]>
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 | */ | |
56 | static int prune_super(struct shrinker *shrink, struct shrink_control *sc) | |
57 | { | |
58 | struct super_block *sb; | |
0e1fdafd DC |
59 | int fs_objects = 0; |
60 | int total_objects; | |
b0d40c92 DC |
61 | |
62 | sb = container_of(shrink, struct super_block, s_shrink); | |
63 | ||
64 | /* | |
65 | * Deadlock avoidance. We may hold various FS locks, and we don't want | |
66 | * to recurse into the FS that called us in clear_inode() and friends.. | |
67 | */ | |
68 | if (sc->nr_to_scan && !(sc->gfp_mask & __GFP_FS)) | |
69 | return -1; | |
70 | ||
71 | if (!grab_super_passive(sb)) | |
8e125cd8 | 72 | return -1; |
b0d40c92 | 73 | |
d0407903 | 74 | if (sb->s_op->nr_cached_objects) |
0e1fdafd DC |
75 | fs_objects = sb->s_op->nr_cached_objects(sb); |
76 | ||
77 | total_objects = sb->s_nr_dentry_unused + | |
78 | sb->s_nr_inodes_unused + fs_objects + 1; | |
79 | ||
b0d40c92 | 80 | if (sc->nr_to_scan) { |
0e1fdafd DC |
81 | int dentries; |
82 | int inodes; | |
b0d40c92 | 83 | |
0e1fdafd | 84 | /* proportion the scan between the caches */ |
55f841ce GC |
85 | dentries = mult_frac(sc->nr_to_scan, sb->s_nr_dentry_unused, |
86 | total_objects); | |
87 | inodes = mult_frac(sc->nr_to_scan, sb->s_nr_inodes_unused, | |
88 | total_objects); | |
0e1fdafd | 89 | if (fs_objects) |
55f841ce GC |
90 | fs_objects = mult_frac(sc->nr_to_scan, fs_objects, |
91 | total_objects); | |
0e1fdafd DC |
92 | /* |
93 | * prune the dcache first as the icache is pinned by it, then | |
94 | * prune the icache, followed by the filesystem specific caches | |
95 | */ | |
96 | prune_dcache_sb(sb, dentries); | |
97 | prune_icache_sb(sb, inodes); | |
b0d40c92 | 98 | |
0e1fdafd DC |
99 | if (fs_objects && sb->s_op->free_cached_objects) { |
100 | sb->s_op->free_cached_objects(sb, fs_objects); | |
101 | fs_objects = sb->s_op->nr_cached_objects(sb); | |
102 | } | |
103 | total_objects = sb->s_nr_dentry_unused + | |
104 | sb->s_nr_inodes_unused + fs_objects; | |
b0d40c92 DC |
105 | } |
106 | ||
55f841ce | 107 | total_objects = vfs_pressure_ratio(total_objects); |
b0d40c92 | 108 | drop_super(sb); |
0e1fdafd | 109 | return total_objects; |
b0d40c92 DC |
110 | } |
111 | ||
5accdf82 JK |
112 | static int init_sb_writers(struct super_block *s, struct file_system_type *type) |
113 | { | |
114 | int err; | |
115 | int i; | |
116 | ||
117 | for (i = 0; i < SB_FREEZE_LEVELS; i++) { | |
118 | err = percpu_counter_init(&s->s_writers.counter[i], 0); | |
119 | if (err < 0) | |
120 | goto err_out; | |
121 | lockdep_init_map(&s->s_writers.lock_map[i], sb_writers_name[i], | |
122 | &type->s_writers_key[i], 0); | |
123 | } | |
124 | init_waitqueue_head(&s->s_writers.wait); | |
125 | init_waitqueue_head(&s->s_writers.wait_unfrozen); | |
126 | return 0; | |
127 | err_out: | |
128 | while (--i >= 0) | |
129 | percpu_counter_destroy(&s->s_writers.counter[i]); | |
130 | return err; | |
131 | } | |
132 | ||
133 | static void destroy_sb_writers(struct super_block *s) | |
134 | { | |
135 | int i; | |
136 | ||
137 | for (i = 0; i < SB_FREEZE_LEVELS; i++) | |
138 | percpu_counter_destroy(&s->s_writers.counter[i]); | |
139 | } | |
140 | ||
1da177e4 LT |
141 | /** |
142 | * alloc_super - create new superblock | |
fe2bbc48 | 143 | * @type: filesystem type superblock should belong to |
9249e17f | 144 | * @flags: the mount flags |
1da177e4 LT |
145 | * |
146 | * Allocates and initializes a new &struct super_block. alloc_super() | |
147 | * returns a pointer new superblock or %NULL if allocation had failed. | |
148 | */ | |
9249e17f | 149 | static struct super_block *alloc_super(struct file_system_type *type, int flags) |
1da177e4 | 150 | { |
11b0b5ab | 151 | struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER); |
b87221de | 152 | static const struct super_operations default_op; |
1da177e4 LT |
153 | |
154 | if (s) { | |
7b7a8665 CH |
155 | if (security_sb_alloc(s)) |
156 | goto out_free_sb; | |
157 | ||
6416ccb7 NP |
158 | #ifdef CONFIG_SMP |
159 | s->s_files = alloc_percpu(struct list_head); | |
5accdf82 JK |
160 | if (!s->s_files) |
161 | goto err_out; | |
162 | else { | |
6416ccb7 NP |
163 | int i; |
164 | ||
165 | for_each_possible_cpu(i) | |
166 | INIT_LIST_HEAD(per_cpu_ptr(s->s_files, i)); | |
167 | } | |
168 | #else | |
1da177e4 | 169 | INIT_LIST_HEAD(&s->s_files); |
6416ccb7 | 170 | #endif |
5accdf82 JK |
171 | if (init_sb_writers(s, type)) |
172 | goto err_out; | |
9249e17f | 173 | s->s_flags = flags; |
95f28604 | 174 | s->s_bdi = &default_backing_dev_info; |
a5166169 | 175 | INIT_HLIST_NODE(&s->s_instances); |
ceb5bdc2 | 176 | INIT_HLIST_BL_HEAD(&s->s_anon); |
1da177e4 | 177 | INIT_LIST_HEAD(&s->s_inodes); |
da3bbdd4 | 178 | INIT_LIST_HEAD(&s->s_dentry_lru); |
19156840 | 179 | spin_lock_init(&s->s_dentry_lru_lock); |
98b745c6 | 180 | INIT_LIST_HEAD(&s->s_inode_lru); |
09cc9fc7 | 181 | spin_lock_init(&s->s_inode_lru_lock); |
39f7c4db | 182 | INIT_LIST_HEAD(&s->s_mounts); |
1da177e4 | 183 | init_rwsem(&s->s_umount); |
897c6ff9 | 184 | lockdep_set_class(&s->s_umount, &type->s_umount_key); |
ada723dc PZ |
185 | /* |
186 | * sget() can have s_umount recursion. | |
187 | * | |
188 | * When it cannot find a suitable sb, it allocates a new | |
189 | * one (this one), and tries again to find a suitable old | |
190 | * one. | |
191 | * | |
192 | * In case that succeeds, it will acquire the s_umount | |
193 | * lock of the old one. Since these are clearly distrinct | |
194 | * locks, and this object isn't exposed yet, there's no | |
195 | * risk of deadlocks. | |
196 | * | |
197 | * Annotate this by putting this lock in a different | |
198 | * subclass. | |
199 | */ | |
200 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); | |
b20bd1a5 | 201 | s->s_count = 1; |
1da177e4 | 202 | atomic_set(&s->s_active, 1); |
a11f3a05 | 203 | mutex_init(&s->s_vfs_rename_mutex); |
51ee049e | 204 | lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key); |
d3be915f IM |
205 | mutex_init(&s->s_dquot.dqio_mutex); |
206 | mutex_init(&s->s_dquot.dqonoff_mutex); | |
1da177e4 | 207 | init_rwsem(&s->s_dquot.dqptr_sem); |
1da177e4 | 208 | s->s_maxbytes = MAX_NON_LFS; |
1da177e4 LT |
209 | s->s_op = &default_op; |
210 | s->s_time_gran = 1000000000; | |
c515e1fd | 211 | s->cleancache_poolid = -1; |
b0d40c92 DC |
212 | |
