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); |
f6041567 | 198 | list_lru_init(&s->s_dentry_lru); |
bc3b14cb | 199 | list_lru_init(&s->s_inode_lru); |
39f7c4db | 200 | INIT_LIST_HEAD(&s->s_mounts); |
1da177e4 | 201 | init_rwsem(&s->s_umount); |
897c6ff9 | 202 | lockdep_set_class(&s->s_umount, &type->s_umount_key); |
ada723dc PZ |
203 | /* |
204 | * sget() can have s_umount recursion. | |
205 | * | |
206 | * When it cannot find a suitable sb, it allocates a new | |
207 | * one (this one), and tries again to find a suitable old | |
208 | * one. | |
209 | * | |
210 | * In case that succeeds, it will acquire the s_umount | |
211 | * lock of the old one. Since these are clearly distrinct | |
212 | * locks, and this object isn't exposed yet, there's no | |
213 | * risk of deadlocks. | |
214 | * | |
215 | * Annotate this by putting this lock in a different | |
216 | * subclass. | |
217 | */ | |
218 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); | |
b20bd1a5 | 219 | s->s_count = 1; |
1da177e4 | 220 | atomic_set(&s->s_active, 1); |
a11f3a05 | 221 | mutex_init(&s->s_vfs_rename_mutex); |
51ee049e | 222 | lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key); |
d3be915f IM |
223 | mutex_init(&s->s_dquot.dqio_mutex); |
224 | mutex_init(&s->s_dquot.dqonoff_mutex); | |
1da177e4 | 225 | init_rwsem(&s->s_dquot.dqptr_sem); |
1da177e4 | 226 | s->s_maxbytes = MAX_NON_LFS; |
1da177e4 LT |
227 | s->s_op = &default_op; |
228 | s->s_time_gran = 1000000000; | |
c515e1fd | 229 | s->cleancache_poolid = -1; |
b0d40c92 DC |
230 | |
231 | s->s_shrink.seeks = DEFAULT_SEEKS; | |
0a234c6d DC |
232 | s->s_shrink.scan_objects = super_cache_scan; |
233 | s->s_shrink.count_objects = super_cache_count; | |
8ab47664 | 234 | s->s_shrink.batch = 1024; |
9b17c623 | 235 | s->s_shrink.flags = SHRINKER_NUMA_AWARE; |
1da177e4 LT |
236 | } |
237 | out: | |
238 | return s; | |
5accdf82 JK |
239 | err_out: |
240 | security_sb_free(s); | |
241 | #ifdef CONFIG_SMP | |
242 | if (s->s_files) | |
243 | free_percpu(s->s_files); | |
244 | #endif | |
245 | destroy_sb_writers(s); | |
7b7a8665 | 246 | out_free_sb: |
5accdf82 JK |
247 | kfree(s); |
248 | s = NULL; | |
249 | goto out; | |
1da177e4 LT |
250 | } |
251 | ||
252 | /** | |
253 | * destroy_super - frees a superblock | |
254 | * @s: superblock to free | |
255 | * | |
256 | * Frees a superblock. | |
257 | */ | |
258 | static inline void destroy_super(struct super_block *s) | |
259 | { | |
6416ccb7 NP |
260 | #ifdef CONFIG_SMP |
261 | free_percpu(s->s_files); | |
262 | #endif | |
5accdf82 | 263 | destroy_sb_writers(s); |
1da177e4 | 264 | security_sb_free(s); |
39f7c4db | 265 | WARN_ON(!list_empty(&s->s_mounts)); |
79c0b2df | 266 | kfree(s->s_subtype); |
b3b304a2 | 267 | kfree(s->s_options); |
1da177e4 LT |
268 | kfree(s); |
269 | } | |
270 | ||
271 | /* Superblock refcounting */ | |
272 | ||
273 | /* | |
35cf7ba0 | 274 | * Drop a superblock's refcount. The caller must hold sb_lock. |
1da177e4 | 275 | */ |
f47ec3f2 | 276 | static void __put_super(struct super_block *sb) |
1da177e4 | 277 | { |
1da177e4 | 278 | if (!--sb->s_count) { |
551de6f3 | 279 | list_del_init(&sb->s_list); |
1da177e4 | 280 | destroy_super(sb); |
1da177e4 | 281 | } |
1da177e4 LT |
282 | } |
283 | ||
284 | /** | |
285 | * put_super - drop a temporary reference to superblock | |
286 | * @sb: superblock in question | |
287 | * | |
288 | * Drops a temporary reference, frees superblock if there's no | |
289 | * references left. | |
290 | */ | |
f47ec3f2 | 291 | static void put_super(struct super_block *sb) |
1da177e4 LT |
292 | { |
293 | spin_lock(&sb_lock); | |
294 | __put_super(sb); | |
295 | spin_unlock(&sb_lock); | |
296 | } | |
297 | ||
298 | ||
299 | /** | |
1712ac8f | 300 | * deactivate_locked_super - drop an active reference to superblock |
1da177e4 LT |
301 | * @s: superblock to deactivate |
302 | * | |
1712ac8f AV |
303 | * Drops an active reference to superblock, converting it into a temprory |
304 | * one if there is no other active references left. In that case we | |
1da177e4 LT |
305 | * tell fs driver to shut it down and drop the temporary reference we |
306 | * had just acquired. | |
1712ac8f AV |
307 | * |
308 | * Caller holds exclusive lock on superblock; that lock is released. | |
