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> | |
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> |
5477d0fa | 40 | #include <linux/backing-dev.h> |
1da177e4 | 41 | #include <asm/uaccess.h> |
6d59e7f5 | 42 | #include "internal.h" |
1da177e4 LT |
43 | |
44 | ||
1da177e4 LT |
45 | LIST_HEAD(super_blocks); |
46 | DEFINE_SPINLOCK(sb_lock); | |
47 | ||
48 | /** | |
49 | * alloc_super - create new superblock | |
fe2bbc48 | 50 | * @type: filesystem type superblock should belong to |
1da177e4 LT |
51 | * |
52 | * Allocates and initializes a new &struct super_block. alloc_super() | |
53 | * returns a pointer new superblock or %NULL if allocation had failed. | |
54 | */ | |
cf516249 | 55 | static struct super_block *alloc_super(struct file_system_type *type) |
1da177e4 | 56 | { |
11b0b5ab | 57 | struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER); |
b87221de | 58 | static const struct super_operations default_op; |
1da177e4 LT |
59 | |
60 | if (s) { | |
1da177e4 LT |
61 | if (security_sb_alloc(s)) { |
62 | kfree(s); | |
63 | s = NULL; | |
64 | goto out; | |
65 | } | |
1da177e4 LT |
66 | INIT_LIST_HEAD(&s->s_files); |
67 | INIT_LIST_HEAD(&s->s_instances); | |
68 | INIT_HLIST_HEAD(&s->s_anon); | |
69 | INIT_LIST_HEAD(&s->s_inodes); | |
da3bbdd4 | 70 | INIT_LIST_HEAD(&s->s_dentry_lru); |
1da177e4 | 71 | init_rwsem(&s->s_umount); |
7892f2f4 | 72 | mutex_init(&s->s_lock); |
897c6ff9 | 73 | lockdep_set_class(&s->s_umount, &type->s_umount_key); |
cf516249 IM |
74 | /* |
75 | * The locking rules for s_lock are up to the | |
76 | * filesystem. For example ext3fs has different | |
77 | * lock ordering than usbfs: | |
78 | */ | |
79 | lockdep_set_class(&s->s_lock, &type->s_lock_key); | |
ada723dc PZ |
80 | /* |
81 | * sget() can have s_umount recursion. | |
82 | * | |
83 | * When it cannot find a suitable sb, it allocates a new | |
84 | * one (this one), and tries again to find a suitable old | |
85 | * one. | |
86 | * | |
87 | * In case that succeeds, it will acquire the s_umount | |
88 | * lock of the old one. Since these are clearly distrinct | |
89 | * locks, and this object isn't exposed yet, there's no | |
90 | * risk of deadlocks. | |
91 | * | |
92 | * Annotate this by putting this lock in a different | |
93 | * subclass. | |
94 | */ | |
95 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); | |
b20bd1a5 | 96 | s->s_count = 1; |
1da177e4 | 97 | atomic_set(&s->s_active, 1); |
a11f3a05 | 98 | mutex_init(&s->s_vfs_rename_mutex); |
d3be915f IM |
99 | mutex_init(&s->s_dquot.dqio_mutex); |
100 | mutex_init(&s->s_dquot.dqonoff_mutex); | |
1da177e4 LT |
101 | init_rwsem(&s->s_dquot.dqptr_sem); |
102 | init_waitqueue_head(&s->s_wait_unfrozen); | |
103 | s->s_maxbytes = MAX_NON_LFS; | |
104 | s->dq_op = sb_dquot_ops; | |
105 | s->s_qcop = sb_quotactl_ops; | |
106 | s->s_op = &default_op; | |
107 | s->s_time_gran = 1000000000; | |
108 | } | |
109 | out: | |
110 | return s; | |
111 | } | |
112 | ||
113 | /** | |
114 | * destroy_super - frees a superblock | |
115 | * @s: superblock to free | |
116 | * | |
117 | * Frees a superblock. | |
118 | */ | |
119 | static inline void destroy_super(struct super_block *s) | |
120 | { | |
121 | security_sb_free(s); | |
79c0b2df | 122 | kfree(s->s_subtype); |
b3b304a2 | 123 | kfree(s->s_options); |
1da177e4 LT |
124 | kfree(s); |
125 | } | |
126 | ||
127 | /* Superblock refcounting */ | |
128 | ||
129 | /* | |
130 | * Drop a superblock's refcount. Returns non-zero if the superblock was | |
131 | * destroyed. The caller must hold sb_lock. | |
132 | */ | |
8edd64bd | 133 | int __put_super(struct super_block *sb) |
1da177e4 LT |
134 | { |
135 | int ret = 0; | |
136 | ||
137 | if (!--sb->s_count) { | |
551de6f3 | 138 | list_del_init(&sb->s_list); |
1da177e4 LT |
139 | destroy_super(sb); |
140 | ret = 1; | |
141 | } | |
142 | return ret; | |
143 | } | |
144 | ||
145 | /* | |
146 | * Drop a superblock's refcount. | |
147 | * Returns non-zero if the superblock is about to be destroyed and | |
148 | * at least is already removed from super_blocks list, so if we are | |
149 | * making a loop through super blocks then we need to restart. | |
150 | * The caller must hold sb_lock. | |
151 | */ | |
152 | int __put_super_and_need_restart(struct super_block *sb) | |
153 | { | |
154 | /* check for race with generic_shutdown_super() */ | |
551de6f3 | 155 | if (list_empty(&sb->s_instances)) { |
1da177e4 LT |
