206
74dbbdd7
AV
207/**
208 * deactivate_locked_super - drop an active reference to superblock
209 * @s: superblock to deactivate
210 *
211 * Equivalent of up_write(&s->s_umount); deactivate_super(s);, except that
212 * it does not unlock it until it's all over. As the result, it's safe to
213 * use to dispose of new superblock on ->get_sb() failure exits - nobody
214 * will see the sucker until it's all over. Equivalent using up_write +
215 * deactivate_super is safe for that purpose only if superblock is either
216 * safe to use or has NULL ->s_root when we unlock.
217 */
218void deactivate_locked_super(struct super_block *s)
219{
220 struct file_system_type *fs = s->s_type;
221 if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
222 s->s_count -= S_BIAS-1;
223 spin_unlock(&sb_lock);
224 vfs_dq_off(s, 0);
225 fs->kill_sb(s);
226 put_filesystem(fs);
227 put_super(s);
228 } else {
229 up_write(&s->s_umount);
230 }
231}
232
233EXPORT_SYMBOL(deactivate_locked_super);
234
1da177e4
LT
235/**
236 * grab_super - acquire an active reference
237 * @s: reference we are trying to make active
238 *
239 * Tries to acquire an active reference. grab_super() is used when we
240 * had just found a superblock in super_blocks or fs_type->fs_supers
241 * and want to turn it into a full-blown active reference. grab_super()
242 * is called with sb_lock held and drops it. Returns 1 in case of
243 * success, 0 if we had failed (superblock contents was already dead or
244 * dying when grab_super() had been called).
245 */
9c4dbee7 246static int grab_super(struct super_block *s) __releases(sb_lock)
1da177e4
LT
247{
248 s->s_count++;
249 spin_unlock(&sb_lock);
250 down_write(&s->s_umount);
251 if (s->s_root) {
252 spin_lock(&sb_lock);
253 if (s->s_count > S_BIAS) {
254 atomic_inc(&s->s_active);
255 s->s_count--;
256 spin_unlock(&sb_lock);
257 return 1;
258 }
259 spin_unlock(&sb_lock);
260 }
261 up_write(&s->s_umount);
262 put_super(s);
263 yield();
264 return 0;
265}
266
914e2637
AV
267/*
268 * Superblock locking. We really ought to get rid of these two.
269 */
270void lock_super(struct super_block * sb)
271{
272 get_fs_excl();
273 mutex_lock(&sb->s_lock);
274}
275
276void unlock_super(struct super_block * sb)
277{
278 put_fs_excl();
279 mutex_unlock(&sb->s_lock);
280}
281
282EXPORT_SYMBOL(lock_super);
283EXPORT_SYMBOL(unlock_super);
284
1da177e4
LT
285/**
286 * generic_shutdown_super - common helper for ->kill_sb()
287 * @sb: superblock to kill
288 *
289 * generic_shutdown_super() does all fs-independent work on superblock
290 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
291 * that need destruction out of superblock, call generic_shutdown_super()
292 * and release aforementioned objects. Note: dentries and inodes _are_
293 * taken care of and do not need specific handling.
c636ebdb
DH
294 *
295 * Upon calling this function, the filesystem may no longer alter or
296 * rearrange the set of dentries belonging to this super_block, nor may it
297 * change the attachments of dentries to inodes.