1da177e4 | 309 | */ |
1712ac8f | 310 | void deactivate_locked_super(struct super_block *s) |
1da177e4 LT |
311 | { |
312 | struct file_system_type *fs = s->s_type; | |
b20bd1a5 | 313 | if (atomic_dec_and_test(&s->s_active)) { |
3167760f | 314 | cleancache_invalidate_fs(s); |
1da177e4 | 315 | fs->kill_sb(s); |
b0d40c92 DC |
316 | |
317 | /* caches are now gone, we can safely kill the shrinker now */ | |
318 | unregister_shrinker(&s->s_shrink); | |
1da177e4 LT |
319 | put_filesystem(fs); |
320 | put_super(s); | |
1712ac8f AV |
321 | } else { |
322 | up_write(&s->s_umount); | |
1da177e4 LT |
323 | } |
324 | } | |
325 | ||
1712ac8f | 326 | EXPORT_SYMBOL(deactivate_locked_super); |
1da177e4 | 327 | |
74dbbdd7 | 328 | /** |
1712ac8f | 329 | * deactivate_super - drop an active reference to superblock |
74dbbdd7 AV |
330 | * @s: superblock to deactivate |
331 | * | |
1712ac8f AV |
332 | * Variant of deactivate_locked_super(), except that superblock is *not* |
333 | * locked by caller. If we are going to drop the final active reference, | |
334 | * lock will be acquired prior to that. | |
74dbbdd7 | 335 | */ |
1712ac8f | 336 | void deactivate_super(struct super_block *s) |
74dbbdd7 | 337 | { |
1712ac8f AV |
338 | if (!atomic_add_unless(&s->s_active, -1, 1)) { |
339 | down_write(&s->s_umount); | |
340 | deactivate_locked_super(s); | |
74dbbdd7 AV |
341 | } |
342 | } | |
343 | ||
1712ac8f | 344 | EXPORT_SYMBOL(deactivate_super); |
74dbbdd7 | 345 | |
1da177e4 LT |
346 | /** |
347 | * grab_super - acquire an active reference | |
348 | * @s: reference we are trying to make active | |
349 | * | |
350 | * Tries to acquire an active reference. grab_super() is used when we | |
351 | * had just found a superblock in super_blocks or fs_type->fs_supers | |
352 | * and want to turn it into a full-blown active reference. grab_super() | |
353 | * is called with sb_lock held and drops it. Returns 1 in case of | |
354 | * success, 0 if we had failed (superblock contents was already dead or | |
acfec9a5 AV |
355 | * dying when grab_super() had been called). Note that this is only |
356 | * called for superblocks not in rundown mode (== ones still on ->fs_supers | |
357 | * of their type), so increment of ->s_count is OK here. | |
1da177e4 | 358 | */ |
9c4dbee7 | 359 | static int grab_super(struct super_block *s) __releases(sb_lock) |
1da177e4 LT |
360 | { |
361 | s->s_count++; | |
362 | spin_unlock(&sb_lock); | |
363 | down_write(&s->s_umount); | |
acfec9a5 AV |
364 | if ((s->s_flags & MS_BORN) && atomic_inc_not_zero(&s->s_active)) { |
365 | put_super(s); | |
366 | return 1; | |
367 | } | |
1da177e4 LT |
368 | up_write(&s->s_umount); |
369 | put_super(s); | |
1da177e4 LT |
370 | return 0; |
371 | } | |
372 | ||
12ad3ab6 DC |
373 | /* |
374 | * grab_super_passive - acquire a passive reference | |
331cbdee | 375 | * @sb: reference we are trying to grab |
12ad3ab6 DC |
376 | * |
377 | * Tries to acquire a passive reference. This is used in places where we | |
378 | * cannot take an active reference but we need to ensure that the | |
379 | * superblock does not go away while we are working on it. It returns | |
380 | * false if a reference was not gained, and returns true with the s_umount | |
381 | * lock held in read mode if a reference is gained. On successful return, | |
382 | * the caller must drop the s_umount lock and the passive reference when | |
383 | * done. | |
384 | */ | |
385 | bool grab_super_passive(struct super_block *sb) | |
386 | { | |
387 | spin_lock(&sb_lock); | |
a5166169 | 388 | if (hlist_unhashed(&sb->s_instances)) { |
12ad3ab6 DC |
389 | spin_unlock(&sb_lock); |
390 | return false; | |
391 | } | |
392 | ||
393 | sb->s_count++; | |
394 | spin_unlock(&sb_lock); | |
395 | ||
396 | if (down_read_trylock(&sb->s_umount)) { | |
dabe0dc1 | 397 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
12ad3ab6 DC |
398 | return true; |
399 | up_read(&sb->s_umount); | |
400 | } | |
401 | ||
402 | put_super(sb); | |
403 | return false; | |
404 | } | |
405 | ||
1da177e4 LT |
406 | /** |
407 | * generic_shutdown_super - common helper for ->kill_sb() | |
408 | * @sb: superblock to kill | |
409 | * | |
410 | * generic_shutdown_super() does all fs-independent work on superblock | |
