]>
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> | |
1da177e4 LT |
25 | #include <linux/acct.h> |
26 | #include <linux/blkdev.h> | |
1da177e4 LT |
27 | #include <linux/mount.h> |
28 | #include <linux/security.h> | |
1da177e4 LT |
29 | #include <linux/writeback.h> /* for the emergency remount stuff */ |
30 | #include <linux/idr.h> | |
353ab6e9 | 31 | #include <linux/mutex.h> |
5477d0fa | 32 | #include <linux/backing-dev.h> |
ceb5bdc2 | 33 | #include <linux/rculist_bl.h> |
c515e1fd | 34 | #include <linux/cleancache.h> |
6d59e7f5 | 35 | #include "internal.h" |
1da177e4 LT |
36 | |
37 | ||
1da177e4 LT |
38 | LIST_HEAD(super_blocks); |
39 | DEFINE_SPINLOCK(sb_lock); | |
40 | ||
b0d40c92 DC |
41 | /* |
42 | * One thing we have to be careful of with a per-sb shrinker is that we don't | |
43 | * drop the last active reference to the superblock from within the shrinker. | |
44 | * If that happens we could trigger unregistering the shrinker from within the | |
45 | * shrinker path and that leads to deadlock on the shrinker_rwsem. Hence we | |
46 | * take a passive reference to the superblock to avoid this from occurring. | |
47 | */ | |
48 | static int prune_super(struct shrinker *shrink, struct shrink_control *sc) | |
49 | { | |
50 | struct super_block *sb; | |
0e1fdafd DC |
51 | int fs_objects = 0; |
52 | int total_objects; | |
b0d40c92 DC |
53 | |
54 | sb = container_of(shrink, struct super_block, s_shrink); | |
55 | ||
56 | /* | |
57 | * Deadlock avoidance. We may hold various FS locks, and we don't want | |
58 | * to recurse into the FS that called us in clear_inode() and friends.. | |
59 | */ | |
60 | if (sc->nr_to_scan && !(sc->gfp_mask & __GFP_FS)) | |
61 | return -1; | |
62 | ||
63 | if (!grab_super_passive(sb)) | |
09f363c7 | 64 | return !sc->nr_to_scan ? 0 : -1; |
b0d40c92 | 65 | |
0e1fdafd DC |
66 | if (sb->s_op && sb->s_op->nr_cached_objects) |
67 | fs_objects = sb->s_op->nr_cached_objects(sb); | |
68 | ||
69 | total_objects = sb->s_nr_dentry_unused + | |
70 | sb->s_nr_inodes_unused + fs_objects + 1; | |
71 | ||
b0d40c92 | 72 | if (sc->nr_to_scan) { |
0e1fdafd DC |
73 | int dentries; |
74 | int inodes; | |
b0d40c92 | 75 | |
0e1fdafd DC |
76 | /* proportion the scan between the caches */ |
77 | dentries = (sc->nr_to_scan * sb->s_nr_dentry_unused) / | |
78 | total_objects; | |
79 | inodes = (sc->nr_to_scan * sb->s_nr_inodes_unused) / | |
80 | total_objects; | |
81 | if (fs_objects) | |
82 | fs_objects = (sc->nr_to_scan * fs_objects) / | |
83 | total_objects; | |
84 | /* | |
85 | * prune the dcache first as the icache is pinned by it, then | |
86 | * prune the icache, followed by the filesystem specific caches | |
87 | */ | |
88 | prune_dcache_sb(sb, dentries); | |
89 | prune_icache_sb(sb, inodes); | |
b0d40c92 | 90 | |
0e1fdafd DC |
91 | if (fs_objects && sb->s_op->free_cached_objects) { |
92 | sb->s_op->free_cached_objects(sb, fs_objects); | |
93 | fs_objects = sb->s_op->nr_cached_objects(sb); | |
94 | } | |
95 | total_objects = sb->s_nr_dentry_unused + | |
96 | sb->s_nr_inodes_unused + fs_objects; | |
b0d40c92 DC |
97 | } |
98 | ||
0e1fdafd | 99 | total_objects = (total_objects / 100) * sysctl_vfs_cache_pressure; |
b0d40c92 | 100 | drop_super(sb); |
0e1fdafd | 101 | return total_objects; |
b0d40c92 DC |
102 | } |
103 | ||
1da177e4 LT |
104 | /** |
105 | * alloc_super - create new superblock | |
fe2bbc48 | 106 | * @type: filesystem type superblock should belong to |
1da177e4 LT |
107 | * |
108 | * Allocates and initializes a new &struct super_block. alloc_super() | |
109 | * returns a pointer new superblock or %NULL if allocation had failed. | |
110 | */ | |
cf516249 | 111 | static struct super_block *alloc_super(struct file_system_type *type) |
1da177e4 | 112 | { |
11b0b5ab | 113 | struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER); |
b87221de | 114 | static const struct super_operations default_op; |
1da177e4 LT |
115 | |
116 | if (s) { | |
1da177e4 LT |
117 | if (security_sb_alloc(s)) { |
118 | kfree(s); | |
119 | s = NULL; | |
120 | goto out; | |
121 | } | |
6416ccb7 NP |
122 | #ifdef CONFIG_SMP |
123 | s->s_files = alloc_percpu(struct list_head); | |
124 | if (!s->s_files) { | |
125 | security_sb_free(s); | |
126 | kfree(s); | |
127 | s = NULL; | |
128 | goto out; | |
129 | } else { | |
130 | int i; | |
131 | ||
132 | for_each_possible_cpu(i) | |
133 | INIT_LIST_HEAD(per_cpu_ptr(s->s_files, i)); | |
134 | } | |
135 | #else | |
1da177e4 | 136 | INIT_LIST_HEAD(&s->s_files); |
6416ccb7 | 137 | #endif |
95f28604 | 138 | s->s_bdi = &default_backing_dev_info; |
a5166169 | 139 | INIT_HLIST_NODE(&s->s_instances); |
ceb5bdc2 | 140 | INIT_HLIST_BL_HEAD(&s->s_anon); |
1da177e4 | 141 | INIT_LIST_HEAD(&s->s_inodes); |
da3bbdd4 | 142 | INIT_LIST_HEAD(&s->s_dentry_lru); |
98b745c6 | 143 | INIT_LIST_HEAD(&s->s_inode_lru); |
09cc9fc7 | 144 | spin_lock_init(&s->s_inode_lru_lock); |
39f7c4db | 145 | INIT_LIST_HEAD(&s->s_mounts); |
1da177e4 | 146 | init_rwsem(&s->s_umount); |
7892f2f4 | 147 | mutex_init(&s->s_lock); |
897c6ff9 | 148 | lockdep_set_class(&s->s_umount, &type->s_umount_key); |
cf516249 IM |
149 | /* |
150 | * The locking rules for s_lock are up to the | |
151 | * filesystem. For example ext3fs has different | |
152 | * lock ordering than usbfs: | |
153 | */ | |
154 | lockdep_set_class(&s->s_lock, &type->s_lock_key); | |
ada723dc PZ |
155 | /* |
156 | * sget() can have s_umount recursion. | |
157 | * | |
158 | * When it cannot find a suitable sb, it allocates a new | |
159 | * one (this one), and tries again to find a suitable old | |
160 | * one. | |
161 | * | |
162 | * In case that succeeds, it will acquire the s_umount | |