213 | s->s_shrink.seeks = DEFAULT_SEEKS; | |
214 | s->s_shrink.shrink = prune_super; | |
8ab47664 | 215 | s->s_shrink.batch = 1024; |
1da177e4 LT |
216 | } |
217 | out: | |
218 | return s; | |
5accdf82 JK |
219 | err_out: |
220 | security_sb_free(s); | |
221 | #ifdef CONFIG_SMP | |
222 | if (s->s_files) | |
223 | free_percpu(s->s_files); | |
224 | #endif | |
225 | destroy_sb_writers(s); | |
7b7a8665 | 226 | out_free_sb: |
5accdf82 JK |
227 | kfree(s); |
228 | s = NULL; | |
229 | goto out; | |
1da177e4 LT |
230 | } |
231 | ||
232 | /** | |
233 | * destroy_super - frees a superblock | |
234 | * @s: superblock to free | |
235 | * | |
236 | * Frees a superblock. | |
237 | */ | |
238 | static inline void destroy_super(struct super_block *s) | |
239 | { | |
6416ccb7 NP |
240 | #ifdef CONFIG_SMP |
241 | free_percpu(s->s_files); | |
242 | #endif | |
5accdf82 | 243 | destroy_sb_writers(s); |
1da177e4 | 244 | security_sb_free(s); |
39f7c4db | 245 | WARN_ON(!list_empty(&s->s_mounts)); |
79c0b2df | 246 | kfree(s->s_subtype); |
b3b304a2 | 247 | kfree(s->s_options); |
1da177e4 LT |
248 | kfree(s); |
249 | } | |
250 | ||
251 | /* Superblock refcounting */ | |
252 | ||
253 | /* | |
35cf7ba0 | 254 | * Drop a superblock's refcount. The caller must hold sb_lock. |
1da177e4 | 255 | */ |
f47ec3f2 | 256 | static void __put_super(struct super_block *sb) |
1da177e4 | 257 | { |
1da177e4 | 258 | if (!--sb->s_count) { |
551de6f3 | 259 | list_del_init(&sb->s_list); |
1da177e4 | 260 | destroy_super(sb); |
1da177e4 | 261 | } |
1da177e4 LT |
262 | } |
263 | ||
264 | /** | |
265 | * put_super - drop a temporary reference to superblock | |
266 | * @sb: superblock in question | |
267 | * | |
268 | * Drops a temporary reference, frees superblock if there's no | |
269 | * references left. | |
270 | */ | |
f47ec3f2 | 271 | static void put_super(struct super_block *sb) |
1da177e4 LT |
272 | { |
273 | spin_lock(&sb_lock); | |
274 | __put_super(sb); | |
275 | spin_unlock(&sb_lock); | |
276 | } | |
277 | ||
278 | ||
279 | /** | |
1712ac8f | 280 | * deactivate_locked_super - drop an active reference to superblock |
1da177e4 LT |
281 | * @s: superblock to deactivate |
282 | * | |
1712ac8f AV |
283 | * Drops an active reference to superblock, converting it into a temprory |
284 | * one if there is no other active references left. In that case we | |
1da177e4 LT |
285 | * tell fs driver to shut it down and drop the temporary reference we |
286 | * had just acquired. | |
1712ac8f AV |
287 | * |
288 | * Caller holds exclusive lock on superblock; that lock is released. | |
1da177e4 | 289 | */ |
1712ac8f | 290 | void deactivate_locked_super(struct super_block *s) |
1da177e4 LT |
291 | { |
292 | struct file_system_type *fs = s->s_type; | |
b20bd1a5 | 293 | if (atomic_dec_and_test(&s->s_active)) { |
3167760f | 294 | cleancache_invalidate_fs(s); |
1da177e4 | 295 | fs->kill_sb(s); |
b0d40c92 DC |
296 | |
297 | /* caches are now gone, we can safely kill the shrinker now */ | |
298 | unregister_shrinker(&s->s_shrink); | |
1da177e4 LT |
299 | put_filesystem(fs); |
300 | put_super(s); | |
1712ac8f AV |
301 | } else { |
302 | up_write(&s->s_umount); | |
1da177e4 LT |
303 | } |
304 | } | |
305 | ||
1712ac8f | 306 | EXPORT_SYMBOL(deactivate_locked_super); |
1da177e4 | 307 | |
74dbbdd7 | 308 | /** |
1712ac8f | 309 | * deactivate_super - drop an active reference to superblock |
74dbbdd7 AV |
310 | * @s: superblock to deactivate |
311 | * | |
1712ac8f AV |
312 | * Variant of deactivate_locked_super(), except that superblock is *not* |
313 | * locked by caller. If we are going to drop the final active reference, | |
314 | * lock will be acquired prior to that. | |
74dbbdd7 | 315 | */ |
1712ac8f | 316 | void deactivate_super(struct super_block *s) |
74dbbdd7 | 317 | { |
1712ac8f AV |
318 | if (!atomic_add_unless(&s->s_active, -1, 1)) { |
319 | down_write(&s->s_umount); | |
320 | deactivate_locked_super(s); | |
74dbbdd7 AV |
321 | } |
322 | } | |
323 | ||
1712ac8f | 324 | EXPORT_SYMBOL(deactivate_super); |
74dbbdd7 | 325 | |
1da177e4 LT |
326 | /** |
327 | * grab_super - acquire an active reference | |
328 | * @s: reference we are trying to make active | |
329 | * | |
330 | * Tries to acquire an active reference. grab_super() is used when we | |
331 | * had just found a superblock in super_blocks or fs_type->fs_supers | |
332 | * and want to turn it into a full-blown active reference. grab_super() | |
333 | * is called with sb_lock held and drops it. Returns 1 in case of | |
334 | * success, 0 if we had failed (superblock contents was already dead or | |
acfec9a5 AV |
335 | * dying when grab_super() had been called). Note that this is only |
336 | * called for superblocks not in rundown mode (== ones still on ->fs_supers | |
337 | * of their type), so increment of ->s_count is OK here. | |
1da177e4 | 338 | */ |
9c4dbee7 | 339 | static int grab_super(struct super_block *s) __releases(sb_lock) |
1da177e4 LT |
340 | { |
341 | s->s_count++; | |
342 | spin_unlock(&sb_lock); | |
343 | down_write(&s->s_umount); | |
acfec9a5 AV |
344 | if ((s->s_flags & MS_BORN) && atomic_inc_not_zero(&s->s_active)) { |
345 | put_super(s); | |
346 | return 1; | |
347 | } | |
1da177e4 LT |
348 | up_write(&s->s_umount); |
349 | put_super(s); | |
1da177e4 LT |
350 | return 0; |
351 | } | |
352 | ||
12ad3ab6 DC |
353 | /* |
354 | * grab_super_passive - acquire a passive reference | |
331cbdee | 355 | * @sb: reference we are trying to grab |
12ad3ab6 DC |
356 | * |
357 | * Tries to acquire a passive reference. This is used in places where we | |
358 | * cannot take an active reference but we need to ensure that the | |
359 | * superblock does not go away while we are working on it. It returns | |
360 | * false if a reference was not gained, and returns true with the s_umount | |
361 | * lock held in read mode if a reference is gained. On successful return, | |
362 | * the caller must drop the s_umount lock and the passive reference when | |
363 | * done. | |
364 | */ | |
365 | bool grab_super_passive(struct super_block *sb) | |
366 | { | |
367 | spin_lock(&sb_lock); | |
a5166169 | 368 | if (hlist_unhashed(&sb->s_instances)) { |
12ad3ab6 DC |
369 | spin_unlock(&sb_lock); |
370 | return false; | |
371 | } | |
372 | ||
373 | sb->s_count++; | |
374 | spin_unlock(&sb_lock); | |
375 | ||
