Commit | Line | Data |
---|---|---|
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> | |
1da177e4 LT |
34 | #include <linux/writeback.h> /* for the emergency remount stuff */ |
35 | #include <linux/idr.h> | |
36 | #include <linux/kobject.h> | |
353ab6e9 | 37 | #include <linux/mutex.h> |
49e0d02c | 38 | #include <linux/file.h> |
5477d0fa | 39 | #include <linux/backing-dev.h> |
1da177e4 | 40 | #include <asm/uaccess.h> |
6d59e7f5 | 41 | #include "internal.h" |
1da177e4 LT |
42 | |
43 | ||
1da177e4 LT |
44 | LIST_HEAD(super_blocks); |
45 | DEFINE_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 | 54 | static 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); |
b87221de | 57 | static const struct super_operations default_op; |
1da177e4 LT |
58 | |
59 | if (s) { | |
1da177e4 LT |
60 | if (security_sb_alloc(s)) { |
61 | kfree(s); | |
62 | s = NULL; | |
63 | goto out; | |
64 | } | |
1da177e4 LT |
65 | INIT_LIST_HEAD(&s->s_files); |
66 | INIT_LIST_HEAD(&s->s_instances); | |
67 | INIT_HLIST_HEAD(&s->s_anon); | |
68 | INIT_LIST_HEAD(&s->s_inodes); | |
da3bbdd4 | 69 | INIT_LIST_HEAD(&s->s_dentry_lru); |
1da177e4 | 70 | init_rwsem(&s->s_umount); |
7892f2f4 | 71 | mutex_init(&s->s_lock); |
897c6ff9 | 72 | lockdep_set_class(&s->s_umount, &type->s_umount_key); |
cf516249 IM |
73 | /* |
74 | * The locking rules for s_lock are up to the | |
75 | * filesystem. For example ext3fs has different | |
76 | * lock ordering than usbfs: | |
77 | */ | |
78 | lockdep_set_class(&s->s_lock, &type->s_lock_key); | |
ada723dc PZ |
79 | /* |
80 | * sget() can have s_umount recursion. | |
81 | * | |
82 | * When it cannot find a suitable sb, it allocates a new | |
83 | * one (this one), and tries again to find a suitable old | |
84 | * one. | |
85 | * | |
86 | * In case that succeeds, it will acquire the s_umount | |
87 | * lock of the old one. Since these are clearly distrinct | |
88 | * locks, and this object isn't exposed yet, there's no | |
89 | * risk of deadlocks. | |
90 | * | |
91 | * Annotate this by putting this lock in a different | |
92 | * subclass. | |
93 | */ | |
94 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); | |
b20bd1a5 | 95 | s->s_count = 1; |
1da177e4 | 96 | atomic_set(&s->s_active, 1); |
a11f3a05 | 97 | mutex_init(&s->s_vfs_rename_mutex); |
d3be915f IM |
98 | mutex_init(&s->s_dquot.dqio_mutex); |
99 | mutex_init(&s->s_dquot.dqonoff_mutex); | |
1da177e4 LT |
100 | init_rwsem(&s->s_dquot.dqptr_sem); |
101 | init_waitqueue_head(&s->s_wait_unfrozen); | |
102 | s->s_maxbytes = MAX_NON_LFS; | |
103 | s->dq_op = sb_dquot_ops; | |
104 | s->s_qcop = sb_quotactl_ops; | |
105 | s->s_op = &default_op; | |
106 | s->s_time_gran = 1000000000; | |
107 | } | |
108 | out: | |
109 | return s; | |
110 | } | |
111 | ||
112 | /** | |
113 | * destroy_super - frees a superblock | |
114 | * @s: superblock to free | |
115 | * | |
116 | * Frees a superblock. | |
117 | */ | |
118 | static inline void destroy_super(struct super_block *s) | |
119 | { | |
120 | security_sb_free(s); | |
79c0b2df | 121 | kfree(s->s_subtype); |
b3b304a2 | 122 | kfree(s->s_options); |
1da177e4 LT |
123 | kfree(s); |
124 | } | |
125 | ||
126 | /* Superblock refcounting */ | |
127 | ||
128 | /* | |
35cf7ba0 | 129 | * Drop a superblock's refcount. The caller must hold sb_lock. |
1da177e4 | 130 | */ |
35cf7ba0 | 131 | void __put_super(struct super_block *sb) |
1da177e4 | 132 | { |
1da177e4 | 133 | if (!--sb->s_count) { |
551de6f3 | 134 | list_del_init(&sb->s_list); |
1da177e4 | 135 | destroy_super(sb); |
1da177e4 | 136 | } |
1da177e4 LT |
137 | } |
138 | ||
139 | /** | |
140 | * put_super - drop a temporary reference to superblock | |
141 | * @sb: superblock in question | |
142 | * | |
143 | * Drops a temporary reference, frees superblock if there's no | |
144 | * references left. | |
145 | */ | |
03ba3782 | 146 | void put_super(struct super_block *sb) |
1da177e4 LT |
147 | { |
148 | spin_lock(&sb_lock); | |
149 | __put_super(sb); | |
150 | spin_unlock(&sb_lock); | |
151 | } | |
152 | ||
153 | ||
154 | /** | |
1712ac8f | 155 | * deactivate_locked_super - drop an active reference to superblock |
1da177e4 LT |
156 | * @s: superblock to deactivate |
157 | * | |
1712ac8f AV |
158 | * Drops an active reference to superblock, converting it into a temprory |