163 | * lock of the old one. Since these are clearly distrinct | |
164 | * locks, and this object isn't exposed yet, there's no | |
165 | * risk of deadlocks. | |
166 | * | |
167 | * Annotate this by putting this lock in a different | |
168 | * subclass. | |
169 | */ | |
170 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); | |
b20bd1a5 | 171 | s->s_count = 1; |
1da177e4 | 172 | atomic_set(&s->s_active, 1); |
a11f3a05 | 173 | mutex_init(&s->s_vfs_rename_mutex); |
51ee049e | 174 | lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key); |
d3be915f IM |
175 | mutex_init(&s->s_dquot.dqio_mutex); |
176 | mutex_init(&s->s_dquot.dqonoff_mutex); | |
1da177e4 LT |
177 | init_rwsem(&s->s_dquot.dqptr_sem); |
178 | init_waitqueue_head(&s->s_wait_unfrozen); | |
179 | s->s_maxbytes = MAX_NON_LFS; | |
1da177e4 LT |
180 | s->s_op = &default_op; |
181 | s->s_time_gran = 1000000000; | |
c515e1fd | 182 | s->cleancache_poolid = -1; |
b0d40c92 DC |
183 | |
184 | s->s_shrink.seeks = DEFAULT_SEEKS; | |
185 | s->s_shrink.shrink = prune_super; | |
8ab47664 | 186 | s->s_shrink.batch = 1024; |
1da177e4 LT |
187 | } |
188 | out: | |
189 | return s; | |
190 | } | |
191 | ||
192 | /** | |
193 | * destroy_super - frees a superblock | |
194 | * @s: superblock to free | |
195 | * | |
196 | * Frees a superblock. | |
197 | */ | |
198 | static inline void destroy_super(struct super_block *s) | |
199 | { | |
6416ccb7 NP |
200 | #ifdef CONFIG_SMP |
201 | free_percpu(s->s_files); | |
202 | #endif | |
1da177e4 | 203 | security_sb_free(s); |
39f7c4db | 204 | WARN_ON(!list_empty(&s->s_mounts)); |
79c0b2df | 205 | kfree(s->s_subtype); |
b3b304a2 | 206 | kfree(s->s_options); |
1da177e4 LT |
207 | kfree(s); |
208 | } | |
209 | ||
210 | /* Superblock refcounting */ | |
211 | ||
212 | /* | |
35cf7ba0 | 213 | * Drop a superblock's refcount. The caller must hold sb_lock. |
1da177e4 | 214 | */ |
f47ec3f2 | 215 | static void __put_super(struct super_block *sb) |
1da177e4 | 216 | { |
1da177e4 | 217 | if (!--sb->s_count) { |
551de6f3 | 218 | list_del_init(&sb->s_list); |
1da177e4 | 219 | destroy_super(sb); |
1da177e4 | 220 | } |
1da177e4 LT |
221 | } |
222 | ||
223 | /** | |
224 | * put_super - drop a temporary reference to superblock | |
225 | * @sb: superblock in question | |
226 | * | |
227 | * Drops a temporary reference, frees superblock if there's no | |
228 | * references left. | |
229 | */ | |
f47ec3f2 | 230 | static void put_super(struct super_block *sb) |
1da177e4 LT |
231 | { |
232 | spin_lock(&sb_lock); | |
233 | __put_super(sb); | |
234 | spin_unlock(&sb_lock); | |
235 | } | |
236 | ||
237 | ||
238 | /** | |
1712ac8f | 239 | * deactivate_locked_super - drop an active reference to superblock |
1da177e4 LT |
240 | * @s: superblock to deactivate |
241 | * | |
1712ac8f AV |
242 | * Drops an active reference to superblock, converting it into a temprory |
243 | * one if there is no other active references left. In that case we | |
1da177e4 LT |
244 | * tell fs driver to shut it down and drop the temporary reference we |
245 | * had just acquired. | |
1712ac8f AV |
246 | * |
247 | * Caller holds exclusive lock on superblock; that lock is released. | |
1da177e4 | 248 | */ |
1712ac8f | 249 | void deactivate_locked_super(struct super_block *s) |
1da177e4 LT |
250 | { |
251 | struct file_system_type *fs = s->s_type; | |
b20bd1a5 | 252 | if (atomic_dec_and_test(&s->s_active)) { |
c515e1fd | 253 | cleancache_flush_fs(s); |
1da177e4 | 254 | fs->kill_sb(s); |
b0d40c92 DC |
255 | |
256 | /* caches are now gone, we can safely kill the shrinker now */ | |
257 | unregister_shrinker(&s->s_shrink); | |
258 | ||
d863b50a BH |
259 | /* |
260 | * We need to call rcu_barrier so all the delayed rcu free | |
261 | * inodes are flushed before we release the fs module. | |
262 | */ | |
263 | rcu_barrier(); | |
1da177e4 LT |
264 | put_filesystem(fs); |
265 | put_super(s); | |
1712ac8f AV |
266 | } else { |
267 | up_write(&s->s_umount); | |
1da177e4 LT |
268 | } |
269 | } | |
270 | ||
1712ac8f | 271 | EXPORT_SYMBOL(deactivate_locked_super); |
1da177e4 | 272 | |
74dbbdd7 | 273 | /** |
1712ac8f | 274 | * deactivate_super - drop an active reference to superblock |
74dbbdd7 AV |
275 | * @s: superblock to deactivate |
276 | * | |
1712ac8f AV |
277 | * Variant of deactivate_locked_super(), except that superblock is *not* |
278 | * locked by caller. If we are going to drop the final active reference, | |
279 | * lock will be acquired prior to that. | |
74dbbdd7 | 280 | */ |
1712ac8f | 281 | void deactivate_super(struct super_block *s) |
74dbbdd7 | 282 | { |
1712ac8f AV |
283 | if (!atomic_add_unless(&s->s_active, -1, 1)) { |
284 | down_write(&s->s_umount); | |
285 | deactivate_locked_super(s); | |
74dbbdd7 AV |
286 | } |
287 | } | |
288 | ||
1712ac8f | 289 | EXPORT_SYMBOL(deactivate_super); |
74dbbdd7 | 290 | |
1da177e4 LT |
291 | /** |
292 | * grab_super - acquire an active reference | |
293 | * @s: reference we are trying to make active | |
294 | * | |
295 | * Tries to acquire an active reference. grab_super() is used when we | |
296 | * had just found a superblock in super_blocks or fs_type->fs_supers | |
297 | * and want to turn it into a full-blown active reference. grab_super() | |
298 | * is called with sb_lock held and drops it. Returns 1 in case of | |
299 | * success, 0 if we had failed (superblock contents was already dead or | |
300 | * dying when grab_super() had been called). | |
301 | */ | |
9c4dbee7 | 302 | static int grab_super(struct super_block *s) __releases(sb_lock) |
1da177e4 | 303 | { |
b20bd1a5 AV |
304 | if (atomic_inc_not_zero(&s->s_active)) { |
305 | spin_unlock(&sb_lock); | |
b20bd1a5 AV |
306 | return 1; |
