]>
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 | |
0e1fdafd DC |
74 | if (sb->s_op && sb->s_op->nr_cached_objects) |
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 DC |
84 | /* proportion the scan between the caches */ |
85 | dentries = (sc->nr_to_scan * sb->s_nr_dentry_unused) / | |
86 | total_objects; | |
87 | inodes = (sc->nr_to_scan * sb->s_nr_inodes_unused) / | |
88 | total_objects; | |
89 | if (fs_objects) | |
90 | fs_objects = (sc->nr_to_scan * fs_objects) / | |
91 | total_objects; | |
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 | ||
0e1fdafd | 107 | total_objects = (total_objects / 100) * sysctl_vfs_cache_pressure; |
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) { | |
1da177e4 | 155 | if (security_sb_alloc(s)) { |
5accdf82 JK |
156 | /* |
157 | * We cannot call security_sb_free() without | |
158 | * security_sb_alloc() succeeding. So bail out manually | |
159 | */ | |
1da177e4 LT |
160 | kfree(s); |
161 | s = NULL; | |
162 | goto out; | |
163 | } | |
6416ccb7 NP |
164 | #ifdef CONFIG_SMP |
165 | s->s_files = alloc_percpu(struct list_head); | |
5accdf82 JK |
166 | if (!s->s_files) |
167 | goto err_out; | |
168 | else { | |
6416ccb7 NP |
169 | int i; |
170 | ||
171 | for_each_possible_cpu(i) | |
172 | INIT_LIST_HEAD(per_cpu_ptr(s->s_files, i)); | |
173 | } | |
174 | #else | |
1da177e4 | 175 | INIT_LIST_HEAD(&s->s_files); |
6416ccb7 | 176 | #endif |
5accdf82 JK |
177 | if (init_sb_writers(s, type)) |
178 | goto err_out; | |
9249e17f | 179 | s->s_flags = flags; |
95f28604 | 180 | s->s_bdi = &default_backing_dev_info; |
a5166169 | 181 | INIT_HLIST_NODE(&s->s_instances); |
ceb5bdc2 | 182 | INIT_HLIST_BL_HEAD(&s->s_anon); |
1da177e4 | 183 | INIT_LIST_HEAD(&s->s_inodes); |
da3bbdd4 | 184 | INIT_LIST_HEAD(&s->s_dentry_lru); |
98b745c6 | 185 | INIT_LIST_HEAD(&s->s_inode_lru); |
09cc9fc7 | 186 | spin_lock_init(&s->s_inode_lru_lock); |
39f7c4db | 187 | INIT_LIST_HEAD(&s->s_mounts); |
1da177e4 | 188 | init_rwsem(&s->s_umount); |
7892f2f4 | 189 | mutex_init(&s->s_lock); |
897c6ff9 | 190 | lockdep_set_class(&s->s_umount, &type->s_umount_key); |
cf516249 IM |
191 | /* |
192 | * The locking rules for s_lock are up to the | |
193 | * filesystem. For example ext3fs has different | |
194 | * lock ordering than usbfs: | |
195 | */ | |
196 | lockdep_set_class(&s->s_lock, &type->s_lock_key); | |
ada723dc PZ |
197 | /* |
198 | * sget() can have s_umount recursion. | |
199 | * | |
200 | * When it cannot find a suitable sb, it allocates a new | |
201 | * one (this one), and tries again to find a suitable old | |
202 | * one. | |
203 | * | |
204 | * In case that succeeds, it will acquire the s_umount | |
205 | * lock of the old one. Since these are clearly distrinct | |
206 | * locks, and this object isn't exposed yet, there's no | |
207 | * risk of deadlocks. | |
208 | * | |
209 | * Annotate this by putting this lock in a different | |
210 | * subclass. | |
211 | */ | |
212 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); | |
b20bd1a5 | 213 | s->s_count = 1; |
1da177e4 | 214 | atomic_set(&s->s_active, 1); |
a11f3a05 | 215 | mutex_init(&s->s_vfs_rename_mutex); |
51ee049e | 216 | lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key); |
d3be915f IM |
217 | mutex_init(&s->s_dquot.dqio_mutex); |
218 | mutex_init(&s->s_dquot.dqonoff_mutex); | |
1da177e4 | 219 | init_rwsem(&s->s_dquot.dqptr_sem); |