376 | if (down_read_trylock(&sb->s_umount)) { | |
dabe0dc1 | 377 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
12ad3ab6 DC |
378 | return true; |
379 | up_read(&sb->s_umount); | |
380 | } | |
381 | ||
382 | put_super(sb); | |
383 | return false; | |
384 | } | |
385 | ||
1da177e4 LT |
386 | /** |
387 | * generic_shutdown_super - common helper for ->kill_sb() | |
388 | * @sb: superblock to kill | |
389 | * | |
390 | * generic_shutdown_super() does all fs-independent work on superblock | |
391 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects | |
392 | * that need destruction out of superblock, call generic_shutdown_super() | |
393 | * and release aforementioned objects. Note: dentries and inodes _are_ | |
394 | * taken care of and do not need specific handling. | |
c636ebdb DH |
395 | * |
396 | * Upon calling this function, the filesystem may no longer alter or | |
397 | * rearrange the set of dentries belonging to this super_block, nor may it | |
398 | * change the attachments of dentries to inodes. | |
1da177e4 LT |
399 | */ |
400 | void generic_shutdown_super(struct super_block *sb) | |
401 | { | |
ee9b6d61 | 402 | const struct super_operations *sop = sb->s_op; |
1da177e4 | 403 | |
c636ebdb DH |
404 | if (sb->s_root) { |
405 | shrink_dcache_for_umount(sb); | |
60b0680f | 406 | sync_filesystem(sb); |
1da177e4 | 407 | sb->s_flags &= ~MS_ACTIVE; |
efaee192 | 408 | |
63997e98 AV |
409 | fsnotify_unmount_inodes(&sb->s_inodes); |
410 | ||
411 | evict_inodes(sb); | |
1da177e4 | 412 | |
7b7a8665 CH |
413 | if (sb->s_dio_done_wq) { |
414 | destroy_workqueue(sb->s_dio_done_wq); | |
415 | sb->s_dio_done_wq = NULL; | |
416 | } | |
417 | ||
1da177e4 LT |
418 | if (sop->put_super) |
419 | sop->put_super(sb); | |
420 | ||
63997e98 | 421 | if (!list_empty(&sb->s_inodes)) { |
7b4fe29e DJ |
422 | printk("VFS: Busy inodes after unmount of %s. " |
423 | "Self-destruct in 5 seconds. Have a nice day...\n", | |
424 | sb->s_id); | |
1da177e4 | 425 | } |
1da177e4 LT |
426 | } |
427 | spin_lock(&sb_lock); | |
428 | /* should be initialized for __put_super_and_need_restart() */ | |
a5166169 | 429 | hlist_del_init(&sb->s_instances); |
1da177e4 LT |
430 | spin_unlock(&sb_lock); |
431 | up_write(&sb->s_umount); | |
432 | } | |
433 | ||
434 | EXPORT_SYMBOL(generic_shutdown_super); | |
435 | ||
436 | /** | |
437 | * sget - find or create a superblock | |
438 | * @type: filesystem type superblock should belong to | |
439 | * @test: comparison callback | |
440 | * @set: setup callback | |
9249e17f | 441 | * @flags: mount flags |
1da177e4 LT |
442 | * @data: argument to each of them |
443 | */ | |
444 | struct super_block *sget(struct file_system_type *type, | |
445 | int (*test)(struct super_block *,void *), | |
446 | int (*set)(struct super_block *,void *), | |
9249e17f | 447 | int flags, |
1da177e4 LT |
448 | void *data) |
449 | { | |
450 | struct super_block *s = NULL; | |
d4730127 | 451 | struct super_block *old; |
1da177e4 LT |
452 | int err; |
453 | ||
454 | retry: | |
455 | spin_lock(&sb_lock); | |
d4730127 | 456 | if (test) { |
b67bfe0d | 457 | hlist_for_each_entry(old, &type->fs_supers, s_instances) { |
d4730127 MK |
458 | if (!test(old, data)) |
459 | continue; | |
460 | if (!grab_super(old)) | |
461 | goto retry; | |
a3cfbb53 LZ |
462 | if (s) { |
463 | up_write(&s->s_umount); | |
d4730127 | 464 | destroy_super(s); |
7a4dec53 | 465 | s = NULL; |
a3cfbb53 | 466 | } |
d4730127 MK |
467 | return old; |
468 | } | |
1da177e4 LT |
469 | } |
470 | if (!s) { | |
471 | spin_unlock(&sb_lock); | |
9249e17f | 472 | s = alloc_super(type, flags); |
1da177e4 LT |
473 | if (!s) |
474 | return ERR_PTR(-ENOMEM); | |
475 | goto retry; | |
476 | } | |
477 | ||
478 | err = set(s, data); | |
479 | if (err) { | |
480 | spin_unlock(&sb_lock); | |
a3cfbb53 | 481 | up_write(&s->s_umount); |
1da177e4 LT |
482 | destroy_super(s); |
483 | return ERR_PTR(err); | |
484 | } | |
485 | s->s_type = type; | |
486 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); | |
487 | list_add_tail(&s->s_list, &super_blocks); | |
a5166169 | 488 | hlist_add_head(&s->s_instances, &type->fs_supers); |
1da177e4 LT |
489 | spin_unlock(&sb_lock); |
490 | get_filesystem(type); | |
b0d40c92 | 491 | register_shrinker(&s->s_shrink); |
1da177e4 LT |
492 | return s; |
493 | } | |
494 | ||
495 | EXPORT_SYMBOL(sget); | |
496 | ||
497 | void drop_super(struct super_block *sb) | |
498 | { | |
499 | up_read(&sb->s_umount); | |
500 | put_super(sb); | |
501 | } | |
502 | ||
503 | EXPORT_SYMBOL(drop_super); | |
504 | ||
01a05b33 AV |
505 | /** |
506 | * iterate_supers - call function for all active superblocks | |
507 | * @f: function to call | |
508 | * @arg: argument to pass to it | |
509 | * | |
510 | * Scans the superblock list and calls given function, passing it | |
511 | * locked superblock and given argument. | |
512 | */ | |
513 | void iterate_supers(void (*f)(struct super_block *, void *), void *arg) | |
514 | { | |
dca33252 | 515 | struct super_block *sb, *p = NULL; |
01a05b33 AV |
516 | |
517 | spin_lock(&sb_lock); | |
dca33252 | 518 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 519 | if (hlist_unhashed(&sb->s_instances)) |
01a05b33 AV |
520 | continue; |
521 | sb->s_count++; | |
522 | spin_unlock(&sb_lock); | |
523 | ||
524 | down_read(&sb->s_umount); | |
dabe0dc1 | 525 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
01a05b33 AV |
526 | f(sb, arg); |
527 | up_read(&sb->s_umount); | |
528 | ||
529 | spin_lock(&sb_lock); | |
dca33252 AV |
530 | if (p) |
531 | __put_super(p); | |
532 | p = sb; | |
01a05b33 | 533 | } |
dca33252 AV |
534 | if (p) |
535 | __put_super(p); | |
01a05b33 AV |
536 | spin_unlock(&sb_lock); |
537 | } | |
538 | ||
43e15cdb AV |
539 | /** |
540 | * iterate_supers_type - call function for superblocks of given type | |
541 | * @type: fs type | |
542 | * @f: function to call | |
543 | * @arg: argument to pass to it | |
544 | * | |
545 | * Scans the superblock list and calls given function, passing it | |
546 | * locked superblock and given argument. | |
547 | */ | |
548 | void iterate_supers_type(struct file_system_type *type, | |
549 | void (*f)(struct super_block *, void *), void *arg) | |
550 | { | |
551 | struct super_block *sb, *p = NULL; | |
552 | ||
553 | spin_lock(&sb_lock); | |
b67bfe0d | 554 | hlist_for_each_entry(sb, &type->fs_supers, s_instances) { |
