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 | ||
630d9c47 | 23 | #include <linux/export.h> |
1da177e4 | 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> |
40401530 | 35 | #include <linux/fsnotify.h> |
5accdf82 | 36 | #include <linux/lockdep.h> |
6d59e7f5 | 37 | #include "internal.h" |
1da177e4 LT |
38 | |
39 | ||
1da177e4 LT |
40 | LIST_HEAD(super_blocks); |
41 | DEFINE_SPINLOCK(sb_lock); | |
42 | ||
5accdf82 JK |
43 | static char *sb_writers_name[SB_FREEZE_LEVELS] = { |
44 | "sb_writers", | |
45 | "sb_pagefaults", | |
46 | "sb_internal", | |
47 | }; | |
48 | ||
b0d40c92 DC |
49 | /* |
50 | * One thing we have to be careful of with a per-sb shrinker is that we don't | |
51 | * drop the last active reference to the superblock from within the shrinker. | |
52 | * If that happens we could trigger unregistering the shrinker from within the | |
53 | * shrinker path and that leads to deadlock on the shrinker_rwsem. Hence we | |
54 | * take a passive reference to the superblock to avoid this from occurring. | |
55 | */ | |
0a234c6d DC |
56 | static unsigned long super_cache_scan(struct shrinker *shrink, |
57 | struct shrink_control *sc) | |
b0d40c92 DC |
58 | { |
59 | struct super_block *sb; | |
0a234c6d DC |
60 | long fs_objects = 0; |
61 | long total_objects; | |
62 | long freed = 0; | |
63 | long dentries; | |
64 | long inodes; | |
b0d40c92 DC |
65 | |
66 | sb = container_of(shrink, struct super_block, s_shrink); | |
67 | ||
68 | /* | |
69 | * Deadlock avoidance. We may hold various FS locks, and we don't want | |
70 | * to recurse into the FS that called us in clear_inode() and friends.. | |
71 | */ | |
0a234c6d DC |
72 | if (!(sc->gfp_mask & __GFP_FS)) |
73 | return SHRINK_STOP; | |
b0d40c92 DC |
74 | |
75 | if (!grab_super_passive(sb)) | |
0a234c6d | 76 | return SHRINK_STOP; |
b0d40c92 | 77 | |
d0407903 | 78 | if (sb->s_op->nr_cached_objects) |
9b17c623 | 79 | fs_objects = sb->s_op->nr_cached_objects(sb, sc->nid); |
0e1fdafd | 80 | |
9b17c623 DC |
81 | inodes = list_lru_count_node(&sb->s_inode_lru, sc->nid); |
82 | dentries = list_lru_count_node(&sb->s_dentry_lru, sc->nid); | |
f6041567 | 83 | total_objects = dentries + inodes + fs_objects + 1; |
0e1fdafd | 84 | |
0a234c6d | 85 | /* proportion the scan between the caches */ |
f6041567 | 86 | dentries = mult_frac(sc->nr_to_scan, dentries, total_objects); |
bc3b14cb | 87 | inodes = mult_frac(sc->nr_to_scan, inodes, total_objects); |
b0d40c92 | 88 | |
0a234c6d DC |
89 | /* |
90 | * prune the dcache first as the icache is pinned by it, then | |
91 | * prune the icache, followed by the filesystem specific caches | |
92 | */ | |
9b17c623 DC |
93 | freed = prune_dcache_sb(sb, dentries, sc->nid); |
94 | freed += prune_icache_sb(sb, inodes, sc->nid); | |
0a234c6d DC |
95 | |
96 | if (fs_objects) { | |
97 | fs_objects = mult_frac(sc->nr_to_scan, fs_objects, | |
98 | total_objects); | |
9b17c623 DC |
99 | freed += sb->s_op->free_cached_objects(sb, fs_objects, |
100 | sc->nid); | |
b0d40c92 DC |
101 | } |
102 | ||
0a234c6d DC |
103 | drop_super(sb); |
104 | return freed; | |
105 | } | |
106 | ||
107 | static unsigned long super_cache_count(struct shrinker *shrink, | |
108 | struct shrink_control *sc) | |
109 | { | |
110 | struct super_block *sb; | |
111 | long total_objects = 0; | |
112 | ||
113 | sb = container_of(shrink, struct super_block, s_shrink); | |
114 | ||
115 | if (!grab_super_passive(sb)) | |
116 | return 0; | |
117 | ||
118 | if (sb->s_op && sb->s_op->nr_cached_objects) | |
9b17c623 DC |
119 | total_objects = sb->s_op->nr_cached_objects(sb, |
120 | sc->nid); | |
0a234c6d | 121 | |
9b17c623 DC |
122 | total_objects += list_lru_count_node(&sb->s_dentry_lru, |
123 | sc->nid); | |
124 | total_objects += list_lru_count_node(&sb->s_inode_lru, | |
125 | sc->nid); | |
0a234c6d | 126 | |
55f841ce | 127 | total_objects = vfs_pressure_ratio(total_objects); |
b0d40c92 | 128 | drop_super(sb); |
0e1fdafd | 129 | return total_objects; |
b0d40c92 DC |
130 | } |
131 | ||
5accdf82 JK |
132 | static int init_sb_writers(struct super_block *s, struct file_system_type *type) |
133 | { | |
134 | int err; | |
135 | int i; | |
136 | ||
137 | for (i = 0; i < SB_FREEZE_LEVELS; i++) { | |
138 | err = percpu_counter_init(&s->s_writers.counter[i], 0); | |
139 | if (err < 0) | |
140 | goto err_out; | |
141 | lockdep_init_map(&s->s_writers.lock_map[i], sb_writers_name[i], | |
142 | &type->s_writers_key[i], 0); | |
143 | } | |
144 | init_waitqueue_head(&s->s_writers.wait); | |
145 | init_waitqueue_head(&s->s_writers.wait_unfrozen); | |
146 | return 0; | |
147 | err_out: | |
148 | while (--i >= 0) | |
149 | percpu_counter_destroy(&s->s_writers.counter[i]); | |
150 | return err; | |
151 | } | |
152 | ||
153 | static void destroy_sb_writers(struct super_block *s) | |
154 | { | |
155 | int i; | |
156 | ||
157 | for (i = 0; i < SB_FREEZE_LEVELS; i++) | |
158 | percpu_counter_destroy(&s->s_writers.counter[i]); | |
159 | } | |
160 | ||
1da177e4 LT |
161 | /** |
162 | * alloc_super - create new superblock | |
fe2bbc48 | 163 | * @type: filesystem type superblock should belong to |
9249e17f | 164 | * @flags: the mount flags |
1da177e4 LT |
165 | * |
166 | * Allocates and initializes a new &struct super_block. alloc_super() | |
167 | * returns a pointer new superblock or %NULL if allocation had failed. | |
168 | */ | |
9249e17f | 169 | static struct super_block *alloc_super(struct file_system_type *type, int flags) |
1da177e4 | 170 | { |
11b0b5ab | 171 | struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER); |
b87221de | 172 | static const struct super_operations default_op; |
1da177e4 LT |
173 | |
174 | if (s) { | |
7b7a8665 CH |
175 | if (security_sb_alloc(s)) |
176 | goto out_free_sb; | |
177 | ||
6416ccb7 NP |
178 | #ifdef CONFIG_SMP |
179 | s->s_files = alloc_percpu(struct list_head); | |
5accdf82 JK |
180 | if (!s->s_files) |
181 | goto err_out; | |
182 | else { | |
6416ccb7 NP |
183 | int i; |
184 | ||
185 | for_each_possible_cpu(i) | |
186 | INIT_LIST_HEAD(per_cpu_ptr(s->s_files, i)); | |
187 | } | |
188 | #else | |
1da177e4 | 189 | INIT_LIST_HEAD(&s->s_files); |
6416ccb7 | 190 | #endif |
5accdf82 JK |
191 | if (init_sb_writers(s, type)) |
192 | goto err_out; | |
9249e17f | 193 | s->s_flags = flags; |
95f28604 | 194 | s->s_bdi = &default_backing_dev_info; |
a5166169 | 195 | INIT_HLIST_NODE(&s->s_instances); |
ceb5bdc2 | 196 | INIT_HLIST_BL_HEAD(&s->s_anon); |
1da177e4 | 197 | INIT_LIST_HEAD(&s->s_inodes); |
5ca302c8 GC |
198 | |
199 | if (list_lru_init(&s->s_dentry_lru)) | |
200 | goto err_out; | |