156 | /* super block is removed, need to restart... */ |
157 | __put_super(sb); | |
158 | return 1; | |
159 | } | |
160 | /* can't be the last, since s_list is still in use */ | |
161 | sb->s_count--; | |
162 | BUG_ON(sb->s_count == 0); | |
163 | return 0; | |
164 | } | |
165 | ||
166 | /** | |
167 | * put_super - drop a temporary reference to superblock | |
168 | * @sb: superblock in question | |
169 | * | |
170 | * Drops a temporary reference, frees superblock if there's no | |
171 | * references left. | |
172 | */ | |
03ba3782 | 173 | void put_super(struct super_block *sb) |
1da177e4 LT |
174 | { |
175 | spin_lock(&sb_lock); | |
176 | __put_super(sb); | |
177 | spin_unlock(&sb_lock); | |
178 | } | |
179 | ||
180 | ||
181 | /** | |
1712ac8f | 182 | * deactivate_locked_super - drop an active reference to superblock |
1da177e4 LT |
183 | * @s: superblock to deactivate |
184 | * | |
1712ac8f AV |
185 | * Drops an active reference to superblock, converting it into a temprory |
186 | * one if there is no other active references left. In that case we | |
1da177e4 LT |
187 | * tell fs driver to shut it down and drop the temporary reference we |
188 | * had just acquired. | |
1712ac8f AV |
189 | * |
190 | * Caller holds exclusive lock on superblock; that lock is released. | |
1da177e4 | 191 | */ |
1712ac8f | 192 | void deactivate_locked_super(struct super_block *s) |
1da177e4 LT |
193 | { |
194 | struct file_system_type *fs = s->s_type; | |
b20bd1a5 | 195 | if (atomic_dec_and_test(&s->s_active)) { |
9e3509e2 | 196 | vfs_dq_off(s, 0); |
1da177e4 LT |
197 | fs->kill_sb(s); |
198 | put_filesystem(fs); | |
199 | put_super(s); | |
1712ac8f AV |
200 | } else { |
201 | up_write(&s->s_umount); | |
1da177e4 LT |
202 | } |
203 | } | |
204 | ||
1712ac8f | 205 | EXPORT_SYMBOL(deactivate_locked_super); |
1da177e4 | 206 | |
74dbbdd7 | 207 | /** |
1712ac8f | 208 | * deactivate_super - drop an active reference to superblock |
74dbbdd7 AV |
209 | * @s: superblock to deactivate |
210 | * | |
1712ac8f AV |
211 | * Variant of deactivate_locked_super(), except that superblock is *not* |
212 | * locked by caller. If we are going to drop the final active reference, | |
213 | * lock will be acquired prior to that. | |
74dbbdd7 | 214 | */ |
1712ac8f | 215 | void deactivate_super(struct super_block *s) |
74dbbdd7 | 216 | { |
1712ac8f AV |
217 | if (!atomic_add_unless(&s->s_active, -1, 1)) { |
218 | down_write(&s->s_umount); | |
219 | deactivate_locked_super(s); | |
74dbbdd7 AV |
220 | } |
221 | } | |
222 | ||
1712ac8f | 223 | EXPORT_SYMBOL(deactivate_super); |
74dbbdd7 | 224 | |
1da177e4 LT |
225 | /** |
226 | * grab_super - acquire an active reference | |
227 | * @s: reference we are trying to make active | |
228 | * | |
229 | * Tries to acquire an active reference. grab_super() is used when we | |
230 | * had just found a superblock in super_blocks or fs_type->fs_supers | |
231 | * and want to turn it into a full-blown active reference. grab_super() | |
232 | * is called with sb_lock held and drops it. Returns 1 in case of | |
233 | * success, 0 if we had failed (superblock contents was already dead or | |
234 | * dying when grab_super() had been called). | |
235 | */ | |
9c4dbee7 | 236 | static int grab_super(struct super_block *s) __releases(sb_lock) |
1da177e4 | 237 | { |
b20bd1a5 AV |
238 | if (atomic_inc_not_zero(&s->s_active)) { |
239 | spin_unlock(&sb_lock); | |
240 | down_write(&s->s_umount); | |
241 | return 1; | |
242 | } | |
243 | /* it's going away */ | |
1da177e4 LT |
244 | s->s_count++; |
245 | spin_unlock(&sb_lock); | |
1712ac8f | 246 | /* wait for it to die */ |
1da177e4 | 247 | down_write(&s->s_umount); |
1da177e4 LT |
248 | up_write(&s->s_umount); |
249 | put_super(s); | |
1da177e4 LT |
250 | return 0; |
251 | } | |
252 | ||
914e2637 AV |
253 | /* |
254 | * Superblock locking. We really ought to get rid of these two. | |
255 | */ | |
256 | void lock_super(struct super_block * sb) | |
257 | { | |
258 | get_fs_excl(); | |
259 | mutex_lock(&sb->s_lock); | |
260 | } | |
261 | ||
262 | void unlock_super(struct super_block * sb) | |
263 | { | |
264 | put_fs_excl(); | |
265 | mutex_unlock(&sb->s_lock); | |
266 | } | |
267 | ||
268 | EXPORT_SYMBOL(lock_super); | |
269 | EXPORT_SYMBOL(unlock_super); | |
270 | ||
1da177e4 LT |
271 | /** |
272 | * generic_shutdown_super - common helper for ->kill_sb() | |
273 | * @sb: superblock to kill | |
274 | * | |