43e15cdb AV |
555 | sb->s_count++; |
556 | spin_unlock(&sb_lock); | |
557 | ||
558 | down_read(&sb->s_umount); | |
dabe0dc1 | 559 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
43e15cdb AV |
560 | f(sb, arg); |
561 | up_read(&sb->s_umount); | |
562 | ||
563 | spin_lock(&sb_lock); | |
564 | if (p) | |
565 | __put_super(p); | |
566 | p = sb; | |
567 | } | |
568 | if (p) | |
569 | __put_super(p); | |
570 | spin_unlock(&sb_lock); | |
571 | } | |
572 | ||
573 | EXPORT_SYMBOL(iterate_supers_type); | |
574 | ||
1da177e4 LT |
575 | /** |
576 | * get_super - get the superblock of a device | |
577 | * @bdev: device to get the superblock for | |
578 | * | |
579 | * Scans the superblock list and finds the superblock of the file system | |
580 | * mounted on the device given. %NULL is returned if no match is found. | |
581 | */ | |
582 | ||
df40c01a | 583 | struct super_block *get_super(struct block_device *bdev) |
1da177e4 | 584 | { |
618f0636 KK |
585 | struct super_block *sb; |
586 | ||
1da177e4 LT |
587 | if (!bdev) |
588 | return NULL; | |
618f0636 | 589 | |
1da177e4 | 590 | spin_lock(&sb_lock); |
618f0636 KK |
591 | rescan: |
592 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 593 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 594 | continue; |
618f0636 KK |
595 | if (sb->s_bdev == bdev) { |
596 | sb->s_count++; | |
1da177e4 | 597 | spin_unlock(&sb_lock); |
618f0636 | 598 | down_read(&sb->s_umount); |
df40c01a | 599 | /* still alive? */ |
dabe0dc1 | 600 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
618f0636 KK |
601 | return sb; |
602 | up_read(&sb->s_umount); | |
df40c01a | 603 | /* nope, got unmounted */ |
618f0636 | 604 | spin_lock(&sb_lock); |
df40c01a AV |
605 | __put_super(sb); |
606 | goto rescan; | |
1da177e4 LT |
607 | } |
608 | } | |
609 | spin_unlock(&sb_lock); | |
610 | return NULL; | |
611 | } | |
612 | ||
613 | EXPORT_SYMBOL(get_super); | |
4504230a | 614 | |
6b6dc836 JK |
615 | /** |
616 | * get_super_thawed - get thawed superblock of a device | |
617 | * @bdev: device to get the superblock for | |
618 | * | |
619 | * Scans the superblock list and finds the superblock of the file system | |
620 | * mounted on the device. The superblock is returned once it is thawed | |
621 | * (or immediately if it was not frozen). %NULL is returned if no match | |
622 | * is found. | |
623 | */ | |
624 | struct super_block *get_super_thawed(struct block_device *bdev) | |
625 | { | |
626 | while (1) { | |
627 | struct super_block *s = get_super(bdev); | |
5accdf82 | 628 | if (!s || s->s_writers.frozen == SB_UNFROZEN) |
6b6dc836 JK |
629 | return s; |
630 | up_read(&s->s_umount); | |
5accdf82 JK |
631 | wait_event(s->s_writers.wait_unfrozen, |
632 | s->s_writers.frozen == SB_UNFROZEN); | |
6b6dc836 JK |
633 | put_super(s); |
634 | } | |
635 | } | |
636 | EXPORT_SYMBOL(get_super_thawed); | |
637 | ||
4504230a CH |
638 | /** |
639 | * get_active_super - get an active reference to the superblock of a device | |
640 | * @bdev: device to get the superblock for | |
641 | * | |
642 | * Scans the superblock list and finds the superblock of the file system | |
643 | * mounted on the device given. Returns the superblock with an active | |
d3f21473 | 644 | * reference or %NULL if none was found. |
4504230a CH |
645 | */ |
646 | struct super_block *get_active_super(struct block_device *bdev) | |
647 | { | |
648 | struct super_block *sb; | |
649 | ||
650 | if (!bdev) | |
651 | return NULL; | |
652 | ||
1494583d | 653 | restart: |
4504230a CH |
654 | spin_lock(&sb_lock); |
655 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 656 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 657 | continue; |
1494583d | 658 | if (sb->s_bdev == bdev) { |
acfec9a5 | 659 | if (!grab_super(sb)) |
1494583d | 660 | goto restart; |
acfec9a5 AV |
661 | up_write(&sb->s_umount); |
662 | return sb; | |
1494583d | 663 | } |
4504230a CH |
664 | } |
665 | spin_unlock(&sb_lock); | |
666 | return NULL; | |
667 | } | |
1da177e4 | 668 | |
df40c01a | 669 | struct super_block *user_get_super(dev_t dev) |
1da177e4 | 670 | { |
618f0636 | 671 | struct super_block *sb; |
1da177e4 | 672 | |
1da177e4 | 673 | spin_lock(&sb_lock); |
618f0636 KK |
674 | rescan: |
675 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 676 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 677 | continue; |
618f0636 KK |
678 | if (sb->s_dev == dev) { |
679 | sb->s_count++; | |
1da177e4 | 680 | spin_unlock(&sb_lock); |
618f0636 | 681 | down_read(&sb->s_umount); |
df40c01a | 682 | /* still alive? */ |
dabe0dc1 | 683 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
618f0636 KK |
684 | return sb; |
685 | up_read(&sb->s_umount); | |
df40c01a | 686 | /* nope, got unmounted */ |
618f0636 | 687 | spin_lock(&sb_lock); |
df40c01a AV |
688 | __put_super(sb); |
689 | goto rescan; | |
1da177e4 LT |
690 | } |
691 | } | |
692 | spin_unlock(&sb_lock); | |
693 | return NULL; | |
694 | } | |
695 | ||
1da177e4 LT |
696 | /** |
697 | * do_remount_sb - asks filesystem to change mount options. | |
698 | * @sb: superblock in question | |
699 | * @flags: numeric part of options | |
700 | * @data: the rest of options | |
701 | * @force: whether or not to force the change | |
702 | * | |
703 | * Alters the mount options of a mounted file system. | |
704 | */ | |
705 | int do_remount_sb(struct super_block *sb, int flags, void *data, int force) | |
706 | { | |
707 | int retval; | |
c79d967d | 708 | int remount_ro; |
4504230a | 709 | |
5accdf82 | 710 | if (sb->s_writers.frozen != SB_UNFROZEN) |
4504230a CH |
711 | return -EBUSY; |
712 | ||
9361401e | 713 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
714 | if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev)) |
715 | return -EACCES; | |
9361401e | 716 | #endif |
4504230a | 717 | |
1da177e4 LT |
718 | if (flags & MS_RDONLY) |
719 | acct_auto_close(sb); | |
720 | shrink_dcache_sb(sb); | |
60b0680f | 721 | sync_filesystem(sb); |
1da177e4 | 722 | |
d208bbdd | 723 | remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY); |
d208bbdd | 724 | |
1da177e4 LT |
725 | /* If we are remounting RDONLY and current sb is read/write, |
726 | make sure there are no rw files opened */ | |
d208bbdd | 727 | if (remount_ro) { |
4ed5e82f | 728 | if (force) { |
1da177e4 | 729 | mark_files_ro(sb); |