1da177e4
LT
298 */
299void generic_shutdown_super(struct super_block *sb)
300{
ee9b6d61 301 const struct super_operations *sop = sb->s_op;
1da177e4 302
efaee192 303
c636ebdb
DH
304 if (sb->s_root) {
305 shrink_dcache_for_umount(sb);
60b0680f 306 sync_filesystem(sb);
a9e220f8 307 get_fs_excl();
1da177e4 308 sb->s_flags &= ~MS_ACTIVE;
efaee192 309
1da177e4
LT
310 /* bad name - it should be evict_inodes() */
311 invalidate_inodes(sb);
1da177e4 312
1da177e4
LT
313 if (sop->put_super)
314 sop->put_super(sb);
315
316 /* Forget any remaining inodes */
317 if (invalidate_inodes(sb)) {
7b4fe29e
DJ
318 printk("VFS: Busy inodes after unmount of %s. "
319 "Self-destruct in 5 seconds. Have a nice day...\n",
320 sb->s_id);
1da177e4 321 }
a9e220f8 322 put_fs_excl();
1da177e4
LT
323 }
324 spin_lock(&sb_lock);
325 /* should be initialized for __put_super_and_need_restart() */
326 list_del_init(&sb->s_list);
327 list_del(&sb->s_instances);
328 spin_unlock(&sb_lock);
329 up_write(&sb->s_umount);
330}
331
332EXPORT_SYMBOL(generic_shutdown_super);
333
334/**
335 * sget - find or create a superblock
336 * @type: filesystem type superblock should belong to
337 * @test: comparison callback
338 * @set: setup callback
339 * @data: argument to each of them
340 */
341struct super_block *sget(struct file_system_type *type,
342 int (*test)(struct super_block *,void *),
343 int (*set)(struct super_block *,void *),
344 void *data)
345{
346 struct super_block *s = NULL;
d4730127 347 struct super_block *old;
1da177e4
LT
348 int err;
349
350retry:
351 spin_lock(&sb_lock);
d4730127
MK
352 if (test) {
353 list_for_each_entry(old, &type->fs_supers, s_instances) {
354 if (!test(old, data))
355 continue;
356 if (!grab_super(old))
357 goto retry;
a3cfbb53
LZ
358 if (s) {
359 up_write(&s->s_umount);
d4730127 360 destroy_super(s);
a3cfbb53 361 }
d4730127
MK
362 return old;
363 }
1da177e4
LT
364 }
365 if (!s) {
366 spin_unlock(&sb_lock);
cf516249 367 s = alloc_super(type);
1da177e4
LT
368 if (!s)
369 return ERR_PTR(-ENOMEM);
370 goto retry;
371 }
372
373 err = set(s, data);
374 if (err) {
375 spin_unlock(&sb_lock);
a3cfbb53 376 up_write(&s->s_umount);
1da177e4
LT
377 destroy_super(s);
378 return ERR_PTR(err);
379 }
380 s->s_type = type;
381 strlcpy(s->s_id, type->name, sizeof(s->s_id));
382 list_add_tail(&s->s_list, &super_blocks);
383 list_add(&s->s_instances, &type->fs_supers);
384 spin_unlock(&sb_lock);
385 get_filesystem(type);
386 return s;
387}
388
389EXPORT_SYMBOL(sget);
390
391void drop_super(struct super_block *sb)
392{
393 up_read(&sb->s_umount);
394 put_super(sb);
395}
396
397EXPORT_SYMBOL(drop_super);
398
e5004753
CH
399/**
400 * sync_supers - helper for periodic superblock writeback
401 *
402 * Call the write_super method if present on all dirty superblocks in
403 * the system. This is for the periodic writeback used by most older
404 * filesystems. For data integrity superblock writeback use
405 * sync_filesystems() instead.
406 *
1da177e4
LT
407 * Note: check the dirty flag before waiting, so we don't
408 * hold up the sync while mounting a device. (The newly
409 * mounted device won't need syncing.)
410 */
411void sync_supers(void)
412{
618f0636
KK
413 struct super_block *sb;
414
1da177e4 415 spin_lock(&sb_lock);
618f0636
KK
416restart:
417 list_for_each_entry(sb, &super_blocks, s_list) {
e5004753 418 if (sb->s_op->write_super && sb->s_dirt) {
1da177e4
LT
419 sb->s_count++;
420 spin_unlock(&sb_lock);
e5004753 421
1da177e4 422 down_read(&sb->s_umount);
e5004753
CH
423 if (sb->s_root && sb->s_dirt)
424 sb->s_op->write_super(sb);
618f0636 425 up_read(&sb->s_umount);
e5004753 426
618f0636
KK
427 spin_lock(&sb_lock);
428 if (__put_super_and_need_restart(sb))
429 goto restart;
430 }
431 }
1da177e4
LT
432 spin_unlock(&sb_lock);
433}
434
1da177e4
LT
435/**
436 * get_super - get the superblock of a device
437 * @bdev: device to get the superblock for
438 *
439 * Scans the superblock list and finds the superblock of the file system
440 * mounted on the device given. %NULL is returned if no match is found.