411 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects | |
412 | * that need destruction out of superblock, call generic_shutdown_super() | |
413 | * and release aforementioned objects. Note: dentries and inodes _are_ | |
414 | * taken care of and do not need specific handling. | |
c636ebdb DH |
415 | * |
416 | * Upon calling this function, the filesystem may no longer alter or | |
417 | * rearrange the set of dentries belonging to this super_block, nor may it | |
418 | * change the attachments of dentries to inodes. | |
1da177e4 LT |
419 | */ |
420 | void generic_shutdown_super(struct super_block *sb) | |
421 | { | |
ee9b6d61 | 422 | const struct super_operations *sop = sb->s_op; |
1da177e4 | 423 | |
c636ebdb DH |
424 | if (sb->s_root) { |
425 | shrink_dcache_for_umount(sb); | |
60b0680f | 426 | sync_filesystem(sb); |
1da177e4 | 427 | sb->s_flags &= ~MS_ACTIVE; |
efaee192 | 428 | |
63997e98 AV |
429 | fsnotify_unmount_inodes(&sb->s_inodes); |
430 | ||
431 | evict_inodes(sb); | |
1da177e4 | 432 | |
7b7a8665 CH |
433 | if (sb->s_dio_done_wq) { |
434 | destroy_workqueue(sb->s_dio_done_wq); | |
435 | sb->s_dio_done_wq = NULL; | |
436 | } | |
437 | ||
1da177e4 LT |
438 | if (sop->put_super) |
439 | sop->put_super(sb); | |
440 | ||
63997e98 | 441 | if (!list_empty(&sb->s_inodes)) { |
7b4fe29e DJ |
442 | printk("VFS: Busy inodes after unmount of %s. " |
443 | "Self-destruct in 5 seconds. Have a nice day...\n", | |
444 | sb->s_id); | |
1da177e4 | 445 | } |
1da177e4 LT |
446 | } |
447 | spin_lock(&sb_lock); | |
448 | /* should be initialized for __put_super_and_need_restart() */ | |
a5166169 | 449 | hlist_del_init(&sb->s_instances); |
1da177e4 LT |
450 | spin_unlock(&sb_lock); |
451 | up_write(&sb->s_umount); | |
452 | } | |
453 | ||
454 | EXPORT_SYMBOL(generic_shutdown_super); | |
455 | ||
456 | /** | |
457 | * sget - find or create a superblock | |
458 | * @type: filesystem type superblock should belong to | |
459 | * @test: comparison callback | |
460 | * @set: setup callback | |
9249e17f | 461 | * @flags: mount flags |
1da177e4 LT |
462 | * @data: argument to each of them |
463 | */ | |
464 | struct super_block *sget(struct file_system_type *type, | |
465 | int (*test)(struct super_block *,void *), | |
466 | int (*set)(struct super_block *,void *), | |
9249e17f | 467 | int flags, |
1da177e4 LT |
468 | void *data) |
469 | { | |
470 | struct super_block *s = NULL; | |
d4730127 | 471 | struct super_block *old; |
1da177e4 LT |
472 | int err; |
473 | ||
474 | retry: | |
475 | spin_lock(&sb_lock); | |
d4730127 | 476 | if (test) { |
b67bfe0d | 477 | hlist_for_each_entry(old, &type->fs_supers, s_instances) { |
d4730127 MK |
478 | if (!test(old, data)) |
479 | continue; | |
480 | if (!grab_super(old)) | |
481 | goto retry; | |
a3cfbb53 LZ |
482 | if (s) { |
483 | up_write(&s->s_umount); | |
d4730127 | 484 | destroy_super(s); |
7a4dec53 | 485 | s = NULL; |
a3cfbb53 | 486 | } |
d4730127 MK |
487 | return old; |
488 | } | |
1da177e4 LT |
489 | } |
490 | if (!s) { | |
491 | spin_unlock(&sb_lock); | |
9249e17f | 492 | s = alloc_super(type, flags); |
1da177e4 LT |
493 | if (!s) |
494 | return ERR_PTR(-ENOMEM); | |
495 | goto retry; | |
496 | } | |
497 | ||
498 | err = set(s, data); | |
499 | if (err) { | |
500 | spin_unlock(&sb_lock); | |
a3cfbb53 | 501 | up_write(&s->s_umount); |
1da177e4 LT |
502 | destroy_super(s); |
503 | return ERR_PTR(err); | |
504 | } | |
505 | s->s_type = type; | |
506 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); | |
507 | list_add_tail(&s->s_list, &super_blocks); | |
a5166169 | 508 | hlist_add_head(&s->s_instances, &type->fs_supers); |
1da177e4 LT |
509 | spin_unlock(&sb_lock); |
510 | get_filesystem(type); | |
b0d40c92 | 511 | register_shrinker(&s->s_shrink); |
1da177e4 LT |
512 | return s; |
513 | } | |
514 | ||
515 | EXPORT_SYMBOL(sget); | |
516 | ||
517 | void drop_super(struct super_block *sb) | |
518 | { | |
519 | up_read(&sb->s_umount); | |
520 | put_super(sb); | |
521 | } | |
522 | ||
523 | EXPORT_SYMBOL(drop_super); | |
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; | |
572 | ||
573 | spin_lock(&sb_lock); | |
b67bfe0d | 574 | hlist_for_each_entry(sb, &type->fs_supers, s_instances) { |
43e15cdb AV |
575 | sb->s_count++; |
576 | spin_unlock(&sb_lock); | |
577 | ||