4ed5e82f MS |
730 | } else { |
731 | retval = sb_prepare_remount_readonly(sb); | |
732 | if (retval) | |
733 | return retval; | |
4ed5e82f | 734 | } |
1da177e4 LT |
735 | } |
736 | ||
737 | if (sb->s_op->remount_fs) { | |
1da177e4 | 738 | retval = sb->s_op->remount_fs(sb, &flags, data); |
2833eb2b MS |
739 | if (retval) { |
740 | if (!force) | |
4ed5e82f | 741 | goto cancel_readonly; |
2833eb2b MS |
742 | /* If forced remount, go ahead despite any errors */ |
743 | WARN(1, "forced remount of a %s fs returned %i\n", | |
744 | sb->s_type->name, retval); | |
745 | } | |
1da177e4 LT |
746 | } |
747 | sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK); | |
4ed5e82f MS |
748 | /* Needs to be ordered wrt mnt_is_readonly() */ |
749 | smp_wmb(); | |
750 | sb->s_readonly_remount = 0; | |
c79d967d | 751 | |
d208bbdd NP |
752 | /* |
753 | * Some filesystems modify their metadata via some other path than the | |
754 | * bdev buffer cache (eg. use a private mapping, or directories in | |
755 | * pagecache, etc). Also file data modifications go via their own | |
756 | * mappings. So If we try to mount readonly then copy the filesystem | |
757 | * from bdev, we could get stale data, so invalidate it to give a best | |
758 | * effort at coherency. | |
759 | */ | |
760 | if (remount_ro && sb->s_bdev) | |
761 | invalidate_bdev(sb->s_bdev); | |
1da177e4 | 762 | return 0; |
4ed5e82f MS |
763 | |
764 | cancel_readonly: | |
765 | sb->s_readonly_remount = 0; | |
766 | return retval; | |
1da177e4 LT |
767 | } |
768 | ||
a2a9537a | 769 | static void do_emergency_remount(struct work_struct *work) |
1da177e4 | 770 | { |
dca33252 | 771 | struct super_block *sb, *p = NULL; |
1da177e4 LT |
772 | |
773 | spin_lock(&sb_lock); | |
dca33252 | 774 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 775 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 776 | continue; |
1da177e4 LT |
777 | sb->s_count++; |
778 | spin_unlock(&sb_lock); | |
443b94ba | 779 | down_write(&sb->s_umount); |
dabe0dc1 AV |
780 | if (sb->s_root && sb->s_bdev && (sb->s_flags & MS_BORN) && |
781 | !(sb->s_flags & MS_RDONLY)) { | |
1da177e4 | 782 | /* |
1da177e4 LT |
783 | * What lock protects sb->s_flags?? |
784 | */ | |
1da177e4 | 785 | do_remount_sb(sb, MS_RDONLY, NULL, 1); |
1da177e4 | 786 | } |
443b94ba | 787 | up_write(&sb->s_umount); |
1da177e4 | 788 | spin_lock(&sb_lock); |
dca33252 AV |
789 | if (p) |
790 | __put_super(p); | |
791 | p = sb; | |
1da177e4 | 792 | } |
dca33252 AV |
793 | if (p) |
794 | __put_super(p); | |
1da177e4 | 795 | spin_unlock(&sb_lock); |
a2a9537a | 796 | kfree(work); |
1da177e4 LT |
797 | printk("Emergency Remount complete\n"); |
798 | } | |
799 | ||
800 | void emergency_remount(void) | |
801 | { | |
a2a9537a JA |
802 | struct work_struct *work; |
803 | ||
804 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | |
805 | if (work) { | |
806 | INIT_WORK(work, do_emergency_remount); | |
807 | schedule_work(work); | |
808 | } | |
1da177e4 LT |
809 | } |
810 | ||
811 | /* | |
812 | * Unnamed block devices are dummy devices used by virtual | |
813 | * filesystems which don't use real block-devices. -- jrs | |
814 | */ | |
815 | ||
ad76cbc6 | 816 | static DEFINE_IDA(unnamed_dev_ida); |
1da177e4 | 817 | static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */ |
c63e09ec | 818 | static int unnamed_dev_start = 0; /* don't bother trying below it */ |
1da177e4 | 819 | |
0ee5dc67 | 820 | int get_anon_bdev(dev_t *p) |
1da177e4 LT |
821 | { |
822 | int dev; | |
823 | int error; | |
824 | ||
825 | retry: | |
ad76cbc6 | 826 | if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0) |
1da177e4 LT |
827 | return -ENOMEM; |
828 | spin_lock(&unnamed_dev_lock); | |
c63e09ec | 829 | error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev); |
f21f6220 AV |
830 | if (!error) |
831 | unnamed_dev_start = dev + 1; | |
1da177e4 LT |
832 | spin_unlock(&unnamed_dev_lock); |
833 | if (error == -EAGAIN) | |
834 | /* We raced and lost with another CPU. */ | |
835 | goto retry; | |
836 | else if (error) | |
837 | return -EAGAIN; | |
838 | ||
e8c8d1bc | 839 | if (dev == (1 << MINORBITS)) { |
1da177e4 | 840 | spin_lock(&unnamed_dev_lock); |
ad76cbc6 | 841 | ida_remove(&unnamed_dev_ida, dev); |
f21f6220 AV |
842 | if (unnamed_dev_start > dev) |
843 | unnamed_dev_start = dev; | |
1da177e4 LT |
844 | spin_unlock(&unnamed_dev_lock); |
845 | return -EMFILE; | |
846 | } | |
0ee5dc67 | 847 | *p = MKDEV(0, dev & MINORMASK); |
1da177e4 LT |
848 | return 0; |
849 | } | |
0ee5dc67 | 850 | EXPORT_SYMBOL(get_anon_bdev); |
1da177e4 | 851 | |
0ee5dc67 | 852 | void free_anon_bdev(dev_t dev) |
1da177e4 | 853 | { |
0ee5dc67 | 854 | int slot = MINOR(dev); |
1da177e4 | 855 | spin_lock(&unnamed_dev_lock); |
ad76cbc6 | 856 | ida_remove(&unnamed_dev_ida, slot); |
c63e09ec AV |
857 | if (slot < unnamed_dev_start) |
858 | unnamed_dev_start = slot; | |
1da177e4 LT |
859 | spin_unlock(&unnamed_dev_lock); |
860 | } | |
0ee5dc67 AV |
861 | EXPORT_SYMBOL(free_anon_bdev); |
862 | ||
863 | int set_anon_super(struct super_block *s, void *data) | |
864 | { | |
865 | int error = get_anon_bdev(&s->s_dev); | |
866 | if (!error) | |
867 | s->s_bdi = &noop_backing_dev_info; | |
868 | return error; | |
869 | } | |
870 | ||
871 | EXPORT_SYMBOL(set_anon_super); | |
872 | ||
873 | void kill_anon_super(struct super_block *sb) | |
874 | { | |
875 | dev_t dev = sb->s_dev; | |
876 | generic_shutdown_super(sb); | |
877 | free_anon_bdev(dev); | |
878 | } | |
1da177e4 LT |
879 | |
880 | EXPORT_SYMBOL(kill_anon_super); | |
881 | ||
1da177e4 LT |
882 | void kill_litter_super(struct super_block *sb) |
883 | { | |
884 | if (sb->s_root) | |
885 | d_genocide(sb->s_root); | |
886 | kill_anon_super(sb); | |
887 | } | |
888 | ||
889 | EXPORT_SYMBOL(kill_litter_super); | |
890 | ||
909e6d94 SH |
891 | static int ns_test_super(struct super_block *sb, void *data) |
892 | { | |
893 | return sb->s_fs_info == data; | |
894 | } | |
895 | ||
896 | static int ns_set_super(struct super_block *sb, void *data) | |
897 | { | |
898 | sb->s_fs_info = data; | |
899 | return set_anon_super(sb, NULL); | |
900 | } | |
901 | ||
ceefda69 AV |
902 | struct dentry *mount_ns(struct file_system_type *fs_type, int flags, |
903 | void *data, int (*fill_super)(struct super_block *, void *, int)) | |
909e6d94 SH |
904 | { |