275 | * generic_shutdown_super() does all fs-independent work on superblock | |
276 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects | |
277 | * that need destruction out of superblock, call generic_shutdown_super() | |
278 | * and release aforementioned objects. Note: dentries and inodes _are_ | |
279 | * taken care of and do not need specific handling. | |
c636ebdb DH |
280 | * |
281 | * Upon calling this function, the filesystem may no longer alter or | |
282 | * rearrange the set of dentries belonging to this super_block, nor may it | |
283 | * change the attachments of dentries to inodes. | |
1da177e4 LT |
284 | */ |
285 | void generic_shutdown_super(struct super_block *sb) | |
286 | { | |
ee9b6d61 | 287 | const struct super_operations *sop = sb->s_op; |
1da177e4 | 288 | |
efaee192 | 289 | |
c636ebdb DH |
290 | if (sb->s_root) { |
291 | shrink_dcache_for_umount(sb); | |
60b0680f | 292 | sync_filesystem(sb); |
a9e220f8 | 293 | get_fs_excl(); |
1da177e4 | 294 | sb->s_flags &= ~MS_ACTIVE; |
efaee192 | 295 | |
1da177e4 LT |
296 | /* bad name - it should be evict_inodes() */ |
297 | invalidate_inodes(sb); | |
1da177e4 | 298 | |
1da177e4 LT |
299 | if (sop->put_super) |
300 | sop->put_super(sb); | |
301 | ||
302 | /* Forget any remaining inodes */ | |
303 | if (invalidate_inodes(sb)) { | |
7b4fe29e DJ |
304 | printk("VFS: Busy inodes after unmount of %s. " |
305 | "Self-destruct in 5 seconds. Have a nice day...\n", | |
306 | sb->s_id); | |
1da177e4 | 307 | } |
a9e220f8 | 308 | put_fs_excl(); |
1da177e4 LT |
309 | } |
310 | spin_lock(&sb_lock); | |
311 | /* should be initialized for __put_super_and_need_restart() */ | |
551de6f3 | 312 | list_del_init(&sb->s_instances); |
1da177e4 LT |
313 | spin_unlock(&sb_lock); |
314 | up_write(&sb->s_umount); | |
315 | } | |
316 | ||
317 | EXPORT_SYMBOL(generic_shutdown_super); | |
318 | ||
319 | /** | |
320 | * sget - find or create a superblock | |
321 | * @type: filesystem type superblock should belong to | |
322 | * @test: comparison callback | |
323 | * @set: setup callback | |
324 | * @data: argument to each of them | |
325 | */ | |
326 | struct super_block *sget(struct file_system_type *type, | |
327 | int (*test)(struct super_block *,void *), | |
328 | int (*set)(struct super_block *,void *), | |
329 | void *data) | |
330 | { | |
331 | struct super_block *s = NULL; | |
d4730127 | 332 | struct super_block *old; |
1da177e4 LT |
333 | int err; |
334 | ||
335 | retry: | |
336 | spin_lock(&sb_lock); | |
d4730127 MK |
337 | if (test) { |
338 | list_for_each_entry(old, &type->fs_supers, s_instances) { | |
339 | if (!test(old, data)) | |
340 | continue; | |
341 | if (!grab_super(old)) | |
342 | goto retry; | |
a3cfbb53 LZ |
343 | if (s) { |
344 | up_write(&s->s_umount); | |
d4730127 | 345 | destroy_super(s); |
a3cfbb53 | 346 | } |
d4730127 MK |
347 | return old; |
348 | } | |
1da177e4 LT |
349 | } |
350 | if (!s) { | |
351 | spin_unlock(&sb_lock); | |
cf516249 | 352 | s = alloc_super(type); |
1da177e4 LT |
353 | if (!s) |
354 | return ERR_PTR(-ENOMEM); | |
355 | goto retry; | |
356 | } | |
357 | ||
358 | err = set(s, data); | |
359 | if (err) { | |
360 | spin_unlock(&sb_lock); | |
a3cfbb53 | 361 | up_write(&s->s_umount); |
1da177e4 LT |
362 | destroy_super(s); |
363 | return ERR_PTR(err); | |
364 | } | |
365 | s->s_type = type; | |
366 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); | |
367 | list_add_tail(&s->s_list, &super_blocks); | |
368 | list_add(&s->s_instances, &type->fs_supers); | |
369 | spin_unlock(&sb_lock); | |
370 | get_filesystem(type); | |
371 | return s; | |
372 | } | |
373 | ||
374 | EXPORT_SYMBOL(sget); | |
375 | ||
376 | void drop_super(struct super_block *sb) | |
377 | { | |
378 | up_read(&sb->s_umount); | |
379 | put_super(sb); | |
380 | } | |
381 | ||
382 | EXPORT_SYMBOL(drop_super); | |
383 | ||
e5004753 CH |
384 | /** |
385 | * sync_supers - helper for periodic superblock writeback | |
386 | * | |
387 | * Call the write_super method if present on all dirty superblocks in | |
388 | * the system. This is for the periodic writeback used by most older | |
389 | * filesystems. For data integrity superblock writeback use | |
390 | * sync_filesystems() instead. | |
391 | * | |
1da177e4 LT |
392 | * Note: check the dirty flag before waiting, so we don't |
393 | * hold up the sync while mounting a device. (The newly | |
394 | * mounted device won't need syncing.) | |