1da177e4 | 220 | s->s_maxbytes = MAX_NON_LFS; |
1da177e4 LT |
221 | s->s_op = &default_op; |
222 | s->s_time_gran = 1000000000; | |
c515e1fd | 223 | s->cleancache_poolid = -1; |
b0d40c92 DC |
224 | |
225 | s->s_shrink.seeks = DEFAULT_SEEKS; | |
226 | s->s_shrink.shrink = prune_super; | |
8ab47664 | 227 | s->s_shrink.batch = 1024; |
1da177e4 LT |
228 | } |
229 | out: | |
230 | return s; | |
5accdf82 JK |
231 | err_out: |
232 | security_sb_free(s); | |
233 | #ifdef CONFIG_SMP | |
234 | if (s->s_files) | |
235 | free_percpu(s->s_files); | |
236 | #endif | |
237 | destroy_sb_writers(s); | |
238 | kfree(s); | |
239 | s = NULL; | |
240 | goto out; | |
1da177e4 LT |
241 | } |
242 | ||
243 | /** | |
244 | * destroy_super - frees a superblock | |
245 | * @s: superblock to free | |
246 | * | |
247 | * Frees a superblock. | |
248 | */ | |
249 | static inline void destroy_super(struct super_block *s) | |
250 | { | |
6416ccb7 NP |
251 | #ifdef CONFIG_SMP |
252 | free_percpu(s->s_files); | |
253 | #endif | |
5accdf82 | 254 | destroy_sb_writers(s); |
1da177e4 | 255 | security_sb_free(s); |
39f7c4db | 256 | WARN_ON(!list_empty(&s->s_mounts)); |
79c0b2df | 257 | kfree(s->s_subtype); |
b3b304a2 | 258 | kfree(s->s_options); |
1da177e4 LT |
259 | kfree(s); |
260 | } | |
261 | ||
262 | /* Superblock refcounting */ | |
263 | ||
264 | /* | |
35cf7ba0 | 265 | * Drop a superblock's refcount. The caller must hold sb_lock. |
1da177e4 | 266 | */ |
f47ec3f2 | 267 | static void __put_super(struct super_block *sb) |
1da177e4 | 268 | { |
1da177e4 | 269 | if (!--sb->s_count) { |
551de6f3 | 270 | list_del_init(&sb->s_list); |
1da177e4 | 271 | destroy_super(sb); |
1da177e4 | 272 | } |
1da177e4 LT |
273 | } |
274 | ||
275 | /** | |
276 | * put_super - drop a temporary reference to superblock | |
277 | * @sb: superblock in question | |
278 | * | |
279 | * Drops a temporary reference, frees superblock if there's no | |
280 | * references left. | |
281 | */ | |
f47ec3f2 | 282 | static void put_super(struct super_block *sb) |
1da177e4 LT |
283 | { |
284 | spin_lock(&sb_lock); | |
285 | __put_super(sb); | |
286 | spin_unlock(&sb_lock); | |
287 | } | |
288 | ||
289 | ||
290 | /** | |
1712ac8f | 291 | * deactivate_locked_super - drop an active reference to superblock |
1da177e4 LT |
292 | * @s: superblock to deactivate |
293 | * | |
1712ac8f AV |
294 | * Drops an active reference to superblock, converting it into a temprory |
295 | * one if there is no other active references left. In that case we | |
1da177e4 LT |
296 | * tell fs driver to shut it down and drop the temporary reference we |
297 | * had just acquired. | |
1712ac8f AV |
298 | * |
299 | * Caller holds exclusive lock on superblock; that lock is released. | |
1da177e4 | 300 | */ |
1712ac8f | 301 | void deactivate_locked_super(struct super_block *s) |
1da177e4 LT |
302 | { |
303 | struct file_system_type *fs = s->s_type; | |
b20bd1a5 | 304 | if (atomic_dec_and_test(&s->s_active)) { |
3167760f | 305 | cleancache_invalidate_fs(s); |
1da177e4 | 306 | fs->kill_sb(s); |
b0d40c92 DC |
307 | |
308 | /* caches are now gone, we can safely kill the shrinker now */ | |
309 | unregister_shrinker(&s->s_shrink); | |
310 | ||
d863b50a BH |
311 | /* |
312 | * We need to call rcu_barrier so all the delayed rcu free | |
313 | * inodes are flushed before we release the fs module. | |
314 | */ | |
315 | rcu_barrier(); | |
