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 | ||
7eb5e882 AV |
132 | /** |
133 | * destroy_super - frees a superblock | |
134 | * @s: superblock to free | |
135 | * | |
136 | * Frees a superblock. | |
137 | */ | |
138 | static void destroy_super(struct super_block *s) | |
5accdf82 JK |
139 | { |
140 | int i; | |
7eb5e882 AV |
141 | list_lru_destroy(&s->s_dentry_lru); |
142 | list_lru_destroy(&s->s_inode_lru); | |
143 | #ifdef CONFIG_SMP | |
144 | free_percpu(s->s_files); | |
145 | #endif | |
5accdf82 JK |
146 | for (i = 0; i < SB_FREEZE_LEVELS; i++) |
147 | percpu_counter_destroy(&s->s_writers.counter[i]); | |
7eb5e882 AV |
148 | security_sb_free(s); |
149 | WARN_ON(!list_empty(&s->s_mounts)); | |
150 | kfree(s->s_subtype); | |
151 | kfree(s->s_options); | |
e2fec7c3 | 152 | kfree_rcu(s, rcu); |
5accdf82 JK |
153 | } |
154 | ||
1da177e4 LT |
155 | /** |
156 | * alloc_super - create new superblock | |
fe2bbc48 | 157 | * @type: filesystem type superblock should belong to |
9249e17f | 158 | * @flags: the mount flags |
1da177e4 LT |
159 | * |
160 | * Allocates and initializes a new &struct super_block. alloc_super() | |
161 | * returns a pointer new superblock or %NULL if allocation had failed. | |
162 | */ | |
9249e17f | 163 | static struct super_block *alloc_super(struct file_system_type *type, int flags) |
1da177e4 | 164 | { |
11b0b5ab | 165 | struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER); |
b87221de | 166 | static const struct super_operations default_op; |
7eb5e882 AV |
167 | int i; |
168 | ||
169 | if (!s) | |
170 | return NULL; | |
1da177e4 | 171 | |
7eb5e882 AV |
172 | if (security_sb_alloc(s)) |
173 | goto fail; | |
7b7a8665 | 174 | |
6416ccb7 | 175 | #ifdef CONFIG_SMP |
7eb5e882 AV |
176 | s->s_files = alloc_percpu(struct list_head); |
177 | if (!s->s_files) | |
178 | goto fail; | |
179 | for_each_possible_cpu(i) | |
180 | INIT_LIST_HEAD(per_cpu_ptr(s->s_files, i)); | |
6416ccb7 | 181 | #else |
7eb5e882 | 182 | INIT_LIST_HEAD(&s->s_files); |
6416ccb7 | 183 | #endif |
7eb5e882 AV |
184 | for (i = 0; i < SB_FREEZE_LEVELS; i++) { |
185 | if (percpu_counter_init(&s->s_writers.counter[i], 0) < 0) | |
186 | goto fail; | |
187 | lockdep_init_map(&s->s_writers.lock_map[i], sb_writers_name[i], | |
188 | &type->s_writers_key[i], 0); | |
1da177e4 | 189 | } |
7eb5e882 AV |
190 | init_waitqueue_head(&s->s_writers.wait); |
191 | init_waitqueue_head(&s->s_writers.wait_unfrozen); | |
192 | s->s_flags = flags; | |
193 | s->s_bdi = &default_backing_dev_info; | |
194 | INIT_HLIST_NODE(&s->s_instances); | |
195 | INIT_HLIST_BL_HEAD(&s->s_anon); | |
196 | INIT_LIST_HEAD(&s->s_inodes); | |
197 | ||
198 | if (list_lru_init(&s->s_dentry_lru)) | |
199 | goto fail; | |
200 | if (list_lru_init(&s->s_inode_lru)) | |
201 | goto fail; | |
202 | ||
203 | INIT_LIST_HEAD(&s->s_mounts); | |
204 | init_rwsem(&s->s_umount); | |
205 | lockdep_set_class(&s->s_umount, &type->s_umount_key); | |
206 | /* | |
207 | * sget() can have s_umount recursion. | |
208 | * | |
209 | * When it cannot find a suitable sb, it allocates a new | |
210 | * one (this one), and tries again to find a suitable old | |
211 | * one. | |
212 | * | |
213 | * In case that succeeds, it will acquire the s_umount | |
214 | * lock of the old one. Since these are clearly distrinct | |
215 | * locks, and this object isn't exposed yet, there's no | |
216 | * risk of deadlocks. | |
217 | * | |
218 | * Annotate this by putting this lock in a different | |
219 | * subclass. | |
220 | */ | |
221 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); | |
222 | s->s_count = 1; | |
223 | atomic_set(&s->s_active, 1); | |
224 | mutex_init(&s->s_vfs_rename_mutex); | |
225 | lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key); | |
226 | mutex_init(&s->s_dquot.dqio_mutex); | |
227 | mutex_init(&s->s_dquot.dqonoff_mutex); | |
228 | init_rwsem(&s->s_dquot.dqptr_sem); | |
229 | s->s_maxbytes = MAX_NON_LFS; | |
230 | s->s_op = &default_op; | |
231 | s->s_time_gran = 1000000000; | |
232 | s->cleancache_poolid = -1; | |
233 | ||
234 | s->s_shrink.seeks = DEFAULT_SEEKS; | |
235 | s->s_shrink.scan_objects = super_cache_scan; | |
236 | s->s_shrink.count_objects = super_cache_count; | |
237 | s->s_shrink.batch = 1024; | |
238 | s->s_shrink.flags = SHRINKER_NUMA_AWARE; | |
1da177e4 | 239 | return s; |
5ca302c8 | 240 | |
7eb5e882 AV |
241 | fail: |
242 | destroy_super(s); | |
243 | return NULL; | |
1da177e4 LT |
244 | } |
245 | ||
246 | /* Superblock refcounting */ | |
247 | ||
248 | /* | |
35cf7ba0 | 249 | * Drop a superblock's refcount. The caller must hold sb_lock. |
1da177e4 | 250 | */ |
f47ec3f2 | 251 | static void __put_super(struct super_block *sb) |
1da177e4 | 252 | { |
1da177e4 | 253 | if (!--sb->s_count) { |
551de6f3 | 254 | list_del_init(&sb->s_list); |
1da177e4 | 255 | destroy_super(sb); |
1da177e4 | 256 | } |
1da177e4 LT |
257 | } |
258 | ||
259 | /** | |
260 | * put_super - drop a temporary reference to superblock | |
261 | * @sb: superblock in question | |
262 | * | |
263 | * Drops a temporary reference, frees superblock if there's no | |
264 | * references left. | |
265 | */ | |
f47ec3f2 | 266 | static void put_super(struct super_block *sb) |
1da177e4 LT |
267 | { |
268 | spin_lock(&sb_lock); | |
269 | __put_super(sb); | |
270 | spin_unlock(&sb_lock); | |
271 | } | |
272 | ||
273 | ||
274 | /** | |
1712ac8f | 275 | * deactivate_locked_super - drop an active reference to superblock |
1da177e4 LT |
276 | * @s: superblock to deactivate |
277 | * | |
1712ac8f AV |
278 | * Drops an active reference to superblock, converting it into a temprory |
279 | * one if there is no other active references left. In that case we | |
1da177e4 LT |
280 | * tell fs driver to shut it down and drop the temporary reference we |
281 | * had just acquired. | |
1712ac8f AV |
282 | * |
283 | * Caller holds exclusive lock on superblock; that lock is released. | |
1da177e4 | 284 | */ |