395 | */ | |
396 | void sync_supers(void) | |
397 | { | |
6754af64 | 398 | struct super_block *sb, *n; |
618f0636 | 399 | |
1da177e4 | 400 | spin_lock(&sb_lock); |
6754af64 | 401 | list_for_each_entry_safe(sb, n, &super_blocks, s_list) { |
551de6f3 AV |
402 | if (list_empty(&sb->s_instances)) |
403 | continue; | |
e5004753 | 404 | if (sb->s_op->write_super && sb->s_dirt) { |
1da177e4 LT |
405 | sb->s_count++; |
406 | spin_unlock(&sb_lock); | |
e5004753 | 407 | |
1da177e4 | 408 | down_read(&sb->s_umount); |
e5004753 CH |
409 | if (sb->s_root && sb->s_dirt) |
410 | sb->s_op->write_super(sb); | |
618f0636 | 411 | up_read(&sb->s_umount); |
e5004753 | 412 | |
618f0636 | 413 | spin_lock(&sb_lock); |
6754af64 | 414 | __put_super(sb); |
618f0636 KK |
415 | } |
416 | } | |
1da177e4 LT |
417 | spin_unlock(&sb_lock); |
418 | } | |
419 | ||
1da177e4 LT |
420 | /** |
421 | * get_super - get the superblock of a device | |
422 | * @bdev: device to get the superblock for | |
423 | * | |
424 | * Scans the superblock list and finds the superblock of the file system | |
425 | * mounted on the device given. %NULL is returned if no match is found. | |
426 | */ | |
427 | ||
df40c01a | 428 | struct super_block *get_super(struct block_device *bdev) |
1da177e4 | 429 | { |
618f0636 KK |
430 | struct super_block *sb; |
431 | ||
1da177e4 LT |
432 | if (!bdev) |
433 | return NULL; | |
618f0636 | 434 | |
1da177e4 | 435 | spin_lock(&sb_lock); |
618f0636 KK |
436 | rescan: |
437 | list_for_each_entry(sb, &super_blocks, s_list) { | |
551de6f3 AV |
438 | if (list_empty(&sb->s_instances)) |
439 | continue; | |
618f0636 KK |
440 | if (sb->s_bdev == bdev) { |
441 | sb->s_count++; | |
1da177e4 | 442 | spin_unlock(&sb_lock); |
618f0636 | 443 | down_read(&sb->s_umount); |
df40c01a | 444 | /* still alive? */ |
618f0636 KK |
445 | if (sb->s_root) |
446 | return sb; | |
447 | up_read(&sb->s_umount); | |
df40c01a | 448 | /* nope, got unmounted */ |
618f0636 | 449 | spin_lock(&sb_lock); |
df40c01a AV |
450 | __put_super(sb); |
451 | goto rescan; | |
1da177e4 LT |
452 | } |
453 | } | |
454 | spin_unlock(&sb_lock); | |
455 | return NULL; | |
456 | } | |
457 | ||
458 | EXPORT_SYMBOL(get_super); | |
4504230a CH |
459 | |
460 | /** | |
461 | * get_active_super - get an active reference to the superblock of a device | |
462 | * @bdev: device to get the superblock for | |
463 | * | |
464 | * Scans the superblock list and finds the superblock of the file system | |
465 | * mounted on the device given. Returns the superblock with an active | |
466 | * reference and s_umount held exclusively or %NULL if none was found. | |
467 | */ | |
468 | struct super_block *get_active_super(struct block_device *bdev) | |
469 | { | |
470 | struct super_block *sb; | |
471 | ||
472 | if (!bdev) | |
473 | return NULL; | |
474 | ||
1494583d | 475 | restart: |
4504230a CH |
476 | spin_lock(&sb_lock); |
477 | list_for_each_entry(sb, &super_blocks, s_list) { | |
551de6f3 AV |
478 | if (list_empty(&sb->s_instances)) |
479 | continue; | |
1494583d AV |
480 | if (sb->s_bdev == bdev) { |
481 | if (grab_super(sb)) /* drops sb_lock */ | |
482 | return sb; | |
483 | else | |
484 | goto restart; | |
485 | } | |
4504230a CH |
486 | } |
487 | spin_unlock(&sb_lock); | |
488 | return NULL; | |
489 | } | |
1da177e4 | 490 | |
df40c01a | 491 | struct super_block *user_get_super(dev_t dev) |
1da177e4 | 492 | { |
618f0636 | 493 | struct super_block *sb; |
1da177e4 | 494 | |
1da177e4 | 495 | spin_lock(&sb_lock); |
618f0636 KK |
496 | rescan: |
497 | list_for_each_entry(sb, &super_blocks, s_list) { | |
551de6f3 AV |
498 | if (list_empty(&sb->s_instances)) |
499 | continue; | |
618f0636 KK |
500 | if (sb->s_dev == dev) { |
501 | sb->s_count++; | |
1da177e4 | 502 | spin_unlock(&sb_lock); |
618f0636 | 503 | down_read(&sb->s_umount); |
df40c01a | 504 | /* still alive? */ |
618f0636 KK |
505 | if (sb->s_root) |
506 | return sb; | |
507 | up_read(&sb->s_umount); | |
df40c01a | 508 | /* nope, got unmounted */ |
618f0636 | 509 | spin_lock(&sb_lock); |
df40c01a AV |
510 | __put_super(sb); |
511 | goto rescan; | |
1da177e4 LT |
512 | } |
513 | } | |
514 | spin_unlock(&sb_lock); | |
515 | return NULL; | |
516 | } | |
517 | ||
257ac264 | 518 | SYSCALL_DEFINE2(ustat, unsigned, dev, struct ustat __user *, ubuf) |
1da177e4 LT |
519 | { |
520 | struct super_block *s; | |
521 | struct ustat tmp; | |
522 | struct kstatfs sbuf; | |