905 | struct super_block *sb; | |
906 | ||
9249e17f | 907 | sb = sget(fs_type, ns_test_super, ns_set_super, flags, data); |
909e6d94 | 908 | if (IS_ERR(sb)) |
ceefda69 | 909 | return ERR_CAST(sb); |
909e6d94 SH |
910 | |
911 | if (!sb->s_root) { | |
912 | int err; | |
909e6d94 SH |
913 | err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0); |
914 | if (err) { | |
74dbbdd7 | 915 | deactivate_locked_super(sb); |
ceefda69 | 916 | return ERR_PTR(err); |
909e6d94 SH |
917 | } |
918 | ||
919 | sb->s_flags |= MS_ACTIVE; | |
920 | } | |
921 | ||
ceefda69 | 922 | return dget(sb->s_root); |
909e6d94 SH |
923 | } |
924 | ||
ceefda69 | 925 | EXPORT_SYMBOL(mount_ns); |
909e6d94 | 926 | |
9361401e | 927 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
928 | static int set_bdev_super(struct super_block *s, void *data) |
929 | { | |
930 | s->s_bdev = data; | |
931 | s->s_dev = s->s_bdev->bd_dev; | |
32a88aa1 JA |
932 | |
933 | /* | |
934 | * We set the bdi here to the queue backing, file systems can | |
935 | * overwrite this in ->fill_super() | |
936 | */ | |
937 | s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info; | |
1da177e4 LT |
938 | return 0; |
939 | } | |
940 | ||
941 | static int test_bdev_super(struct super_block *s, void *data) | |
942 | { | |
943 | return (void *)s->s_bdev == data; | |
944 | } | |
945 | ||
152a0836 | 946 | struct dentry *mount_bdev(struct file_system_type *fs_type, |
1da177e4 | 947 | int flags, const char *dev_name, void *data, |
152a0836 | 948 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
949 | { |
950 | struct block_device *bdev; | |
951 | struct super_block *s; | |
d4d77629 | 952 | fmode_t mode = FMODE_READ | FMODE_EXCL; |
1da177e4 LT |
953 | int error = 0; |
954 | ||
30c40d2c AV |
955 | if (!(flags & MS_RDONLY)) |
956 | mode |= FMODE_WRITE; | |
957 | ||
d4d77629 | 958 | bdev = blkdev_get_by_path(dev_name, mode, fs_type); |
1da177e4 | 959 | if (IS_ERR(bdev)) |
152a0836 | 960 | return ERR_CAST(bdev); |
1da177e4 LT |
961 | |
962 | /* | |
963 | * once the super is inserted into the list by sget, s_umount | |
964 | * will protect the lockfs code from trying to start a snapshot | |
965 | * while we are mounting | |
966 | */ | |
4fadd7bb CH |
967 | mutex_lock(&bdev->bd_fsfreeze_mutex); |
968 | if (bdev->bd_fsfreeze_count > 0) { | |
969 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
970 | error = -EBUSY; | |
971 | goto error_bdev; | |
972 | } | |
9249e17f DH |
973 | s = sget(fs_type, test_bdev_super, set_bdev_super, flags | MS_NOSEC, |
974 | bdev); | |
4fadd7bb | 975 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
1da177e4 | 976 | if (IS_ERR(s)) |
454e2398 | 977 | goto error_s; |
1da177e4 LT |
978 | |
979 | if (s->s_root) { | |
980 | if ((flags ^ s->s_flags) & MS_RDONLY) { | |
74dbbdd7 | 981 | deactivate_locked_super(s); |
454e2398 DH |
982 | error = -EBUSY; |
983 | goto error_bdev; | |
1da177e4 | 984 | } |
454e2398 | 985 | |
4f331f01 TH |
986 | /* |
987 | * s_umount nests inside bd_mutex during | |
e525fd89 TH |
988 | * __invalidate_device(). blkdev_put() acquires |
989 | * bd_mutex and can't be called under s_umount. Drop | |
990 | * s_umount temporarily. This is safe as we're | |
991 | * holding an active reference. | |
4f331f01 TH |
992 | */ |
993 | up_write(&s->s_umount); | |
d4d77629 | 994 | blkdev_put(bdev, mode); |
4f331f01 | 995 | down_write(&s->s_umount); |
1da177e4 LT |
996 | } else { |
997 | char b[BDEVNAME_SIZE]; | |
998 | ||
30c40d2c | 999 | s->s_mode = mode; |
1da177e4 | 1000 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); |
e78c9a00 | 1001 | sb_set_blocksize(s, block_size(bdev)); |
9b04c997 | 1002 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1003 | if (error) { |
74dbbdd7 | 1004 | deactivate_locked_super(s); |
454e2398 | 1005 | goto error; |
fa675765 | 1006 | } |
454e2398 DH |
1007 | |
1008 | s->s_flags |= MS_ACTIVE; | |
87d8fe1e | 1009 | bdev->bd_super = s; |
1da177e4 LT |
1010 | } |
1011 | ||
152a0836 | 1012 | return dget(s->s_root); |
1da177e4 | 1013 | |
454e2398 DH |
1014 | error_s: |
1015 | error = PTR_ERR(s); | |
1016 | error_bdev: | |
d4d77629 | 1017 | blkdev_put(bdev, mode); |
454e2398 | 1018 | error: |
152a0836 AV |
1019 | return ERR_PTR(error); |
1020 | } | |
1021 | EXPORT_SYMBOL(mount_bdev); | |
1022 | ||
1da177e4 LT |
1023 | void kill_block_super(struct super_block *sb) |
1024 | { | |
1025 | struct block_device *bdev = sb->s_bdev; | |
30c40d2c | 1026 | fmode_t mode = sb->s_mode; |
1da177e4 | 1027 | |
ddbaaf30 | 1028 | bdev->bd_super = NULL; |
1da177e4 LT |
1029 | generic_shutdown_super(sb); |
1030 | sync_blockdev(bdev); | |
d4d77629 | 1031 | WARN_ON_ONCE(!(mode & FMODE_EXCL)); |
e525fd89 | 1032 | blkdev_put(bdev, mode | FMODE_EXCL); |
1da177e4 LT |
1033 | } |
1034 | ||
1035 | EXPORT_SYMBOL(kill_block_super); | |
9361401e | 1036 | #endif |
1da177e4 | 1037 | |
3c26ff6e | 1038 | struct dentry *mount_nodev(struct file_system_type *fs_type, |
1da177e4 | 1039 | int flags, void *data, |
3c26ff6e | 1040 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1041 | { |
1042 | int error; | |
9249e17f | 1043 | struct super_block *s = sget(fs_type, NULL, set_anon_super, flags, NULL); |
1da177e4 LT |
1044 | |
1045 | if (IS_ERR(s)) | |
3c26ff6e | 1046 | return ERR_CAST(s); |
1da177e4 | 1047 | |
9b04c997 | 1048 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1049 | if (error) { |
74dbbdd7 | 1050 | deactivate_locked_super(s); |
3c26ff6e | 1051 | return ERR_PTR(error); |
1da177e4 LT |
1052 | } |
1053 | s->s_flags |= MS_ACTIVE; | |
3c26ff6e | 1054 | return dget(s->s_root); |
1da177e4 | 1055 | } |
3c26ff6e AV |
1056 | EXPORT_SYMBOL(mount_nodev); |
1057 | ||
1da177e4 LT |
1058 | static int compare_single(struct super_block *s, void *p) |
1059 | { | |
1060 | return 1; | |
1061 | } | |
1062 | ||
fc14f2fe | 1063 | struct dentry *mount_single(struct file_system_type *fs_type, |
1da177e4 | 1064 | int flags, void *data, |
fc14f2fe | 1065 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1066 | { |
1067 | struct super_block *s; | |
1068 | int error; | |
1069 | ||
9249e17f | 1070 | s = sget(fs_type, compare_single, set_anon_super, flags, NULL); |
1da177e4 | 1071 | if (IS_ERR(s)) |
fc14f2fe | 1072 | return ERR_CAST(s); |
1da177e4 | 1073 | if (!s->s_root) { |
9b04c997 | 1074 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1075 | if (error) { |