1712ac8f | 285 | void deactivate_locked_super(struct super_block *s) |
1da177e4 LT |
286 | { |
287 | struct file_system_type *fs = s->s_type; | |
b20bd1a5 | 288 | if (atomic_dec_and_test(&s->s_active)) { |
3167760f | 289 | cleancache_invalidate_fs(s); |
1da177e4 | 290 | fs->kill_sb(s); |
b0d40c92 DC |
291 | |
292 | /* caches are now gone, we can safely kill the shrinker now */ | |
293 | unregister_shrinker(&s->s_shrink); | |
f5e1dd34 | 294 | |
1da177e4 LT |
295 | put_filesystem(fs); |
296 | put_super(s); | |
1712ac8f AV |
297 | } else { |
298 | up_write(&s->s_umount); | |
1da177e4 LT |
299 | } |
300 | } | |
301 | ||
1712ac8f | 302 | EXPORT_SYMBOL(deactivate_locked_super); |
1da177e4 | 303 | |
74dbbdd7 | 304 | /** |
1712ac8f | 305 | * deactivate_super - drop an active reference to superblock |
74dbbdd7 AV |
306 | * @s: superblock to deactivate |
307 | * | |
1712ac8f AV |
308 | * Variant of deactivate_locked_super(), except that superblock is *not* |
309 | * locked by caller. If we are going to drop the final active reference, | |
310 | * lock will be acquired prior to that. | |
74dbbdd7 | 311 | */ |
1712ac8f | 312 | void deactivate_super(struct super_block *s) |
74dbbdd7 | 313 | { |
1712ac8f AV |
314 | if (!atomic_add_unless(&s->s_active, -1, 1)) { |
315 | down_write(&s->s_umount); | |
316 | deactivate_locked_super(s); | |
74dbbdd7 AV |
317 | } |
318 | } | |
319 | ||
1712ac8f | 320 | EXPORT_SYMBOL(deactivate_super); |
74dbbdd7 | 321 | |
1da177e4 LT |
322 | /** |
323 | * grab_super - acquire an active reference | |
324 | * @s: reference we are trying to make active | |
325 | * | |
326 | * Tries to acquire an active reference. grab_super() is used when we | |
327 | * had just found a superblock in super_blocks or fs_type->fs_supers | |
328 | * and want to turn it into a full-blown active reference. grab_super() | |
329 | * is called with sb_lock held and drops it. Returns 1 in case of | |
330 | * success, 0 if we had failed (superblock contents was already dead or | |
acfec9a5 AV |
331 | * dying when grab_super() had been called). Note that this is only |
332 | * called for superblocks not in rundown mode (== ones still on ->fs_supers | |
333 | * of their type), so increment of ->s_count is OK here. | |
1da177e4 | 334 | */ |
9c4dbee7 | 335 | static int grab_super(struct super_block *s) __releases(sb_lock) |
1da177e4 LT |
336 | { |
337 | s->s_count++; | |
338 | spin_unlock(&sb_lock); | |
339 | down_write(&s->s_umount); | |
acfec9a5 AV |
340 | if ((s->s_flags & MS_BORN) && atomic_inc_not_zero(&s->s_active)) { |
341 | put_super(s); | |
342 | return 1; | |
343 | } | |
1da177e4 LT |
344 | up_write(&s->s_umount); |
345 | put_super(s); | |
1da177e4 LT |
346 | return 0; |
347 | } | |
348 | ||
12ad3ab6 DC |
349 | /* |
350 | * grab_super_passive - acquire a passive reference | |
331cbdee | 351 | * @sb: reference we are trying to grab |
12ad3ab6 DC |
352 | * |
353 | * Tries to acquire a passive reference. This is used in places where we | |
354 | * cannot take an active reference but we need to ensure that the | |
355 | * superblock does not go away while we are working on it. It returns | |
356 | * false if a reference was not gained, and returns true with the s_umount | |
357 | * lock held in read mode if a reference is gained. On successful return, | |
358 | * the caller must drop the s_umount lock and the passive reference when | |
359 | * done. | |
360 | */ | |
361 | bool grab_super_passive(struct super_block *sb) | |
362 | { | |
363 | spin_lock(&sb_lock); | |
a5166169 | 364 | if (hlist_unhashed(&sb->s_instances)) { |
12ad3ab6 DC |
365 | spin_unlock(&sb_lock); |
366 | return false; | |
367 | } | |
368 | ||
369 | sb->s_count++; | |
370 | spin_unlock(&sb_lock); | |
371 | ||
372 | if (down_read_trylock(&sb->s_umount)) { | |
dabe0dc1 | 373 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
12ad3ab6 DC |
374 | return true; |
375 | up_read(&sb->s_umount); | |
376 | } | |
377 | ||
378 | put_super(sb); | |
379 | return false; | |
380 | } | |
381 | ||
1da177e4 LT |
382 | /** |
383 | * generic_shutdown_super - common helper for ->kill_sb() | |
384 | * @sb: superblock to kill | |
385 | * | |
386 | * generic_shutdown_super() does all fs-independent work on superblock | |
387 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects | |
388 | * that need destruction out of superblock, call generic_shutdown_super() | |
389 | * and release aforementioned objects. Note: dentries and inodes _are_ | |
390 | * taken care of and do not need specific handling. | |
c636ebdb DH |
391 | * |
392 | * Upon calling this function, the filesystem may no longer alter or | |
393 | * rearrange the set of dentries belonging to this super_block, nor may it | |
394 | * change the attachments of dentries to inodes. | |
1da177e4 LT |
395 | */ |
396 | void generic_shutdown_super(struct super_block *sb) | |
397 | { | |
ee9b6d61 | 398 | const struct super_operations *sop = sb->s_op; |
1da177e4 | 399 | |
c636ebdb DH |
400 | if (sb->s_root) { |
401 | shrink_dcache_for_umount(sb); | |
60b0680f | 402 | sync_filesystem(sb); |
1da177e4 | 403 | sb->s_flags &= ~MS_ACTIVE; |
efaee192 | 404 | |
63997e98 AV |
405 | fsnotify_unmount_inodes(&sb->s_inodes); |
406 | ||
407 | evict_inodes(sb); | |
1da177e4 | 408 | |
7b7a8665 CH |
409 | if (sb->s_dio_done_wq) { |
410 | destroy_workqueue(sb->s_dio_done_wq); | |
411 | sb->s_dio_done_wq = NULL; | |
412 | } | |
413 | ||
1da177e4 LT |
414 | if (sop->put_super) |
415 | sop->put_super(sb); | |
416 | ||
63997e98 | 417 | if (!list_empty(&sb->s_inodes)) { |
7b4fe29e DJ |
418 | printk("VFS: Busy inodes after unmount of %s. " |
419 | "Self-destruct in 5 seconds. Have a nice day...\n", | |
420 | sb->s_id); | |
1da177e4 | 421 | } |
1da177e4 LT |
422 | } |
423 | spin_lock(&sb_lock); | |
424 | /* should be initialized for __put_super_and_need_restart() */ | |
a5166169 | 425 | hlist_del_init(&sb->s_instances); |
1da177e4 LT |
426 | spin_unlock(&sb_lock); |
427 | up_write(&sb->s_umount); | |