307 | } | |
308 | /* it's going away */ | |
1da177e4 LT |
309 | s->s_count++; |
310 | spin_unlock(&sb_lock); | |
1712ac8f | 311 | /* wait for it to die */ |
1da177e4 | 312 | down_write(&s->s_umount); |
1da177e4 LT |
313 | up_write(&s->s_umount); |
314 | put_super(s); | |
1da177e4 LT |
315 | return 0; |
316 | } | |
317 | ||
12ad3ab6 DC |
318 | /* |
319 | * grab_super_passive - acquire a passive reference | |
320 | * @s: reference we are trying to grab | |
321 | * | |
322 | * Tries to acquire a passive reference. This is used in places where we | |
323 | * cannot take an active reference but we need to ensure that the | |
324 | * superblock does not go away while we are working on it. It returns | |
325 | * false if a reference was not gained, and returns true with the s_umount | |
326 | * lock held in read mode if a reference is gained. On successful return, | |
327 | * the caller must drop the s_umount lock and the passive reference when | |
328 | * done. | |
329 | */ | |
330 | bool grab_super_passive(struct super_block *sb) | |
331 | { | |
332 | spin_lock(&sb_lock); | |
a5166169 | 333 | if (hlist_unhashed(&sb->s_instances)) { |
12ad3ab6 DC |
334 | spin_unlock(&sb_lock); |
335 | return false; | |
336 | } | |
337 | ||
338 | sb->s_count++; | |
339 | spin_unlock(&sb_lock); | |
340 | ||
341 | if (down_read_trylock(&sb->s_umount)) { | |
dabe0dc1 | 342 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
12ad3ab6 DC |
343 | return true; |
344 | up_read(&sb->s_umount); | |
345 | } | |
346 | ||
347 | put_super(sb); | |
348 | return false; | |
349 | } | |
350 | ||
914e2637 AV |
351 | /* |
352 | * Superblock locking. We really ought to get rid of these two. | |
353 | */ | |
354 | void lock_super(struct super_block * sb) | |
355 | { | |
914e2637 AV |
356 | mutex_lock(&sb->s_lock); |
357 | } | |
358 | ||
359 | void unlock_super(struct super_block * sb) | |
360 | { | |
914e2637 AV |
361 | mutex_unlock(&sb->s_lock); |
362 | } | |
363 | ||
364 | EXPORT_SYMBOL(lock_super); | |
365 | EXPORT_SYMBOL(unlock_super); | |
366 | ||
1da177e4 LT |
367 | /** |
368 | * generic_shutdown_super - common helper for ->kill_sb() | |
369 | * @sb: superblock to kill | |
370 | * | |
371 | * generic_shutdown_super() does all fs-independent work on superblock | |
372 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects | |
373 | * that need destruction out of superblock, call generic_shutdown_super() | |
374 | * and release aforementioned objects. Note: dentries and inodes _are_ | |
375 | * taken care of and do not need specific handling. | |
c636ebdb DH |
376 | * |
377 | * Upon calling this function, the filesystem may no longer alter or | |
378 | * rearrange the set of dentries belonging to this super_block, nor may it | |
379 | * change the attachments of dentries to inodes. | |
1da177e4 LT |
380 | */ |
381 | void generic_shutdown_super(struct super_block *sb) | |
382 | { | |
ee9b6d61 | 383 | const struct super_operations *sop = sb->s_op; |
1da177e4 | 384 | |
c636ebdb DH |
385 | if (sb->s_root) { |
386 | shrink_dcache_for_umount(sb); | |
60b0680f | 387 | sync_filesystem(sb); |
1da177e4 | 388 | sb->s_flags &= ~MS_ACTIVE; |
efaee192 | 389 | |
63997e98 AV |
390 | fsnotify_unmount_inodes(&sb->s_inodes); |
391 | ||
392 | evict_inodes(sb); | |
1da177e4 | 393 | |
1da177e4 LT |
394 | if (sop->put_super) |
395 | sop->put_super(sb); | |
396 | ||
63997e98 | 397 | if (!list_empty(&sb->s_inodes)) { |
7b4fe29e DJ |
398 | printk("VFS: Busy inodes after unmount of %s. " |
399 | "Self-destruct in 5 seconds. Have a nice day...\n", | |
400 | sb->s_id); | |
1da177e4 | 401 | } |
1da177e4 LT |
402 | } |
403 | spin_lock(&sb_lock); | |
404 | /* should be initialized for __put_super_and_need_restart() */ | |
a5166169 | 405 | hlist_del_init(&sb->s_instances); |
1da177e4 LT |
406 | spin_unlock(&sb_lock); |
407 | up_write(&sb->s_umount); | |
408 | } | |
409 | ||
410 | EXPORT_SYMBOL(generic_shutdown_super); | |
411 | ||
412 | /** | |
413 | * sget - find or create a superblock | |
414 | * @type: filesystem type superblock should belong to | |
415 | * @test: comparison callback | |
416 | * @set: setup callback | |
417 | * @data: argument to each of them | |
418 | */ | |
419 | struct super_block *sget(struct file_system_type *type, | |
420 | int (*test)(struct super_block *,void *), | |
421 | int (*set)(struct super_block *,void *), | |
422 | void *data) | |
423 | { | |
424 | struct super_block *s = NULL; | |
a5166169 | 425 | struct hlist_node *node; |
d4730127 | 426 | struct super_block *old; |
1da177e4 LT |
427 | int err; |
428 | ||
429 | retry: | |
430 | spin_lock(&sb_lock); | |
d4730127 | 431 | if (test) { |
a5166169 | 432 | hlist_for_each_entry(old, node, &type->fs_supers, s_instances) { |
d4730127 MK |
433 | if (!test(old, data)) |
434 | continue; | |
435 | if (!grab_super(old)) | |
436 | goto retry; | |
a3cfbb53 LZ |
437 | if (s) { |
438 | up_write(&s->s_umount); | |
d4730127 | 439 | destroy_super(s); |
7a4dec53 | 440 | s = NULL; |
a3cfbb53 | 441 | } |
d3f21473 | 442 | down_write(&old->s_umount); |
7a4dec53 AV |
443 | if (unlikely(!(old->s_flags & MS_BORN))) { |
444 | deactivate_locked_super(old); | |
445 | goto retry; | |
446 | } | |
d4730127 MK |
447 | return old; |
448 | } | |
1da177e4 LT |
449 | } |
450 | if (!s) { | |
451 | spin_unlock(&sb_lock); | |
cf516249 | 452 | s = alloc_super(type); |
1da177e4 LT |
453 | if (!s) |
454 | return ERR_PTR(-ENOMEM); | |
455 | goto retry; | |
456 | } | |
457 | ||
458 | err = set(s, data); | |
459 | if (err) { | |
460 | spin_unlock(&sb_lock); | |
a3cfbb53 | 461 | up_write(&s->s_umount); |
1da177e4 LT |
462 | destroy_super(s); |
463 | return ERR_PTR(err); | |
464 | } | |