523 | int err = -EINVAL; | |
524 | ||
525 | s = user_get_super(new_decode_dev(dev)); | |
526 | if (s == NULL) | |
527 | goto out; | |
726c3342 | 528 | err = vfs_statfs(s->s_root, &sbuf); |
1da177e4 LT |
529 | drop_super(s); |
530 | if (err) | |
531 | goto out; | |
532 | ||
533 | memset(&tmp,0,sizeof(struct ustat)); | |
534 | tmp.f_tfree = sbuf.f_bfree; | |
535 | tmp.f_tinode = sbuf.f_ffree; | |
536 | ||
537 | err = copy_to_user(ubuf,&tmp,sizeof(struct ustat)) ? -EFAULT : 0; | |
538 | out: | |
539 | return err; | |
540 | } | |
541 | ||
1da177e4 LT |
542 | /** |
543 | * do_remount_sb - asks filesystem to change mount options. | |
544 | * @sb: superblock in question | |
545 | * @flags: numeric part of options | |
546 | * @data: the rest of options | |
547 | * @force: whether or not to force the change | |
548 | * | |
549 | * Alters the mount options of a mounted file system. | |
550 | */ | |
551 | int do_remount_sb(struct super_block *sb, int flags, void *data, int force) | |
552 | { | |
553 | int retval; | |
d208bbdd | 554 | int remount_rw, remount_ro; |
4504230a CH |
555 | |
556 | if (sb->s_frozen != SB_UNFROZEN) | |
557 | return -EBUSY; | |
558 | ||
9361401e | 559 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
560 | if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev)) |
561 | return -EACCES; | |
9361401e | 562 | #endif |
4504230a | 563 | |
1da177e4 LT |
564 | if (flags & MS_RDONLY) |
565 | acct_auto_close(sb); | |
566 | shrink_dcache_sb(sb); | |
60b0680f | 567 | sync_filesystem(sb); |
1da177e4 | 568 | |
d208bbdd NP |
569 | remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY); |
570 | remount_rw = !(flags & MS_RDONLY) && (sb->s_flags & MS_RDONLY); | |
571 | ||
1da177e4 LT |
572 | /* If we are remounting RDONLY and current sb is read/write, |
573 | make sure there are no rw files opened */ | |
d208bbdd | 574 | if (remount_ro) { |
1da177e4 LT |
575 | if (force) |
576 | mark_files_ro(sb); | |
b0895513 | 577 | else if (!fs_may_remount_ro(sb)) |
1da177e4 | 578 | return -EBUSY; |
9e3509e2 | 579 | retval = vfs_dq_off(sb, 1); |
b0895513 | 580 | if (retval < 0 && retval != -ENOSYS) |
0ff5af83 | 581 | return -EBUSY; |
1da177e4 LT |
582 | } |
583 | ||
584 | if (sb->s_op->remount_fs) { | |
1da177e4 | 585 | retval = sb->s_op->remount_fs(sb, &flags, data); |
b0895513 | 586 | if (retval) |
1da177e4 LT |
587 | return retval; |
588 | } | |
589 | sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK); | |
0ff5af83 | 590 | if (remount_rw) |
9e3509e2 | 591 | vfs_dq_quota_on_remount(sb); |
d208bbdd NP |
592 | /* |
593 | * Some filesystems modify their metadata via some other path than the | |
594 | * bdev buffer cache (eg. use a private mapping, or directories in | |
595 | * pagecache, etc). Also file data modifications go via their own | |
596 | * mappings. So If we try to mount readonly then copy the filesystem | |
597 | * from bdev, we could get stale data, so invalidate it to give a best | |
598 | * effort at coherency. | |
599 | */ | |
600 | if (remount_ro && sb->s_bdev) | |
601 | invalidate_bdev(sb->s_bdev); | |
1da177e4 LT |
602 | return 0; |
603 | } | |
604 | ||
a2a9537a | 605 | static void do_emergency_remount(struct work_struct *work) |
1da177e4 | 606 | { |
e7fe0585 | 607 | struct super_block *sb, *n; |
1da177e4 LT |
608 | |
609 | spin_lock(&sb_lock); | |
e7fe0585 | 610 | list_for_each_entry_safe(sb, n, &super_blocks, s_list) { |
551de6f3 AV |
611 | if (list_empty(&sb->s_instances)) |
612 | continue; | |
1da177e4 LT |
613 | sb->s_count++; |
614 | spin_unlock(&sb_lock); | |
443b94ba | 615 | down_write(&sb->s_umount); |
1da177e4 LT |
616 | if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) { |
617 | /* | |
1da177e4 LT |
618 | * What lock protects sb->s_flags?? |
619 | */ | |
1da177e4 | 620 | do_remount_sb(sb, MS_RDONLY, NULL, 1); |
1da177e4 | 621 | } |
443b94ba | 622 | up_write(&sb->s_umount); |
1da177e4 | 623 | spin_lock(&sb_lock); |
e7fe0585 | 624 | __put_super(sb); |
1da177e4 LT |
625 | } |
626 | spin_unlock(&sb_lock); | |
a2a9537a | 627 | kfree(work); |
1da177e4 LT |
628 | printk("Emergency Remount complete\n"); |
629 | } | |
630 | ||
631 | void emergency_remount(void) | |
632 | { | |
a2a9537a JA |
633 | struct work_struct *work; |
634 | ||
635 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | |
636 | if (work) { | |
637 | INIT_WORK(work, do_emergency_remount); | |
638 | schedule_work(work); | |
639 | } | |
1da177e4 LT |