201 | if (list_lru_init(&s->s_inode_lru)) | |
202 | goto err_out_dentry_lru; | |
203 | ||
39f7c4db | 204 | INIT_LIST_HEAD(&s->s_mounts); |
1da177e4 | 205 | init_rwsem(&s->s_umount); |
897c6ff9 | 206 | lockdep_set_class(&s->s_umount, &type->s_umount_key); |
ada723dc PZ |
207 | /* |
208 | * sget() can have s_umount recursion. | |
209 | * | |
210 | * When it cannot find a suitable sb, it allocates a new | |
211 | * one (this one), and tries again to find a suitable old | |
212 | * one. | |
213 | * | |
214 | * In case that succeeds, it will acquire the s_umount | |
215 | * lock of the old one. Since these are clearly distrinct | |
216 | * locks, and this object isn't exposed yet, there's no | |
217 | * risk of deadlocks. | |
218 | * | |
219 | * Annotate this by putting this lock in a different | |
220 | * subclass. | |
221 | */ | |
222 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); | |
b20bd1a5 | 223 | s->s_count = 1; |
1da177e4 | 224 | atomic_set(&s->s_active, 1); |
a11f3a05 | 225 | mutex_init(&s->s_vfs_rename_mutex); |
51ee049e | 226 | lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key); |
d3be915f IM |
227 | mutex_init(&s->s_dquot.dqio_mutex); |
228 | mutex_init(&s->s_dquot.dqonoff_mutex); | |
1da177e4 | 229 | init_rwsem(&s->s_dquot.dqptr_sem); |
1da177e4 | 230 | s->s_maxbytes = MAX_NON_LFS; |
1da177e4 LT |
231 | s->s_op = &default_op; |
232 | s->s_time_gran = 1000000000; | |
c515e1fd | 233 | s->cleancache_poolid = -1; |
b0d40c92 DC |
234 | |
235 | s->s_shrink.seeks = DEFAULT_SEEKS; | |
0a234c6d DC |
236 | s->s_shrink.scan_objects = super_cache_scan; |
237 | s->s_shrink.count_objects = super_cache_count; | |
8ab47664 | 238 | s->s_shrink.batch = 1024; |
9b17c623 | 239 | s->s_shrink.flags = SHRINKER_NUMA_AWARE; |
1da177e4 LT |
240 | } |
241 | out: | |
242 | return s; | |
5ca302c8 GC |
243 | |
244 | err_out_dentry_lru: | |
245 | list_lru_destroy(&s->s_dentry_lru); | |
5accdf82 JK |
246 | err_out: |
247 | security_sb_free(s); | |
248 | #ifdef CONFIG_SMP | |
249 | if (s->s_files) | |
250 | free_percpu(s->s_files); | |
251 | #endif | |
252 | destroy_sb_writers(s); | |
7b7a8665 | 253 | out_free_sb: |
5accdf82 JK |
254 | kfree(s); |
255 | s = NULL; | |
256 | goto out; | |
1da177e4 LT |
257 | } |
258 | ||
259 | /** | |
260 | * destroy_super - frees a superblock | |
261 | * @s: superblock to free | |
262 | * | |
263 | * Frees a superblock. | |
264 | */ | |
265 | static inline void destroy_super(struct super_block *s) | |
266 | { | |
6416ccb7 NP |
267 | #ifdef CONFIG_SMP |
268 | free_percpu(s->s_files); | |
269 | #endif | |
5accdf82 | 270 | destroy_sb_writers(s); |
1da177e4 | 271 | security_sb_free(s); |
39f7c4db | 272 | WARN_ON(!list_empty(&s->s_mounts)); |
79c0b2df | 273 | kfree(s->s_subtype); |
b3b304a2 | 274 | kfree(s->s_options); |
1da177e4 LT |
275 | kfree(s); |
276 | } | |
277 | ||
278 | /* Superblock refcounting */ | |
279 | ||
280 | /* | |
35cf7ba0 | 281 | * Drop a superblock's refcount. The caller must hold sb_lock. |
1da177e4 | 282 | */ |
f47ec3f2 | 283 | static void __put_super(struct super_block *sb) |
1da177e4 | 284 | { |
1da177e4 | 285 | if (!--sb->s_count) { |
551de6f3 | 286 | list_del_init(&sb->s_list); |
1da177e4 | 287 | destroy_super(sb); |
1da177e4 | 288 | } |
1da177e4 LT |
289 | } |
290 | ||
291 | /** | |
292 | * put_super - drop a temporary reference to superblock | |
293 | * @sb: superblock in question | |
294 | * | |
295 | * Drops a temporary reference, frees superblock if there's no | |
296 | * references left. | |
297 | */ | |
f47ec3f2 | 298 | static void put_super(struct super_block *sb) |
1da177e4 LT |
299 | { |
300 | spin_lock(&sb_lock); | |
301 | __put_super(sb); | |
302 | spin_unlock(&sb_lock); | |
303 | } | |
304 | ||
305 | ||
306 | /** | |
1712ac8f | 307 | * deactivate_locked_super - drop an active reference to superblock |
1da177e4 LT |
308 | * @s: superblock to deactivate |
309 | * | |
1712ac8f AV |
310 | * Drops an active reference to superblock, converting it into a temprory |
311 | * one if there is no other active references left. In that case we | |
1da177e4 LT |
312 | * tell fs driver to shut it down and drop the temporary reference we |
313 | * had just acquired. | |
1712ac8f AV |
314 | * |
315 | * Caller holds exclusive lock on superblock; that lock is released. | |
1da177e4 | 316 | */ |
1712ac8f | 317 | void deactivate_locked_super(struct super_block *s) |
1da177e4 LT |
318 | { |
319 | struct file_system_type *fs = s->s_type; | |
b20bd1a5 | 320 | if (atomic_dec_and_test(&s->s_active)) { |
3167760f | 321 | cleancache_invalidate_fs(s); |
1da177e4 | 322 | fs->kill_sb(s); |
b0d40c92 DC |
323 | |
324 | /* caches are now gone, we can safely kill the shrinker now */ | |
325 | unregister_shrinker(&s->s_shrink); | |
1da177e4 LT |
326 | put_filesystem(fs); |
327 | put_super(s); | |
1712ac8f AV |
328 | } else { |
329 | up_write(&s->s_umount); | |
1da177e4 LT |
330 | } |
331 | } | |
332 | ||
1712ac8f | 333 | EXPORT_SYMBOL(deactivate_locked_super); |
1da177e4 | 334 | |
74dbbdd7 | 335 | /** |
1712ac8f | 336 | * deactivate_super - drop an active reference to superblock |
74dbbdd7 AV |
337 | * @s: superblock to deactivate |
338 | * | |
1712ac8f AV |
339 | * Variant of deactivate_locked_super(), except that superblock is *not* |
340 | * locked by caller. If we are going to drop the final active reference, | |
341 | * lock will be acquired prior to that. | |
74dbbdd7 | 342 | */ |
1712ac8f | 343 | void deactivate_super(struct super_block *s) |
74dbbdd7 | 344 | { |
1712ac8f AV |
345 | if (!atomic_add_unless(&s->s_active, -1, 1)) { |
346 | down_write(&s->s_umount); | |
347 | deactivate_locked_super(s); | |
74dbbdd7 AV |
348 | } |
349 | } | |
350 | ||
1712ac8f | 351 | EXPORT_SYMBOL(deactivate_super); |
74dbbdd7 | 352 | |
1da177e4 LT |
353 | /** |
354 | * grab_super - acquire an active reference | |
355 | * @s: reference we are trying to make active | |
356 | * | |
357 | * Tries to acquire an active reference. grab_super() is used when we | |
358 | * had just found a superblock in super_blocks or fs_type->fs_supers | |
359 | * and want to turn it into a full-blown active reference. grab_super() | |
360 | * is called with sb_lock held and drops it. Returns 1 in case of | |
361 | * success, 0 if we had failed (superblock contents was already dead or | |
acfec9a5 AV |
362 | * dying when grab_super() had been called). Note that this is only |
363 | * called for superblocks not in rundown mode (== ones still on ->fs_supers | |
364 | * of their type), so increment of ->s_count is OK here. | |
1da177e4 | 365 | */ |