465 | s->s_type = type; | |
466 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); | |
467 | list_add_tail(&s->s_list, &super_blocks); | |
a5166169 | 468 | hlist_add_head(&s->s_instances, &type->fs_supers); |
1da177e4 LT |
469 | spin_unlock(&sb_lock); |
470 | get_filesystem(type); | |
b0d40c92 | 471 | register_shrinker(&s->s_shrink); |
1da177e4 LT |
472 | return s; |
473 | } | |
474 | ||
475 | EXPORT_SYMBOL(sget); | |
476 | ||
477 | void drop_super(struct super_block *sb) | |
478 | { | |
479 | up_read(&sb->s_umount); | |
480 | put_super(sb); | |
481 | } | |
482 | ||
483 | EXPORT_SYMBOL(drop_super); | |
484 | ||
e5004753 CH |
485 | /** |
486 | * sync_supers - helper for periodic superblock writeback | |
487 | * | |
488 | * Call the write_super method if present on all dirty superblocks in | |
489 | * the system. This is for the periodic writeback used by most older | |
490 | * filesystems. For data integrity superblock writeback use | |
491 | * sync_filesystems() instead. | |
492 | * | |
1da177e4 LT |
493 | * Note: check the dirty flag before waiting, so we don't |
494 | * hold up the sync while mounting a device. (The newly | |
495 | * mounted device won't need syncing.) | |
496 | */ | |
497 | void sync_supers(void) | |
498 | { | |
dca33252 | 499 | struct super_block *sb, *p = NULL; |
618f0636 | 500 | |
1da177e4 | 501 | spin_lock(&sb_lock); |
dca33252 | 502 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 503 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 504 | continue; |
e5004753 | 505 | if (sb->s_op->write_super && sb->s_dirt) { |
1da177e4 LT |
506 | sb->s_count++; |
507 | spin_unlock(&sb_lock); | |
e5004753 | 508 | |
1da177e4 | 509 | down_read(&sb->s_umount); |
dabe0dc1 | 510 | if (sb->s_root && sb->s_dirt && (sb->s_flags & MS_BORN)) |
e5004753 | 511 | sb->s_op->write_super(sb); |
618f0636 | 512 | up_read(&sb->s_umount); |
e5004753 | 513 | |
618f0636 | 514 | spin_lock(&sb_lock); |
dca33252 AV |
515 | if (p) |
516 | __put_super(p); | |
517 | p = sb; | |
618f0636 KK |
518 | } |
519 | } | |
dca33252 AV |
520 | if (p) |
521 | __put_super(p); | |
1da177e4 LT |
522 | spin_unlock(&sb_lock); |
523 | } | |
524 | ||
01a05b33 AV |
525 | /** |
526 | * iterate_supers - call function for all active superblocks | |
527 | * @f: function to call | |
528 | * @arg: argument to pass to it | |
529 | * | |
530 | * Scans the superblock list and calls given function, passing it | |
531 | * locked superblock and given argument. | |
532 | */ | |
533 | void iterate_supers(void (*f)(struct super_block *, void *), void *arg) | |
534 | { | |
dca33252 | 535 | struct super_block *sb, *p = NULL; |
01a05b33 AV |
536 | |
537 | spin_lock(&sb_lock); | |
dca33252 | 538 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 539 | if (hlist_unhashed(&sb->s_instances)) |
01a05b33 AV |
540 | continue; |
541 | sb->s_count++; | |
542 | spin_unlock(&sb_lock); | |
543 | ||
544 | down_read(&sb->s_umount); | |
dabe0dc1 | 545 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
01a05b33 AV |
546 | f(sb, arg); |
547 | up_read(&sb->s_umount); | |
548 | ||
549 | spin_lock(&sb_lock); | |
dca33252 AV |
550 | if (p) |
551 | __put_super(p); | |
552 | p = sb; | |
01a05b33 | 553 | } |
dca33252 AV |
554 | if (p) |
555 | __put_super(p); | |
01a05b33 AV |
556 | spin_unlock(&sb_lock); |
557 | } | |
558 | ||
43e15cdb AV |
559 | /** |
560 | * iterate_supers_type - call function for superblocks of given type | |
561 | * @type: fs type | |
562 | * @f: function to call | |
563 | * @arg: argument to pass to it | |
564 | * | |
565 | * Scans the superblock list and calls given function, passing it | |
566 | * locked superblock and given argument. | |
567 | */ | |
568 | void iterate_supers_type(struct file_system_type *type, | |
569 | void (*f)(struct super_block *, void *), void *arg) | |
570 | { | |
571 | struct super_block *sb, *p = NULL; | |
a5166169 | 572 | struct hlist_node *node; |
43e15cdb AV |
573 | |
574 | spin_lock(&sb_lock); | |
a5166169 | 575 | hlist_for_each_entry(sb, node, &type->fs_supers, s_instances) { |
43e15cdb AV |
576 | sb->s_count++; |
577 | spin_unlock(&sb_lock); | |
578 | ||
579 | down_read(&sb->s_umount); | |
dabe0dc1 | 580 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
43e15cdb AV |
581 | f(sb, arg); |
582 | up_read(&sb->s_umount); | |
583 | ||
584 | spin_lock(&sb_lock); | |
585 | if (p) | |
586 | __put_super(p); | |
587 | p = sb; | |
588 | } | |
589 | if (p) | |
590 | __put_super(p); | |
591 | spin_unlock(&sb_lock); | |
592 | } | |
593 | ||
594 | EXPORT_SYMBOL(iterate_supers_type); | |
595 | ||
1da177e4 LT |
596 | /** |
597 | * get_super - get the superblock of a device | |
598 | * @bdev: device to get the superblock for | |
599 | * | |
600 | * Scans the superblock list and finds the superblock of the file system | |
601 | * mounted on the device given. %NULL is returned if no match is found. | |
602 | */ | |
603 | ||
df40c01a | 604 | struct super_block *get_super(struct block_device *bdev) |
1da177e4 | 605 | { |
618f0636 KK |
606 | struct super_block *sb; |
607 | ||
1da177e4 LT |
608 | if (!bdev) |
609 | return NULL; | |
618f0636 | 610 | |
1da177e4 | 611 | spin_lock(&sb_lock); |
618f0636 KK |
612 | rescan: |
613 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 614 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 615 | continue; |
618f0636 KK |
616 | if (sb->s_bdev == bdev) { |
617 | sb->s_count++; | |
1da177e4 | 618 | spin_unlock(&sb_lock); |
618f0636 | 619 | down_read(&sb->s_umount); |
df40c01a | 620 | /* still alive? */ |
dabe0dc1 | 621 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
618f0636 KK |
622 | return sb; |
623 | up_read(&sb->s_umount); | |