428 | } | |
429 | ||
430 | EXPORT_SYMBOL(generic_shutdown_super); | |
431 | ||
432 | /** | |
433 | * sget - find or create a superblock | |
434 | * @type: filesystem type superblock should belong to | |
435 | * @test: comparison callback | |
436 | * @set: setup callback | |
9249e17f | 437 | * @flags: mount flags |
1da177e4 LT |
438 | * @data: argument to each of them |
439 | */ | |
440 | struct super_block *sget(struct file_system_type *type, | |
441 | int (*test)(struct super_block *,void *), | |
442 | int (*set)(struct super_block *,void *), | |
9249e17f | 443 | int flags, |
1da177e4 LT |
444 | void *data) |
445 | { | |
446 | struct super_block *s = NULL; | |
d4730127 | 447 | struct super_block *old; |
1da177e4 LT |
448 | int err; |
449 | ||
450 | retry: | |
451 | spin_lock(&sb_lock); | |
d4730127 | 452 | if (test) { |
b67bfe0d | 453 | hlist_for_each_entry(old, &type->fs_supers, s_instances) { |
d4730127 MK |
454 | if (!test(old, data)) |
455 | continue; | |
456 | if (!grab_super(old)) | |
457 | goto retry; | |
a3cfbb53 LZ |
458 | if (s) { |
459 | up_write(&s->s_umount); | |
d4730127 | 460 | destroy_super(s); |
7a4dec53 | 461 | s = NULL; |
a3cfbb53 | 462 | } |
d4730127 MK |
463 | return old; |
464 | } | |
1da177e4 LT |
465 | } |
466 | if (!s) { | |
467 | spin_unlock(&sb_lock); | |
9249e17f | 468 | s = alloc_super(type, flags); |
1da177e4 LT |
469 | if (!s) |
470 | return ERR_PTR(-ENOMEM); | |
471 | goto retry; | |
472 | } | |
473 | ||
474 | err = set(s, data); | |
475 | if (err) { | |
476 | spin_unlock(&sb_lock); | |
a3cfbb53 | 477 | up_write(&s->s_umount); |
1da177e4 LT |
478 | destroy_super(s); |
479 | return ERR_PTR(err); | |
480 | } | |
481 | s->s_type = type; | |
482 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); | |
483 | list_add_tail(&s->s_list, &super_blocks); | |
a5166169 | 484 | hlist_add_head(&s->s_instances, &type->fs_supers); |
1da177e4 LT |
485 | spin_unlock(&sb_lock); |
486 | get_filesystem(type); | |
b0d40c92 | 487 | register_shrinker(&s->s_shrink); |
1da177e4 LT |
488 | return s; |
489 | } | |
490 | ||
491 | EXPORT_SYMBOL(sget); | |
492 | ||
493 | void drop_super(struct super_block *sb) | |
494 | { | |
495 | up_read(&sb->s_umount); | |
496 | put_super(sb); | |
497 | } | |
498 | ||
499 | EXPORT_SYMBOL(drop_super); | |
500 | ||
01a05b33 AV |
501 | /** |
502 | * iterate_supers - call function for all active superblocks | |
503 | * @f: function to call | |
504 | * @arg: argument to pass to it | |
505 | * | |
506 | * Scans the superblock list and calls given function, passing it | |
507 | * locked superblock and given argument. | |
508 | */ | |
509 | void iterate_supers(void (*f)(struct super_block *, void *), void *arg) | |
510 | { | |
dca33252 | 511 | struct super_block *sb, *p = NULL; |
01a05b33 AV |
512 | |
513 | spin_lock(&sb_lock); | |
dca33252 | 514 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 515 | if (hlist_unhashed(&sb->s_instances)) |
01a05b33 AV |
516 | continue; |
517 | sb->s_count++; | |
518 | spin_unlock(&sb_lock); | |
519 | ||
520 | down_read(&sb->s_umount); | |
dabe0dc1 | 521 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
01a05b33 AV |
522 | f(sb, arg); |
523 | up_read(&sb->s_umount); | |
524 | ||
525 | spin_lock(&sb_lock); | |
dca33252 AV |
526 | if (p) |
527 | __put_super(p); | |
528 | p = sb; | |
01a05b33 | 529 | } |
dca33252 AV |
530 | if (p) |
531 | __put_super(p); | |
01a05b33 AV |
532 | spin_unlock(&sb_lock); |
533 | } | |
534 | ||
43e15cdb AV |
535 | /** |
536 | * iterate_supers_type - call function for superblocks of given type | |
537 | * @type: fs type | |
538 | * @f: function to call | |
539 | * @arg: argument to pass to it | |
540 | * | |
541 | * Scans the superblock list and calls given function, passing it | |
542 | * locked superblock and given argument. | |
543 | */ | |
544 | void iterate_supers_type(struct file_system_type *type, | |
545 | void (*f)(struct super_block *, void *), void *arg) | |
546 | { | |
547 | struct super_block *sb, *p = NULL; | |
548 | ||
549 | spin_lock(&sb_lock); | |
b67bfe0d | 550 | hlist_for_each_entry(sb, &type->fs_supers, s_instances) { |
43e15cdb AV |
551 | sb->s_count++; |
552 | spin_unlock(&sb_lock); | |
553 | ||
554 | down_read(&sb->s_umount); | |
dabe0dc1 | 555 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
43e15cdb AV |
556 | f(sb, arg); |
557 | up_read(&sb->s_umount); | |
558 | ||
559 | spin_lock(&sb_lock); | |
560 | if (p) | |
561 | __put_super(p); | |
562 | p = sb; | |
563 | } | |
564 | if (p) | |
565 | __put_super(p); | |
566 | spin_unlock(&sb_lock); | |
567 | } | |
568 | ||
569 | EXPORT_SYMBOL(iterate_supers_type); | |
570 | ||
1da177e4 LT |
571 | /** |
572 | * get_super - get the superblock of a device | |
573 | * @bdev: device to get the superblock for | |
574 | * | |
575 | * Scans the superblock list and finds the superblock of the file system | |
576 | * mounted on the device given. %NULL is returned if no match is found. | |
577 | */ | |
578 | ||
df40c01a | 579 | struct super_block *get_super(struct block_device *bdev) |
1da177e4 | 580 | { |
618f0636 KK |
581 | struct super_block *sb; |
582 | ||
1da177e4 LT |
583 | if (!bdev) |
584 | return NULL; | |
618f0636 | 585 | |
1da177e4 | 586 | spin_lock(&sb_lock); |
618f0636 KK |
587 | rescan: |
588 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 589 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 590 | continue; |
618f0636 KK |
591 | if (sb->s_bdev == bdev) { |
592 | sb->s_count++; | |
1da177e4 | 593 | spin_unlock(&sb_lock); |
618f0636 | 594 | down_read(&sb->s_umount); |
df40c01a | 595 | /* still alive? */ |
dabe0dc1 | 596 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
618f0636 KK |
597 | return sb; |
598 | up_read(&sb->s_umount); | |
df40c01a | 599 | /* nope, got unmounted */ |
618f0636 | 600 | spin_lock(&sb_lock); |
df40c01a AV |
601 | __put_super(sb); |
602 | goto rescan; | |
1da177e4 LT |
603 | } |
604 | } | |
605 | spin_unlock(&sb_lock); | |