9c4dbee7 | 366 | static int grab_super(struct super_block *s) __releases(sb_lock) |
1da177e4 LT |
367 | { |
368 | s->s_count++; | |
369 | spin_unlock(&sb_lock); | |
370 | down_write(&s->s_umount); | |
acfec9a5 AV |
371 | if ((s->s_flags & MS_BORN) && atomic_inc_not_zero(&s->s_active)) { |
372 | put_super(s); | |
373 | return 1; | |
374 | } | |
1da177e4 LT |
375 | up_write(&s->s_umount); |
376 | put_super(s); | |
1da177e4 LT |
377 | return 0; |
378 | } | |
379 | ||
12ad3ab6 DC |
380 | /* |
381 | * grab_super_passive - acquire a passive reference | |
331cbdee | 382 | * @sb: reference we are trying to grab |
12ad3ab6 DC |
383 | * |
384 | * Tries to acquire a passive reference. This is used in places where we | |
385 | * cannot take an active reference but we need to ensure that the | |
386 | * superblock does not go away while we are working on it. It returns | |
387 | * false if a reference was not gained, and returns true with the s_umount | |
388 | * lock held in read mode if a reference is gained. On successful return, | |
389 | * the caller must drop the s_umount lock and the passive reference when | |
390 | * done. | |
391 | */ | |
392 | bool grab_super_passive(struct super_block *sb) | |
393 | { | |
394 | spin_lock(&sb_lock); | |
a5166169 | 395 | if (hlist_unhashed(&sb->s_instances)) { |
12ad3ab6 DC |
396 | spin_unlock(&sb_lock); |
397 | return false; | |
398 | } | |
399 | ||
400 | sb->s_count++; | |
401 | spin_unlock(&sb_lock); | |
402 | ||
403 | if (down_read_trylock(&sb->s_umount)) { | |
dabe0dc1 | 404 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
12ad3ab6 DC |
405 | return true; |
406 | up_read(&sb->s_umount); | |
407 | } | |
408 | ||
409 | put_super(sb); | |
410 | return false; | |
411 | } | |
412 | ||
1da177e4 LT |
413 | /** |
414 | * generic_shutdown_super - common helper for ->kill_sb() | |
415 | * @sb: superblock to kill | |
416 | * | |
417 | * generic_shutdown_super() does all fs-independent work on superblock | |
418 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects | |
419 | * that need destruction out of superblock, call generic_shutdown_super() | |
420 | * and release aforementioned objects. Note: dentries and inodes _are_ | |
421 | * taken care of and do not need specific handling. | |
c636ebdb DH |
422 | * |
423 | * Upon calling this function, the filesystem may no longer alter or | |
424 | * rearrange the set of dentries belonging to this super_block, nor may it | |
425 | * change the attachments of dentries to inodes. | |
1da177e4 LT |
426 | */ |
427 | void generic_shutdown_super(struct super_block *sb) | |
428 | { | |
ee9b6d61 | 429 | const struct super_operations *sop = sb->s_op; |
1da177e4 | 430 | |
c636ebdb DH |
431 | if (sb->s_root) { |
432 | shrink_dcache_for_umount(sb); | |
60b0680f | 433 | sync_filesystem(sb); |
1da177e4 | 434 | sb->s_flags &= ~MS_ACTIVE; |
efaee192 | 435 | |
63997e98 AV |
436 | fsnotify_unmount_inodes(&sb->s_inodes); |
437 | ||
438 | evict_inodes(sb); | |
1da177e4 | 439 | |
7b7a8665 CH |
440 | if (sb->s_dio_done_wq) { |
441 | destroy_workqueue(sb->s_dio_done_wq); | |
442 | sb->s_dio_done_wq = NULL; | |
443 | } | |
444 | ||
1da177e4 LT |
445 | if (sop->put_super) |
446 | sop->put_super(sb); | |
447 | ||
63997e98 | 448 | if (!list_empty(&sb->s_inodes)) { |
7b4fe29e DJ |
449 | printk("VFS: Busy inodes after unmount of %s. " |
450 | "Self-destruct in 5 seconds. Have a nice day...\n", | |
451 | sb->s_id); | |
1da177e4 | 452 | } |
1da177e4 LT |
453 | } |
454 | spin_lock(&sb_lock); | |
455 | /* should be initialized for __put_super_and_need_restart() */ | |
a5166169 | 456 | hlist_del_init(&sb->s_instances); |
1da177e4 LT |
457 | spin_unlock(&sb_lock); |
458 | up_write(&sb->s_umount); | |
459 | } | |
460 | ||
461 | EXPORT_SYMBOL(generic_shutdown_super); | |
462 | ||
463 | /** | |
464 | * sget - find or create a superblock | |
465 | * @type: filesystem type superblock should belong to | |
466 | * @test: comparison callback | |
467 | * @set: setup callback | |
9249e17f | 468 | * @flags: mount flags |
1da177e4 LT |
469 | * @data: argument to each of them |
470 | */ | |
471 | struct super_block *sget(struct file_system_type *type, | |
472 | int (*test)(struct super_block *,void *), | |
473 | int (*set)(struct super_block *,void *), | |
9249e17f | 474 | int flags, |
1da177e4 LT |
475 | void *data) |
476 | { | |
477 | struct super_block *s = NULL; | |
d4730127 | 478 | struct super_block *old; |
1da177e4 LT |
479 | int err; |
480 | ||
481 | retry: | |
482 | spin_lock(&sb_lock); | |
d4730127 | 483 | if (test) { |
b67bfe0d | 484 | hlist_for_each_entry(old, &type->fs_supers, s_instances) { |
d4730127 MK |
485 | if (!test(old, data)) |
486 | continue; | |
487 | if (!grab_super(old)) | |
488 | goto retry; | |
a3cfbb53 LZ |
489 | if (s) { |
490 | up_write(&s->s_umount); | |
d4730127 | 491 | destroy_super(s); |
7a4dec53 | 492 | s = NULL; |
a3cfbb53 | 493 | } |
d4730127 MK |
494 | return old; |
495 | } | |
1da177e4 LT |
496 | } |
497 | if (!s) { | |
498 | spin_unlock(&sb_lock); | |
9249e17f | 499 | s = alloc_super(type, flags); |
1da177e4 LT |
500 | if (!s) |
501 | return ERR_PTR(-ENOMEM); | |
502 | goto retry; | |
503 | } | |
504 | ||
505 | err = set(s, data); | |
506 | if (err) { | |
507 | spin_unlock(&sb_lock); | |
a3cfbb53 | 508 | up_write(&s->s_umount); |
1da177e4 LT |
509 | destroy_super(s); |
510 | return ERR_PTR(err); | |
511 | } | |
512 | s->s_type = type; | |
513 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); | |
514 | list_add_tail(&s->s_list, &super_blocks); | |
a5166169 | 515 | hlist_add_head(&s->s_instances, &type->fs_supers); |
1da177e4 LT |
516 | spin_unlock(&sb_lock); |
517 | get_filesystem(type); | |
b0d40c92 | 518 | register_shrinker(&s->s_shrink); |
1da177e4 LT |
519 | return s; |
520 | } | |
521 | ||
522 | EXPORT_SYMBOL(sget); | |
523 | ||
524 | void drop_super(struct super_block *sb) | |
525 | { | |
526 | up_read(&sb->s_umount); | |
527 | put_super(sb); | |
528 | } | |
529 | ||
530 | EXPORT_SYMBOL(drop_super); | |
531 | ||
01a05b33 AV |
532 | /** |
533 | * iterate_supers - call function for all active superblocks | |
534 | * @f: function to call | |
535 | * @arg: argument to pass to it | |
536 | * | |
537 | * Scans the superblock list and calls given function, passing it | |
538 | * locked superblock and given argument. | |
539 | */ | |
540 | void iterate_supers(void (*f)(struct super_block *, void *), void *arg) | |
541 | { | |
dca33252 | 542 | struct super_block *sb, *p = NULL; |
01a05b33 AV |
543 | |
544 | spin_lock(&sb_lock); | |