df40c01a | 624 | /* nope, got unmounted */ |
618f0636 | 625 | spin_lock(&sb_lock); |
df40c01a AV |
626 | __put_super(sb); |
627 | goto rescan; | |
1da177e4 LT |
628 | } |
629 | } | |
630 | spin_unlock(&sb_lock); | |
631 | return NULL; | |
632 | } | |
633 | ||
634 | EXPORT_SYMBOL(get_super); | |
4504230a | 635 | |
6b6dc836 JK |
636 | /** |
637 | * get_super_thawed - get thawed superblock of a device | |
638 | * @bdev: device to get the superblock for | |
639 | * | |
640 | * Scans the superblock list and finds the superblock of the file system | |
641 | * mounted on the device. The superblock is returned once it is thawed | |
642 | * (or immediately if it was not frozen). %NULL is returned if no match | |
643 | * is found. | |
644 | */ | |
645 | struct super_block *get_super_thawed(struct block_device *bdev) | |
646 | { | |
647 | while (1) { | |
648 | struct super_block *s = get_super(bdev); | |
649 | if (!s || s->s_frozen == SB_UNFROZEN) | |
650 | return s; | |
651 | up_read(&s->s_umount); | |
652 | vfs_check_frozen(s, SB_FREEZE_WRITE); | |
653 | put_super(s); | |
654 | } | |
655 | } | |
656 | EXPORT_SYMBOL(get_super_thawed); | |
657 | ||
4504230a CH |
658 | /** |
659 | * get_active_super - get an active reference to the superblock of a device | |
660 | * @bdev: device to get the superblock for | |
661 | * | |
662 | * Scans the superblock list and finds the superblock of the file system | |
663 | * mounted on the device given. Returns the superblock with an active | |
d3f21473 | 664 | * reference or %NULL if none was found. |
4504230a CH |
665 | */ |
666 | struct super_block *get_active_super(struct block_device *bdev) | |
667 | { | |
668 | struct super_block *sb; | |
669 | ||
670 | if (!bdev) | |
671 | return NULL; | |
672 | ||
1494583d | 673 | restart: |
4504230a CH |
674 | spin_lock(&sb_lock); |
675 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 676 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 677 | continue; |
1494583d AV |
678 | if (sb->s_bdev == bdev) { |
679 | if (grab_super(sb)) /* drops sb_lock */ | |
680 | return sb; | |
681 | else | |
682 | goto restart; | |
683 | } | |
4504230a CH |
684 | } |
685 | spin_unlock(&sb_lock); | |
686 | return NULL; | |
687 | } | |
1da177e4 | 688 | |
df40c01a | 689 | struct super_block *user_get_super(dev_t dev) |
1da177e4 | 690 | { |
618f0636 | 691 | struct super_block *sb; |
1da177e4 | 692 | |
1da177e4 | 693 | spin_lock(&sb_lock); |
618f0636 KK |
694 | rescan: |
695 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 696 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 697 | continue; |
618f0636 KK |
698 | if (sb->s_dev == dev) { |
699 | sb->s_count++; | |
1da177e4 | 700 | spin_unlock(&sb_lock); |
618f0636 | 701 | down_read(&sb->s_umount); |
df40c01a | 702 | /* still alive? */ |
dabe0dc1 | 703 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
618f0636 KK |
704 | return sb; |
705 | up_read(&sb->s_umount); | |
df40c01a | 706 | /* nope, got unmounted */ |
618f0636 | 707 | spin_lock(&sb_lock); |
df40c01a AV |
708 | __put_super(sb); |
709 | goto rescan; | |
1da177e4 LT |
710 | } |
711 | } | |
712 | spin_unlock(&sb_lock); | |
713 | return NULL; | |
714 | } | |
715 | ||
1da177e4 LT |
716 | /** |
717 | * do_remount_sb - asks filesystem to change mount options. | |
718 | * @sb: superblock in question | |
719 | * @flags: numeric part of options | |
720 | * @data: the rest of options | |
721 | * @force: whether or not to force the change | |
722 | * | |
723 | * Alters the mount options of a mounted file system. | |
724 | */ | |
725 | int do_remount_sb(struct super_block *sb, int flags, void *data, int force) | |
726 | { | |
727 | int retval; | |
c79d967d | 728 | int remount_ro; |
4504230a CH |
729 | |
730 | if (sb->s_frozen != SB_UNFROZEN) | |
731 | return -EBUSY; | |
732 | ||
9361401e | 733 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
734 | if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev)) |
735 | return -EACCES; | |
9361401e | 736 | #endif |
4504230a | 737 | |
1da177e4 LT |
738 | if (flags & MS_RDONLY) |
739 | acct_auto_close(sb); | |
740 | shrink_dcache_sb(sb); | |
60b0680f | 741 | sync_filesystem(sb); |
1da177e4 | 742 | |
d208bbdd | 743 | remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY); |
d208bbdd | 744 | |
1da177e4 LT |
745 | /* If we are remounting RDONLY and current sb is read/write, |
746 | make sure there are no rw files opened */ | |
d208bbdd | 747 | if (remount_ro) { |
4ed5e82f | 748 | if (force) { |
1da177e4 | 749 | mark_files_ro(sb); |
4ed5e82f MS |
750 | } else { |
751 | retval = sb_prepare_remount_readonly(sb); | |
752 | if (retval) | |
753 | return retval; | |
4ed5e82f | 754 | } |
1da177e4 LT |
755 | } |
756 | ||
757 | if (sb->s_op->remount_fs) { | |
1da177e4 | 758 | retval = sb->s_op->remount_fs(sb, &flags, data); |
2833eb2b MS |
759 | if (retval) { |
760 | if (!force) | |
4ed5e82f | 761 | goto cancel_readonly; |
2833eb2b MS |
762 | /* If forced remount, go ahead despite any errors */ |
763 | WARN(1, "forced remount of a %s fs returned %i\n", | |
764 | sb->s_type->name, retval); | |
765 | } | |
1da177e4 LT |
766 | } |
767 | sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK); | |
4ed5e82f MS |
768 | /* Needs to be ordered wrt mnt_is_readonly() */ |
769 | smp_wmb(); | |
770 | sb->s_readonly_remount = 0; | |
c79d967d | 771 | |
d208bbdd NP |
772 | /* |
773 | * Some filesystems modify their metadata via some other path than the | |
774 | * bdev buffer cache (eg. use a private mapping, or directories in | |
775 | * pagecache, etc). Also file data modifications go via their own | |
776 | * mappings. So If we try to mount readonly then copy the filesystem | |