159 | * one if there is no other active references left. In that case we | |
1da177e4 LT |
160 | * tell fs driver to shut it down and drop the temporary reference we |
161 | * had just acquired. | |
1712ac8f AV |
162 | * |
163 | * Caller holds exclusive lock on superblock; that lock is released. | |
1da177e4 | 164 | */ |
1712ac8f | 165 | void deactivate_locked_super(struct super_block *s) |
1da177e4 LT |
166 | { |
167 | struct file_system_type *fs = s->s_type; | |
b20bd1a5 | 168 | if (atomic_dec_and_test(&s->s_active)) { |
9e3509e2 | 169 | vfs_dq_off(s, 0); |
1da177e4 LT |
170 | fs->kill_sb(s); |
171 | put_filesystem(fs); | |
172 | put_super(s); | |
1712ac8f AV |
173 | } else { |
174 | up_write(&s->s_umount); | |
1da177e4 LT |
175 | } |
176 | } | |
177 | ||
1712ac8f | 178 | EXPORT_SYMBOL(deactivate_locked_super); |
1da177e4 | 179 | |
74dbbdd7 | 180 | /** |
1712ac8f | 181 | * deactivate_super - drop an active reference to superblock |
74dbbdd7 AV |
182 | * @s: superblock to deactivate |
183 | * | |
1712ac8f AV |
184 | * Variant of deactivate_locked_super(), except that superblock is *not* |
185 | * locked by caller. If we are going to drop the final active reference, | |
186 | * lock will be acquired prior to that. | |
74dbbdd7 | 187 | */ |
1712ac8f | 188 | void deactivate_super(struct super_block *s) |
74dbbdd7 | 189 | { |
1712ac8f AV |
190 | if (!atomic_add_unless(&s->s_active, -1, 1)) { |
191 | down_write(&s->s_umount); | |
192 | deactivate_locked_super(s); | |
74dbbdd7 AV |
193 | } |
194 | } | |
195 | ||
1712ac8f | 196 | EXPORT_SYMBOL(deactivate_super); |
74dbbdd7 | 197 | |
1da177e4 LT |
198 | /** |
199 | * grab_super - acquire an active reference | |
200 | * @s: reference we are trying to make active | |
201 | * | |
202 | * Tries to acquire an active reference. grab_super() is used when we | |
203 | * had just found a superblock in super_blocks or fs_type->fs_supers | |
204 | * and want to turn it into a full-blown active reference. grab_super() | |
205 | * is called with sb_lock held and drops it. Returns 1 in case of | |
206 | * success, 0 if we had failed (superblock contents was already dead or | |
207 | * dying when grab_super() had been called). | |
208 | */ | |
9c4dbee7 | 209 | static int grab_super(struct super_block *s) __releases(sb_lock) |
1da177e4 | 210 | { |
b20bd1a5 AV |
211 | if (atomic_inc_not_zero(&s->s_active)) { |
212 | spin_unlock(&sb_lock); | |
213 | down_write(&s->s_umount); | |
214 | return 1; | |
215 | } | |
216 | /* it's going away */ | |
1da177e4 LT |
217 | s->s_count++; |
218 | spin_unlock(&sb_lock); | |
1712ac8f | 219 | /* wait for it to die */ |
1da177e4 | 220 | down_write(&s->s_umount); |
1da177e4 LT |
221 | up_write(&s->s_umount); |
222 | put_super(s); | |
1da177e4 LT |
223 | return 0; |
224 | } | |
225 | ||
914e2637 AV |
226 | /* |
227 | * Superblock locking. We really ought to get rid of these two. | |
228 | */ | |
229 | void lock_super(struct super_block * sb) | |
230 | { | |
231 | get_fs_excl(); | |
232 | mutex_lock(&sb->s_lock); | |
233 | } | |
234 | ||
235 | void unlock_super(struct super_block * sb) | |
236 | { | |
237 | put_fs_excl(); | |
238 | mutex_unlock(&sb->s_lock); | |
239 | } | |
240 | ||
241 | EXPORT_SYMBOL(lock_super); | |
242 | EXPORT_SYMBOL(unlock_super); | |
243 | ||
1da177e4 LT |
244 | /** |
245 | * generic_shutdown_super - common helper for ->kill_sb() | |
246 | * @sb: superblock to kill | |
247 | * | |
248 | * generic_shutdown_super() does all fs-independent work on superblock | |
249 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects | |
250 | * that need destruction out of superblock, call generic_shutdown_super() | |
251 | * and release aforementioned objects. Note: dentries and inodes _are_ | |
252 | * taken care of and do not need specific handling. | |
c636ebdb DH |
253 | * |
254 | * Upon calling this function, the filesystem may no longer alter or | |
255 | * rearrange the set of dentries belonging to this super_block, nor may it | |
256 | * change the attachments of dentries to inodes. | |
1da177e4 LT |
257 | */ |
258 | void generic_shutdown_super(struct super_block *sb) | |
259 | { | |
ee9b6d61 | 260 | const struct super_operations *sop = sb->s_op; |
1da177e4 | 261 | |
efaee192 | 262 | |
c636ebdb DH |
263 | if (sb->s_root) { |
264 | shrink_dcache_for_umount(sb); | |