dca33252 | 545 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 546 | if (hlist_unhashed(&sb->s_instances)) |
01a05b33 AV |
547 | continue; |
548 | sb->s_count++; | |
549 | spin_unlock(&sb_lock); | |
550 | ||
551 | down_read(&sb->s_umount); | |
dabe0dc1 | 552 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
01a05b33 AV |
553 | f(sb, arg); |
554 | up_read(&sb->s_umount); | |
555 | ||
556 | spin_lock(&sb_lock); | |
dca33252 AV |
557 | if (p) |
558 | __put_super(p); | |
559 | p = sb; | |
01a05b33 | 560 | } |
dca33252 AV |
561 | if (p) |
562 | __put_super(p); | |
01a05b33 AV |
563 | spin_unlock(&sb_lock); |
564 | } | |
565 | ||
43e15cdb AV |
566 | /** |
567 | * iterate_supers_type - call function for superblocks of given type | |
568 | * @type: fs type | |
569 | * @f: function to call | |
570 | * @arg: argument to pass to it | |
571 | * | |
572 | * Scans the superblock list and calls given function, passing it | |
573 | * locked superblock and given argument. | |
574 | */ | |
575 | void iterate_supers_type(struct file_system_type *type, | |
576 | void (*f)(struct super_block *, void *), void *arg) | |
577 | { | |
578 | struct super_block *sb, *p = NULL; | |
579 | ||
580 | spin_lock(&sb_lock); | |
b67bfe0d | 581 | hlist_for_each_entry(sb, &type->fs_supers, s_instances) { |
43e15cdb AV |
582 | sb->s_count++; |
583 | spin_unlock(&sb_lock); | |
584 | ||
585 | down_read(&sb->s_umount); | |
dabe0dc1 | 586 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
43e15cdb AV |
587 | f(sb, arg); |
588 | up_read(&sb->s_umount); | |
589 | ||
590 | spin_lock(&sb_lock); | |
591 | if (p) | |
592 | __put_super(p); | |
593 | p = sb; | |
594 | } | |
595 | if (p) | |
596 | __put_super(p); | |
597 | spin_unlock(&sb_lock); | |
598 | } | |
599 | ||
600 | EXPORT_SYMBOL(iterate_supers_type); | |
601 | ||
1da177e4 LT |
602 | /** |
603 | * get_super - get the superblock of a device | |
604 | * @bdev: device to get the superblock for | |
605 | * | |
606 | * Scans the superblock list and finds the superblock of the file system | |
607 | * mounted on the device given. %NULL is returned if no match is found. | |
608 | */ | |
609 | ||
df40c01a | 610 | struct super_block *get_super(struct block_device *bdev) |
1da177e4 | 611 | { |
618f0636 KK |
612 | struct super_block *sb; |
613 | ||
1da177e4 LT |
614 | if (!bdev) |
615 | return NULL; | |
618f0636 | 616 | |
1da177e4 | 617 | spin_lock(&sb_lock); |
618f0636 KK |
618 | rescan: |
619 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 620 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 621 | continue; |
618f0636 KK |
622 | if (sb->s_bdev == bdev) { |
623 | sb->s_count++; | |
1da177e4 | 624 | spin_unlock(&sb_lock); |
618f0636 | 625 | down_read(&sb->s_umount); |
df40c01a | 626 | /* still alive? */ |
dabe0dc1 | 627 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
618f0636 KK |
628 | return sb; |
629 | up_read(&sb->s_umount); | |
df40c01a | 630 | /* nope, got unmounted */ |
618f0636 | 631 | spin_lock(&sb_lock); |
df40c01a AV |
632 | __put_super(sb); |
633 | goto rescan; | |
1da177e4 LT |
634 | } |
635 | } | |
636 | spin_unlock(&sb_lock); | |
637 | return NULL; | |
638 | } | |
639 | ||
640 | EXPORT_SYMBOL(get_super); | |
4504230a | 641 | |
6b6dc836 JK |
642 | /** |
643 | * get_super_thawed - get thawed superblock of a device | |
644 | * @bdev: device to get the superblock for | |
645 | * | |
646 | * Scans the superblock list and finds the superblock of the file system | |
647 | * mounted on the device. The superblock is returned once it is thawed | |
648 | * (or immediately if it was not frozen). %NULL is returned if no match | |
649 | * is found. | |
650 | */ | |
651 | struct super_block *get_super_thawed(struct block_device *bdev) | |
652 | { | |
653 | while (1) { | |
654 | struct super_block *s = get_super(bdev); | |
5accdf82 | 655 | if (!s || s->s_writers.frozen == SB_UNFROZEN) |
6b6dc836 JK |
656 | return s; |
657 | up_read(&s->s_umount); | |
5accdf82 JK |
658 | wait_event(s->s_writers.wait_unfrozen, |
659 | s->s_writers.frozen == SB_UNFROZEN); | |
6b6dc836 JK |
660 | put_super(s); |
661 | } | |
662 | } | |
663 | EXPORT_SYMBOL(get_super_thawed); | |
664 | ||
4504230a CH |
665 | /** |
666 | * get_active_super - get an active reference to the superblock of a device | |
667 | * @bdev: device to get the superblock for | |
668 | * | |
669 | * Scans the superblock list and finds the superblock of the file system | |
670 | * mounted on the device given. Returns the superblock with an active | |
d3f21473 | 671 | * reference or %NULL if none was found. |
4504230a CH |
672 | */ |
673 | struct super_block *get_active_super(struct block_device *bdev) | |
674 | { | |
675 | struct super_block *sb; | |
676 | ||
677 | if (!bdev) | |
678 | return NULL; | |
679 | ||
1494583d | 680 | restart: |
4504230a CH |
681 | spin_lock(&sb_lock); |
682 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 683 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 684 | continue; |
1494583d | 685 | if (sb->s_bdev == bdev) { |
acfec9a5 | 686 | if (!grab_super(sb)) |
1494583d | 687 | goto restart; |
acfec9a5 AV |
688 | up_write(&sb->s_umount); |
689 | return sb; | |
1494583d | 690 | } |
4504230a CH |
691 | } |
692 | spin_unlock(&sb_lock); | |
693 | return NULL; | |
694 | } | |
1da177e4 | 695 | |
df40c01a | 696 | struct super_block *user_get_super(dev_t dev) |
1da177e4 | 697 | { |
618f0636 | 698 | struct super_block *sb; |
1da177e4 | 699 | |
1da177e4 | 700 | spin_lock(&sb_lock); |
618f0636 KK |
701 | rescan: |
702 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 703 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 704 | continue; |
618f0636 KK |
705 | if (sb->s_dev == dev) { |
706 | sb->s_count++; | |
1da177e4 | 707 | spin_unlock(&sb_lock); |
618f0636 | 708 | down_read(&sb->s_umount); |
df40c01a | 709 | /* still alive? */ |
dabe0dc1 | 710 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
618f0636 KK |
711 | return sb; |
712 | up_read(&sb->s_umount); | |
df40c01a | 713 | /* nope, got unmounted */ |
618f0636 | 714 | spin_lock(&sb_lock); |
df40c01a AV |
715 | __put_super(sb); |
716 | goto rescan; | |
1da177e4 LT |
717 | } |
718 | } | |
719 | spin_unlock(&sb_lock); | |
720 | return NULL; | |
721 | } | |
722 | ||
1da177e4 LT |
723 | /** |
724 | * do_remount_sb - asks filesystem to change mount options. | |
725 | * @sb: superblock in question | |
726 | * @flags: numeric part of options | |
727 | * @data: the rest of options | |
728 | * @force: whether or not to force the change | |
729 | * | |
730 | * Alters the mount options of a mounted file system. | |
731 | */ | |