777 | * from bdev, we could get stale data, so invalidate it to give a best | |
778 | * effort at coherency. | |
779 | */ | |
780 | if (remount_ro && sb->s_bdev) | |
781 | invalidate_bdev(sb->s_bdev); | |
1da177e4 | 782 | return 0; |
4ed5e82f MS |
783 | |
784 | cancel_readonly: | |
785 | sb->s_readonly_remount = 0; | |
786 | return retval; | |
1da177e4 LT |
787 | } |
788 | ||
a2a9537a | 789 | static void do_emergency_remount(struct work_struct *work) |
1da177e4 | 790 | { |
dca33252 | 791 | struct super_block *sb, *p = NULL; |
1da177e4 LT |
792 | |
793 | spin_lock(&sb_lock); | |
dca33252 | 794 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 795 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 796 | continue; |
1da177e4 LT |
797 | sb->s_count++; |
798 | spin_unlock(&sb_lock); | |
443b94ba | 799 | down_write(&sb->s_umount); |
dabe0dc1 AV |
800 | if (sb->s_root && sb->s_bdev && (sb->s_flags & MS_BORN) && |
801 | !(sb->s_flags & MS_RDONLY)) { | |
1da177e4 | 802 | /* |
1da177e4 LT |
803 | * What lock protects sb->s_flags?? |
804 | */ | |
1da177e4 | 805 | do_remount_sb(sb, MS_RDONLY, NULL, 1); |
1da177e4 | 806 | } |
443b94ba | 807 | up_write(&sb->s_umount); |
1da177e4 | 808 | spin_lock(&sb_lock); |
dca33252 AV |
809 | if (p) |
810 | __put_super(p); | |
811 | p = sb; | |
1da177e4 | 812 | } |
dca33252 AV |
813 | if (p) |
814 | __put_super(p); | |
1da177e4 | 815 | spin_unlock(&sb_lock); |
a2a9537a | 816 | kfree(work); |
1da177e4 LT |
817 | printk("Emergency Remount complete\n"); |
818 | } | |
819 | ||
820 | void emergency_remount(void) | |
821 | { | |
a2a9537a JA |
822 | struct work_struct *work; |
823 | ||
824 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | |
825 | if (work) { | |
826 | INIT_WORK(work, do_emergency_remount); | |
827 | schedule_work(work); | |
828 | } | |
1da177e4 LT |
829 | } |
830 | ||
831 | /* | |
832 | * Unnamed block devices are dummy devices used by virtual | |
833 | * filesystems which don't use real block-devices. -- jrs | |
834 | */ | |
835 | ||
ad76cbc6 | 836 | static DEFINE_IDA(unnamed_dev_ida); |
1da177e4 | 837 | static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */ |
c63e09ec | 838 | static int unnamed_dev_start = 0; /* don't bother trying below it */ |
1da177e4 | 839 | |
0ee5dc67 | 840 | int get_anon_bdev(dev_t *p) |
1da177e4 LT |
841 | { |
842 | int dev; | |
843 | int error; | |
844 | ||
845 | retry: | |
ad76cbc6 | 846 | if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0) |
1da177e4 LT |
847 | return -ENOMEM; |
848 | spin_lock(&unnamed_dev_lock); | |
c63e09ec | 849 | error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev); |
f21f6220 AV |
850 | if (!error) |
851 | unnamed_dev_start = dev + 1; | |
1da177e4 LT |
852 | spin_unlock(&unnamed_dev_lock); |
853 | if (error == -EAGAIN) | |
854 | /* We raced and lost with another CPU. */ | |
855 | goto retry; | |
856 | else if (error) | |
857 | return -EAGAIN; | |
858 | ||
859 | if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) { | |
860 | spin_lock(&unnamed_dev_lock); | |
ad76cbc6 | 861 | ida_remove(&unnamed_dev_ida, dev); |
f21f6220 AV |
862 | if (unnamed_dev_start > dev) |
863 | unnamed_dev_start = dev; | |
1da177e4 LT |
864 | spin_unlock(&unnamed_dev_lock); |
865 | return -EMFILE; | |
866 | } | |
0ee5dc67 | 867 | *p = MKDEV(0, dev & MINORMASK); |
1da177e4 LT |
868 | return 0; |
869 | } | |
0ee5dc67 | 870 | EXPORT_SYMBOL(get_anon_bdev); |
1da177e4 | 871 | |
0ee5dc67 | 872 | void free_anon_bdev(dev_t dev) |
1da177e4 | 873 | { |
0ee5dc67 | 874 | int slot = MINOR(dev); |
1da177e4 | 875 | spin_lock(&unnamed_dev_lock); |
ad76cbc6 | 876 | ida_remove(&unnamed_dev_ida, slot); |
c63e09ec AV |
877 | if (slot < unnamed_dev_start) |
878 | unnamed_dev_start = slot; | |
1da177e4 LT |
879 | spin_unlock(&unnamed_dev_lock); |
880 | } | |
0ee5dc67 AV |
881 | EXPORT_SYMBOL(free_anon_bdev); |
882 | ||
883 | int set_anon_super(struct super_block *s, void *data) | |
884 | { | |
885 | int error = get_anon_bdev(&s->s_dev); | |
886 | if (!error) | |
887 | s->s_bdi = &noop_backing_dev_info; | |
888 | return error; | |
889 | } | |
890 | ||
891 | EXPORT_SYMBOL(set_anon_super); | |
892 | ||
893 | void kill_anon_super(struct super_block *sb) | |
894 | { | |
895 | dev_t dev = sb->s_dev; | |
896 | generic_shutdown_super(sb); | |
897 | free_anon_bdev(dev); | |
898 | } | |
1da177e4 LT |
899 | |
900 | EXPORT_SYMBOL(kill_anon_super); | |
901 | ||
1da177e4 LT |
902 | void kill_litter_super(struct super_block *sb) |
903 | { | |
904 | if (sb->s_root) | |
905 | d_genocide(sb->s_root); | |
906 | kill_anon_super(sb); | |
907 | } | |
908 | ||
909 | EXPORT_SYMBOL(kill_litter_super); | |
910 | ||
909e6d94 SH |
911 | static int ns_test_super(struct super_block *sb, void *data) |
912 | { | |
913 | return sb->s_fs_info == data; | |
914 | } | |
915 | ||
916 | static int ns_set_super(struct super_block *sb, void *data) | |
917 | { | |
918 | sb->s_fs_info = data; | |
919 | return set_anon_super(sb, NULL); | |
920 | } | |
921 | ||
ceefda69 AV |
922 | struct dentry *mount_ns(struct file_system_type *fs_type, int flags, |
923 | void *data, int (*fill_super)(struct super_block *, void *, int)) | |
909e6d94 SH |
924 | { |
925 | struct super_block *sb; | |
926 | ||
927 | sb = sget(fs_type, ns_test_super, ns_set_super, data); | |
928 | if (IS_ERR(sb)) | |
ceefda69 | 929 | return ERR_CAST(sb); |
909e6d94 SH |
930 | |
931 | if (!sb->s_root) { | |
932 | int err; | |
933 | sb->s_flags = flags; | |
934 | err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0); | |
935 | if (err) { | |
74dbbdd7 | 936 | deactivate_locked_super(sb); |
ceefda69 | 937 | return ERR_PTR(err); |
909e6d94 SH |
938 | } |
939 | ||