606 | return NULL; | |
607 | } | |
608 | ||
609 | EXPORT_SYMBOL(get_super); | |
4504230a | 610 | |
6b6dc836 JK |
611 | /** |
612 | * get_super_thawed - get thawed superblock of a device | |
613 | * @bdev: device to get the superblock for | |
614 | * | |
615 | * Scans the superblock list and finds the superblock of the file system | |
616 | * mounted on the device. The superblock is returned once it is thawed | |
617 | * (or immediately if it was not frozen). %NULL is returned if no match | |
618 | * is found. | |
619 | */ | |
620 | struct super_block *get_super_thawed(struct block_device *bdev) | |
621 | { | |
622 | while (1) { | |
623 | struct super_block *s = get_super(bdev); | |
5accdf82 | 624 | if (!s || s->s_writers.frozen == SB_UNFROZEN) |
6b6dc836 JK |
625 | return s; |
626 | up_read(&s->s_umount); | |
5accdf82 JK |
627 | wait_event(s->s_writers.wait_unfrozen, |
628 | s->s_writers.frozen == SB_UNFROZEN); | |
6b6dc836 JK |
629 | put_super(s); |
630 | } | |
631 | } | |
632 | EXPORT_SYMBOL(get_super_thawed); | |
633 | ||
4504230a CH |
634 | /** |
635 | * get_active_super - get an active reference to the superblock of a device | |
636 | * @bdev: device to get the superblock for | |
637 | * | |
638 | * Scans the superblock list and finds the superblock of the file system | |
639 | * mounted on the device given. Returns the superblock with an active | |
d3f21473 | 640 | * reference or %NULL if none was found. |
4504230a CH |
641 | */ |
642 | struct super_block *get_active_super(struct block_device *bdev) | |
643 | { | |
644 | struct super_block *sb; | |
645 | ||
646 | if (!bdev) | |
647 | return NULL; | |
648 | ||
1494583d | 649 | restart: |
4504230a CH |
650 | spin_lock(&sb_lock); |
651 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 652 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 653 | continue; |
1494583d | 654 | if (sb->s_bdev == bdev) { |
acfec9a5 | 655 | if (!grab_super(sb)) |
1494583d | 656 | goto restart; |
acfec9a5 AV |
657 | up_write(&sb->s_umount); |
658 | return sb; | |
1494583d | 659 | } |
4504230a CH |
660 | } |
661 | spin_unlock(&sb_lock); | |
662 | return NULL; | |
663 | } | |
1da177e4 | 664 | |
df40c01a | 665 | struct super_block *user_get_super(dev_t dev) |
1da177e4 | 666 | { |
618f0636 | 667 | struct super_block *sb; |
1da177e4 | 668 | |
1da177e4 | 669 | spin_lock(&sb_lock); |
618f0636 KK |
670 | rescan: |
671 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 672 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 673 | continue; |
618f0636 KK |
674 | if (sb->s_dev == dev) { |
675 | sb->s_count++; | |
1da177e4 | 676 | spin_unlock(&sb_lock); |
618f0636 | 677 | down_read(&sb->s_umount); |
df40c01a | 678 | /* still alive? */ |
dabe0dc1 | 679 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
618f0636 KK |
680 | return sb; |
681 | up_read(&sb->s_umount); | |
df40c01a | 682 | /* nope, got unmounted */ |
618f0636 | 683 | spin_lock(&sb_lock); |
df40c01a AV |
684 | __put_super(sb); |
685 | goto rescan; | |
1da177e4 LT |
686 | } |
687 | } | |
688 | spin_unlock(&sb_lock); | |
689 | return NULL; | |
690 | } | |
691 | ||
1da177e4 LT |
692 | /** |
693 | * do_remount_sb - asks filesystem to change mount options. | |
694 | * @sb: superblock in question | |
695 | * @flags: numeric part of options | |
696 | * @data: the rest of options | |
697 | * @force: whether or not to force the change | |
698 | * | |
699 | * Alters the mount options of a mounted file system. | |
700 | */ | |
701 | int do_remount_sb(struct super_block *sb, int flags, void *data, int force) | |
702 | { | |
703 | int retval; | |
c79d967d | 704 | int remount_ro; |
4504230a | 705 | |
5accdf82 | 706 | if (sb->s_writers.frozen != SB_UNFROZEN) |
4504230a CH |
707 | return -EBUSY; |
708 | ||
9361401e | 709 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
710 | if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev)) |
711 | return -EACCES; | |
9361401e | 712 | #endif |
4504230a | 713 | |
1da177e4 LT |
714 | if (flags & MS_RDONLY) |
715 | acct_auto_close(sb); | |
716 | shrink_dcache_sb(sb); | |
60b0680f | 717 | sync_filesystem(sb); |
1da177e4 | 718 | |
d208bbdd | 719 | remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY); |
d208bbdd | 720 | |
1da177e4 LT |
721 | /* If we are remounting RDONLY and current sb is read/write, |
722 | make sure there are no rw files opened */ | |
d208bbdd | 723 | if (remount_ro) { |
4ed5e82f | 724 | if (force) { |
1da177e4 | 725 | mark_files_ro(sb); |
4ed5e82f MS |
726 | } else { |
727 | retval = sb_prepare_remount_readonly(sb); | |
728 | if (retval) | |
729 | return retval; | |
4ed5e82f | 730 | } |
1da177e4 LT |
731 | } |
732 | ||
733 | if (sb->s_op->remount_fs) { | |
1da177e4 | 734 | retval = sb->s_op->remount_fs(sb, &flags, data); |
2833eb2b MS |
735 | if (retval) { |
736 | if (!force) | |
4ed5e82f | 737 | goto cancel_readonly; |
2833eb2b MS |
738 | /* If forced remount, go ahead despite any errors */ |
739 | WARN(1, "forced remount of a %s fs returned %i\n", | |
740 | sb->s_type->name, retval); | |
741 | } | |
1da177e4 LT |
742 | } |
743 | sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK); | |
4ed5e82f MS |
744 | /* Needs to be ordered wrt mnt_is_readonly() */ |
745 | smp_wmb(); | |
746 | sb->s_readonly_remount = 0; | |
c79d967d | 747 | |
d208bbdd NP |
748 | /* |
749 | * Some filesystems modify their metadata via some other path than the | |
750 | * bdev buffer cache (eg. use a private mapping, or directories in | |
751 | * pagecache, etc). Also file data modifications go via their own | |
752 | * mappings. So If we try to mount readonly then copy the filesystem | |
753 | * from bdev, we could get stale data, so invalidate it to give a best | |
754 | * effort at coherency. | |
755 | */ | |
756 | if (remount_ro && sb->s_bdev) | |
757 | invalidate_bdev(sb->s_bdev); | |
1da177e4 | 758 | return 0; |
4ed5e82f MS |
759 | |
760 | cancel_readonly: | |
761 | sb->s_readonly_remount = 0; | |