578 | down_read(&sb->s_umount); | |
dabe0dc1 | 579 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
43e15cdb AV |
580 | f(sb, arg); |
581 | up_read(&sb->s_umount); | |
582 | ||
583 | spin_lock(&sb_lock); | |
584 | if (p) | |
585 | __put_super(p); | |
586 | p = sb; | |
587 | } | |
588 | if (p) | |
589 | __put_super(p); | |
590 | spin_unlock(&sb_lock); | |
591 | } | |
592 | ||
593 | EXPORT_SYMBOL(iterate_supers_type); | |
594 | ||
1da177e4 LT |
595 | /** |
596 | * get_super - get the superblock of a device | |
597 | * @bdev: device to get the superblock for | |
598 | * | |
599 | * Scans the superblock list and finds the superblock of the file system | |
600 | * mounted on the device given. %NULL is returned if no match is found. | |
601 | */ | |
602 | ||
df40c01a | 603 | struct super_block *get_super(struct block_device *bdev) |
1da177e4 | 604 | { |
618f0636 KK |
605 | struct super_block *sb; |
606 | ||
1da177e4 LT |
607 | if (!bdev) |
608 | return NULL; | |
618f0636 | 609 | |
1da177e4 | 610 | spin_lock(&sb_lock); |
618f0636 KK |
611 | rescan: |
612 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 613 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 614 | continue; |
618f0636 KK |
615 | if (sb->s_bdev == bdev) { |
616 | sb->s_count++; | |
1da177e4 | 617 | spin_unlock(&sb_lock); |
618f0636 | 618 | down_read(&sb->s_umount); |
df40c01a | 619 | /* still alive? */ |
dabe0dc1 | 620 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
618f0636 KK |
621 | return sb; |
622 | up_read(&sb->s_umount); | |
df40c01a | 623 | /* nope, got unmounted */ |
618f0636 | 624 | spin_lock(&sb_lock); |
df40c01a AV |
625 | __put_super(sb); |
626 | goto rescan; | |
1da177e4 LT |
627 | } |
628 | } | |
629 | spin_unlock(&sb_lock); | |
630 | return NULL; | |
631 | } | |
632 | ||
633 | EXPORT_SYMBOL(get_super); | |
4504230a | 634 | |
6b6dc836 JK |
635 | /** |
636 | * get_super_thawed - get thawed superblock of a device | |
637 | * @bdev: device to get the superblock for | |
638 | * | |
639 | * Scans the superblock list and finds the superblock of the file system | |
640 | * mounted on the device. The superblock is returned once it is thawed | |
641 | * (or immediately if it was not frozen). %NULL is returned if no match | |
642 | * is found. | |
643 | */ | |
644 | struct super_block *get_super_thawed(struct block_device *bdev) | |
645 | { | |
646 | while (1) { | |
647 | struct super_block *s = get_super(bdev); | |
5accdf82 | 648 | if (!s || s->s_writers.frozen == SB_UNFROZEN) |
6b6dc836 JK |
649 | return s; |
650 | up_read(&s->s_umount); | |
5accdf82 JK |
651 | wait_event(s->s_writers.wait_unfrozen, |
652 | s->s_writers.frozen == SB_UNFROZEN); | |
6b6dc836 JK |
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 | 678 | if (sb->s_bdev == bdev) { |
acfec9a5 | 679 | if (!grab_super(sb)) |
1494583d | 680 | goto restart; |
acfec9a5 AV |
681 | up_write(&sb->s_umount); |
682 | return sb; | |
1494583d | 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 | 729 | |
5accdf82 | 730 | if (sb->s_writers.frozen != SB_UNFROZEN) |
4504230a CH |
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 | ||
e8c8d1bc | 859 | if (dev == (1 << MINORBITS)) { |
1da177e4 | 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 | ||
9249e17f | 927 | sb = sget(fs_type, ns_test_super, ns_set_super, flags, data); |
909e6d94 | 928 | if (IS_ERR(sb)) |
ceefda69 | 929 | return ERR_CAST(sb); |
909e6d94 SH |
930 | |
931 | if (!sb->s_root) { | |
932 | int err; | |
909e6d94 SH |
933 | err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0); |
934 | if (err) { | |
74dbbdd7 | 935 | deactivate_locked_super(sb); |
ceefda69 | 936 | return ERR_PTR(err); |
909e6d94 SH |
937 | } |
938 | ||
939 | sb->s_flags |= MS_ACTIVE; | |
940 | } | |
941 | ||
ceefda69 | 942 | return dget(sb->s_root); |
909e6d94 SH |
943 | } |
944 | ||
ceefda69 | 945 | EXPORT_SYMBOL(mount_ns); |
909e6d94 | 946 | |
9361401e | 947 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
948 | static int set_bdev_super(struct super_block *s, void *data) |
949 | { | |
950 | s->s_bdev = data; | |
951 | s->s_dev = s->s_bdev->bd_dev; | |
32a88aa1 JA |
952 | |
953 | /* | |
954 | * We set the bdi here to the queue backing, file systems can | |
955 | * overwrite this in ->fill_super() | |