732 | int do_remount_sb(struct super_block *sb, int flags, void *data, int force) | |
733 | { | |
734 | int retval; | |
c79d967d | 735 | int remount_ro; |
4504230a | 736 | |
5accdf82 | 737 | if (sb->s_writers.frozen != SB_UNFROZEN) |
4504230a CH |
738 | return -EBUSY; |
739 | ||
9361401e | 740 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
741 | if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev)) |
742 | return -EACCES; | |
9361401e | 743 | #endif |
4504230a | 744 | |
1da177e4 LT |
745 | if (flags & MS_RDONLY) |
746 | acct_auto_close(sb); | |
747 | shrink_dcache_sb(sb); | |
60b0680f | 748 | sync_filesystem(sb); |
1da177e4 | 749 | |
d208bbdd | 750 | remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY); |
d208bbdd | 751 | |
1da177e4 LT |
752 | /* If we are remounting RDONLY and current sb is read/write, |
753 | make sure there are no rw files opened */ | |
d208bbdd | 754 | if (remount_ro) { |
4ed5e82f | 755 | if (force) { |
1da177e4 | 756 | mark_files_ro(sb); |
4ed5e82f MS |
757 | } else { |
758 | retval = sb_prepare_remount_readonly(sb); | |
759 | if (retval) | |
760 | return retval; | |
4ed5e82f | 761 | } |
1da177e4 LT |
762 | } |
763 | ||
764 | if (sb->s_op->remount_fs) { | |
1da177e4 | 765 | retval = sb->s_op->remount_fs(sb, &flags, data); |
2833eb2b MS |
766 | if (retval) { |
767 | if (!force) | |
4ed5e82f | 768 | goto cancel_readonly; |
2833eb2b MS |
769 | /* If forced remount, go ahead despite any errors */ |
770 | WARN(1, "forced remount of a %s fs returned %i\n", | |
771 | sb->s_type->name, retval); | |
772 | } | |
1da177e4 LT |
773 | } |
774 | sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK); | |
4ed5e82f MS |
775 | /* Needs to be ordered wrt mnt_is_readonly() */ |
776 | smp_wmb(); | |
777 | sb->s_readonly_remount = 0; | |
c79d967d | 778 | |
d208bbdd NP |
779 | /* |
780 | * Some filesystems modify their metadata via some other path than the | |
781 | * bdev buffer cache (eg. use a private mapping, or directories in | |
782 | * pagecache, etc). Also file data modifications go via their own | |
783 | * mappings. So If we try to mount readonly then copy the filesystem | |
784 | * from bdev, we could get stale data, so invalidate it to give a best | |
785 | * effort at coherency. | |
786 | */ | |
787 | if (remount_ro && sb->s_bdev) | |
788 | invalidate_bdev(sb->s_bdev); | |
1da177e4 | 789 | return 0; |
4ed5e82f MS |
790 | |
791 | cancel_readonly: | |
792 | sb->s_readonly_remount = 0; | |
793 | return retval; | |
1da177e4 LT |
794 | } |
795 | ||
a2a9537a | 796 | static void do_emergency_remount(struct work_struct *work) |
1da177e4 | 797 | { |
dca33252 | 798 | struct super_block *sb, *p = NULL; |
1da177e4 LT |
799 | |
800 | spin_lock(&sb_lock); | |
dca33252 | 801 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 802 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 803 | continue; |
1da177e4 LT |
804 | sb->s_count++; |
805 | spin_unlock(&sb_lock); | |
443b94ba | 806 | down_write(&sb->s_umount); |
dabe0dc1 AV |
807 | if (sb->s_root && sb->s_bdev && (sb->s_flags & MS_BORN) && |
808 | !(sb->s_flags & MS_RDONLY)) { | |
1da177e4 | 809 | /* |
1da177e4 LT |
810 | * What lock protects sb->s_flags?? |
811 | */ | |
1da177e4 | 812 | do_remount_sb(sb, MS_RDONLY, NULL, 1); |
1da177e4 | 813 | } |
443b94ba | 814 | up_write(&sb->s_umount); |
1da177e4 | 815 | spin_lock(&sb_lock); |
dca33252 AV |
816 | if (p) |
817 | __put_super(p); | |
818 | p = sb; | |
1da177e4 | 819 | } |
dca33252 AV |
820 | if (p) |
821 | __put_super(p); | |
1da177e4 | 822 | spin_unlock(&sb_lock); |
a2a9537a | 823 | kfree(work); |
1da177e4 LT |
824 | printk("Emergency Remount complete\n"); |
825 | } | |
826 | ||
827 | void emergency_remount(void) | |
828 | { | |
a2a9537a JA |
829 | struct work_struct *work; |
830 | ||
831 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | |
832 | if (work) { | |
833 | INIT_WORK(work, do_emergency_remount); | |
834 | schedule_work(work); | |
835 | } | |
1da177e4 LT |
836 | } |
837 | ||
838 | /* | |
839 | * Unnamed block devices are dummy devices used by virtual | |
840 | * filesystems which don't use real block-devices. -- jrs | |
841 | */ | |
842 | ||
ad76cbc6 | 843 | static DEFINE_IDA(unnamed_dev_ida); |
1da177e4 | 844 | static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */ |
c63e09ec | 845 | static int unnamed_dev_start = 0; /* don't bother trying below it */ |
1da177e4 | 846 | |
0ee5dc67 | 847 | int get_anon_bdev(dev_t *p) |
1da177e4 LT |
848 | { |
849 | int dev; | |
850 | int error; | |
851 | ||
852 | retry: | |
ad76cbc6 | 853 | if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0) |
1da177e4 LT |
854 | return -ENOMEM; |
855 | spin_lock(&unnamed_dev_lock); | |
c63e09ec | 856 | error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev); |
f21f6220 AV |
857 | if (!error) |
858 | unnamed_dev_start = dev + 1; | |
1da177e4 LT |
859 | spin_unlock(&unnamed_dev_lock); |
860 | if (error == -EAGAIN) | |
861 | /* We raced and lost with another CPU. */ | |
862 | goto retry; | |
863 | else if (error) | |
864 | return -EAGAIN; | |
865 | ||
e8c8d1bc | 866 | if (dev == (1 << MINORBITS)) { |
1da177e4 | 867 | spin_lock(&unnamed_dev_lock); |
ad76cbc6 | 868 | ida_remove(&unnamed_dev_ida, dev); |
f21f6220 AV |
869 | if (unnamed_dev_start > dev) |
870 | unnamed_dev_start = dev; | |
1da177e4 LT |
871 | spin_unlock(&unnamed_dev_lock); |
872 | return -EMFILE; | |
873 | } | |
0ee5dc67 | 874 | *p = MKDEV(0, dev & MINORMASK); |
1da177e4 LT |
875 | return 0; |
876 | } | |
0ee5dc67 | 877 | EXPORT_SYMBOL(get_anon_bdev); |
1da177e4 | 878 | |
0ee5dc67 | 879 | void free_anon_bdev(dev_t dev) |
1da177e4 | 880 | { |
0ee5dc67 | 881 | int slot = MINOR(dev); |
1da177e4 | 882 | spin_lock(&unnamed_dev_lock); |
ad76cbc6 | 883 | ida_remove(&unnamed_dev_ida, slot); |
c63e09ec AV |
884 | if (slot < unnamed_dev_start) |
885 | unnamed_dev_start = slot; | |
1da177e4 LT |
886 | spin_unlock(&unnamed_dev_lock); |
887 | } | |
0ee5dc67 AV |
888 | EXPORT_SYMBOL(free_anon_bdev); |
889 | ||
890 | int set_anon_super(struct super_block *s, void *data) | |
891 | { | |
892 | int error = get_anon_bdev(&s->s_dev); | |
893 | if (!error) | |
894 | s->s_bdi = &noop_backing_dev_info; | |
895 | return error; | |
896 | } | |
897 | ||
898 | EXPORT_SYMBOL(set_anon_super); | |
899 | ||
900 | void kill_anon_super(struct super_block *sb) | |
901 | { | |
902 | dev_t dev = sb->s_dev; | |
903 | generic_shutdown_super(sb); | |
904 | free_anon_bdev(dev); | |
905 | } | |
1da177e4 LT |
906 | |
907 | EXPORT_SYMBOL(kill_anon_super); | |
908 | ||
1da177e4 LT |
909 | void kill_litter_super(struct super_block *sb) |