762 | return retval; | |
1da177e4 LT |
763 | } |
764 | ||
a2a9537a | 765 | static void do_emergency_remount(struct work_struct *work) |
1da177e4 | 766 | { |
dca33252 | 767 | struct super_block *sb, *p = NULL; |
1da177e4 LT |
768 | |
769 | spin_lock(&sb_lock); | |
dca33252 | 770 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 771 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 772 | continue; |
1da177e4 LT |
773 | sb->s_count++; |
774 | spin_unlock(&sb_lock); | |
443b94ba | 775 | down_write(&sb->s_umount); |
dabe0dc1 AV |
776 | if (sb->s_root && sb->s_bdev && (sb->s_flags & MS_BORN) && |
777 | !(sb->s_flags & MS_RDONLY)) { | |
1da177e4 | 778 | /* |
1da177e4 LT |
779 | * What lock protects sb->s_flags?? |
780 | */ | |
1da177e4 | 781 | do_remount_sb(sb, MS_RDONLY, NULL, 1); |
1da177e4 | 782 | } |
443b94ba | 783 | up_write(&sb->s_umount); |
1da177e4 | 784 | spin_lock(&sb_lock); |
dca33252 AV |
785 | if (p) |
786 | __put_super(p); | |
787 | p = sb; | |
1da177e4 | 788 | } |
dca33252 AV |
789 | if (p) |
790 | __put_super(p); | |
1da177e4 | 791 | spin_unlock(&sb_lock); |
a2a9537a | 792 | kfree(work); |
1da177e4 LT |
793 | printk("Emergency Remount complete\n"); |
794 | } | |
795 | ||
796 | void emergency_remount(void) | |
797 | { | |
a2a9537a JA |
798 | struct work_struct *work; |
799 | ||
800 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | |
801 | if (work) { | |
802 | INIT_WORK(work, do_emergency_remount); | |
803 | schedule_work(work); | |
804 | } | |
1da177e4 LT |
805 | } |
806 | ||
807 | /* | |
808 | * Unnamed block devices are dummy devices used by virtual | |
809 | * filesystems which don't use real block-devices. -- jrs | |
810 | */ | |
811 | ||
ad76cbc6 | 812 | static DEFINE_IDA(unnamed_dev_ida); |
1da177e4 | 813 | static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */ |
c63e09ec | 814 | static int unnamed_dev_start = 0; /* don't bother trying below it */ |
1da177e4 | 815 | |
0ee5dc67 | 816 | int get_anon_bdev(dev_t *p) |
1da177e4 LT |
817 | { |
818 | int dev; | |
819 | int error; | |
820 | ||
821 | retry: | |
ad76cbc6 | 822 | if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0) |
1da177e4 LT |
823 | return -ENOMEM; |
824 | spin_lock(&unnamed_dev_lock); | |
c63e09ec | 825 | error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev); |
f21f6220 AV |
826 | if (!error) |
827 | unnamed_dev_start = dev + 1; | |
1da177e4 LT |
828 | spin_unlock(&unnamed_dev_lock); |
829 | if (error == -EAGAIN) | |
830 | /* We raced and lost with another CPU. */ | |
831 | goto retry; | |
832 | else if (error) | |
833 | return -EAGAIN; | |
834 | ||
e8c8d1bc | 835 | if (dev == (1 << MINORBITS)) { |
1da177e4 | 836 | spin_lock(&unnamed_dev_lock); |
ad76cbc6 | 837 | ida_remove(&unnamed_dev_ida, dev); |
f21f6220 AV |
838 | if (unnamed_dev_start > dev) |
839 | unnamed_dev_start = dev; | |
1da177e4 LT |
840 | spin_unlock(&unnamed_dev_lock); |
841 | return -EMFILE; | |
842 | } | |
0ee5dc67 | 843 | *p = MKDEV(0, dev & MINORMASK); |
1da177e4 LT |
844 | return 0; |
845 | } | |
0ee5dc67 | 846 | EXPORT_SYMBOL(get_anon_bdev); |
1da177e4 | 847 | |
0ee5dc67 | 848 | void free_anon_bdev(dev_t dev) |
1da177e4 | 849 | { |
0ee5dc67 | 850 | int slot = MINOR(dev); |
1da177e4 | 851 | spin_lock(&unnamed_dev_lock); |
ad76cbc6 | 852 | ida_remove(&unnamed_dev_ida, slot); |
c63e09ec AV |
853 | if (slot < unnamed_dev_start) |
854 | unnamed_dev_start = slot; | |
1da177e4 LT |
855 | spin_unlock(&unnamed_dev_lock); |
856 | } | |
0ee5dc67 AV |
857 | EXPORT_SYMBOL(free_anon_bdev); |
858 | ||
859 | int set_anon_super(struct super_block *s, void *data) | |
860 | { | |
861 | int error = get_anon_bdev(&s->s_dev); | |
862 | if (!error) | |
863 | s->s_bdi = &noop_backing_dev_info; | |
864 | return error; | |
865 | } | |
866 | ||
867 | EXPORT_SYMBOL(set_anon_super); | |
868 | ||
869 | void kill_anon_super(struct super_block *sb) | |
870 | { | |
871 | dev_t dev = sb->s_dev; | |
872 | generic_shutdown_super(sb); | |
873 | free_anon_bdev(dev); | |
874 | } | |
1da177e4 LT |
875 | |
876 | EXPORT_SYMBOL(kill_anon_super); | |
877 | ||
1da177e4 LT |
878 | void kill_litter_super(struct super_block *sb) |
879 | { | |
880 | if (sb->s_root) | |
881 | d_genocide(sb->s_root); | |
882 | kill_anon_super(sb); | |
883 | } | |
884 | ||
885 | EXPORT_SYMBOL(kill_litter_super); | |
886 | ||
909e6d94 SH |
887 | static int ns_test_super(struct super_block *sb, void *data) |
888 | { | |
889 | return sb->s_fs_info == data; | |
890 | } | |
891 | ||
892 | static int ns_set_super(struct super_block *sb, void *data) | |
893 | { | |
894 | sb->s_fs_info = data; | |
895 | return set_anon_super(sb, NULL); | |
896 | } | |
897 | ||
ceefda69 AV |
898 | struct dentry *mount_ns(struct file_system_type *fs_type, int flags, |
899 | void *data, int (*fill_super)(struct super_block *, void *, int)) | |
909e6d94 SH |
900 | { |
901 | struct super_block *sb; | |
902 | ||
9249e17f | 903 | sb = sget(fs_type, ns_test_super, ns_set_super, flags, data); |
909e6d94 | 904 | if (IS_ERR(sb)) |
ceefda69 | 905 | return ERR_CAST(sb); |
909e6d94 SH |
906 | |
907 | if (!sb->s_root) { | |
908 | int err; | |
909e6d94 SH |
909 | err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0); |
910 | if (err) { | |
74dbbdd7 | 911 | deactivate_locked_super(sb); |
ceefda69 | 912 | return ERR_PTR(err); |
909e6d94 SH |
913 | } |
914 | ||
915 | sb->s_flags |= MS_ACTIVE; | |
916 | } | |
917 | ||
ceefda69 | 918 | return dget(sb->s_root); |
909e6d94 SH |
919 | } |
920 | ||
ceefda69 | 921 | EXPORT_SYMBOL(mount_ns); |
909e6d94 | 922 | |
9361401e | 923 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
924 | static int set_bdev_super(struct super_block *s, void *data) |
925 | { | |
926 | s->s_bdev = data; | |
927 | s->s_dev = s->s_bdev->bd_dev; | |
32a88aa1 JA |
928 | |
929 | /* | |
930 | * We set the bdi here to the queue backing, file systems can | |
931 | * overwrite this in ->fill_super() | |
932 | */ | |
933 | s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info; | |