956 | */ | |
957 | s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info; | |
1da177e4 LT |
958 | return 0; |
959 | } | |
960 | ||
961 | static int test_bdev_super(struct super_block *s, void *data) | |
962 | { | |
963 | return (void *)s->s_bdev == data; | |
964 | } | |
965 | ||
152a0836 | 966 | struct dentry *mount_bdev(struct file_system_type *fs_type, |
1da177e4 | 967 | int flags, const char *dev_name, void *data, |
152a0836 | 968 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
969 | { |
970 | struct block_device *bdev; | |
971 | struct super_block *s; | |
d4d77629 | 972 | fmode_t mode = FMODE_READ | FMODE_EXCL; |
1da177e4 LT |
973 | int error = 0; |
974 | ||
30c40d2c AV |
975 | if (!(flags & MS_RDONLY)) |
976 | mode |= FMODE_WRITE; | |
977 | ||
d4d77629 | 978 | bdev = blkdev_get_by_path(dev_name, mode, fs_type); |
1da177e4 | 979 | if (IS_ERR(bdev)) |
152a0836 | 980 | return ERR_CAST(bdev); |
1da177e4 LT |
981 | |
982 | /* | |
983 | * once the super is inserted into the list by sget, s_umount | |
984 | * will protect the lockfs code from trying to start a snapshot | |
985 | * while we are mounting | |
986 | */ | |
4fadd7bb CH |
987 | mutex_lock(&bdev->bd_fsfreeze_mutex); |
988 | if (bdev->bd_fsfreeze_count > 0) { | |
989 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
990 | error = -EBUSY; | |
991 | goto error_bdev; | |
992 | } | |
9249e17f DH |
993 | s = sget(fs_type, test_bdev_super, set_bdev_super, flags | MS_NOSEC, |
994 | 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 | ||
30c40d2c | 1019 | s->s_mode = mode; |
1da177e4 | 1020 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); |
e78c9a00 | 1021 | sb_set_blocksize(s, block_size(bdev)); |
9b04c997 | 1022 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1023 | if (error) { |
74dbbdd7 | 1024 | deactivate_locked_super(s); |
454e2398 | 1025 | goto error; |
fa675765 | 1026 | } |
454e2398 DH |
1027 | |
1028 | s->s_flags |= MS_ACTIVE; | |
87d8fe1e | 1029 | bdev->bd_super = s; |
1da177e4 LT |
1030 | } |
1031 | ||
152a0836 | 1032 | return dget(s->s_root); |
1da177e4 | 1033 | |
454e2398 DH |
1034 | error_s: |
1035 | error = PTR_ERR(s); | |
1036 | error_bdev: | |
d4d77629 | 1037 | blkdev_put(bdev, mode); |
454e2398 | 1038 | error: |
152a0836 AV |
1039 | return ERR_PTR(error); |
1040 | } | |
1041 | EXPORT_SYMBOL(mount_bdev); | |
1042 | ||
1da177e4 LT |
1043 | void kill_block_super(struct super_block *sb) |
1044 | { | |
1045 | struct block_device *bdev = sb->s_bdev; | |
30c40d2c | 1046 | fmode_t mode = sb->s_mode; |
1da177e4 | 1047 | |
ddbaaf30 | 1048 | bdev->bd_super = NULL; |
1da177e4 LT |
1049 | generic_shutdown_super(sb); |
1050 | sync_blockdev(bdev); | |
d4d77629 | 1051 | WARN_ON_ONCE(!(mode & FMODE_EXCL)); |
e525fd89 | 1052 | blkdev_put(bdev, mode | FMODE_EXCL); |
1da177e4 LT |
1053 | } |
1054 | ||
1055 | EXPORT_SYMBOL(kill_block_super); | |
9361401e | 1056 | #endif |
1da177e4 | 1057 | |
3c26ff6e | 1058 | struct dentry *mount_nodev(struct file_system_type *fs_type, |
1da177e4 | 1059 | int flags, void *data, |
3c26ff6e | 1060 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1061 | { |
1062 | int error; | |
9249e17f | 1063 | struct super_block *s = sget(fs_type, NULL, set_anon_super, flags, NULL); |
1da177e4 LT |
1064 | |
1065 | if (IS_ERR(s)) | |
3c26ff6e | 1066 | return ERR_CAST(s); |
1da177e4 | 1067 | |
9b04c997 | 1068 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1069 | if (error) { |
74dbbdd7 | 1070 | deactivate_locked_super(s); |
3c26ff6e | 1071 | return ERR_PTR(error); |
1da177e4 LT |
1072 | } |
1073 | s->s_flags |= MS_ACTIVE; | |
3c26ff6e | 1074 | return dget(s->s_root); |
1da177e4 | 1075 | } |
3c26ff6e AV |
1076 | EXPORT_SYMBOL(mount_nodev); |
1077 | ||
1da177e4 LT |
1078 | static int compare_single(struct super_block *s, void *p) |
1079 | { | |
1080 | return 1; | |
1081 | } | |
1082 | ||
fc14f2fe | 1083 | struct dentry *mount_single(struct file_system_type *fs_type, |
1da177e4 | 1084 | int flags, void *data, |
fc14f2fe | 1085 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1086 | { |
1087 | struct super_block *s; | |
1088 | int error; | |
1089 | ||
9249e17f | 1090 | s = sget(fs_type, compare_single, set_anon_super, flags, NULL); |
1da177e4 | 1091 | if (IS_ERR(s)) |