910 | { | |
911 | if (sb->s_root) | |
912 | d_genocide(sb->s_root); | |
913 | kill_anon_super(sb); | |
914 | } | |
915 | ||
916 | EXPORT_SYMBOL(kill_litter_super); | |
917 | ||
909e6d94 SH |
918 | static int ns_test_super(struct super_block *sb, void *data) |
919 | { | |
920 | return sb->s_fs_info == data; | |
921 | } | |
922 | ||
923 | static int ns_set_super(struct super_block *sb, void *data) | |
924 | { | |
925 | sb->s_fs_info = data; | |
926 | return set_anon_super(sb, NULL); | |
927 | } | |
928 | ||
ceefda69 AV |
929 | struct dentry *mount_ns(struct file_system_type *fs_type, int flags, |
930 | void *data, int (*fill_super)(struct super_block *, void *, int)) | |
909e6d94 SH |
931 | { |
932 | struct super_block *sb; | |
933 | ||
9249e17f | 934 | sb = sget(fs_type, ns_test_super, ns_set_super, flags, data); |
909e6d94 | 935 | if (IS_ERR(sb)) |
ceefda69 | 936 | return ERR_CAST(sb); |
909e6d94 SH |
937 | |
938 | if (!sb->s_root) { | |
939 | int err; | |
909e6d94 SH |
940 | err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0); |
941 | if (err) { | |
74dbbdd7 | 942 | deactivate_locked_super(sb); |
ceefda69 | 943 | return ERR_PTR(err); |
909e6d94 SH |
944 | } |
945 | ||
946 | sb->s_flags |= MS_ACTIVE; | |
947 | } | |
948 | ||
ceefda69 | 949 | return dget(sb->s_root); |
909e6d94 SH |
950 | } |
951 | ||
ceefda69 | 952 | EXPORT_SYMBOL(mount_ns); |
909e6d94 | 953 | |
9361401e | 954 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
955 | static int set_bdev_super(struct super_block *s, void *data) |
956 | { | |
957 | s->s_bdev = data; | |
958 | s->s_dev = s->s_bdev->bd_dev; | |
32a88aa1 JA |
959 | |
960 | /* | |
961 | * We set the bdi here to the queue backing, file systems can | |
962 | * overwrite this in ->fill_super() | |
963 | */ | |
964 | s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info; | |
1da177e4 LT |
965 | return 0; |
966 | } | |
967 | ||
968 | static int test_bdev_super(struct super_block *s, void *data) | |
969 | { | |
970 | return (void *)s->s_bdev == data; | |
971 | } | |
972 | ||
152a0836 | 973 | struct dentry *mount_bdev(struct file_system_type *fs_type, |
1da177e4 | 974 | int flags, const char *dev_name, void *data, |
152a0836 | 975 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
976 | { |
977 | struct block_device *bdev; | |
978 | struct super_block *s; | |
d4d77629 | 979 | fmode_t mode = FMODE_READ | FMODE_EXCL; |
1da177e4 LT |
980 | int error = 0; |
981 | ||
30c40d2c AV |
982 | if (!(flags & MS_RDONLY)) |
983 | mode |= FMODE_WRITE; | |
984 | ||
d4d77629 | 985 | bdev = blkdev_get_by_path(dev_name, mode, fs_type); |
1da177e4 | 986 | if (IS_ERR(bdev)) |
152a0836 | 987 | return ERR_CAST(bdev); |
1da177e4 LT |
988 | |
989 | /* | |
990 | * once the super is inserted into the list by sget, s_umount | |
991 | * will protect the lockfs code from trying to start a snapshot | |
992 | * while we are mounting | |
993 | */ | |
4fadd7bb CH |
994 | mutex_lock(&bdev->bd_fsfreeze_mutex); |
995 | if (bdev->bd_fsfreeze_count > 0) { | |
996 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
997 | error = -EBUSY; | |
998 | goto error_bdev; | |
999 | } | |
9249e17f DH |
1000 | s = sget(fs_type, test_bdev_super, set_bdev_super, flags | MS_NOSEC, |
1001 | bdev); | |
4fadd7bb | 1002 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
1da177e4 | 1003 | if (IS_ERR(s)) |
454e2398 | 1004 | goto error_s; |
1da177e4 LT |
1005 | |
1006 | if (s->s_root) { | |
1007 | if ((flags ^ s->s_flags) & MS_RDONLY) { | |
74dbbdd7 | 1008 | deactivate_locked_super(s); |
454e2398 DH |
1009 | error = -EBUSY; |
1010 | goto error_bdev; | |
1da177e4 | 1011 | } |
454e2398 | 1012 | |
4f331f01 TH |
1013 | /* |
1014 | * s_umount nests inside bd_mutex during | |
e525fd89 TH |
1015 | * __invalidate_device(). blkdev_put() acquires |
1016 | * bd_mutex and can't be called under s_umount. Drop | |
1017 | * s_umount temporarily. This is safe as we're | |
1018 | * holding an active reference. | |
4f331f01 TH |
1019 | */ |
1020 | up_write(&s->s_umount); | |
d4d77629 | 1021 | blkdev_put(bdev, mode); |
4f331f01 | 1022 | down_write(&s->s_umount); |
1da177e4 LT |
1023 | } else { |
1024 | char b[BDEVNAME_SIZE]; | |
1025 | ||
30c40d2c | 1026 | s->s_mode = mode; |
1da177e4 | 1027 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); |
e78c9a00 | 1028 | sb_set_blocksize(s, block_size(bdev)); |
9b04c997 | 1029 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1030 | if (error) { |
74dbbdd7 | 1031 | deactivate_locked_super(s); |
454e2398 | 1032 | goto error; |
fa675765 | 1033 | } |
454e2398 DH |
1034 | |
1035 | s->s_flags |= MS_ACTIVE; | |
87d8fe1e | 1036 | bdev->bd_super = s; |
1da177e4 LT |
1037 | } |
1038 | ||
152a0836 | 1039 | return dget(s->s_root); |
1da177e4 | 1040 | |
454e2398 DH |
1041 | error_s: |
1042 | error = PTR_ERR(s); | |
1043 | error_bdev: | |
d4d77629 | 1044 | blkdev_put(bdev, mode); |
454e2398 | 1045 | error: |
152a0836 AV |
1046 | return ERR_PTR(error); |
1047 | } | |
1048 | EXPORT_SYMBOL(mount_bdev); | |
1049 | ||
1da177e4 LT |
1050 | void kill_block_super(struct super_block *sb) |
1051 | { | |
1052 | struct block_device *bdev = sb->s_bdev; | |
30c40d2c | 1053 | fmode_t mode = sb->s_mode; |
1da177e4 | 1054 | |
ddbaaf30 | 1055 | bdev->bd_super = NULL; |
1da177e4 LT |
1056 | generic_shutdown_super(sb); |
1057 | sync_blockdev(bdev); | |
d4d77629 | 1058 | WARN_ON_ONCE(!(mode & FMODE_EXCL)); |
e525fd89 | 1059 | blkdev_put(bdev, mode | FMODE_EXCL); |
1da177e4 LT |
1060 | } |
1061 | ||
1062 | EXPORT_SYMBOL(kill_block_super); | |
9361401e | 1063 | #endif |
1da177e4 | 1064 | |
3c26ff6e | 1065 | struct dentry *mount_nodev(struct file_system_type *fs_type, |
1da177e4 | 1066 | int flags, void *data, |
3c26ff6e | 1067 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1068 | { |
1069 | int error; | |
9249e17f | 1070 | struct super_block *s = sget(fs_type, NULL, set_anon_super, flags, NULL); |
1da177e4 LT |
1071 | |
1072 | if (IS_ERR(s)) | |
3c26ff6e | 1073 | return ERR_CAST(s); |
1da177e4 | 1074 | |
9b04c997 | 1075 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1076 | if (error) { |
74dbbdd7 | 1077 | deactivate_locked_super(s); |
3c26ff6e | 1078 | return ERR_PTR(error); |
1da177e4 LT |
1079 | } |
1080 | s->s_flags |= MS_ACTIVE; | |
3c26ff6e | 1081 | return dget(s->s_root); |
1da177e4 | 1082 | } |
3c26ff6e AV |
1083 | EXPORT_SYMBOL(mount_nodev); |
1084 | ||
1da177e4 LT |
1085 | static int compare_single(struct super_block *s, void *p) |
1086 | { | |
1087 | return 1; | |
1088 | } | |
1089 | ||
fc14f2fe | 1090 | struct dentry *mount_single(struct file_system_type *fs_type, |