441 */
442
443struct super_block * get_super(struct block_device *bdev)
444{
618f0636
KK
445 struct super_block *sb;
446
1da177e4
LT
447 if (!bdev)
448 return NULL;
618f0636 449
1da177e4 450 spin_lock(&sb_lock);
618f0636
KK
451rescan:
452 list_for_each_entry(sb, &super_blocks, s_list) {
453 if (sb->s_bdev == bdev) {
454 sb->s_count++;
1da177e4 455 spin_unlock(&sb_lock);
618f0636
KK
456 down_read(&sb->s_umount);
457 if (sb->s_root)
458 return sb;
459 up_read(&sb->s_umount);
460 /* restart only when sb is no longer on the list */
461 spin_lock(&sb_lock);
462 if (__put_super_and_need_restart(sb))
463 goto rescan;
1da177e4
LT
464 }
465 }
466 spin_unlock(&sb_lock);
467 return NULL;
468}
469
470EXPORT_SYMBOL(get_super);
471
472struct super_block * user_get_super(dev_t dev)
473{
618f0636 474 struct super_block *sb;
1da177e4 475
1da177e4 476 spin_lock(&sb_lock);
618f0636
KK
477rescan:
478 list_for_each_entry(sb, &super_blocks, s_list) {
479 if (sb->s_dev == dev) {
480 sb->s_count++;
1da177e4 481 spin_unlock(&sb_lock);
618f0636
KK
482 down_read(&sb->s_umount);
483 if (sb->s_root)
484 return sb;
485 up_read(&sb->s_umount);
486 /* restart only when sb is no longer on the list */
487 spin_lock(&sb_lock);
488 if (__put_super_and_need_restart(sb))
489 goto rescan;
1da177e4
LT
490 }
491 }
492 spin_unlock(&sb_lock);
493 return NULL;
494}
495
257ac264 496SYSCALL_DEFINE2(ustat, unsigned, dev, struct ustat __user *, ubuf)
1da177e4
LT
497{
498 struct super_block *s;
499 struct ustat tmp;
500 struct kstatfs sbuf;
501 int err = -EINVAL;
502
503 s = user_get_super(new_decode_dev(dev));
504 if (s == NULL)
505 goto out;
726c3342 506 err = vfs_statfs(s->s_root, &sbuf);
1da177e4
LT
507 drop_super(s);
508 if (err)
509 goto out;
510
511 memset(&tmp,0,sizeof(struct ustat));
512 tmp.f_tfree = sbuf.f_bfree;
513 tmp.f_tinode = sbuf.f_ffree;
514
515 err = copy_to_user(ubuf,&tmp,sizeof(struct ustat)) ? -EFAULT : 0;
516out:
517 return err;
518}
519
1da177e4
LT
520/**
521 * do_remount_sb - asks filesystem to change mount options.
522 * @sb: superblock in question
523 * @flags: numeric part of options
524 * @data: the rest of options
525 * @force: whether or not to force the change
526 *
527 * Alters the mount options of a mounted file system.
528 */
529int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
530{
531 int retval;
0ff5af83 532 int remount_rw;
1da177e4 533
9361401e 534#ifdef CONFIG_BLOCK
1da177e4
LT
535 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
536 return -EACCES;
9361401e 537#endif
1da177e4
LT
538 if (flags & MS_RDONLY)
539 acct_auto_close(sb);
540 shrink_dcache_sb(sb);
60b0680f 541 sync_filesystem(sb);
1da177e4
LT
542
543 /* If we are remounting RDONLY and current sb is read/write,
544 make sure there are no rw files opened */
545 if ((flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY)) {
546 if (force)
547 mark_files_ro(sb);
b0895513 548 else if (!fs_may_remount_ro(sb))
1da177e4 549 return -EBUSY;
9e3509e2 550 retval = vfs_dq_off(sb, 1);
b0895513 551 if (retval < 0 && retval != -ENOSYS)
0ff5af83 552 return -EBUSY;
1da177e4 553 }
0ff5af83 554 remount_rw = !(flags & MS_RDONLY) && (sb->s_flags & MS_RDONLY);
1da177e4
LT
555
556 if (sb->s_op->remount_fs) {
1da177e4 557 retval = sb->s_op->remount_fs(sb, &flags, data);
b0895513 558 if (retval)
1da177e4
LT
559 return retval;
560 }
561 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
0ff5af83 562 if (remount_rw)
9e3509e2 563 vfs_dq_quota_on_remount(sb);
1da177e4
LT
564 return 0;
565}
566
a2a9537a 567static void do_emergency_remount(struct work_struct *work)
1da177e4
LT
568{
569 struct super_block *sb;
570
571 spin_lock(&sb_lock);
572 list_for_each_entry(sb, &super_blocks, s_list) {
573 sb->s_count++;
574 spin_unlock(&sb_lock);
443b94ba 575 down_write(&sb->s_umount);
1da177e4
LT
576 if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) {
577 /*
578 * ->remount_fs needs lock_kernel().