fc14f2fe | 1092 | return ERR_CAST(s); |
1da177e4 | 1093 | if (!s->s_root) { |
9b04c997 | 1094 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1095 | if (error) { |
74dbbdd7 | 1096 | deactivate_locked_super(s); |
fc14f2fe | 1097 | return ERR_PTR(error); |
1da177e4 LT |
1098 | } |
1099 | s->s_flags |= MS_ACTIVE; | |
9329d1be KS |
1100 | } else { |
1101 | do_remount_sb(s, flags, data, 0); | |
1da177e4 | 1102 | } |
fc14f2fe AV |
1103 | return dget(s->s_root); |
1104 | } | |
1105 | EXPORT_SYMBOL(mount_single); | |
1106 | ||
9d412a43 AV |
1107 | struct dentry * |
1108 | mount_fs(struct file_system_type *type, int flags, const char *name, void *data) | |
1da177e4 | 1109 | { |
c96e41e9 | 1110 | struct dentry *root; |
9d412a43 | 1111 | struct super_block *sb; |
1da177e4 | 1112 | char *secdata = NULL; |
9d412a43 | 1113 | int error = -ENOMEM; |
8089352a | 1114 | |
e0007529 | 1115 | if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) { |
1da177e4 | 1116 | secdata = alloc_secdata(); |
454e2398 | 1117 | if (!secdata) |
9d412a43 | 1118 | goto out; |
1da177e4 | 1119 | |
e0007529 | 1120 | error = security_sb_copy_data(data, secdata); |
454e2398 | 1121 | if (error) |
1da177e4 | 1122 | goto out_free_secdata; |
1da177e4 LT |
1123 | } |
1124 | ||
1a102ff9 AV |
1125 | root = type->mount(type, flags, name, data); |
1126 | if (IS_ERR(root)) { | |
1127 | error = PTR_ERR(root); | |
1128 | goto out_free_secdata; | |
c96e41e9 | 1129 | } |
9d412a43 AV |
1130 | sb = root->d_sb; |
1131 | BUG_ON(!sb); | |
1132 | WARN_ON(!sb->s_bdi); | |
6c510389 | 1133 | WARN_ON(sb->s_bdi == &default_backing_dev_info); |
9d412a43 | 1134 | sb->s_flags |= MS_BORN; |
454e2398 | 1135 | |
9d412a43 | 1136 | error = security_sb_kern_mount(sb, flags, secdata); |
5129a469 JE |
1137 | if (error) |
1138 | goto out_sb; | |
454e2398 | 1139 | |
42cb56ae JL |
1140 | /* |
1141 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE | |
1142 | * but s_maxbytes was an unsigned long long for many releases. Throw | |
1143 | * this warning for a little while to try and catch filesystems that | |
4358b567 | 1144 | * violate this rule. |
42cb56ae | 1145 | */ |
9d412a43 AV |
1146 | WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " |
1147 | "negative value (%lld)\n", type->name, sb->s_maxbytes); | |
42cb56ae | 1148 | |
9d412a43 | 1149 | up_write(&sb->s_umount); |
8680e22f | 1150 | free_secdata(secdata); |
9d412a43 | 1151 | return root; |
1da177e4 | 1152 | out_sb: |
9d412a43 AV |
1153 | dput(root); |
1154 | deactivate_locked_super(sb); | |
1da177e4 LT |
1155 | out_free_secdata: |
1156 | free_secdata(secdata); | |
1da177e4 | 1157 | out: |
454e2398 | 1158 | return ERR_PTR(error); |
1da177e4 LT |
1159 | } |
1160 | ||
5accdf82 JK |
1161 | /* |
1162 | * This is an internal function, please use sb_end_{write,pagefault,intwrite} | |
1163 | * instead. | |
1164 | */ | |
1165 | void __sb_end_write(struct super_block *sb, int level) | |
1166 | { | |
1167 | percpu_counter_dec(&sb->s_writers.counter[level-1]); | |
1168 | /* | |
1169 | * Make sure s_writers are updated before we wake up waiters in | |
1170 | * freeze_super(). | |
1171 | */ | |
1172 | smp_mb(); | |
1173 | if (waitqueue_active(&sb->s_writers.wait)) | |
1174 | wake_up(&sb->s_writers.wait); | |
1175 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _RET_IP_); | |
1176 | } | |
1177 | EXPORT_SYMBOL(__sb_end_write); | |
1178 | ||
1179 | #ifdef CONFIG_LOCKDEP | |
1180 | /* | |
1181 | * We want lockdep to tell us about possible deadlocks with freezing but | |
1182 | * it's it bit tricky to properly instrument it. Getting a freeze protection | |
1183 | * works as getting a read lock but there are subtle problems. XFS for example | |
1184 | * gets freeze protection on internal level twice in some cases, which is OK | |
1185 | * only because we already hold a freeze protection also on higher level. Due | |
1186 | * to these cases we have to tell lockdep we are doing trylock when we | |
1187 | * already hold a freeze protection for a higher freeze level. | |
1188 | */ | |
1189 | static void acquire_freeze_lock(struct super_block *sb, int level, bool trylock, | |
1190 | unsigned long ip) | |
1191 | { | |
1192 | int i; | |
1193 | ||
1194 | if (!trylock) { | |
1195 | for (i = 0; i < level - 1; i++) | |