940 | sb->s_flags |= MS_ACTIVE; | |
941 | } | |
942 | ||
ceefda69 | 943 | return dget(sb->s_root); |
909e6d94 SH |
944 | } |
945 | ||
ceefda69 | 946 | EXPORT_SYMBOL(mount_ns); |
909e6d94 | 947 | |
9361401e | 948 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
949 | static int set_bdev_super(struct super_block *s, void *data) |
950 | { | |
951 | s->s_bdev = data; | |
952 | s->s_dev = s->s_bdev->bd_dev; | |
32a88aa1 JA |
953 | |
954 | /* | |
955 | * We set the bdi here to the queue backing, file systems can | |
956 | * overwrite this in ->fill_super() | |
957 | */ | |
958 | s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info; | |
1da177e4 LT |
959 | return 0; |
960 | } | |
961 | ||
962 | static int test_bdev_super(struct super_block *s, void *data) | |
963 | { | |
964 | return (void *)s->s_bdev == data; | |
965 | } | |
966 | ||
152a0836 | 967 | struct dentry *mount_bdev(struct file_system_type *fs_type, |
1da177e4 | 968 | int flags, const char *dev_name, void *data, |
152a0836 | 969 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
970 | { |
971 | struct block_device *bdev; | |
972 | struct super_block *s; | |
d4d77629 | 973 | fmode_t mode = FMODE_READ | FMODE_EXCL; |
1da177e4 LT |
974 | int error = 0; |
975 | ||
30c40d2c AV |
976 | if (!(flags & MS_RDONLY)) |
977 | mode |= FMODE_WRITE; | |
978 | ||
d4d77629 | 979 | bdev = blkdev_get_by_path(dev_name, mode, fs_type); |
1da177e4 | 980 | if (IS_ERR(bdev)) |
152a0836 | 981 | return ERR_CAST(bdev); |
1da177e4 LT |
982 | |
983 | /* | |
984 | * once the super is inserted into the list by sget, s_umount | |
985 | * will protect the lockfs code from trying to start a snapshot | |
986 | * while we are mounting | |
987 | */ | |
4fadd7bb CH |
988 | mutex_lock(&bdev->bd_fsfreeze_mutex); |
989 | if (bdev->bd_fsfreeze_count > 0) { | |
990 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
991 | error = -EBUSY; | |
992 | goto error_bdev; | |
993 | } | |
1da177e4 | 994 | s = sget(fs_type, test_bdev_super, set_bdev_super, bdev); |
4fadd7bb | 995 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
1da177e4 | 996 | if (IS_ERR(s)) |
454e2398 | 997 | goto error_s; |
1da177e4 LT |
998 | |
999 | if (s->s_root) { | |
1000 | if ((flags ^ s->s_flags) & MS_RDONLY) { | |
74dbbdd7 | 1001 | deactivate_locked_super(s); |
454e2398 DH |
1002 | error = -EBUSY; |
1003 | goto error_bdev; | |
1da177e4 | 1004 | } |
454e2398 | 1005 | |
4f331f01 TH |
1006 | /* |
1007 | * s_umount nests inside bd_mutex during | |
e525fd89 TH |
1008 | * __invalidate_device(). blkdev_put() acquires |
1009 | * bd_mutex and can't be called under s_umount. Drop | |
1010 | * s_umount temporarily. This is safe as we're | |
1011 | * holding an active reference. | |
4f331f01 TH |
1012 | */ |
1013 | up_write(&s->s_umount); | |
d4d77629 | 1014 | blkdev_put(bdev, mode); |
4f331f01 | 1015 | down_write(&s->s_umount); |
1da177e4 LT |
1016 | } else { |
1017 | char b[BDEVNAME_SIZE]; | |
1018 | ||
9e1f1de0 | 1019 | s->s_flags = flags | MS_NOSEC; |
30c40d2c | 1020 | s->s_mode = mode; |
1da177e4 | 1021 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); |
e78c9a00 | 1022 | sb_set_blocksize(s, block_size(bdev)); |
9b04c997 | 1023 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1024 | if (error) { |
74dbbdd7 | 1025 | deactivate_locked_super(s); |
454e2398 | 1026 | goto error; |
fa675765 | 1027 | } |
454e2398 DH |
1028 | |
1029 | s->s_flags |= MS_ACTIVE; | |
87d8fe1e | 1030 | bdev->bd_super = s; |
1da177e4 LT |
1031 | } |
1032 | ||
152a0836 | 1033 | return dget(s->s_root); |
1da177e4 | 1034 | |
454e2398 DH |
1035 | error_s: |
1036 | error = PTR_ERR(s); | |
1037 | error_bdev: | |
d4d77629 | 1038 | blkdev_put(bdev, mode); |
454e2398 | 1039 | error: |
152a0836 AV |
1040 | return ERR_PTR(error); |
1041 | } | |
1042 | EXPORT_SYMBOL(mount_bdev); | |
1043 | ||
1da177e4 LT |
1044 | void kill_block_super(struct super_block *sb) |
1045 | { | |
1046 | struct block_device *bdev = sb->s_bdev; | |
30c40d2c | 1047 | fmode_t mode = sb->s_mode; |
1da177e4 | 1048 | |
ddbaaf30 | 1049 | bdev->bd_super = NULL; |
1da177e4 LT |
1050 | generic_shutdown_super(sb); |
1051 | sync_blockdev(bdev); | |
d4d77629 | 1052 | WARN_ON_ONCE(!(mode & FMODE_EXCL)); |
e525fd89 | 1053 | blkdev_put(bdev, mode | FMODE_EXCL); |
1da177e4 LT |
1054 | } |
1055 | ||
1056 | EXPORT_SYMBOL(kill_block_super); | |
9361401e | 1057 | #endif |
1da177e4 | 1058 | |
3c26ff6e | 1059 | struct dentry *mount_nodev(struct file_system_type *fs_type, |
1da177e4 | 1060 | int flags, void *data, |
3c26ff6e | 1061 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1062 | { |
1063 | int error; | |
1064 | struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL); | |
1065 | ||
1066 | if (IS_ERR(s)) | |
3c26ff6e | 1067 | return ERR_CAST(s); |
1da177e4 LT |
1068 | |
1069 | s->s_flags = flags; | |
1070 | ||
9b04c997 | 1071 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1072 | if (error) { |
74dbbdd7 | 1073 | deactivate_locked_super(s); |
3c26ff6e | 1074 | return ERR_PTR(error); |
1da177e4 LT |
1075 | } |
1076 | s->s_flags |= MS_ACTIVE; | |
3c26ff6e | 1077 | return dget(s->s_root); |
1da177e4 | 1078 | } |
3c26ff6e AV |
1079 | EXPORT_SYMBOL(mount_nodev); |
1080 | ||
1da177e4 LT |
1081 | static int compare_single(struct super_block *s, void *p) |
1082 | { | |
1083 | return 1; | |
1084 | } | |
1085 | ||
fc14f2fe | 1086 | struct dentry *mount_single(struct file_system_type *fs_type, |
1da177e4 | 1087 | int flags, void *data, |
fc14f2fe | 1088 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1089 | { |
1090 | struct super_block *s; | |
1091 | int error; | |
1092 | ||
1093 | s = sget(fs_type, compare_single, set_anon_super, NULL); | |