1da177e4 LT |
934 | return 0; |
935 | } | |
936 | ||
937 | static int test_bdev_super(struct super_block *s, void *data) | |
938 | { | |
939 | return (void *)s->s_bdev == data; | |
940 | } | |
941 | ||
152a0836 | 942 | struct dentry *mount_bdev(struct file_system_type *fs_type, |
1da177e4 | 943 | int flags, const char *dev_name, void *data, |
152a0836 | 944 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
945 | { |
946 | struct block_device *bdev; | |
947 | struct super_block *s; | |
d4d77629 | 948 | fmode_t mode = FMODE_READ | FMODE_EXCL; |
1da177e4 LT |
949 | int error = 0; |
950 | ||
30c40d2c AV |
951 | if (!(flags & MS_RDONLY)) |
952 | mode |= FMODE_WRITE; | |
953 | ||
d4d77629 | 954 | bdev = blkdev_get_by_path(dev_name, mode, fs_type); |
1da177e4 | 955 | if (IS_ERR(bdev)) |
152a0836 | 956 | return ERR_CAST(bdev); |
1da177e4 LT |
957 | |
958 | /* | |
959 | * once the super is inserted into the list by sget, s_umount | |
960 | * will protect the lockfs code from trying to start a snapshot | |
961 | * while we are mounting | |
962 | */ | |
4fadd7bb CH |
963 | mutex_lock(&bdev->bd_fsfreeze_mutex); |
964 | if (bdev->bd_fsfreeze_count > 0) { | |
965 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
966 | error = -EBUSY; | |
967 | goto error_bdev; | |
968 | } | |
9249e17f DH |
969 | s = sget(fs_type, test_bdev_super, set_bdev_super, flags | MS_NOSEC, |
970 | bdev); | |
4fadd7bb | 971 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
1da177e4 | 972 | if (IS_ERR(s)) |
454e2398 | 973 | goto error_s; |
1da177e4 LT |
974 | |
975 | if (s->s_root) { | |
976 | if ((flags ^ s->s_flags) & MS_RDONLY) { | |
74dbbdd7 | 977 | deactivate_locked_super(s); |
454e2398 DH |
978 | error = -EBUSY; |
979 | goto error_bdev; | |
1da177e4 | 980 | } |
454e2398 | 981 | |
4f331f01 TH |
982 | /* |
983 | * s_umount nests inside bd_mutex during | |
e525fd89 TH |
984 | * __invalidate_device(). blkdev_put() acquires |
985 | * bd_mutex and can't be called under s_umount. Drop | |
986 | * s_umount temporarily. This is safe as we're | |
987 | * holding an active reference. | |
4f331f01 TH |
988 | */ |
989 | up_write(&s->s_umount); | |
d4d77629 | 990 | blkdev_put(bdev, mode); |
4f331f01 | 991 | down_write(&s->s_umount); |
1da177e4 LT |
992 | } else { |
993 | char b[BDEVNAME_SIZE]; | |
994 | ||
30c40d2c | 995 | s->s_mode = mode; |
1da177e4 | 996 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); |
e78c9a00 | 997 | sb_set_blocksize(s, block_size(bdev)); |
9b04c997 | 998 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 999 | if (error) { |
74dbbdd7 | 1000 | deactivate_locked_super(s); |
454e2398 | 1001 | goto error; |
fa675765 | 1002 | } |
454e2398 DH |
1003 | |
1004 | s->s_flags |= MS_ACTIVE; | |
87d8fe1e | 1005 | bdev->bd_super = s; |
1da177e4 LT |
1006 | } |
1007 | ||
152a0836 | 1008 | return dget(s->s_root); |
1da177e4 | 1009 | |
454e2398 DH |
1010 | error_s: |
1011 | error = PTR_ERR(s); | |
1012 | error_bdev: | |
d4d77629 | 1013 | blkdev_put(bdev, mode); |
454e2398 | 1014 | error: |
152a0836 AV |
1015 | return ERR_PTR(error); |
1016 | } | |
1017 | EXPORT_SYMBOL(mount_bdev); | |
1018 | ||
1da177e4 LT |
1019 | void kill_block_super(struct super_block *sb) |
1020 | { | |
1021 | struct block_device *bdev = sb->s_bdev; | |
30c40d2c | 1022 | fmode_t mode = sb->s_mode; |
1da177e4 | 1023 | |
ddbaaf30 | 1024 | bdev->bd_super = NULL; |
1da177e4 LT |
1025 | generic_shutdown_super(sb); |
1026 | sync_blockdev(bdev); | |
d4d77629 | 1027 | WARN_ON_ONCE(!(mode & FMODE_EXCL)); |
e525fd89 | 1028 | blkdev_put(bdev, mode | FMODE_EXCL); |
1da177e4 LT |
1029 | } |
1030 | ||
1031 | EXPORT_SYMBOL(kill_block_super); | |
9361401e | 1032 | #endif |
1da177e4 | 1033 | |
3c26ff6e | 1034 | struct dentry *mount_nodev(struct file_system_type *fs_type, |
1da177e4 | 1035 | int flags, void *data, |
3c26ff6e | 1036 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1037 | { |
1038 | int error; | |
9249e17f | 1039 | struct super_block *s = sget(fs_type, NULL, set_anon_super, flags, NULL); |
1da177e4 LT |
1040 | |
1041 | if (IS_ERR(s)) | |
3c26ff6e | 1042 | return ERR_CAST(s); |
1da177e4 | 1043 | |
9b04c997 | 1044 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1045 | if (error) { |
74dbbdd7 | 1046 | deactivate_locked_super(s); |
3c26ff6e | 1047 | return ERR_PTR(error); |
1da177e4 LT |
1048 | } |
1049 | s->s_flags |= MS_ACTIVE; | |
3c26ff6e | 1050 | return dget(s->s_root); |
1da177e4 | 1051 | } |
3c26ff6e AV |
1052 | EXPORT_SYMBOL(mount_nodev); |
1053 | ||
1da177e4 LT |
1054 | static int compare_single(struct super_block *s, void *p) |
1055 | { | |
1056 | return 1; | |
1057 | } | |
1058 | ||
fc14f2fe | 1059 | struct dentry *mount_single(struct file_system_type *fs_type, |
1da177e4 | 1060 | int flags, void *data, |
fc14f2fe | 1061 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1062 | { |
1063 | struct super_block *s; | |
1064 | int error; | |
1065 | ||
9249e17f | 1066 | s = sget(fs_type, compare_single, set_anon_super, flags, NULL); |
1da177e4 | 1067 | if (IS_ERR(s)) |
fc14f2fe | 1068 | return ERR_CAST(s); |
1da177e4 | 1069 | if (!s->s_root) { |
9b04c997 | 1070 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1071 | if (error) { |
74dbbdd7 | 1072 | deactivate_locked_super(s); |
fc14f2fe | 1073 | return ERR_PTR(error); |
1da177e4 LT |
1074 | } |
1075 | s->s_flags |= MS_ACTIVE; | |
9329d1be KS |
1076 | } else { |
1077 | do_remount_sb(s, flags, data, 0); | |
1da177e4 | 1078 | } |
fc14f2fe AV |
1079 | return dget(s->s_root); |
1080 | } | |
1081 | EXPORT_SYMBOL(mount_single); | |
1082 | ||
9d412a43 AV |
1083 | struct dentry * |
1084 | mount_fs(struct file_system_type *type, int flags, const char *name, void *data) | |
1da177e4 | 1085 | { |
c96e41e9 | 1086 | struct dentry *root; |
9d412a43 | 1087 | struct super_block *sb; |
1da177e4 | 1088 | char *secdata = NULL; |
9d412a43 | 1089 | int error = -ENOMEM; |
8089352a | 1090 | |