1da177e4 | 1091 | int flags, void *data, |
fc14f2fe | 1092 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1093 | { |
1094 | struct super_block *s; | |
1095 | int error; | |
1096 | ||
9249e17f | 1097 | s = sget(fs_type, compare_single, set_anon_super, flags, NULL); |
1da177e4 | 1098 | if (IS_ERR(s)) |
fc14f2fe | 1099 | return ERR_CAST(s); |
1da177e4 | 1100 | if (!s->s_root) { |
9b04c997 | 1101 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1102 | if (error) { |
74dbbdd7 | 1103 | deactivate_locked_super(s); |
fc14f2fe | 1104 | return ERR_PTR(error); |
1da177e4 LT |
1105 | } |
1106 | s->s_flags |= MS_ACTIVE; | |
9329d1be KS |
1107 | } else { |
1108 | do_remount_sb(s, flags, data, 0); | |
1da177e4 | 1109 | } |
fc14f2fe AV |
1110 | return dget(s->s_root); |
1111 | } | |
1112 | EXPORT_SYMBOL(mount_single); | |
1113 | ||
9d412a43 AV |
1114 | struct dentry * |
1115 | mount_fs(struct file_system_type *type, int flags, const char *name, void *data) | |
1da177e4 | 1116 | { |
c96e41e9 | 1117 | struct dentry *root; |
9d412a43 | 1118 | struct super_block *sb; |
1da177e4 | 1119 | char *secdata = NULL; |
9d412a43 | 1120 | int error = -ENOMEM; |
8089352a | 1121 | |
e0007529 | 1122 | if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) { |
1da177e4 | 1123 | secdata = alloc_secdata(); |
454e2398 | 1124 | if (!secdata) |
9d412a43 | 1125 | goto out; |
1da177e4 | 1126 | |
e0007529 | 1127 | error = security_sb_copy_data(data, secdata); |
454e2398 | 1128 | if (error) |
1da177e4 | 1129 | goto out_free_secdata; |
1da177e4 LT |
1130 | } |
1131 | ||
1a102ff9 AV |
1132 | root = type->mount(type, flags, name, data); |
1133 | if (IS_ERR(root)) { | |
1134 | error = PTR_ERR(root); | |
1135 | goto out_free_secdata; | |
c96e41e9 | 1136 | } |
9d412a43 AV |
1137 | sb = root->d_sb; |
1138 | BUG_ON(!sb); | |
1139 | WARN_ON(!sb->s_bdi); | |
6c510389 | 1140 | WARN_ON(sb->s_bdi == &default_backing_dev_info); |
9d412a43 | 1141 | sb->s_flags |= MS_BORN; |
454e2398 | 1142 | |
9d412a43 | 1143 | error = security_sb_kern_mount(sb, flags, secdata); |
5129a469 JE |
1144 | if (error) |
1145 | goto out_sb; | |
454e2398 | 1146 | |
42cb56ae JL |
1147 | /* |
1148 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE | |
1149 | * but s_maxbytes was an unsigned long long for many releases. Throw | |
1150 | * this warning for a little while to try and catch filesystems that | |
4358b567 | 1151 | * violate this rule. |
42cb56ae | 1152 | */ |
9d412a43 AV |
1153 | WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " |
1154 | "negative value (%lld)\n", type->name, sb->s_maxbytes); | |
42cb56ae | 1155 | |
9d412a43 | 1156 | up_write(&sb->s_umount); |
8680e22f | 1157 | free_secdata(secdata); |
9d412a43 | 1158 | return root; |
1da177e4 | 1159 | out_sb: |
9d412a43 AV |
1160 | dput(root); |
1161 | deactivate_locked_super(sb); | |
1da177e4 LT |
1162 | out_free_secdata: |
1163 | free_secdata(secdata); | |
1da177e4 | 1164 | out: |
454e2398 | 1165 | return ERR_PTR(error); |
1da177e4 LT |
1166 | } |
1167 | ||
5accdf82 JK |
1168 | /* |
1169 | * This is an internal function, please use sb_end_{write,pagefault,intwrite} | |
1170 | * instead. | |
1171 | */ | |
1172 | void __sb_end_write(struct super_block *sb, int level) | |
1173 | { | |
1174 | percpu_counter_dec(&sb->s_writers.counter[level-1]); | |
1175 | /* | |
1176 | * Make sure s_writers are updated before we wake up waiters in | |
1177 | * freeze_super(). | |
1178 | */ | |
1179 | smp_mb(); | |
1180 | if (waitqueue_active(&sb->s_writers.wait)) | |
1181 | wake_up(&sb->s_writers.wait); | |
1182 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _RET_IP_); | |
1183 | } | |
1184 | EXPORT_SYMBOL(__sb_end_write); | |
1185 | ||
1186 | #ifdef CONFIG_LOCKDEP | |
1187 | /* | |
1188 | * We want lockdep to tell us about possible deadlocks with freezing but | |
1189 | * it's it bit tricky to properly instrument it. Getting a freeze protection | |
1190 | * works as getting a read lock but there are subtle problems. XFS for example | |
1191 | * gets freeze protection on internal level twice in some cases, which is OK | |
1192 | * only because we already hold a freeze protection also on higher level. Due | |
1193 | * to these cases we have to tell lockdep we are doing trylock when we | |
1194 | * already hold a freeze protection for a higher freeze level. | |
1195 | */ | |
1196 | static void acquire_freeze_lock(struct super_block *sb, int level, bool trylock, | |
1197 | unsigned long ip) | |
1198 | { | |
1199 | int i; | |
1200 | ||
1201 | if (!trylock) { | |
1202 | for (i = 0; i < level - 1; i++) | |
1203 | if (lock_is_held(&sb->s_writers.lock_map[i])) { | |
1204 | trylock = true; | |
1205 | break; | |
1206 | } | |
1207 | } | |
1208 | rwsem_acquire_read(&sb->s_writers.lock_map[level-1], 0, trylock, ip); | |
1209 | } | |
1210 | #endif | |
1211 | ||
1212 | /* | |
1213 | * This is an internal function, please use sb_start_{write,pagefault,intwrite} | |
1214 | * instead. | |
1215 | */ | |
1216 | int __sb_start_write(struct super_block *sb, int level, bool wait) | |
1217 | { | |
1218 | retry: | |
1219 | if (unlikely(sb->s_writers.frozen >= level)) { | |
1220 | if (!wait) | |
1221 | return 0; | |
1222 | wait_event(sb->s_writers.wait_unfrozen, | |
1223 | sb->s_writers.frozen < level); | |
1224 | } | |
1225 | ||
1226 | #ifdef CONFIG_LOCKDEP | |
1227 | acquire_freeze_lock(sb, level, !wait, _RET_IP_); | |
1228 | #endif | |
1229 | percpu_counter_inc(&sb->s_writers.counter[level-1]); | |
1230 | /* | |
1231 | * Make sure counter is updated before we check for frozen. | |
1232 | * freeze_super() first sets frozen and then checks the counter. | |
1233 | */ | |
1234 | smp_mb(); | |
1235 | if (unlikely(sb->s_writers.frozen >= level)) { | |
1236 | __sb_end_write(sb, level); | |
1237 | goto retry; | |
1238 | } | |
1239 | return 1; | |
1240 | } | |
1241 | EXPORT_SYMBOL(__sb_start_write); | |
1242 | ||
1243 | /** | |
1244 | * sb_wait_write - wait until all writers to given file system finish | |
1245 | * @sb: the super for which we wait | |
1246 | * @level: type of writers we wait for (normal vs page fault) | |
1247 | * | |
1248 | * This function waits until there are no writers of given type to given file | |
1249 | * system. Caller of this function should make sure there can be no new writers | |
1250 | * of type @level before calling this function. Otherwise this function can | |
1251 | * livelock. | |
1252 | */ | |
1253 | static void sb_wait_write(struct super_block *sb, int level) | |
1254 | { | |
1255 | s64 writers; | |
1256 | ||
1257 | /* | |