1094 | if (IS_ERR(s)) | |
fc14f2fe | 1095 | return ERR_CAST(s); |
1da177e4 LT |
1096 | if (!s->s_root) { |
1097 | s->s_flags = flags; | |
9b04c997 | 1098 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1099 | if (error) { |
74dbbdd7 | 1100 | deactivate_locked_super(s); |
fc14f2fe | 1101 | return ERR_PTR(error); |
1da177e4 LT |
1102 | } |
1103 | s->s_flags |= MS_ACTIVE; | |
9329d1be KS |
1104 | } else { |
1105 | do_remount_sb(s, flags, data, 0); | |
1da177e4 | 1106 | } |
fc14f2fe AV |
1107 | return dget(s->s_root); |
1108 | } | |
1109 | EXPORT_SYMBOL(mount_single); | |
1110 | ||
9d412a43 AV |
1111 | struct dentry * |
1112 | mount_fs(struct file_system_type *type, int flags, const char *name, void *data) | |
1da177e4 | 1113 | { |
c96e41e9 | 1114 | struct dentry *root; |
9d412a43 | 1115 | struct super_block *sb; |
1da177e4 | 1116 | char *secdata = NULL; |
9d412a43 | 1117 | int error = -ENOMEM; |
8089352a | 1118 | |
e0007529 | 1119 | if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) { |
1da177e4 | 1120 | secdata = alloc_secdata(); |
454e2398 | 1121 | if (!secdata) |
9d412a43 | 1122 | goto out; |
1da177e4 | 1123 | |
e0007529 | 1124 | error = security_sb_copy_data(data, secdata); |
454e2398 | 1125 | if (error) |
1da177e4 | 1126 | goto out_free_secdata; |
1da177e4 LT |
1127 | } |
1128 | ||
1a102ff9 AV |
1129 | root = type->mount(type, flags, name, data); |
1130 | if (IS_ERR(root)) { | |
1131 | error = PTR_ERR(root); | |
1132 | goto out_free_secdata; | |
c96e41e9 | 1133 | } |
9d412a43 AV |
1134 | sb = root->d_sb; |
1135 | BUG_ON(!sb); | |
1136 | WARN_ON(!sb->s_bdi); | |
6c510389 | 1137 | WARN_ON(sb->s_bdi == &default_backing_dev_info); |
9d412a43 | 1138 | sb->s_flags |= MS_BORN; |
454e2398 | 1139 | |
9d412a43 | 1140 | error = security_sb_kern_mount(sb, flags, secdata); |
5129a469 JE |
1141 | if (error) |
1142 | goto out_sb; | |
454e2398 | 1143 | |
42cb56ae JL |
1144 | /* |
1145 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE | |
1146 | * but s_maxbytes was an unsigned long long for many releases. Throw | |
1147 | * this warning for a little while to try and catch filesystems that | |
4358b567 | 1148 | * violate this rule. |
42cb56ae | 1149 | */ |
9d412a43 AV |
1150 | WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " |
1151 | "negative value (%lld)\n", type->name, sb->s_maxbytes); | |
42cb56ae | 1152 | |
9d412a43 | 1153 | up_write(&sb->s_umount); |
8680e22f | 1154 | free_secdata(secdata); |
9d412a43 | 1155 | return root; |
1da177e4 | 1156 | out_sb: |
9d412a43 AV |
1157 | dput(root); |
1158 | deactivate_locked_super(sb); | |
1da177e4 LT |
1159 | out_free_secdata: |
1160 | free_secdata(secdata); | |
1da177e4 | 1161 | out: |
454e2398 | 1162 | return ERR_PTR(error); |
1da177e4 LT |
1163 | } |
1164 | ||
18e9e510 | 1165 | /** |
7000d3c4 RD |
1166 | * freeze_super - lock the filesystem and force it into a consistent state |
1167 | * @sb: the super to lock | |
18e9e510 JB |
1168 | * |
1169 | * Syncs the super to make sure the filesystem is consistent and calls the fs's | |
1170 | * freeze_fs. Subsequent calls to this without first thawing the fs will return | |
1171 | * -EBUSY. | |
1172 | */ | |
1173 | int freeze_super(struct super_block *sb) | |
1174 | { | |
1175 | int ret; | |
1176 | ||
1177 | atomic_inc(&sb->s_active); | |
1178 | down_write(&sb->s_umount); | |
1179 | if (sb->s_frozen) { | |
1180 | deactivate_locked_super(sb); | |
1181 | return -EBUSY; | |
1182 | } | |
1183 | ||
dabe0dc1 AV |
1184 | if (!(sb->s_flags & MS_BORN)) { |
1185 | up_write(&sb->s_umount); | |
1186 | return 0; /* sic - it's "nothing to do" */ | |
1187 | } | |
1188 | ||
18e9e510 JB |
1189 | if (sb->s_flags & MS_RDONLY) { |
1190 | sb->s_frozen = SB_FREEZE_TRANS; | |
1191 | smp_wmb(); | |
1192 | up_write(&sb->s_umount); | |
1193 | return 0; | |
1194 | } | |
1195 | ||
1196 | sb->s_frozen = SB_FREEZE_WRITE; | |
1197 | smp_wmb(); | |
1198 | ||
1199 | sync_filesystem(sb); | |
1200 | ||
1201 | sb->s_frozen = SB_FREEZE_TRANS; | |
1202 | smp_wmb(); | |
1203 | ||
1204 | sync_blockdev(sb->s_bdev); | |
1205 | if (sb->s_op->freeze_fs) { | |
1206 | ret = sb->s_op->freeze_fs(sb); | |
1207 | if (ret) { | |
1208 | printk(KERN_ERR | |
1209 | "VFS:Filesystem freeze failed\n"); | |
1210 | sb->s_frozen = SB_UNFROZEN; | |
e1616300 KM |
1211 | smp_wmb(); |
1212 | wake_up(&sb->s_wait_unfrozen); | |
18e9e510 JB |
1213 | deactivate_locked_super(sb); |
1214 | return ret; | |
1215 | } | |
1216 | } | |
1217 | up_write(&sb->s_umount); | |
1218 | return 0; | |
1219 | } | |
1220 | EXPORT_SYMBOL(freeze_super); | |
1221 | ||
1222 | /** | |
1223 | * thaw_super -- unlock filesystem | |
1224 | * @sb: the super to thaw | |
1225 | * | |
1226 | * Unlocks the filesystem and marks it writeable again after freeze_super(). | |
1227 | */ | |
1228 | int thaw_super(struct super_block *sb) | |
1229 | { | |
1230 | int error; | |
1231 | ||
1232 | down_write(&sb->s_umount); | |
1233 | if (sb->s_frozen == SB_UNFROZEN) { | |
1234 | up_write(&sb->s_umount); | |
1235 | return -EINVAL; | |
1236 | } | |
1237 | ||
1238 | if (sb->s_flags & MS_RDONLY) | |
1239 | goto out; | |
1240 | ||
1241 | if (sb->s_op->unfreeze_fs) { | |
1242 | error = sb->s_op->unfreeze_fs(sb); | |
1243 | if (error) { | |
1244 | printk(KERN_ERR | |
1245 | "VFS:Filesystem thaw failed\n"); | |
1246 | sb->s_frozen = SB_FREEZE_TRANS; | |
1247 | up_write(&sb->s_umount); | |
1248 | return error; | |
1249 | } | |
1250 | } | |
1251 | ||
1252 | out: | |
1253 | sb->s_frozen = SB_UNFROZEN; | |
1254 | smp_wmb(); | |
1255 | wake_up(&sb->s_wait_unfrozen); | |
1256 | deactivate_locked_super(sb); | |
1257 | ||
1258 | return 0; | |
1259 | } | |
1260 | EXPORT_SYMBOL(thaw_super); |