e0007529 | 1091 | if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) { |
1da177e4 | 1092 | secdata = alloc_secdata(); |
454e2398 | 1093 | if (!secdata) |
9d412a43 | 1094 | goto out; |
1da177e4 | 1095 | |
e0007529 | 1096 | error = security_sb_copy_data(data, secdata); |
454e2398 | 1097 | if (error) |
1da177e4 | 1098 | goto out_free_secdata; |
1da177e4 LT |
1099 | } |
1100 | ||
1a102ff9 AV |
1101 | root = type->mount(type, flags, name, data); |
1102 | if (IS_ERR(root)) { | |
1103 | error = PTR_ERR(root); | |
1104 | goto out_free_secdata; | |
c96e41e9 | 1105 | } |
9d412a43 AV |
1106 | sb = root->d_sb; |
1107 | BUG_ON(!sb); | |
1108 | WARN_ON(!sb->s_bdi); | |
6c510389 | 1109 | WARN_ON(sb->s_bdi == &default_backing_dev_info); |
9d412a43 | 1110 | sb->s_flags |= MS_BORN; |
454e2398 | 1111 | |
9d412a43 | 1112 | error = security_sb_kern_mount(sb, flags, secdata); |
5129a469 JE |
1113 | if (error) |
1114 | goto out_sb; | |
454e2398 | 1115 | |
42cb56ae JL |
1116 | /* |
1117 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE | |
1118 | * but s_maxbytes was an unsigned long long for many releases. Throw | |
1119 | * this warning for a little while to try and catch filesystems that | |
4358b567 | 1120 | * violate this rule. |
42cb56ae | 1121 | */ |
9d412a43 AV |
1122 | WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " |
1123 | "negative value (%lld)\n", type->name, sb->s_maxbytes); | |
42cb56ae | 1124 | |
9d412a43 | 1125 | up_write(&sb->s_umount); |
8680e22f | 1126 | free_secdata(secdata); |
9d412a43 | 1127 | return root; |
1da177e4 | 1128 | out_sb: |
9d412a43 AV |
1129 | dput(root); |
1130 | deactivate_locked_super(sb); | |
1da177e4 LT |
1131 | out_free_secdata: |
1132 | free_secdata(secdata); | |
1da177e4 | 1133 | out: |
454e2398 | 1134 | return ERR_PTR(error); |
1da177e4 LT |
1135 | } |
1136 | ||
5accdf82 JK |
1137 | /* |
1138 | * This is an internal function, please use sb_end_{write,pagefault,intwrite} | |
1139 | * instead. | |
1140 | */ | |
1141 | void __sb_end_write(struct super_block *sb, int level) | |
1142 | { | |
1143 | percpu_counter_dec(&sb->s_writers.counter[level-1]); | |
1144 | /* | |
1145 | * Make sure s_writers are updated before we wake up waiters in | |
1146 | * freeze_super(). | |
1147 | */ | |
1148 | smp_mb(); | |
1149 | if (waitqueue_active(&sb->s_writers.wait)) | |
1150 | wake_up(&sb->s_writers.wait); | |
1151 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _RET_IP_); | |
1152 | } | |
1153 | EXPORT_SYMBOL(__sb_end_write); | |
1154 | ||
1155 | #ifdef CONFIG_LOCKDEP | |
1156 | /* | |
1157 | * We want lockdep to tell us about possible deadlocks with freezing but | |
1158 | * it's it bit tricky to properly instrument it. Getting a freeze protection | |
1159 | * works as getting a read lock but there are subtle problems. XFS for example | |
1160 | * gets freeze protection on internal level twice in some cases, which is OK | |
1161 | * only because we already hold a freeze protection also on higher level. Due | |
1162 | * to these cases we have to tell lockdep we are doing trylock when we | |
1163 | * already hold a freeze protection for a higher freeze level. | |
1164 | */ | |
1165 | static void acquire_freeze_lock(struct super_block *sb, int level, bool trylock, | |
1166 | unsigned long ip) | |
1167 | { | |
1168 | int i; | |
1169 | ||
1170 | if (!trylock) { | |
1171 | for (i = 0; i < level - 1; i++) | |
1172 | if (lock_is_held(&sb->s_writers.lock_map[i])) { | |
1173 | trylock = true; | |
1174 | break; | |
1175 | } | |
1176 | } | |
1177 | rwsem_acquire_read(&sb->s_writers.lock_map[level-1], 0, trylock, ip); | |
1178 | } | |
1179 | #endif | |
1180 | ||
1181 | /* | |
1182 | * This is an internal function, please use sb_start_{write,pagefault,intwrite} | |
1183 | * instead. | |
1184 | */ | |
1185 | int __sb_start_write(struct super_block *sb, int level, bool wait) | |
1186 | { | |
1187 | retry: | |
1188 | if (unlikely(sb->s_writers.frozen >= level)) { | |
1189 | if (!wait) | |
1190 | return 0; | |
1191 | wait_event(sb->s_writers.wait_unfrozen, | |
1192 | sb->s_writers.frozen < level); | |
1193 | } | |
1194 | ||
1195 | #ifdef CONFIG_LOCKDEP | |
1196 | acquire_freeze_lock(sb, level, !wait, _RET_IP_); | |
1197 | #endif | |
1198 | percpu_counter_inc(&sb->s_writers.counter[level-1]); | |
1199 | /* | |
1200 | * Make sure counter is updated before we check for frozen. | |
1201 | * freeze_super() first sets frozen and then checks the counter. | |
1202 | */ | |
1203 | smp_mb(); | |
1204 | if (unlikely(sb->s_writers.frozen >= level)) { | |
1205 | __sb_end_write(sb, level); | |
1206 | goto retry; | |
1207 | } | |
1208 | return 1; | |
1209 | } | |
1210 | EXPORT_SYMBOL(__sb_start_write); | |
1211 | ||
1212 | /** | |
1213 | * sb_wait_write - wait until all writers to given file system finish | |
1214 | * @sb: the super for which we wait | |
1215 | * @level: type of writers we wait for (normal vs page fault) | |
1216 | * | |
1217 | * This function waits until there are no writers of given type to given file | |
1218 | * system. Caller of this function should make sure there can be no new writers | |
1219 | * of type @level before calling this function. Otherwise this function can | |
1220 | * livelock. | |
1221 | */ | |
1222 | static void sb_wait_write(struct super_block *sb, int level) | |
1223 | { | |
1224 | s64 writers; | |
1225 | ||
1226 | /* | |
1227 | * We just cycle-through lockdep here so that it does not complain | |
1228 | * about returning with lock to userspace | |
1229 | */ | |
1230 | rwsem_acquire(&sb->s_writers.lock_map[level-1], 0, 0, _THIS_IP_); | |
1231 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _THIS_IP_); | |
1232 | ||
1233 | do { | |
1234 | DEFINE_WAIT(wait); | |
1235 | ||
1236 | /* | |
1237 | * We use a barrier in prepare_to_wait() to separate setting | |
1238 | * of frozen and checking of the counter | |
1239 | */ | |
1240 | prepare_to_wait(&sb->s_writers.wait, &wait, | |
1241 | TASK_UNINTERRUPTIBLE); | |
1242 | ||