640 | } |
641 | ||
642 | /* | |
643 | * Unnamed block devices are dummy devices used by virtual | |
644 | * filesystems which don't use real block-devices. -- jrs | |
645 | */ | |
646 | ||
ad76cbc6 | 647 | static DEFINE_IDA(unnamed_dev_ida); |
1da177e4 | 648 | static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */ |
c63e09ec | 649 | static int unnamed_dev_start = 0; /* don't bother trying below it */ |
1da177e4 LT |
650 | |
651 | int set_anon_super(struct super_block *s, void *data) | |
652 | { | |
653 | int dev; | |
654 | int error; | |
655 | ||
656 | retry: | |
ad76cbc6 | 657 | if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0) |
1da177e4 LT |
658 | return -ENOMEM; |
659 | spin_lock(&unnamed_dev_lock); | |
c63e09ec | 660 | error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev); |
f21f6220 AV |
661 | if (!error) |
662 | unnamed_dev_start = dev + 1; | |
1da177e4 LT |
663 | spin_unlock(&unnamed_dev_lock); |
664 | if (error == -EAGAIN) | |
665 | /* We raced and lost with another CPU. */ | |
666 | goto retry; | |
667 | else if (error) | |
668 | return -EAGAIN; | |
669 | ||
670 | if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) { | |
671 | spin_lock(&unnamed_dev_lock); | |
ad76cbc6 | 672 | ida_remove(&unnamed_dev_ida, dev); |
f21f6220 AV |
673 | if (unnamed_dev_start > dev) |
674 | unnamed_dev_start = dev; | |
1da177e4 LT |
675 | spin_unlock(&unnamed_dev_lock); |
676 | return -EMFILE; | |
677 | } | |
678 | s->s_dev = MKDEV(0, dev & MINORMASK); | |
5129a469 | 679 | s->s_bdi = &noop_backing_dev_info; |
1da177e4 LT |
680 | return 0; |
681 | } | |
682 | ||
683 | EXPORT_SYMBOL(set_anon_super); | |
684 | ||
685 | void kill_anon_super(struct super_block *sb) | |
686 | { | |
687 | int slot = MINOR(sb->s_dev); | |
688 | ||
689 | generic_shutdown_super(sb); | |
690 | spin_lock(&unnamed_dev_lock); | |
ad76cbc6 | 691 | ida_remove(&unnamed_dev_ida, slot); |
c63e09ec AV |
692 | if (slot < unnamed_dev_start) |
693 | unnamed_dev_start = slot; | |
1da177e4 LT |
694 | spin_unlock(&unnamed_dev_lock); |
695 | } | |
696 | ||
697 | EXPORT_SYMBOL(kill_anon_super); | |
698 | ||
1da177e4 LT |
699 | void kill_litter_super(struct super_block *sb) |
700 | { | |
701 | if (sb->s_root) | |
702 | d_genocide(sb->s_root); | |
703 | kill_anon_super(sb); | |
704 | } | |
705 | ||
706 | EXPORT_SYMBOL(kill_litter_super); | |
707 | ||
909e6d94 SH |
708 | static int ns_test_super(struct super_block *sb, void *data) |
709 | { | |
710 | return sb->s_fs_info == data; | |
711 | } | |
712 | ||
713 | static int ns_set_super(struct super_block *sb, void *data) | |
714 | { | |
715 | sb->s_fs_info = data; | |
716 | return set_anon_super(sb, NULL); | |
717 | } | |
718 | ||
719 | int get_sb_ns(struct file_system_type *fs_type, int flags, void *data, | |
720 | int (*fill_super)(struct super_block *, void *, int), | |
721 | struct vfsmount *mnt) | |
722 | { | |
723 | struct super_block *sb; | |
724 | ||
725 | sb = sget(fs_type, ns_test_super, ns_set_super, data); | |
726 | if (IS_ERR(sb)) | |
727 | return PTR_ERR(sb); | |
728 | ||
729 | if (!sb->s_root) { | |
730 | int err; | |
731 | sb->s_flags = flags; | |
732 | err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0); | |
733 | if (err) { | |
74dbbdd7 | 734 | deactivate_locked_super(sb); |
909e6d94 SH |
735 | return err; |
736 | } | |
737 | ||
738 | sb->s_flags |= MS_ACTIVE; | |
739 | } | |
740 | ||
741 | simple_set_mnt(mnt, sb); | |
742 | return 0; | |
743 | } | |
744 | ||
745 | EXPORT_SYMBOL(get_sb_ns); | |
746 | ||
9361401e | 747 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
748 | static int set_bdev_super(struct super_block *s, void *data) |
749 | { | |
750 | s->s_bdev = data; | |
751 | s->s_dev = s->s_bdev->bd_dev; | |
32a88aa1 JA |
752 | |
753 | /* | |
754 | * We set the bdi here to the queue backing, file systems can | |
755 | * overwrite this in ->fill_super() | |
756 | */ | |
757 | s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info; | |
1da177e4 LT |
758 | return 0; |
759 | } | |
760 | ||
761 | static int test_bdev_super(struct super_block *s, void *data) | |
762 | { | |
763 | return (void *)s->s_bdev == data; | |
764 | } | |
765 | ||
454e2398 | 766 | int get_sb_bdev(struct file_system_type *fs_type, |
1da177e4 | 767 | int flags, const char *dev_name, void *data, |
454e2398 DH |
768 | int (*fill_super)(struct super_block *, void *, int), |
769 | struct vfsmount *mnt) | |