60b0680f | 265 | sync_filesystem(sb); |
a9e220f8 | 266 | get_fs_excl(); |
1da177e4 | 267 | sb->s_flags &= ~MS_ACTIVE; |
efaee192 | 268 | |
1da177e4 LT |
269 | /* bad name - it should be evict_inodes() */ |
270 | invalidate_inodes(sb); | |
1da177e4 | 271 | |
1da177e4 LT |
272 | if (sop->put_super) |
273 | sop->put_super(sb); | |
274 | ||
275 | /* Forget any remaining inodes */ | |
276 | if (invalidate_inodes(sb)) { | |
7b4fe29e DJ |
277 | printk("VFS: Busy inodes after unmount of %s. " |
278 | "Self-destruct in 5 seconds. Have a nice day...\n", | |
279 | sb->s_id); | |
1da177e4 | 280 | } |
a9e220f8 | 281 | put_fs_excl(); |
1da177e4 LT |
282 | } |
283 | spin_lock(&sb_lock); | |
284 | /* should be initialized for __put_super_and_need_restart() */ | |
551de6f3 | 285 | list_del_init(&sb->s_instances); |
1da177e4 LT |
286 | spin_unlock(&sb_lock); |
287 | up_write(&sb->s_umount); | |
288 | } | |
289 | ||
290 | EXPORT_SYMBOL(generic_shutdown_super); | |
291 | ||
292 | /** | |
293 | * sget - find or create a superblock | |
294 | * @type: filesystem type superblock should belong to | |
295 | * @test: comparison callback | |
296 | * @set: setup callback | |
297 | * @data: argument to each of them | |
298 | */ | |
299 | struct super_block *sget(struct file_system_type *type, | |
300 | int (*test)(struct super_block *,void *), | |
301 | int (*set)(struct super_block *,void *), | |
302 | void *data) | |
303 | { | |
304 | struct super_block *s = NULL; | |
d4730127 | 305 | struct super_block *old; |
1da177e4 LT |
306 | int err; |
307 | ||
308 | retry: | |
309 | spin_lock(&sb_lock); | |
d4730127 MK |
310 | if (test) { |
311 | list_for_each_entry(old, &type->fs_supers, s_instances) { | |
312 | if (!test(old, data)) | |
313 | continue; | |
314 | if (!grab_super(old)) | |
315 | goto retry; | |
a3cfbb53 LZ |
316 | if (s) { |
317 | up_write(&s->s_umount); | |
d4730127 | 318 | destroy_super(s); |
a3cfbb53 | 319 | } |
d4730127 MK |
320 | return old; |
321 | } | |
1da177e4 LT |
322 | } |
323 | if (!s) { | |
324 | spin_unlock(&sb_lock); | |
cf516249 | 325 | s = alloc_super(type); |
1da177e4 LT |
326 | if (!s) |
327 | return ERR_PTR(-ENOMEM); | |
328 | goto retry; | |
329 | } | |
330 | ||
331 | err = set(s, data); | |
332 | if (err) { | |
333 | spin_unlock(&sb_lock); | |
a3cfbb53 | 334 | up_write(&s->s_umount); |
1da177e4 LT |
335 | destroy_super(s); |
336 | return ERR_PTR(err); | |
337 | } | |
338 | s->s_type = type; | |
339 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); | |
340 | list_add_tail(&s->s_list, &super_blocks); | |
341 | list_add(&s->s_instances, &type->fs_supers); | |
342 | spin_unlock(&sb_lock); | |
343 | get_filesystem(type); | |
344 | return s; | |
345 | } | |
346 | ||
347 | EXPORT_SYMBOL(sget); | |
348 | ||
349 | void drop_super(struct super_block *sb) | |
350 | { | |
351 | up_read(&sb->s_umount); | |
352 | put_super(sb); | |
353 | } | |
354 | ||
355 | EXPORT_SYMBOL(drop_super); | |
356 | ||
e5004753 CH |
357 | /** |
358 | * sync_supers - helper for periodic superblock writeback | |
359 | * | |
360 | * Call the write_super method if present on all dirty superblocks in | |
361 | * the system. This is for the periodic writeback used by most older | |
362 | * filesystems. For data integrity superblock writeback use | |
363 | * sync_filesystems() instead. | |
364 | * | |
1da177e4 LT |
365 | * Note: check the dirty flag before waiting, so we don't |
366 | * hold up the sync while mounting a device. (The newly | |
367 | * mounted device won't need syncing.) | |
368 | */ | |
369 | void sync_supers(void) | |
370 | { | |
6754af64 | 371 | struct super_block *sb, *n; |
618f0636 | 372 | |
1da177e4 | 373 | spin_lock(&sb_lock); |
6754af64 | 374 | list_for_each_entry_safe(sb, n, &super_blocks, s_list) { |
551de6f3 AV |
375 | if (list_empty(&sb->s_instances)) |
376 | continue; | |
e5004753 | 377 | if (sb->s_op->write_super && sb->s_dirt) { |
1da177e4 LT |
378 | sb->s_count++; |
379 | spin_unlock(&sb_lock); | |
e5004753 | 380 | |
1da177e4 | 381 | down_read(&sb->s_umount); |
e5004753 CH |
382 | if (sb->s_root && sb->s_dirt) |
383 | sb->s_op->write_super(sb); | |
618f0636 | 384 | up_read(&sb->s_umount); |
e5004753 | 385 | |
618f0636 | 386 | spin_lock(&sb_lock); |
6754af64 | 387 | __put_super(sb); |
618f0636 KK |
388 | } |