1258 | * We just cycle-through lockdep here so that it does not complain | |
1259 | * about returning with lock to userspace | |
1260 | */ | |
1261 | rwsem_acquire(&sb->s_writers.lock_map[level-1], 0, 0, _THIS_IP_); | |
1262 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _THIS_IP_); | |
1263 | ||
1264 | do { | |
1265 | DEFINE_WAIT(wait); | |
1266 | ||
1267 | /* | |
1268 | * We use a barrier in prepare_to_wait() to separate setting | |
1269 | * of frozen and checking of the counter | |
1270 | */ | |
1271 | prepare_to_wait(&sb->s_writers.wait, &wait, | |
1272 | TASK_UNINTERRUPTIBLE); | |
1273 | ||
1274 | writers = percpu_counter_sum(&sb->s_writers.counter[level-1]); | |
1275 | if (writers) | |
1276 | schedule(); | |
1277 | ||
1278 | finish_wait(&sb->s_writers.wait, &wait); | |
1279 | } while (writers); | |
1280 | } | |
1281 | ||
18e9e510 | 1282 | /** |
7000d3c4 RD |
1283 | * freeze_super - lock the filesystem and force it into a consistent state |
1284 | * @sb: the super to lock | |
18e9e510 JB |
1285 | * |
1286 | * Syncs the super to make sure the filesystem is consistent and calls the fs's | |
1287 | * freeze_fs. Subsequent calls to this without first thawing the fs will return | |
1288 | * -EBUSY. | |
5accdf82 JK |
1289 | * |
1290 | * During this function, sb->s_writers.frozen goes through these values: | |
1291 | * | |
1292 | * SB_UNFROZEN: File system is normal, all writes progress as usual. | |
1293 | * | |
1294 | * SB_FREEZE_WRITE: The file system is in the process of being frozen. New | |
1295 | * writes should be blocked, though page faults are still allowed. We wait for | |
1296 | * all writes to complete and then proceed to the next stage. | |
1297 | * | |
1298 | * SB_FREEZE_PAGEFAULT: Freezing continues. Now also page faults are blocked | |
1299 | * but internal fs threads can still modify the filesystem (although they | |
1300 | * should not dirty new pages or inodes), writeback can run etc. After waiting | |
1301 | * for all running page faults we sync the filesystem which will clean all | |
1302 | * dirty pages and inodes (no new dirty pages or inodes can be created when | |
1303 | * sync is running). | |
1304 | * | |
1305 | * SB_FREEZE_FS: The file system is frozen. Now all internal sources of fs | |
1306 | * modification are blocked (e.g. XFS preallocation truncation on inode | |
1307 | * reclaim). This is usually implemented by blocking new transactions for | |
1308 | * filesystems that have them and need this additional guard. After all | |
1309 | * internal writers are finished we call ->freeze_fs() to finish filesystem | |
1310 | * freezing. Then we transition to SB_FREEZE_COMPLETE state. This state is | |
1311 | * mostly auxiliary for filesystems to verify they do not modify frozen fs. | |
1312 | * | |
1313 | * sb->s_writers.frozen is protected by sb->s_umount. | |
18e9e510 JB |
1314 | */ |
1315 | int freeze_super(struct super_block *sb) | |
1316 | { | |
1317 | int ret; | |
1318 | ||
1319 | atomic_inc(&sb->s_active); | |
1320 | down_write(&sb->s_umount); | |
5accdf82 | 1321 | if (sb->s_writers.frozen != SB_UNFROZEN) { |
18e9e510 JB |
1322 | deactivate_locked_super(sb); |
1323 | return -EBUSY; | |
1324 | } | |
1325 | ||
dabe0dc1 AV |
1326 | if (!(sb->s_flags & MS_BORN)) { |
1327 | up_write(&sb->s_umount); | |
1328 | return 0; /* sic - it's "nothing to do" */ | |
1329 | } | |
1330 | ||
18e9e510 | 1331 | if (sb->s_flags & MS_RDONLY) { |
5accdf82 JK |
1332 | /* Nothing to do really... */ |
1333 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
18e9e510 JB |
1334 | up_write(&sb->s_umount); |
1335 | return 0; | |
1336 | } | |
1337 | ||
5accdf82 JK |
1338 | /* From now on, no new normal writers can start */ |
1339 | sb->s_writers.frozen = SB_FREEZE_WRITE; | |
1340 | smp_wmb(); | |
1341 | ||
1342 | /* Release s_umount to preserve sb_start_write -> s_umount ordering */ | |
1343 | up_write(&sb->s_umount); | |
1344 | ||
1345 | sb_wait_write(sb, SB_FREEZE_WRITE); | |
1346 | ||
1347 | /* Now we go and block page faults... */ | |
1348 | down_write(&sb->s_umount); | |
1349 | sb->s_writers.frozen = SB_FREEZE_PAGEFAULT; | |
18e9e510 JB |
1350 | smp_wmb(); |
1351 | ||
5accdf82 JK |
1352 | sb_wait_write(sb, SB_FREEZE_PAGEFAULT); |
1353 | ||
1354 | /* All writers are done so after syncing there won't be dirty data */ | |
18e9e510 JB |
1355 | sync_filesystem(sb); |
1356 | ||
5accdf82 JK |
1357 | /* Now wait for internal filesystem counter */ |
1358 | sb->s_writers.frozen = SB_FREEZE_FS; | |
18e9e510 | 1359 | smp_wmb(); |
5accdf82 | 1360 | sb_wait_write(sb, SB_FREEZE_FS); |
18e9e510 | 1361 | |
18e9e510 JB |
1362 | if (sb->s_op->freeze_fs) { |
1363 | ret = sb->s_op->freeze_fs(sb); | |
1364 | if (ret) { | |
1365 | printk(KERN_ERR | |
1366 | "VFS:Filesystem freeze failed\n"); | |
5accdf82 | 1367 | sb->s_writers.frozen = SB_UNFROZEN; |
e1616300 | 1368 | smp_wmb(); |
5accdf82 | 1369 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 JB |
1370 | deactivate_locked_super(sb); |
1371 | return ret; | |
1372 | } | |
1373 | } | |
5accdf82 JK |
1374 | /* |
1375 | * This is just for debugging purposes so that fs can warn if it | |
1376 | * sees write activity when frozen is set to SB_FREEZE_COMPLETE. | |
1377 | */ | |
1378 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
18e9e510 JB |
1379 | up_write(&sb->s_umount); |
1380 | return 0; | |
1381 | } | |
1382 | EXPORT_SYMBOL(freeze_super); | |
1383 | ||
1384 | /** | |
1385 | * thaw_super -- unlock filesystem | |
1386 | * @sb: the super to thaw | |
1387 | * | |
1388 | * Unlocks the filesystem and marks it writeable again after freeze_super(). | |
1389 | */ | |
1390 | int thaw_super(struct super_block *sb) | |
1391 | { | |
1392 | int error; | |
1393 | ||
1394 | down_write(&sb->s_umount); | |
5accdf82 | 1395 | if (sb->s_writers.frozen == SB_UNFROZEN) { |
18e9e510 JB |
1396 | up_write(&sb->s_umount); |
1397 | return -EINVAL; | |
1398 | } | |
1399 | ||
1400 | if (sb->s_flags & MS_RDONLY) | |
1401 | goto out; | |
1402 | ||
1403 | if (sb->s_op->unfreeze_fs) { | |
1404 | error = sb->s_op->unfreeze_fs(sb); | |
1405 | if (error) { | |
1406 | printk(KERN_ERR | |
1407 | "VFS:Filesystem thaw failed\n"); | |
18e9e510 JB |
1408 | up_write(&sb->s_umount); |
1409 | return error; | |
1410 | } | |
1411 | } | |
1412 | ||
1413 | out: | |
5accdf82 | 1414 | sb->s_writers.frozen = SB_UNFROZEN; |
18e9e510 | 1415 | smp_wmb(); |
5accdf82 | 1416 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 JB |
1417 | deactivate_locked_super(sb); |
1418 | ||
1419 | return 0; | |
1420 | } | |
1421 | EXPORT_SYMBOL(thaw_super); |