1da177e4 LT |
770 | { |
771 | struct block_device *bdev; | |
772 | struct super_block *s; | |
30c40d2c | 773 | fmode_t mode = FMODE_READ; |
1da177e4 LT |
774 | int error = 0; |
775 | ||
30c40d2c AV |
776 | if (!(flags & MS_RDONLY)) |
777 | mode |= FMODE_WRITE; | |
778 | ||
779 | bdev = open_bdev_exclusive(dev_name, mode, fs_type); | |
1da177e4 | 780 | if (IS_ERR(bdev)) |
454e2398 | 781 | return PTR_ERR(bdev); |
1da177e4 LT |
782 | |
783 | /* | |
784 | * once the super is inserted into the list by sget, s_umount | |
785 | * will protect the lockfs code from trying to start a snapshot | |
786 | * while we are mounting | |
787 | */ | |
4fadd7bb CH |
788 | mutex_lock(&bdev->bd_fsfreeze_mutex); |
789 | if (bdev->bd_fsfreeze_count > 0) { | |
790 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
791 | error = -EBUSY; | |
792 | goto error_bdev; | |
793 | } | |
1da177e4 | 794 | s = sget(fs_type, test_bdev_super, set_bdev_super, bdev); |
4fadd7bb | 795 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
1da177e4 | 796 | if (IS_ERR(s)) |
454e2398 | 797 | goto error_s; |
1da177e4 LT |
798 | |
799 | if (s->s_root) { | |
800 | if ((flags ^ s->s_flags) & MS_RDONLY) { | |
74dbbdd7 | 801 | deactivate_locked_super(s); |
454e2398 DH |
802 | error = -EBUSY; |
803 | goto error_bdev; | |
1da177e4 | 804 | } |
454e2398 | 805 | |
30c40d2c | 806 | close_bdev_exclusive(bdev, mode); |
1da177e4 LT |
807 | } else { |
808 | char b[BDEVNAME_SIZE]; | |
809 | ||
810 | s->s_flags = flags; | |
30c40d2c | 811 | s->s_mode = mode; |
1da177e4 | 812 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); |
e78c9a00 | 813 | sb_set_blocksize(s, block_size(bdev)); |
9b04c997 | 814 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 815 | if (error) { |
74dbbdd7 | 816 | deactivate_locked_super(s); |
454e2398 | 817 | goto error; |
fa675765 | 818 | } |
454e2398 DH |
819 | |
820 | s->s_flags |= MS_ACTIVE; | |
87d8fe1e | 821 | bdev->bd_super = s; |
1da177e4 LT |
822 | } |
823 | ||
a3ec947c SB |
824 | simple_set_mnt(mnt, s); |
825 | return 0; | |
1da177e4 | 826 | |
454e2398 DH |
827 | error_s: |
828 | error = PTR_ERR(s); | |
829 | error_bdev: | |
30c40d2c | 830 | close_bdev_exclusive(bdev, mode); |
454e2398 DH |
831 | error: |
832 | return error; | |
1da177e4 LT |
833 | } |
834 | ||
835 | EXPORT_SYMBOL(get_sb_bdev); | |
836 | ||
837 | void kill_block_super(struct super_block *sb) | |
838 | { | |
839 | struct block_device *bdev = sb->s_bdev; | |
30c40d2c | 840 | fmode_t mode = sb->s_mode; |
1da177e4 | 841 | |
ddbaaf30 | 842 | bdev->bd_super = NULL; |
1da177e4 LT |
843 | generic_shutdown_super(sb); |
844 | sync_blockdev(bdev); | |
30c40d2c | 845 | close_bdev_exclusive(bdev, mode); |
1da177e4 LT |
846 | } |
847 | ||
848 | EXPORT_SYMBOL(kill_block_super); | |
9361401e | 849 | #endif |
1da177e4 | 850 | |
454e2398 | 851 | int get_sb_nodev(struct file_system_type *fs_type, |
1da177e4 | 852 | int flags, void *data, |
454e2398 DH |
853 | int (*fill_super)(struct super_block *, void *, int), |
854 | struct vfsmount *mnt) | |
1da177e4 LT |
855 | { |
856 | int error; | |
857 | struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL); | |
858 | ||
859 | if (IS_ERR(s)) | |
454e2398 | 860 | return PTR_ERR(s); |
1da177e4 LT |
861 | |
862 | s->s_flags = flags; | |
863 | ||
9b04c997 | 864 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 865 | if (error) { |
74dbbdd7 | 866 | deactivate_locked_super(s); |
454e2398 | 867 | return error; |
1da177e4 LT |
868 | } |
869 | s->s_flags |= MS_ACTIVE; | |
a3ec947c SB |
870 | simple_set_mnt(mnt, s); |
871 | return 0; | |
1da177e4 LT |
872 | } |
873 | ||
874 | EXPORT_SYMBOL(get_sb_nodev); | |
875 | ||
876 | static int compare_single(struct super_block *s, void *p) | |
877 | { | |
878 | return 1; | |
879 | } | |
880 | ||
454e2398 | 881 | int get_sb_single(struct file_system_type *fs_type, |
1da177e4 | 882 | int flags, void *data, |
454e2398 DH |
883 | int (*fill_super)(struct super_block *, void *, int), |
884 | struct vfsmount *mnt) | |
1da177e4 LT |
885 | { |
886 | struct super_block *s; | |
887 | int error; | |
888 | ||
889 | s = sget(fs_type, compare_single, set_anon_super, NULL); | |
890 | if (IS_ERR(s)) | |
454e2398 | 891 | return PTR_ERR(s); |
1da177e4 LT |
892 | if (!s->s_root) { |
893 | s->s_flags = flags; | |
9b04c997 | 894 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 895 | if (error) { |
74dbbdd7 | 896 | deactivate_locked_super(s); |