1243 | writers = percpu_counter_sum(&sb->s_writers.counter[level-1]); | |
1244 | if (writers) | |
1245 | schedule(); | |
1246 | ||
1247 | finish_wait(&sb->s_writers.wait, &wait); | |
1248 | } while (writers); | |
1249 | } | |
1250 | ||
18e9e510 | 1251 | /** |
7000d3c4 RD |
1252 | * freeze_super - lock the filesystem and force it into a consistent state |
1253 | * @sb: the super to lock | |
18e9e510 JB |
1254 | * |
1255 | * Syncs the super to make sure the filesystem is consistent and calls the fs's | |
1256 | * freeze_fs. Subsequent calls to this without first thawing the fs will return | |
1257 | * -EBUSY. | |
5accdf82 JK |
1258 | * |
1259 | * During this function, sb->s_writers.frozen goes through these values: | |
1260 | * | |
1261 | * SB_UNFROZEN: File system is normal, all writes progress as usual. | |
1262 | * | |
1263 | * SB_FREEZE_WRITE: The file system is in the process of being frozen. New | |
1264 | * writes should be blocked, though page faults are still allowed. We wait for | |
1265 | * all writes to complete and then proceed to the next stage. | |
1266 | * | |
1267 | * SB_FREEZE_PAGEFAULT: Freezing continues. Now also page faults are blocked | |
1268 | * but internal fs threads can still modify the filesystem (although they | |
1269 | * should not dirty new pages or inodes), writeback can run etc. After waiting | |
1270 | * for all running page faults we sync the filesystem which will clean all | |
1271 | * dirty pages and inodes (no new dirty pages or inodes can be created when | |
1272 | * sync is running). | |
1273 | * | |
1274 | * SB_FREEZE_FS: The file system is frozen. Now all internal sources of fs | |
1275 | * modification are blocked (e.g. XFS preallocation truncation on inode | |
1276 | * reclaim). This is usually implemented by blocking new transactions for | |
1277 | * filesystems that have them and need this additional guard. After all | |
1278 | * internal writers are finished we call ->freeze_fs() to finish filesystem | |
1279 | * freezing. Then we transition to SB_FREEZE_COMPLETE state. This state is | |
1280 | * mostly auxiliary for filesystems to verify they do not modify frozen fs. | |
1281 | * | |
1282 | * sb->s_writers.frozen is protected by sb->s_umount. | |
18e9e510 JB |
1283 | */ |
1284 | int freeze_super(struct super_block *sb) | |
1285 | { | |
1286 | int ret; | |
1287 | ||
1288 | atomic_inc(&sb->s_active); | |
1289 | down_write(&sb->s_umount); | |
5accdf82 | 1290 | if (sb->s_writers.frozen != SB_UNFROZEN) { |
18e9e510 JB |
1291 | deactivate_locked_super(sb); |
1292 | return -EBUSY; | |
1293 | } | |
1294 | ||
dabe0dc1 AV |
1295 | if (!(sb->s_flags & MS_BORN)) { |
1296 | up_write(&sb->s_umount); | |
1297 | return 0; /* sic - it's "nothing to do" */ | |
1298 | } | |
1299 | ||
18e9e510 | 1300 | if (sb->s_flags & MS_RDONLY) { |
5accdf82 JK |
1301 | /* Nothing to do really... */ |
1302 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
18e9e510 JB |
1303 | up_write(&sb->s_umount); |
1304 | return 0; | |
1305 | } | |
1306 | ||
5accdf82 JK |
1307 | /* From now on, no new normal writers can start */ |
1308 | sb->s_writers.frozen = SB_FREEZE_WRITE; | |
1309 | smp_wmb(); | |
1310 | ||
1311 | /* Release s_umount to preserve sb_start_write -> s_umount ordering */ | |
1312 | up_write(&sb->s_umount); | |
1313 | ||
1314 | sb_wait_write(sb, SB_FREEZE_WRITE); | |
1315 | ||
1316 | /* Now we go and block page faults... */ | |
1317 | down_write(&sb->s_umount); | |
1318 | sb->s_writers.frozen = SB_FREEZE_PAGEFAULT; | |
18e9e510 JB |
1319 | smp_wmb(); |
1320 | ||
5accdf82 JK |
1321 | sb_wait_write(sb, SB_FREEZE_PAGEFAULT); |
1322 | ||
1323 | /* All writers are done so after syncing there won't be dirty data */ | |
18e9e510 JB |
1324 | sync_filesystem(sb); |
1325 | ||
5accdf82 JK |
1326 | /* Now wait for internal filesystem counter */ |
1327 | sb->s_writers.frozen = SB_FREEZE_FS; | |
18e9e510 | 1328 | smp_wmb(); |
5accdf82 | 1329 | sb_wait_write(sb, SB_FREEZE_FS); |
18e9e510 | 1330 | |
18e9e510 JB |
1331 | if (sb->s_op->freeze_fs) { |
1332 | ret = sb->s_op->freeze_fs(sb); | |
1333 | if (ret) { | |
1334 | printk(KERN_ERR | |
1335 | "VFS:Filesystem freeze failed\n"); | |
5accdf82 | 1336 | sb->s_writers.frozen = SB_UNFROZEN; |
e1616300 | 1337 | smp_wmb(); |
5accdf82 | 1338 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 JB |
1339 | deactivate_locked_super(sb); |
1340 | return ret; | |
1341 | } | |
1342 | } | |
5accdf82 JK |
1343 | /* |
1344 | * This is just for debugging purposes so that fs can warn if it | |
1345 | * sees write activity when frozen is set to SB_FREEZE_COMPLETE. | |
1346 | */ | |
1347 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
18e9e510 JB |
1348 | up_write(&sb->s_umount); |
1349 | return 0; | |
1350 | } | |
1351 | EXPORT_SYMBOL(freeze_super); | |
1352 | ||
1353 | /** | |
1354 | * thaw_super -- unlock filesystem | |
1355 | * @sb: the super to thaw | |
1356 | * | |
1357 | * Unlocks the filesystem and marks it writeable again after freeze_super(). | |
1358 | */ | |
1359 | int thaw_super(struct super_block *sb) | |
1360 | { | |
1361 | int error; | |
1362 | ||
1363 | down_write(&sb->s_umount); | |
5accdf82 | 1364 | if (sb->s_writers.frozen == SB_UNFROZEN) { |
18e9e510 JB |
1365 | up_write(&sb->s_umount); |
1366 | return -EINVAL; | |
1367 | } | |
1368 | ||
1369 | if (sb->s_flags & MS_RDONLY) | |
1370 | goto out; | |
1371 | ||
1372 | if (sb->s_op->unfreeze_fs) { | |
1373 | error = sb->s_op->unfreeze_fs(sb); | |
1374 | if (error) { | |
1375 | printk(KERN_ERR | |
1376 | "VFS:Filesystem thaw failed\n"); | |
18e9e510 JB |
1377 | up_write(&sb->s_umount); |
1378 | return error; | |
1379 | } | |
1380 | } | |
1381 | ||
1382 | out: | |
5accdf82 | 1383 | sb->s_writers.frozen = SB_UNFROZEN; |
18e9e510 | 1384 | smp_wmb(); |
5accdf82 | 1385 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 JB |
1386 | deactivate_locked_super(sb); |
1387 | ||
1388 | return 0; | |
1389 | } | |
1390 | EXPORT_SYMBOL(thaw_super); |