389 | } | |
1da177e4 LT |
390 | spin_unlock(&sb_lock); |
391 | } | |
392 | ||
01a05b33 AV |
393 | /** |
394 | * iterate_supers - call function for all active superblocks | |
395 | * @f: function to call | |
396 | * @arg: argument to pass to it | |
397 | * | |
398 | * Scans the superblock list and calls given function, passing it | |
399 | * locked superblock and given argument. | |
400 | */ | |
401 | void iterate_supers(void (*f)(struct super_block *, void *), void *arg) | |
402 | { | |
403 | struct super_block *sb, *n; | |
404 | ||
405 | spin_lock(&sb_lock); | |
406 | list_for_each_entry_safe(sb, n, &super_blocks, s_list) { | |
407 | if (list_empty(&sb->s_instances)) | |
408 | continue; | |
409 | sb->s_count++; | |
410 | spin_unlock(&sb_lock); | |
411 | ||
412 | down_read(&sb->s_umount); | |
413 | if (sb->s_root) | |
414 | f(sb, arg); | |
415 | up_read(&sb->s_umount); | |
416 | ||
417 | spin_lock(&sb_lock); | |
418 | __put_super(sb); | |
419 | } | |
420 | spin_unlock(&sb_lock); | |
421 | } | |
422 | ||
1da177e4 LT |
423 | /** |
424 | * get_super - get the superblock of a device | |
425 | * @bdev: device to get the superblock for | |
426 | * | |
427 | * Scans the superblock list and finds the superblock of the file system | |
428 | * mounted on the device given. %NULL is returned if no match is found. | |
429 | */ | |
430 | ||
df40c01a | 431 | struct super_block *get_super(struct block_device *bdev) |
1da177e4 | 432 | { |
618f0636 KK |
433 | struct super_block *sb; |
434 | ||
1da177e4 LT |
435 | if (!bdev) |
436 | return NULL; | |
618f0636 | 437 | |
1da177e4 | 438 | spin_lock(&sb_lock); |
618f0636 KK |
439 | rescan: |
440 | list_for_each_entry(sb, &super_blocks, s_list) { | |
551de6f3 AV |
441 | if (list_empty(&sb->s_instances)) |
442 | continue; | |
618f0636 KK |
443 | if (sb->s_bdev == bdev) { |
444 | sb->s_count++; | |
1da177e4 | 445 | spin_unlock(&sb_lock); |
618f0636 | 446 | down_read(&sb->s_umount); |
df40c01a | 447 | /* still alive? */ |
618f0636 KK |
448 | if (sb->s_root) |
449 | return sb; | |
450 | up_read(&sb->s_umount); | |
df40c01a | 451 | /* nope, got unmounted */ |
618f0636 | 452 | spin_lock(&sb_lock); |
df40c01a AV |
453 | __put_super(sb); |
454 | goto rescan; | |
1da177e4 LT |
455 | } |
456 | } | |
457 | spin_unlock(&sb_lock); | |
458 | return NULL; | |
459 | } | |
460 | ||
461 | EXPORT_SYMBOL(get_super); | |
4504230a CH |
462 | |
463 | /** | |
464 | * get_active_super - get an active reference to the superblock of a device | |
465 | * @bdev: device to get the superblock for | |
466 | * | |
467 | * Scans the superblock list and finds the superblock of the file system | |
468 | * mounted on the device given. Returns the superblock with an active | |
469 | * reference and s_umount held exclusively or %NULL if none was found. | |
470 | */ | |
471 | struct super_block *get_active_super(struct block_device *bdev) | |
472 | { | |
473 | struct super_block *sb; | |
474 | ||
475 | if (!bdev) | |
476 | return NULL; | |
477 | ||
1494583d | 478 | restart: |
4504230a CH |
479 | spin_lock(&sb_lock); |
480 | list_for_each_entry(sb, &super_blocks, s_list) { | |
551de6f3 AV |
481 | if (list_empty(&sb->s_instances)) |
482 | continue; | |
1494583d AV |
483 | if (sb->s_bdev == bdev) { |
484 | if (grab_super(sb)) /* drops sb_lock */ | |
485 | return sb; | |
486 | else | |
487 | goto restart; | |
488 | } | |
4504230a CH |
489 | } |
490 | spin_unlock(&sb_lock); | |
491 | return NULL; | |
492 | } | |
1da177e4 | 493 | |
df40c01a | 494 | struct super_block *user_get_super(dev_t dev) |
1da177e4 | 495 | { |
618f0636 | 496 | struct super_block *sb; |
1da177e4 | 497 | |
1da177e4 | 498 | spin_lock(&sb_lock); |
618f0636 KK |
499 | rescan: |
500 | list_for_each_entry(sb, &super_blocks, s_list) { | |
551de6f3 AV |
501 | if (list_empty(&sb->s_instances)) |
502 | continue; | |
618f0636 KK |
503 | if (sb->s_dev == dev) { |
504 | sb->s_count++; | |
1da177e4 | 505 | spin_unlock(&sb_lock); |
618f0636 | 506 | down_read(&sb->s_umount); |
df40c01a | 507 | /* still alive? */ |
618f0636 KK |
508 | if (sb->s_root) |
509 | return sb; | |
510 | up_read(&sb->s_umount); | |
df40c01a | 511 | /* nope, got unmounted */ |
618f0636 | 512 | spin_lock(&sb_lock); |
df40c01a AV |
513 | __put_super(sb); |
514 | goto rescan; | |
1da177e4 LT |
515 | } |