74dbbdd7 | 1076 | deactivate_locked_super(s); |
fc14f2fe | 1077 | return ERR_PTR(error); |
1da177e4 LT |
1078 | } |
1079 | s->s_flags |= MS_ACTIVE; | |
9329d1be KS |
1080 | } else { |
1081 | do_remount_sb(s, flags, data, 0); | |
1da177e4 | 1082 | } |
fc14f2fe AV |
1083 | return dget(s->s_root); |
1084 | } | |
1085 | EXPORT_SYMBOL(mount_single); | |
1086 | ||
9d412a43 AV |
1087 | struct dentry * |
1088 | mount_fs(struct file_system_type *type, int flags, const char *name, void *data) | |
1da177e4 | 1089 | { |
c96e41e9 | 1090 | struct dentry *root; |
9d412a43 | 1091 | struct super_block *sb; |
1da177e4 | 1092 | char *secdata = NULL; |
9d412a43 | 1093 | int error = -ENOMEM; |
8089352a | 1094 | |
e0007529 | 1095 | if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) { |
1da177e4 | 1096 | secdata = alloc_secdata(); |
454e2398 | 1097 | if (!secdata) |
9d412a43 | 1098 | goto out; |
1da177e4 | 1099 | |
e0007529 | 1100 | error = security_sb_copy_data(data, secdata); |
454e2398 | 1101 | if (error) |
1da177e4 | 1102 | goto out_free_secdata; |
1da177e4 LT |
1103 | } |
1104 | ||
1a102ff9 AV |
1105 | root = type->mount(type, flags, name, data); |
1106 | if (IS_ERR(root)) { | |
1107 | error = PTR_ERR(root); | |
1108 | goto out_free_secdata; | |
c96e41e9 | 1109 | } |
9d412a43 AV |
1110 | sb = root->d_sb; |
1111 | BUG_ON(!sb); | |
1112 | WARN_ON(!sb->s_bdi); | |
6c510389 | 1113 | WARN_ON(sb->s_bdi == &default_backing_dev_info); |
9d412a43 | 1114 | sb->s_flags |= MS_BORN; |
454e2398 | 1115 | |
9d412a43 | 1116 | error = security_sb_kern_mount(sb, flags, secdata); |
5129a469 JE |
1117 | if (error) |
1118 | goto out_sb; | |
454e2398 | 1119 | |
42cb56ae JL |
1120 | /* |
1121 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE | |
1122 | * but s_maxbytes was an unsigned long long for many releases. Throw | |
1123 | * this warning for a little while to try and catch filesystems that | |
4358b567 | 1124 | * violate this rule. |
42cb56ae | 1125 | */ |
9d412a43 AV |
1126 | WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " |
1127 | "negative value (%lld)\n", type->name, sb->s_maxbytes); | |
42cb56ae | 1128 | |
9d412a43 | 1129 | up_write(&sb->s_umount); |
8680e22f | 1130 | free_secdata(secdata); |
9d412a43 | 1131 | return root; |
1da177e4 | 1132 | out_sb: |
9d412a43 AV |
1133 | dput(root); |
1134 | deactivate_locked_super(sb); | |
1da177e4 LT |
1135 | out_free_secdata: |
1136 | free_secdata(secdata); | |
1da177e4 | 1137 | out: |
454e2398 | 1138 | return ERR_PTR(error); |
1da177e4 LT |
1139 | } |
1140 | ||
5accdf82 JK |
1141 | /* |
1142 | * This is an internal function, please use sb_end_{write,pagefault,intwrite} | |
1143 | * instead. | |
1144 | */ | |
1145 | void __sb_end_write(struct super_block *sb, int level) | |
1146 | { | |
1147 | percpu_counter_dec(&sb->s_writers.counter[level-1]); | |
1148 | /* | |
1149 | * Make sure s_writers are updated before we wake up waiters in | |
1150 | * freeze_super(). | |
1151 | */ | |
1152 | smp_mb(); | |
1153 | if (waitqueue_active(&sb->s_writers.wait)) | |
1154 | wake_up(&sb->s_writers.wait); | |
1155 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _RET_IP_); | |
1156 | } | |
1157 | EXPORT_SYMBOL(__sb_end_write); | |
1158 | ||
1159 | #ifdef CONFIG_LOCKDEP | |
1160 | /* | |
1161 | * We want lockdep to tell us about possible deadlocks with freezing but | |
1162 | * it's it bit tricky to properly instrument it. Getting a freeze protection | |
1163 | * works as getting a read lock but there are subtle problems. XFS for example | |
1164 | * gets freeze protection on internal level twice in some cases, which is OK | |
1165 | * only because we already hold a freeze protection also on higher level. Due | |
1166 | * to these cases we have to tell lockdep we are doing trylock when we | |
1167 | * already hold a freeze protection for a higher freeze level. | |
1168 | */ | |
1169 | static void acquire_freeze_lock(struct super_block *sb, int level, bool trylock, | |
1170 | unsigned long ip) | |
1171 | { | |
1172 | int i; | |
1173 | ||
1174 | if (!trylock) { | |
1175 | for (i = 0; i < level - 1; i++) | |
1176 | if (lock_is_held(&sb->s_writers.lock_map[i])) { | |
1177 | trylock = true; | |
1178 | break; | |
1179 | } | |
1180 | } | |
1181 | rwsem_acquire_read(&sb->s_writers.lock_map[level-1], 0, trylock, ip); | |
1182 | } | |
1183 | #endif | |
1184 | ||
1185 | /* | |
1186 | * This is an internal function, please use sb_start_{write,pagefault,intwrite} | |
1187 | * instead. | |
1188 | */ | |
1189 | int __sb_start_write(struct super_block *sb, int level, bool wait) | |
1190 | { | |
1191 | retry: | |
1192 | if (unlikely(sb->s_writers.frozen >= level)) { | |
1193 | if (!wait) | |
1194 | return 0; | |
1195 | wait_event(sb->s_writers.wait_unfrozen, | |
1196 | sb->s_writers.frozen < level); | |
1197 | } | |
1198 | ||
1199 | #ifdef CONFIG_LOCKDEP | |
1200 | acquire_freeze_lock(sb, level, !wait, _RET_IP_); | |
1201 | #endif | |
1202 | percpu_counter_inc(&sb->s_writers.counter[level-1]); | |
1203 | /* | |
1204 | * Make sure counter is updated before we check for frozen. | |
1205 | * freeze_super() first sets frozen and then checks the counter. | |
1206 | */ | |
1207 | smp_mb(); | |
1208 | if (unlikely(sb->s_writers.frozen >= level)) { | |
1209 | __sb_end_write(sb, level); | |
1210 | goto retry; | |
1211 | } | |
1212 | return 1; | |
1213 | } | |
1214 | EXPORT_SYMBOL(__sb_start_write); | |
1215 | ||
1216 | /** | |
1217 | * sb_wait_write - wait until all writers to given file system finish | |
1218 | * @sb: the super for which we wait | |
1219 | * @level: type of writers we wait for (normal vs page fault) | |
1220 | * | |
1221 | * This function waits until there are no writers of given type to given file | |
1222 | * system. Caller of this function should make sure there can be no new writers | |
1223 | * of type @level before calling this function. Otherwise this function can | |
1224 | * livelock. | |
1225 | */ | |
1226 | static void sb_wait_write(struct super_block *sb, int level) | |
1227 | { | |
1228 | s64 writers; | |
1229 | ||
1230 | /* | |
1231 | * We just cycle-through lockdep here so that it does not complain | |
1232 | * about returning with lock to userspace | |
1233 | */ | |
1234 | rwsem_acquire(&sb->s_writers.lock_map[level-1], 0, 0, _THIS_IP_); | |