454e2398 | 897 | return error; |
1da177e4 LT |
898 | } |
899 | s->s_flags |= MS_ACTIVE; | |
9329d1be KS |
900 | } else { |
901 | do_remount_sb(s, flags, data, 0); | |
1da177e4 | 902 | } |
a3ec947c SB |
903 | simple_set_mnt(mnt, s); |
904 | return 0; | |
1da177e4 LT |
905 | } |
906 | ||
907 | EXPORT_SYMBOL(get_sb_single); | |
908 | ||
909 | struct vfsmount * | |
bb4a58bf | 910 | vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data) |
1da177e4 | 911 | { |
1da177e4 | 912 | struct vfsmount *mnt; |
1da177e4 | 913 | char *secdata = NULL; |
454e2398 | 914 | int error; |
1da177e4 LT |
915 | |
916 | if (!type) | |
917 | return ERR_PTR(-ENODEV); | |
1da177e4 | 918 | |
454e2398 | 919 | error = -ENOMEM; |
1da177e4 LT |
920 | mnt = alloc_vfsmnt(name); |
921 | if (!mnt) | |
922 | goto out; | |
923 | ||
8089352a AV |
924 | if (flags & MS_KERNMOUNT) |
925 | mnt->mnt_flags = MNT_INTERNAL; | |
926 | ||
e0007529 | 927 | if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) { |
1da177e4 | 928 | secdata = alloc_secdata(); |
454e2398 | 929 | if (!secdata) |
1da177e4 | 930 | goto out_mnt; |
1da177e4 | 931 | |
e0007529 | 932 | error = security_sb_copy_data(data, secdata); |
454e2398 | 933 | if (error) |
1da177e4 | 934 | goto out_free_secdata; |
1da177e4 LT |
935 | } |
936 | ||
454e2398 DH |
937 | error = type->get_sb(type, flags, name, data, mnt); |
938 | if (error < 0) | |
1da177e4 | 939 | goto out_free_secdata; |
b4c07bce | 940 | BUG_ON(!mnt->mnt_sb); |
5129a469 | 941 | WARN_ON(!mnt->mnt_sb->s_bdi); |
454e2398 | 942 | |
5129a469 JE |
943 | error = security_sb_kern_mount(mnt->mnt_sb, flags, secdata); |
944 | if (error) | |
945 | goto out_sb; | |
454e2398 | 946 | |
42cb56ae JL |
947 | /* |
948 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE | |
949 | * but s_maxbytes was an unsigned long long for many releases. Throw | |
950 | * this warning for a little while to try and catch filesystems that | |
951 | * violate this rule. This warning should be either removed or | |
952 | * converted to a BUG() in 2.6.34. | |
953 | */ | |
954 | WARN((mnt->mnt_sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " | |
955 | "negative value (%lld)\n", type->name, mnt->mnt_sb->s_maxbytes); | |
956 | ||
454e2398 | 957 | mnt->mnt_mountpoint = mnt->mnt_root; |
1da177e4 | 958 | mnt->mnt_parent = mnt; |
454e2398 | 959 | up_write(&mnt->mnt_sb->s_umount); |
8680e22f | 960 | free_secdata(secdata); |
1da177e4 LT |
961 | return mnt; |
962 | out_sb: | |
454e2398 | 963 | dput(mnt->mnt_root); |
74dbbdd7 | 964 | deactivate_locked_super(mnt->mnt_sb); |
1da177e4 LT |
965 | out_free_secdata: |
966 | free_secdata(secdata); | |
967 | out_mnt: | |
968 | free_vfsmnt(mnt); | |
969 | out: | |
454e2398 | 970 | return ERR_PTR(error); |
1da177e4 LT |
971 | } |
972 | ||
bb4a58bf TM |
973 | EXPORT_SYMBOL_GPL(vfs_kern_mount); |
974 | ||
79c0b2df MS |
975 | static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype) |
976 | { | |
977 | int err; | |
978 | const char *subtype = strchr(fstype, '.'); | |
979 | if (subtype) { | |
980 | subtype++; | |
981 | err = -EINVAL; | |
982 | if (!subtype[0]) | |
983 | goto err; | |
984 | } else | |
985 | subtype = ""; | |
986 | ||
987 | mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL); | |
988 | err = -ENOMEM; | |
989 | if (!mnt->mnt_sb->s_subtype) | |
990 | goto err; | |
991 | return mnt; | |
992 | ||
993 | err: | |
994 | mntput(mnt); | |
995 | return ERR_PTR(err); | |
996 | } | |
997 | ||
bb4a58bf TM |
998 | struct vfsmount * |
999 | do_kern_mount(const char *fstype, int flags, const char *name, void *data) | |
1000 | { | |
1001 | struct file_system_type *type = get_fs_type(fstype); | |
1002 | struct vfsmount *mnt; | |
1003 | if (!type) | |
1004 | return ERR_PTR(-ENODEV); | |
1005 | mnt = vfs_kern_mount(type, flags, name, data); | |
79c0b2df MS |
1006 | if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) && |
1007 | !mnt->mnt_sb->s_subtype) | |
1008 | mnt = fs_set_subtype(mnt, fstype); | |
bb4a58bf TM |
1009 | put_filesystem(type); |
1010 | return mnt; | |
1011 | } | |
8a4e98d9 | 1012 | EXPORT_SYMBOL_GPL(do_kern_mount); |
bb4a58bf | 1013 | |
8bf9725c | 1014 | struct vfsmount *kern_mount_data(struct file_system_type *type, void *data) |
1da177e4 | 1015 | { |
8bf9725c | 1016 | return vfs_kern_mount(type, MS_KERNMOUNT, type->name, data); |
1da177e4 LT |
1017 | } |
1018 | ||
8bf9725c | 1019 | EXPORT_SYMBOL_GPL(kern_mount_data); |