579 *
580 * What lock protects sb->s_flags??
581 */
1da177e4 582 do_remount_sb(sb, MS_RDONLY, NULL, 1);
1da177e4 583 }
443b94ba
AV
584 up_write(&sb->s_umount);
585 put_super(sb);
1da177e4
LT
586 spin_lock(&sb_lock);
587 }
588 spin_unlock(&sb_lock);
a2a9537a 589 kfree(work);
1da177e4
LT
590 printk("Emergency Remount complete\n");
591}
592
593void emergency_remount(void)
594{
a2a9537a
JA
595 struct work_struct *work;
596
597 work = kmalloc(sizeof(*work), GFP_ATOMIC);
598 if (work) {
599 INIT_WORK(work, do_emergency_remount);
600 schedule_work(work);
601 }
1da177e4
LT
602}
603
604/*
605 * Unnamed block devices are dummy devices used by virtual
606 * filesystems which don't use real block-devices. -- jrs
607 */
608
ad76cbc6 609static DEFINE_IDA(unnamed_dev_ida);
1da177e4 610static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
c63e09ec 611static int unnamed_dev_start = 0; /* don't bother trying below it */
1da177e4
LT
612
613int set_anon_super(struct super_block *s, void *data)
614{
615 int dev;
616 int error;
617
618 retry:
ad76cbc6 619 if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0)
1da177e4
LT
620 return -ENOMEM;
621 spin_lock(&unnamed_dev_lock);
c63e09ec
AV
622 error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev);
623 unnamed_dev_start = dev + 1;
1da177e4
LT
624 spin_unlock(&unnamed_dev_lock);
625 if (error == -EAGAIN)
626 /* We raced and lost with another CPU. */
627 goto retry;
628 else if (error)
629 return -EAGAIN;
630
631 if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
632 spin_lock(&unnamed_dev_lock);
ad76cbc6 633 ida_remove(&unnamed_dev_ida, dev);
c63e09ec 634 unnamed_dev_start = dev;
1da177e4
LT
635 spin_unlock(&unnamed_dev_lock);
636 return -EMFILE;
637 }
638 s->s_dev = MKDEV(0, dev & MINORMASK);
639 return 0;
640}
641
642EXPORT_SYMBOL(set_anon_super);
643
644void kill_anon_super(struct super_block *sb)
645{
646 int slot = MINOR(sb->s_dev);
647
648 generic_shutdown_super(sb);
649 spin_lock(&unnamed_dev_lock);
ad76cbc6 650 ida_remove(&unnamed_dev_ida, slot);
c63e09ec
AV
651 if (slot < unnamed_dev_start)
652 unnamed_dev_start = slot;
1da177e4
LT
653 spin_unlock(&unnamed_dev_lock);
654}
655
656EXPORT_SYMBOL(kill_anon_super);
657
1da177e4
LT
658void kill_litter_super(struct super_block *sb)
659{
660 if (sb->s_root)
661 d_genocide(sb->s_root);
662 kill_anon_super(sb);
663}
664
665EXPORT_SYMBOL(kill_litter_super);
666
909e6d94
SH
667static int ns_test_super(struct super_block *sb, void *data)
668{
669 return sb->s_fs_info == data;
670}
671
672static int ns_set_super(struct super_block *sb, void *data)
673{
674 sb->s_fs_info = data;
675 return set_anon_super(sb, NULL);
676}
677
678int get_sb_ns(struct file_system_type *fs_type, int flags, void *data,
679 int (*fill_super)(struct super_block *, void *, int),
680 struct vfsmount *mnt)
681{
682 struct super_block *sb;
683
684 sb = sget(fs_type, ns_test_super, ns_set_super, data);
685 if (IS_ERR(sb))
686 return PTR_ERR(sb);
687
688 if (!sb->s_root) {
689 int err;
690 sb->s_flags = flags;
691 err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
692 if (err) {
74dbbdd7 693 deactivate_locked_super(sb);
909e6d94
SH
694 return err;
695 }
696
697 sb->s_flags |= MS_ACTIVE;
698 }
699
700 simple_set_mnt(mnt, sb);
701 return 0;
702}
703
704EXPORT_SYMBOL(get_sb_ns);
705
9361401e 706#ifdef CONFIG_BLOCK
1da177e4
LT
707static int set_bdev_super(struct super_block *s, void *data)
708{