516 | } | |
517 | spin_unlock(&sb_lock); | |
518 | return NULL; | |
519 | } | |
520 | ||
1da177e4 LT |
521 | /** |
522 | * do_remount_sb - asks filesystem to change mount options. | |
523 | * @sb: superblock in question | |
524 | * @flags: numeric part of options | |
525 | * @data: the rest of options | |
526 | * @force: whether or not to force the change | |
527 | * | |
528 | * Alters the mount options of a mounted file system. | |
529 | */ | |
530 | int do_remount_sb(struct super_block *sb, int flags, void *data, int force) | |
531 | { | |
532 | int retval; | |
d208bbdd | 533 | int remount_rw, remount_ro; |
4504230a CH |
534 | |
535 | if (sb->s_frozen != SB_UNFROZEN) | |
536 | return -EBUSY; | |
537 | ||
9361401e | 538 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
539 | if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev)) |
540 | return -EACCES; | |
9361401e | 541 | #endif |
4504230a | 542 | |
1da177e4 LT |
543 | if (flags & MS_RDONLY) |
544 | acct_auto_close(sb); | |
545 | shrink_dcache_sb(sb); | |
60b0680f | 546 | sync_filesystem(sb); |
1da177e4 | 547 | |
d208bbdd NP |
548 | remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY); |
549 | remount_rw = !(flags & MS_RDONLY) && (sb->s_flags & MS_RDONLY); | |
550 | ||
1da177e4 LT |
551 | /* If we are remounting RDONLY and current sb is read/write, |
552 | make sure there are no rw files opened */ | |
d208bbdd | 553 | if (remount_ro) { |
1da177e4 LT |
554 | if (force) |
555 | mark_files_ro(sb); | |
b0895513 | 556 | else if (!fs_may_remount_ro(sb)) |
1da177e4 | 557 | return -EBUSY; |
9e3509e2 | 558 | retval = vfs_dq_off(sb, 1); |
b0895513 | 559 | if (retval < 0 && retval != -ENOSYS) |
0ff5af83 | 560 | return -EBUSY; |
1da177e4 LT |
561 | } |
562 | ||
563 | if (sb->s_op->remount_fs) { | |
1da177e4 | 564 | retval = sb->s_op->remount_fs(sb, &flags, data); |
b0895513 | 565 | if (retval) |
1da177e4 LT |
566 | return retval; |
567 | } | |
568 | sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK); | |
0ff5af83 | 569 | if (remount_rw) |
9e3509e2 | 570 | vfs_dq_quota_on_remount(sb); |
d208bbdd NP |
571 | /* |
572 | * Some filesystems modify their metadata via some other path than the | |
573 | * bdev buffer cache (eg. use a private mapping, or directories in | |
574 | * pagecache, etc). Also file data modifications go via their own | |
575 | * mappings. So If we try to mount readonly then copy the filesystem | |
576 | * from bdev, we could get stale data, so invalidate it to give a best | |
577 | * effort at coherency. | |
578 | */ | |
579 | if (remount_ro && sb->s_bdev) | |
580 | invalidate_bdev(sb->s_bdev); | |
1da177e4 LT |
581 | return 0; |
582 | } | |
583 | ||
a2a9537a | 584 | static void do_emergency_remount(struct work_struct *work) |
1da177e4 | 585 | { |
e7fe0585 | 586 | struct super_block *sb, *n; |
1da177e4 LT |
587 | |
588 | spin_lock(&sb_lock); | |
e7fe0585 | 589 | list_for_each_entry_safe(sb, n, &super_blocks, s_list) { |
551de6f3 AV |
590 | if (list_empty(&sb->s_instances)) |
591 | continue; | |
1da177e4 LT |
592 | sb->s_count++; |
593 | spin_unlock(&sb_lock); | |
443b94ba | 594 | down_write(&sb->s_umount); |
1da177e4 LT |
595 | if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) { |
596 | /* | |
1da177e4 LT |
597 | * What lock protects sb->s_flags?? |
598 | */ | |
1da177e4 | 599 | do_remount_sb(sb, MS_RDONLY, NULL, 1); |
1da177e4 | 600 | } |
443b94ba | 601 | up_write(&sb->s_umount); |
1da177e4 | 602 | spin_lock(&sb_lock); |
e7fe0585 | 603 | __put_super(sb); |
1da177e4 LT |
604 | } |
605 | spin_unlock(&sb_lock); | |
a2a9537a | 606 | kfree(work); |
1da177e4 LT |
607 | printk("Emergency Remount complete\n"); |
608 | } | |
609 | ||
610 | void emergency_remount(void) | |
611 | { | |
a2a9537a JA |
612 | struct work_struct *work; |
613 | ||
614 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | |
615 | if (work) { | |
616 | INIT_WORK(work, do_emergency_remount); | |
617 | schedule_work(work); | |
618 | } | |
1da177e4 LT |
619 | } |
620 | ||
621 | /* | |
622 | * Unnamed block devices are dummy devices used by virtual | |
623 | * filesystems which don't use real block-devices. -- jrs | |
624 | */ | |
625 | ||
ad76cbc6 | 626 | static DEFINE_IDA(unnamed_dev_ida); |
1da177e4 | 627 | static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */ |
c63e09ec | 628 | static int unnamed_dev_start = 0; /* don't bother trying below it */ |
1da177e4 LT |
629 | |
630 | int set_anon_super(struct super_block *s, void *data) | |
631 | { | |
632 | int dev; | |
633 | int error; | |
634 | ||
635 | retry: | |
ad76cbc6 | 636 | if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0) |
1da177e4 LT |
637 | return -ENOMEM; |
638 | spin_lock(&unnamed_dev_lock); | |
c63e09ec | 639 | error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev); |
f21f6220 AV |
640 | if (!error) |
641 | unnamed_dev_start = dev + 1; | |
1da177e4 LT |
642 | spin_unlock(&unnamed_dev_lock); |
643 | if (error == -EAGAIN) | |
644 | /* We raced and lost with another CPU. */ | |
645 | goto retry; | |
646 | else if (error) | |
647 | return -EAGAIN; | |
648 | ||
649 | if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) { | |
650 | spin_lock(&unnamed_dev_lock); | |
ad76cbc6 | 651 | ida_remove(&unnamed_dev_ida, dev); |
f21f6220 AV |
652 | if (unnamed_dev_start > dev) |
653 | unnamed_dev_start = dev; | |
1da177e4 LT |
654 | spin_unlock(&unnamed_dev_lock); |
655 | return -EMFILE; | |
656 | } | |
657 | s->s_dev = MKDEV(0, dev & MINORMASK); | |
5129a469 | 658 | s->s_bdi = &noop_backing_dev_info; |
1da177e4 LT |
659 | return 0; |
660 | } | |
661 | ||
662 | EXPORT_SYMBOL(set_anon_super); | |
663 | ||
664 | void kill_anon_super(struct super_block *sb) | |
665 | { | |
666 | int slot = MINOR(sb->s_dev); | |
667 | ||
668 | generic_shutdown_super(sb); | |
669 | spin_lock(&unnamed_dev_lock); | |
ad76cbc6 | 670 | ida_remove(&unnamed_dev_ida, slot); |
c63e09ec AV |
671 | if (slot < unnamed_dev_start) |
672 | unnamed_dev_start = slot; | |
1da177e4 LT |
673 | spin_unlock(&unnamed_dev_lock); |
674 | } | |
675 | ||
676 | EXPORT_SYMBOL(kill_anon_super); | |
677 | ||
1da177e4 LT |
678 | void kill_litter_super(struct super_block *sb) |
679 | { | |
680 | if (sb->s_root) | |
681 | d_genocide(sb->s_root); | |
682 | kill_anon_super(sb); | |
683 | } | |
684 | ||
685 | EXPORT_SYMBOL(kill_litter_super); | |
686 | ||
909e6d94 SH |
687 | static int ns_test_super(struct super_block *sb, void *data) |
688 | { | |
689 | return sb->s_fs_info == data; | |
690 | } | |
691 | ||
692 | static int ns_set_super(struct super_block *sb, void *data) | |
693 | { | |
694 | sb->s_fs_info = data; | |
695 | return set_anon_super(sb, NULL); | |
696 | } | |
697 | ||
698 | int get_sb_ns(struct file_system_type *fs_type, int flags, void *data, | |
699 | int (*fill_super)(struct super_block *, void *, int), | |
700 | struct vfsmount *mnt) | |
701 | { | |
702 | struct super_block *sb; | |
703 | ||
704 | sb = sget(fs_type, ns_test_super, ns_set_super, data); | |
705 | if (IS_ERR(sb)) | |
706 | return PTR_ERR(sb); | |
707 | ||
708 | if (!sb->s_root) { | |
709 | int err; | |
710 | sb->s_flags = flags; | |
711 | err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0); | |
712 | if (err) { | |
74dbbdd7 | 713 | deactivate_locked_super(sb); |
909e6d94 SH |
714 | return err; |
715 | } | |
716 | ||
717 | sb->s_flags |= MS_ACTIVE; | |
718 | } | |
719 | ||
720 | simple_set_mnt(mnt, sb); | |
721 | return 0; | |
722 | } | |
723 | ||
724 | EXPORT_SYMBOL(get_sb_ns); | |
725 | ||
9361401e | 726 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
727 | static int set_bdev_super(struct super_block *s, void *data) |
728 | { | |
729 | s->s_bdev = data; | |
730 | s->s_dev = s->s_bdev->bd_dev; | |
32a88aa1 JA |
731 | |
732 | /* | |
733 | * We set the bdi here to the queue backing, file systems can | |
734 | * overwrite this in ->fill_super() | |
735 | */ | |
736 | s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info; | |
1da177e4 LT |
737 | return 0; |
738 | } | |
739 | ||
740 | static int test_bdev_super(struct super_block *s, void *data) | |
741 | { | |
742 | return (void *)s->s_bdev == data; | |
743 | } | |
744 | ||
454e2398 | 745 | int get_sb_bdev(struct file_system_type *fs_type, |
1da177e4 | 746 | int flags, const char *dev_name, void *data, |
454e2398 DH |
747 | int (*fill_super)(struct super_block *, void *, int), |
748 | struct vfsmount *mnt) | |
1da177e4 LT |
749 | { |
750 | struct block_device *bdev; | |
751 | struct super_block *s; | |
30c40d2c | 752 | fmode_t mode = FMODE_READ; |
1da177e4 LT |
753 | int error = 0; |
754 | ||
30c40d2c AV |