1196 | if (lock_is_held(&sb->s_writers.lock_map[i])) { | |
1197 | trylock = true; | |
1198 | break; | |
1199 | } | |
1200 | } | |
1201 | rwsem_acquire_read(&sb->s_writers.lock_map[level-1], 0, trylock, ip); | |
1202 | } | |
1203 | #endif | |
1204 | ||
1205 | /* | |
1206 | * This is an internal function, please use sb_start_{write,pagefault,intwrite} | |
1207 | * instead. | |
1208 | */ | |
1209 | int __sb_start_write(struct super_block *sb, int level, bool wait) | |
1210 | { | |
1211 | retry: | |
1212 | if (unlikely(sb->s_writers.frozen >= level)) { | |
1213 | if (!wait) | |
1214 | return 0; | |
1215 | wait_event(sb->s_writers.wait_unfrozen, | |
1216 | sb->s_writers.frozen < level); | |
1217 | } | |
1218 | ||
1219 | #ifdef CONFIG_LOCKDEP | |
1220 | acquire_freeze_lock(sb, level, !wait, _RET_IP_); | |
1221 | #endif | |
1222 | percpu_counter_inc(&sb->s_writers.counter[level-1]); | |
1223 | /* | |
1224 | * Make sure counter is updated before we check for frozen. | |
1225 | * freeze_super() first sets frozen and then checks the counter. | |
1226 | */ | |
1227 | smp_mb(); | |
1228 | if (unlikely(sb->s_writers.frozen >= level)) { | |
1229 | __sb_end_write(sb, level); | |
1230 | goto retry; | |
1231 | } | |
1232 | return 1; | |
1233 | } | |
1234 | EXPORT_SYMBOL(__sb_start_write); | |
1235 | ||
1236 | /** | |
1237 | * sb_wait_write - wait until all writers to given file system finish | |
1238 | * @sb: the super for which we wait | |
1239 | * @level: type of writers we wait for (normal vs page fault) | |
1240 | * | |
1241 | * This function waits until there are no writers of given type to given file | |
1242 | * system. Caller of this function should make sure there can be no new writers | |
1243 | * of type @level before calling this function. Otherwise this function can | |
1244 | * livelock. | |
1245 | */ | |
1246 | static void sb_wait_write(struct super_block *sb, int level) | |
1247 | { | |
1248 | s64 writers; | |
1249 | ||
1250 | /* | |
1251 | * We just cycle-through lockdep here so that it does not complain | |
1252 | * about returning with lock to userspace | |
1253 | */ | |
1254 | rwsem_acquire(&sb->s_writers.lock_map[level-1], 0, 0, _THIS_IP_); | |
1255 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _THIS_IP_); | |
1256 | ||
1257 | do { | |
1258 | DEFINE_WAIT(wait); | |
1259 | ||
1260 | /* | |
1261 | * We use a barrier in prepare_to_wait() to separate setting | |
1262 | * of frozen and checking of the counter | |
1263 | */ | |
1264 | prepare_to_wait(&sb->s_writers.wait, &wait, | |
1265 | TASK_UNINTERRUPTIBLE); | |
1266 | ||
1267 | writers = percpu_counter_sum(&sb->s_writers.counter[level-1]); | |
1268 | if (writers) | |
1269 | schedule(); | |
1270 | ||
1271 | finish_wait(&sb->s_writers.wait, &wait); | |
1272 | } while (writers); | |
1273 | } | |
1274 | ||
18e9e510 | 1275 | /** |
7000d3c4 RD |
1276 | * freeze_super - lock the filesystem and force it into a consistent state |
1277 | * @sb: the super to lock | |
18e9e510 JB |
1278 | * |
1279 | * Syncs the super to make sure the filesystem is consistent and calls the fs's | |
1280 | * freeze_fs. Subsequent calls to this without first thawing the fs will return | |
1281 | * -EBUSY. | |
5accdf82 JK |
1282 | * |
1283 | * During this function, sb->s_writers.frozen goes through these values: | |
1284 | * | |
1285 | * SB_UNFROZEN: File system is normal, all writes progress as usual. | |
1286 | * | |
1287 | * SB_FREEZE_WRITE: The file system is in the process of being frozen. New | |
1288 | * writes should be blocked, though page faults are still allowed. We wait for | |
1289 | * all writes to complete and then proceed to the next stage. | |
1290 | * | |
1291 | * SB_FREEZE_PAGEFAULT: Freezing continues. Now also page faults are blocked | |
1292 | * but internal fs threads can still modify the filesystem (although they | |
1293 | * should not dirty new pages or inodes), writeback can run etc. After waiting | |
1294 | * for all running page faults we sync the filesystem which will clean all | |
1295 | * dirty pages and inodes (no new dirty pages or inodes can be created when | |
1296 | * sync is running). | |
1297 | * | |
1298 | * SB_FREEZE_FS: The file system is frozen. Now all internal sources of fs | |
1299 | * modification are blocked (e.g. XFS preallocation truncation on inode | |