709 s->s_bdev = data;
710 s->s_dev = s->s_bdev->bd_dev;
711 return 0;
712}
713
714static int test_bdev_super(struct super_block *s, void *data)
715{
716 return (void *)s->s_bdev == data;
717}
718
454e2398 719int get_sb_bdev(struct file_system_type *fs_type,
1da177e4 720 int flags, const char *dev_name, void *data,
454e2398
DH
721 int (*fill_super)(struct super_block *, void *, int),
722 struct vfsmount *mnt)
1da177e4
LT
723{
724 struct block_device *bdev;
725 struct super_block *s;
30c40d2c 726 fmode_t mode = FMODE_READ;
1da177e4
LT
727 int error = 0;
728
30c40d2c
AV
729 if (!(flags & MS_RDONLY))
730 mode |= FMODE_WRITE;
731
732 bdev = open_bdev_exclusive(dev_name, mode, fs_type);
1da177e4 733 if (IS_ERR(bdev))
454e2398 734 return PTR_ERR(bdev);
1da177e4
LT
735
736 /*
737 * once the super is inserted into the list by sget, s_umount
738 * will protect the lockfs code from trying to start a snapshot
739 * while we are mounting
740 */
f73ca1b7 741 down(&bdev->bd_mount_sem);
1da177e4 742 s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
f73ca1b7 743 up(&bdev->bd_mount_sem);
1da177e4 744 if (IS_ERR(s))
454e2398 745 goto error_s;
1da177e4
LT
746
747 if (s->s_root) {
748 if ((flags ^ s->s_flags) & MS_RDONLY) {
74dbbdd7 749 deactivate_locked_super(s);
454e2398
DH
750 error = -EBUSY;
751 goto error_bdev;
1da177e4 752 }
454e2398 753
30c40d2c 754 close_bdev_exclusive(bdev, mode);
1da177e4
LT
755 } else {
756 char b[BDEVNAME_SIZE];
757
758 s->s_flags = flags;
30c40d2c 759 s->s_mode = mode;
1da177e4 760 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
e78c9a00 761 sb_set_blocksize(s, block_size(bdev));
9b04c997 762 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
1da177e4 763 if (error) {
74dbbdd7 764 deactivate_locked_super(s);
454e2398 765 goto error;
fa675765 766 }
454e2398
DH
767
768 s->s_flags |= MS_ACTIVE;
87d8fe1e 769 bdev->bd_super = s;
1da177e4
LT
770 }
771
a3ec947c
SB
772 simple_set_mnt(mnt, s);
773 return 0;
1da177e4 774
454e2398
DH
775error_s:
776 error = PTR_ERR(s);
777error_bdev:
30c40d2c 778 close_bdev_exclusive(bdev, mode);
454e2398
DH
779error:
780 return error;
1da177e4
LT
781}
782
783EXPORT_SYMBOL(get_sb_bdev);
784
785void kill_block_super(struct super_block *sb)
786{
787 struct block_device *bdev = sb->s_bdev;
30c40d2c 788 fmode_t mode = sb->s_mode;
1da177e4 789
ddbaaf30 790 bdev->bd_super = NULL;
1da177e4
LT
791 generic_shutdown_super(sb);
792 sync_blockdev(bdev);
30c40d2c 793 close_bdev_exclusive(bdev, mode);
1da177e4
LT
794}
795
796EXPORT_SYMBOL(kill_block_super);
9361401e 797#endif
1da177e4 798
454e2398 799int get_sb_nodev(struct file_system_type *fs_type,
1da177e4 800 int flags, void *data,
454e2398
DH
801 int (*fill_super)(struct super_block *, void *, int),
802 struct vfsmount *mnt)
1da177e4
LT
803{
804 int error;
805 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
806
807 if (IS_ERR(s))
454e2398 808 return PTR_ERR(s);
1da177e4
LT
809
810 s->s_flags = flags;
811
9b04c997 812 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
1da177e4 813 if (error) {
74dbbdd7 814 deactivate_locked_super(s);
454e2398 815 return error;
1da177e4
LT
816 }
817 s->s_flags |= MS_ACTIVE;
a3ec947c
SB
818 simple_set_mnt(mnt, s);
819 return 0;
1da177e4
LT
820}
821
822EXPORT_SYMBOL(get_sb_nodev);
823
824static int compare_single(struct super_block *s, void *p)
825{
826 return 1;
827}
828
454e2398 829int get_sb_single(struct file_system_type *fs_type,