755 | if (!(flags & MS_RDONLY)) |
756 | mode |= FMODE_WRITE; | |
757 | ||
758 | bdev = open_bdev_exclusive(dev_name, mode, fs_type); | |
1da177e4 | 759 | if (IS_ERR(bdev)) |
454e2398 | 760 | return PTR_ERR(bdev); |
1da177e4 LT |
761 | |
762 | /* | |
763 | * once the super is inserted into the list by sget, s_umount | |
764 | * will protect the lockfs code from trying to start a snapshot | |
765 | * while we are mounting | |
766 | */ | |
4fadd7bb CH |
767 | mutex_lock(&bdev->bd_fsfreeze_mutex); |
768 | if (bdev->bd_fsfreeze_count > 0) { | |
769 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
770 | error = -EBUSY; | |
771 | goto error_bdev; | |
772 | } | |
1da177e4 | 773 | s = sget(fs_type, test_bdev_super, set_bdev_super, bdev); |
4fadd7bb | 774 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
1da177e4 | 775 | if (IS_ERR(s)) |
454e2398 | 776 | goto error_s; |
1da177e4 LT |
777 | |
778 | if (s->s_root) { | |
779 | if ((flags ^ s->s_flags) & MS_RDONLY) { | |
74dbbdd7 | 780 | deactivate_locked_super(s); |
454e2398 DH |
781 | error = -EBUSY; |
782 | goto error_bdev; | |
1da177e4 | 783 | } |
454e2398 | 784 | |
30c40d2c | 785 | close_bdev_exclusive(bdev, mode); |
1da177e4 LT |
786 | } else { |
787 | char b[BDEVNAME_SIZE]; | |
788 | ||
789 | s->s_flags = flags; | |
30c40d2c | 790 | s->s_mode = mode; |
1da177e4 | 791 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); |
e78c9a00 | 792 | sb_set_blocksize(s, block_size(bdev)); |
9b04c997 | 793 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 794 | if (error) { |
74dbbdd7 | 795 | deactivate_locked_super(s); |
454e2398 | 796 | goto error; |
fa675765 | 797 | } |
454e2398 DH |
798 | |
799 | s->s_flags |= MS_ACTIVE; | |
87d8fe1e | 800 | bdev->bd_super = s; |
1da177e4 LT |
801 | } |
802 | ||
a3ec947c SB |
803 | simple_set_mnt(mnt, s); |
804 | return 0; | |
1da177e4 | 805 | |
454e2398 DH |
806 | error_s: |
807 | error = PTR_ERR(s); | |
808 | error_bdev: | |
30c40d2c | 809 | close_bdev_exclusive(bdev, mode); |
454e2398 DH |
810 | error: |
811 | return error; | |
1da177e4 LT |
812 | } |
813 | ||
814 | EXPORT_SYMBOL(get_sb_bdev); | |
815 | ||
816 | void kill_block_super(struct super_block *sb) | |
817 | { | |
818 | struct block_device *bdev = sb->s_bdev; | |
30c40d2c | 819 | fmode_t mode = sb->s_mode; |
1da177e4 | 820 | |
ddbaaf30 | 821 | bdev->bd_super = NULL; |
1da177e4 LT |
822 | generic_shutdown_super(sb); |
823 | sync_blockdev(bdev); | |
30c40d2c | 824 | close_bdev_exclusive(bdev, mode); |
1da177e4 LT |
825 | } |
826 | ||
827 | EXPORT_SYMBOL(kill_block_super); | |
9361401e | 828 | #endif |
1da177e4 | 829 | |
454e2398 | 830 | int get_sb_nodev(struct file_system_type *fs_type, |
1da177e4 | 831 | int flags, void *data, |
454e2398 DH |
832 | int (*fill_super)(struct super_block *, void *, int), |
833 | struct vfsmount *mnt) | |
1da177e4 LT |
834 | { |
835 | int error; | |
836 | struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL); | |
837 | ||
838 | if (IS_ERR(s)) | |
454e2398 | 839 | return PTR_ERR(s); |
1da177e4 LT |
840 | |
841 | s->s_flags = flags; | |
842 | ||
9b04c997 | 843 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 844 | if (error) { |
74dbbdd7 | 845 | deactivate_locked_super(s); |
454e2398 | 846 | return error; |
1da177e4 LT |
847 | } |
848 | s->s_flags |= MS_ACTIVE; | |
a3ec947c SB |
849 | simple_set_mnt(mnt, s); |
850 | return 0; | |
1da177e4 LT |
851 | } |
852 | ||
853 | EXPORT_SYMBOL(get_sb_nodev); | |
854 | ||
855 | static int compare_single(struct super_block *s, void *p) | |
856 | { | |
857 | return 1; | |
858 | } | |
859 | ||
454e2398 | 860 | int get_sb_single(struct file_system_type *fs_type, |
1da177e4 | 861 | int flags, void *data, |
454e2398 DH |
862 | int (*fill_super)(struct super_block *, void *, int), |
863 | struct vfsmount *mnt) | |
1da177e4 LT |
864 | { |
865 | struct super_block *s; | |
866 | int error; | |
867 | ||
868 | s = sget(fs_type, compare_single, set_anon_super, NULL); | |
869 | if (IS_ERR(s)) | |
454e2398 | 870 | return PTR_ERR(s); |
1da177e4 LT |
871 | if (!s->s_root) { |
872 | s->s_flags = flags; | |
9b04c997 | 873 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 874 | if (error) { |
74dbbdd7 | 875 | deactivate_locked_super(s); |
454e2398 | 876 | return error; |
1da177e4 LT |
877 | } |
878 | s->s_flags |= MS_ACTIVE; | |