1300 | * reclaim). This is usually implemented by blocking new transactions for | |
1301 | * filesystems that have them and need this additional guard. After all | |
1302 | * internal writers are finished we call ->freeze_fs() to finish filesystem | |
1303 | * freezing. Then we transition to SB_FREEZE_COMPLETE state. This state is | |
1304 | * mostly auxiliary for filesystems to verify they do not modify frozen fs. | |
1305 | * | |
1306 | * sb->s_writers.frozen is protected by sb->s_umount. | |
18e9e510 JB |
1307 | */ |
1308 | int freeze_super(struct super_block *sb) | |
1309 | { | |
1310 | int ret; | |
1311 | ||
1312 | atomic_inc(&sb->s_active); | |
1313 | down_write(&sb->s_umount); | |
5accdf82 | 1314 | if (sb->s_writers.frozen != SB_UNFROZEN) { |
18e9e510 JB |
1315 | deactivate_locked_super(sb); |
1316 | return -EBUSY; | |
1317 | } | |
1318 | ||
dabe0dc1 AV |
1319 | if (!(sb->s_flags & MS_BORN)) { |
1320 | up_write(&sb->s_umount); | |
1321 | return 0; /* sic - it's "nothing to do" */ | |
1322 | } | |
1323 | ||
18e9e510 | 1324 | if (sb->s_flags & MS_RDONLY) { |
5accdf82 JK |
1325 | /* Nothing to do really... */ |
1326 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
18e9e510 JB |
1327 | up_write(&sb->s_umount); |
1328 | return 0; | |
1329 | } | |
1330 | ||
5accdf82 JK |
1331 | /* From now on, no new normal writers can start */ |
1332 | sb->s_writers.frozen = SB_FREEZE_WRITE; | |
1333 | smp_wmb(); | |
1334 | ||
1335 | /* Release s_umount to preserve sb_start_write -> s_umount ordering */ | |
1336 | up_write(&sb->s_umount); | |
1337 | ||
1338 | sb_wait_write(sb, SB_FREEZE_WRITE); | |
1339 | ||
1340 | /* Now we go and block page faults... */ | |
1341 | down_write(&sb->s_umount); | |
1342 | sb->s_writers.frozen = SB_FREEZE_PAGEFAULT; | |
18e9e510 JB |
1343 | smp_wmb(); |
1344 | ||
5accdf82 JK |
1345 | sb_wait_write(sb, SB_FREEZE_PAGEFAULT); |
1346 | ||
1347 | /* All writers are done so after syncing there won't be dirty data */ | |
18e9e510 JB |
1348 | sync_filesystem(sb); |
1349 | ||
5accdf82 JK |
1350 | /* Now wait for internal filesystem counter */ |
1351 | sb->s_writers.frozen = SB_FREEZE_FS; | |
18e9e510 | 1352 | smp_wmb(); |
5accdf82 | 1353 | sb_wait_write(sb, SB_FREEZE_FS); |
18e9e510 | 1354 | |
18e9e510 JB |
1355 | if (sb->s_op->freeze_fs) { |
1356 | ret = sb->s_op->freeze_fs(sb); | |
1357 | if (ret) { | |
1358 | printk(KERN_ERR | |
1359 | "VFS:Filesystem freeze failed\n"); | |
5accdf82 | 1360 | sb->s_writers.frozen = SB_UNFROZEN; |
e1616300 | 1361 | smp_wmb(); |
5accdf82 | 1362 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 JB |
1363 | deactivate_locked_super(sb); |
1364 | return ret; | |
1365 | } | |
1366 | } | |
5accdf82 JK |
1367 | /* |
1368 | * This is just for debugging purposes so that fs can warn if it | |
1369 | * sees write activity when frozen is set to SB_FREEZE_COMPLETE. | |
1370 | */ | |
1371 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
18e9e510 JB |
1372 | up_write(&sb->s_umount); |
1373 | return 0; | |
1374 | } | |
1375 | EXPORT_SYMBOL(freeze_super); | |
1376 | ||
1377 | /** | |
1378 | * thaw_super -- unlock filesystem | |
1379 | * @sb: the super to thaw | |
1380 | * | |
1381 | * Unlocks the filesystem and marks it writeable again after freeze_super(). | |
1382 | */ | |
1383 | int thaw_super(struct super_block *sb) | |
1384 | { | |
1385 | int error; | |
1386 | ||
1387 | down_write(&sb->s_umount); | |
5accdf82 | 1388 | if (sb->s_writers.frozen == SB_UNFROZEN) { |
18e9e510 JB |
1389 | up_write(&sb->s_umount); |
1390 | return -EINVAL; | |
1391 | } | |
1392 | ||
1393 | if (sb->s_flags & MS_RDONLY) | |
1394 | goto out; | |
1395 | ||
1396 | if (sb->s_op->unfreeze_fs) { | |
1397 | error = sb->s_op->unfreeze_fs(sb); | |
1398 | if (error) { | |
1399 | printk(KERN_ERR | |
1400 | "VFS:Filesystem thaw failed\n"); | |
18e9e510 JB |
1401 | up_write(&sb->s_umount); |
1402 | return error; | |
1403 | } | |
1404 | } | |
1405 | ||
1406 | out: | |
5accdf82 | 1407 | sb->s_writers.frozen = SB_UNFROZEN; |
18e9e510 | 1408 | smp_wmb(); |
5accdf82 | 1409 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 JB |
1410 | deactivate_locked_super(sb); |
1411 | ||
1412 | return 0; | |
1413 | } | |
1414 | EXPORT_SYMBOL(thaw_super); |