1da177e4 830 int flags, void *data,
454e2398
DH
831 int (*fill_super)(struct super_block *, void *, int),
832 struct vfsmount *mnt)
1da177e4
LT
833{
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))
454e2398 839 return PTR_ERR(s);
1da177e4
LT
840 if (!s->s_root) {
841 s->s_flags = flags;
9b04c997 842 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
1da177e4 843 if (error) {
74dbbdd7 844 deactivate_locked_super(s);
454e2398 845 return error;
1da177e4
LT
846 }
847 s->s_flags |= MS_ACTIVE;
848 }
849 do_remount_sb(s, flags, data, 0);
a3ec947c
SB
850 simple_set_mnt(mnt, s);
851 return 0;
1da177e4
LT
852}
853
854EXPORT_SYMBOL(get_sb_single);
855
856struct vfsmount *
bb4a58bf 857vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data)
1da177e4 858{
1da177e4 859 struct vfsmount *mnt;
1da177e4 860 char *secdata = NULL;
454e2398 861 int error;
1da177e4
LT
862
863 if (!type)
864 return ERR_PTR(-ENODEV);
1da177e4 865
454e2398 866 error = -ENOMEM;
1da177e4
LT
867 mnt = alloc_vfsmnt(name);
868 if (!mnt)
869 goto out;
870
e0007529 871 if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) {
1da177e4 872 secdata = alloc_secdata();
454e2398 873 if (!secdata)
1da177e4 874 goto out_mnt;
1da177e4 875
e0007529 876 error = security_sb_copy_data(data, secdata);
454e2398 877 if (error)
1da177e4 878 goto out_free_secdata;
1da177e4
LT
879 }
880
454e2398
DH
881 error = type->get_sb(type, flags, name, data, mnt);
882 if (error < 0)
1da177e4 883 goto out_free_secdata;
b4c07bce 884 BUG_ON(!mnt->mnt_sb);
454e2398 885
12204e24 886 error = security_sb_kern_mount(mnt->mnt_sb, flags, secdata);
1da177e4
LT
887 if (error)
888 goto out_sb;
454e2398
DH
889
890 mnt->mnt_mountpoint = mnt->mnt_root;
1da177e4 891 mnt->mnt_parent = mnt;
454e2398 892 up_write(&mnt->mnt_sb->s_umount);
8680e22f 893 free_secdata(secdata);
1da177e4
LT
894 return mnt;
895out_sb:
454e2398 896 dput(mnt->mnt_root);
74dbbdd7 897 deactivate_locked_super(mnt->mnt_sb);
1da177e4
LT
898out_free_secdata:
899 free_secdata(secdata);
900out_mnt:
901 free_vfsmnt(mnt);
902out:
454e2398 903 return ERR_PTR(error);
1da177e4
LT
904}
905
bb4a58bf
TM
906EXPORT_SYMBOL_GPL(vfs_kern_mount);
907
79c0b2df
MS
908static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype)
909{
910 int err;
911 const char *subtype = strchr(fstype, '.');
912 if (subtype) {
913 subtype++;
914 err = -EINVAL;
915 if (!subtype[0])
916 goto err;
917 } else
918 subtype = "";
919
920 mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL);
921 err = -ENOMEM;
922 if (!mnt->mnt_sb->s_subtype)
923 goto err;
924 return mnt;
925
926 err:
927 mntput(mnt);
928 return ERR_PTR(err);
929}
930
bb4a58bf
TM
931struct vfsmount *
932do_kern_mount(const char *fstype, int flags, const char *name, void *data)
933{
934 struct file_system_type *type = get_fs_type(fstype);
935 struct vfsmount *mnt;
936 if (!type)
937 return ERR_PTR(-ENODEV);
938 mnt = vfs_kern_mount(type, flags, name, data);
79c0b2df
MS
939 if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) &&
940 !mnt->mnt_sb->s_subtype)
941 mnt = fs_set_subtype(mnt, fstype);
bb4a58bf
TM
942 put_filesystem(type);
943 return mnt;
944}
8a4e98d9 945EXPORT_SYMBOL_GPL(do_kern_mount);
bb4a58bf 946
8bf9725c 947struct vfsmount *kern_mount_data(struct file_system_type *type, void *data)
1da177e4 948{
8bf9725c 949 return vfs_kern_mount(type, MS_KERNMOUNT, type->name, data);
1da177e4
LT
950}
951
8bf9725c 952EXPORT_SYMBOL_GPL(kern_mount_data);