9329d1be KS |
879 | } else { |
880 | do_remount_sb(s, flags, data, 0); | |
1da177e4 | 881 | } |
a3ec947c SB |
882 | simple_set_mnt(mnt, s); |
883 | return 0; | |
1da177e4 LT |
884 | } |
885 | ||
886 | EXPORT_SYMBOL(get_sb_single); | |
887 | ||
888 | struct vfsmount * | |
bb4a58bf | 889 | vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data) |
1da177e4 | 890 | { |
1da177e4 | 891 | struct vfsmount *mnt; |
1da177e4 | 892 | char *secdata = NULL; |
454e2398 | 893 | int error; |
1da177e4 LT |
894 | |
895 | if (!type) | |
896 | return ERR_PTR(-ENODEV); | |
1da177e4 | 897 | |
454e2398 | 898 | error = -ENOMEM; |
1da177e4 LT |
899 | mnt = alloc_vfsmnt(name); |
900 | if (!mnt) | |
901 | goto out; | |
902 | ||
8089352a AV |
903 | if (flags & MS_KERNMOUNT) |
904 | mnt->mnt_flags = MNT_INTERNAL; | |
905 | ||
e0007529 | 906 | if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) { |
1da177e4 | 907 | secdata = alloc_secdata(); |
454e2398 | 908 | if (!secdata) |
1da177e4 | 909 | goto out_mnt; |
1da177e4 | 910 | |
e0007529 | 911 | error = security_sb_copy_data(data, secdata); |
454e2398 | 912 | if (error) |
1da177e4 | 913 | goto out_free_secdata; |
1da177e4 LT |
914 | } |
915 | ||
454e2398 DH |
916 | error = type->get_sb(type, flags, name, data, mnt); |
917 | if (error < 0) | |
1da177e4 | 918 | goto out_free_secdata; |
b4c07bce | 919 | BUG_ON(!mnt->mnt_sb); |
5129a469 | 920 | WARN_ON(!mnt->mnt_sb->s_bdi); |
454e2398 | 921 | |
5129a469 JE |
922 | error = security_sb_kern_mount(mnt->mnt_sb, flags, secdata); |
923 | if (error) | |
924 | goto out_sb; | |
454e2398 | 925 | |
42cb56ae JL |
926 | /* |
927 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE | |
928 | * but s_maxbytes was an unsigned long long for many releases. Throw | |
929 | * this warning for a little while to try and catch filesystems that | |
930 | * violate this rule. This warning should be either removed or | |
931 | * converted to a BUG() in 2.6.34. | |
932 | */ | |
933 | WARN((mnt->mnt_sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " | |
934 | "negative value (%lld)\n", type->name, mnt->mnt_sb->s_maxbytes); | |
935 | ||
454e2398 | 936 | mnt->mnt_mountpoint = mnt->mnt_root; |
1da177e4 | 937 | mnt->mnt_parent = mnt; |
454e2398 | 938 | up_write(&mnt->mnt_sb->s_umount); |
8680e22f | 939 | free_secdata(secdata); |
1da177e4 LT |
940 | return mnt; |
941 | out_sb: | |
454e2398 | 942 | dput(mnt->mnt_root); |
74dbbdd7 | 943 | deactivate_locked_super(mnt->mnt_sb); |
1da177e4 LT |
944 | out_free_secdata: |
945 | free_secdata(secdata); | |
946 | out_mnt: | |
947 | free_vfsmnt(mnt); | |
948 | out: | |
454e2398 | 949 | return ERR_PTR(error); |
1da177e4 LT |
950 | } |
951 | ||
bb4a58bf TM |
952 | EXPORT_SYMBOL_GPL(vfs_kern_mount); |
953 | ||
79c0b2df MS |
954 | static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype) |
955 | { | |
956 | int err; | |
957 | const char *subtype = strchr(fstype, '.'); | |
958 | if (subtype) { | |
959 | subtype++; | |
960 | err = -EINVAL; | |
961 | if (!subtype[0]) | |
962 | goto err; | |
963 | } else | |
964 | subtype = ""; | |
965 | ||
966 | mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL); | |
967 | err = -ENOMEM; | |
968 | if (!mnt->mnt_sb->s_subtype) | |
969 | goto err; | |
970 | return mnt; | |
971 | ||
972 | err: | |
973 | mntput(mnt); | |
974 | return ERR_PTR(err); | |
975 | } | |
976 | ||
bb4a58bf TM |
977 | struct vfsmount * |
978 | do_kern_mount(const char *fstype, int flags, const char *name, void *data) | |
979 | { | |
980 | struct file_system_type *type = get_fs_type(fstype); | |
981 | struct vfsmount *mnt; | |
982 | if (!type) | |
983 | return ERR_PTR(-ENODEV); | |
984 | mnt = vfs_kern_mount(type, flags, name, data); | |
79c0b2df MS |
985 | if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) && |
986 | !mnt->mnt_sb->s_subtype) | |
987 | mnt = fs_set_subtype(mnt, fstype); | |
bb4a58bf TM |
988 | put_filesystem(type); |
989 | return mnt; | |
990 | } | |
8a4e98d9 | 991 | EXPORT_SYMBOL_GPL(do_kern_mount); |
bb4a58bf | 992 | |
8bf9725c | 993 | struct vfsmount *kern_mount_data(struct file_system_type *type, void *data) |
1da177e4 | 994 | { |
8bf9725c | 995 | return vfs_kern_mount(type, MS_KERNMOUNT, type->name, data); |
1da177e4 LT |
996 | } |
997 | ||
8bf9725c | 998 | EXPORT_SYMBOL_GPL(kern_mount_data); |