1235 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _THIS_IP_); | |
1236 | ||
1237 | do { | |
1238 | DEFINE_WAIT(wait); | |
1239 | ||
1240 | /* | |
1241 | * We use a barrier in prepare_to_wait() to separate setting | |
1242 | * of frozen and checking of the counter | |
1243 | */ | |
1244 | prepare_to_wait(&sb->s_writers.wait, &wait, | |
1245 | TASK_UNINTERRUPTIBLE); | |
1246 | ||
1247 | writers = percpu_counter_sum(&sb->s_writers.counter[level-1]); | |
1248 | if (writers) | |
1249 | schedule(); | |
1250 | ||
1251 | finish_wait(&sb->s_writers.wait, &wait); | |
1252 | } while (writers); | |
1253 | } | |
1254 | ||
18e9e510 | 1255 | /** |
7000d3c4 RD |
1256 | * freeze_super - lock the filesystem and force it into a consistent state |
1257 | * @sb: the super to lock | |
18e9e510 JB |
1258 | * |
1259 | * Syncs the super to make sure the filesystem is consistent and calls the fs's | |
1260 | * freeze_fs. Subsequent calls to this without first thawing the fs will return | |
1261 | * -EBUSY. | |
5accdf82 JK |
1262 | * |
1263 | * During this function, sb->s_writers.frozen goes through these values: | |
1264 | * | |
1265 | * SB_UNFROZEN: File system is normal, all writes progress as usual. | |
1266 | * | |
1267 | * SB_FREEZE_WRITE: The file system is in the process of being frozen. New | |
1268 | * writes should be blocked, though page faults are still allowed. We wait for | |
1269 | * all writes to complete and then proceed to the next stage. | |
1270 | * | |
1271 | * SB_FREEZE_PAGEFAULT: Freezing continues. Now also page faults are blocked | |
1272 | * but internal fs threads can still modify the filesystem (although they | |
1273 | * should not dirty new pages or inodes), writeback can run etc. After waiting | |
1274 | * for all running page faults we sync the filesystem which will clean all | |
1275 | * dirty pages and inodes (no new dirty pages or inodes can be created when | |
1276 | * sync is running). | |
1277 | * | |
1278 | * SB_FREEZE_FS: The file system is frozen. Now all internal sources of fs | |
1279 | * modification are blocked (e.g. XFS preallocation truncation on inode | |
1280 | * reclaim). This is usually implemented by blocking new transactions for | |
1281 | * filesystems that have them and need this additional guard. After all | |
1282 | * internal writers are finished we call ->freeze_fs() to finish filesystem | |
1283 | * freezing. Then we transition to SB_FREEZE_COMPLETE state. This state is | |
1284 | * mostly auxiliary for filesystems to verify they do not modify frozen fs. | |
1285 | * | |
1286 | * sb->s_writers.frozen is protected by sb->s_umount. | |
18e9e510 JB |
1287 | */ |
1288 | int freeze_super(struct super_block *sb) | |
1289 | { | |
1290 | int ret; | |
1291 | ||
1292 | atomic_inc(&sb->s_active); | |
1293 | down_write(&sb->s_umount); | |
5accdf82 | 1294 | if (sb->s_writers.frozen != SB_UNFROZEN) { |
18e9e510 JB |
1295 | deactivate_locked_super(sb); |
1296 | return -EBUSY; | |
1297 | } | |
1298 | ||
dabe0dc1 AV |
1299 | if (!(sb->s_flags & MS_BORN)) { |
1300 | up_write(&sb->s_umount); | |
1301 | return 0; /* sic - it's "nothing to do" */ | |
1302 | } | |
1303 | ||
18e9e510 | 1304 | if (sb->s_flags & MS_RDONLY) { |
5accdf82 JK |
1305 | /* Nothing to do really... */ |
1306 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
18e9e510 JB |
1307 | up_write(&sb->s_umount); |
1308 | return 0; | |
1309 | } | |
1310 | ||
5accdf82 JK |
1311 | /* From now on, no new normal writers can start */ |
1312 | sb->s_writers.frozen = SB_FREEZE_WRITE; | |
1313 | smp_wmb(); | |
1314 | ||
1315 | /* Release s_umount to preserve sb_start_write -> s_umount ordering */ | |
1316 | up_write(&sb->s_umount); | |
1317 | ||
1318 | sb_wait_write(sb, SB_FREEZE_WRITE); | |
1319 | ||
1320 | /* Now we go and block page faults... */ | |
1321 | down_write(&sb->s_umount); | |
1322 | sb->s_writers.frozen = SB_FREEZE_PAGEFAULT; | |
18e9e510 JB |
1323 | smp_wmb(); |
1324 | ||
5accdf82 JK |
1325 | sb_wait_write(sb, SB_FREEZE_PAGEFAULT); |
1326 | ||
1327 | /* All writers are done so after syncing there won't be dirty data */ | |
18e9e510 JB |
1328 | sync_filesystem(sb); |
1329 | ||
5accdf82 JK |
1330 | /* Now wait for internal filesystem counter */ |
1331 | sb->s_writers.frozen = SB_FREEZE_FS; | |
18e9e510 | 1332 | smp_wmb(); |
5accdf82 | 1333 | sb_wait_write(sb, SB_FREEZE_FS); |
18e9e510 | 1334 | |
18e9e510 JB |
1335 | if (sb->s_op->freeze_fs) { |
1336 | ret = sb->s_op->freeze_fs(sb); | |
1337 | if (ret) { | |
1338 | printk(KERN_ERR | |
1339 | "VFS:Filesystem freeze failed\n"); | |
5accdf82 | 1340 | sb->s_writers.frozen = SB_UNFROZEN; |
e1616300 | 1341 | smp_wmb(); |
5accdf82 | 1342 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 JB |
1343 | deactivate_locked_super(sb); |
1344 | return ret; | |
1345 | } | |
1346 | } | |
5accdf82 JK |
1347 | /* |
1348 | * This is just for debugging purposes so that fs can warn if it | |
1349 | * sees write activity when frozen is set to SB_FREEZE_COMPLETE. | |
1350 | */ | |
1351 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
18e9e510 JB |
1352 | up_write(&sb->s_umount); |
1353 | return 0; | |
1354 | } | |
1355 | EXPORT_SYMBOL(freeze_super); | |
1356 | ||
1357 | /** | |
1358 | * thaw_super -- unlock filesystem | |
1359 | * @sb: the super to thaw | |
1360 | * | |
1361 | * Unlocks the filesystem and marks it writeable again after freeze_super(). | |
1362 | */ | |
1363 | int thaw_super(struct super_block *sb) | |
1364 | { | |
1365 | int error; | |
1366 | ||
1367 | down_write(&sb->s_umount); | |
5accdf82 | 1368 | if (sb->s_writers.frozen == SB_UNFROZEN) { |
18e9e510 JB |
1369 | up_write(&sb->s_umount); |
1370 | return -EINVAL; | |
1371 | } | |
1372 | ||
1373 | if (sb->s_flags & MS_RDONLY) | |
1374 | goto out; | |
1375 | ||
1376 | if (sb->s_op->unfreeze_fs) { | |
1377 | error = sb->s_op->unfreeze_fs(sb); | |
1378 | if (error) { | |
1379 | printk(KERN_ERR | |
1380 | "VFS:Filesystem thaw failed\n"); | |
18e9e510 JB |
1381 | up_write(&sb->s_umount); |
1382 | return error; | |
1383 | } | |
1384 | } | |
1385 | ||
1386 | out: | |
5accdf82 | 1387 | sb->s_writers.frozen = SB_UNFROZEN; |
18e9e510 | 1388 | smp_wmb(); |
5accdf82 | 1389 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 JB |
1390 | deactivate_locked_super(sb); |
1391 | ||
1392 | return 0; | |
1393 | } | |
1394 | EXPORT_SYMBOL(thaw_super); |