1da177e4 LT |
316 | put_filesystem(fs); |
317 | put_super(s); | |
1712ac8f AV |
318 | } else { |
319 | up_write(&s->s_umount); | |
1da177e4 LT |
320 | } |
321 | } | |
322 | ||
1712ac8f | 323 | EXPORT_SYMBOL(deactivate_locked_super); |
1da177e4 | 324 | |
74dbbdd7 | 325 | /** |
1712ac8f | 326 | * deactivate_super - drop an active reference to superblock |
74dbbdd7 AV |
327 | * @s: superblock to deactivate |
328 | * | |
1712ac8f AV |
329 | * Variant of deactivate_locked_super(), except that superblock is *not* |
330 | * locked by caller. If we are going to drop the final active reference, | |
331 | * lock will be acquired prior to that. | |
74dbbdd7 | 332 | */ |
1712ac8f | 333 | void deactivate_super(struct super_block *s) |
74dbbdd7 | 334 | { |
1712ac8f AV |
335 | if (!atomic_add_unless(&s->s_active, -1, 1)) { |
336 | down_write(&s->s_umount); | |
337 | deactivate_locked_super(s); | |
74dbbdd7 AV |
338 | } |
339 | } | |
340 | ||
1712ac8f | 341 | EXPORT_SYMBOL(deactivate_super); |
74dbbdd7 | 342 | |
1da177e4 LT |
343 | /** |
344 | * grab_super - acquire an active reference | |
345 | * @s: reference we are trying to make active | |
346 | * | |
347 | * Tries to acquire an active reference. grab_super() is used when we | |
348 | * had just found a superblock in super_blocks or fs_type->fs_supers | |
349 | * and want to turn it into a full-blown active reference. grab_super() | |
350 | * is called with sb_lock held and drops it. Returns 1 in case of | |
351 | * success, 0 if we had failed (superblock contents was already dead or | |
352 | * dying when grab_super() had been called). | |
353 | */ | |
9c4dbee7 | 354 | static int grab_super(struct super_block *s) __releases(sb_lock) |
1da177e4 | 355 | { |
b20bd1a5 AV |
356 | if (atomic_inc_not_zero(&s->s_active)) { |
357 | spin_unlock(&sb_lock); | |
b20bd1a5 AV |
358 | return 1; |
359 | } | |
360 | /* it's going away */ | |
1da177e4 LT |
361 | s->s_count++; |
362 | spin_unlock(&sb_lock); | |
1712ac8f | 363 | /* wait for it to die */ |
1da177e4 | 364 | down_write(&s->s_umount); |
1da177e4 LT |
365 | up_write(&s->s_umount); |
366 | put_super(s); | |
1da177e4 LT |
367 | return 0; |
368 | } | |
369 | ||
12ad3ab6 DC |
370 | /* |
371 | * grab_super_passive - acquire a passive reference | |
331cbdee | 372 | * @sb: reference we are trying to grab |
12ad3ab6 DC |
373 | * |
374 | * Tries to acquire a passive reference. This is used in places where we | |
375 | * cannot take an active reference but we need to ensure that the | |
376 | * superblock does not go away while we are working on it. It returns | |
377 | * false if a reference was not gained, and returns true with the s_umount | |
378 | * lock held in read mode if a reference is gained. On successful return, | |
379 | * the caller must drop the s_umount lock and the passive reference when | |
380 | * done. | |
381 | */ | |
382 | bool grab_super_passive(struct super_block *sb) | |
383 | { | |
384 | spin_lock(&sb_lock); | |
a5166169 | 385 | if (hlist_unhashed(&sb->s_instances)) { |
12ad3ab6 DC |
386 | spin_unlock(&sb_lock); |
387 | return false; | |
388 | } | |
389 | ||
390 | sb->s_count++; | |
391 | spin_unlock(&sb_lock); | |
392 | ||
393 | if (down_read_trylock(&sb->s_umount)) { | |
dabe0dc1 | 394 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
12ad3ab6 DC |
395 | return true; |
396 | up_read(&sb->s_umount); | |
397 | } | |
398 | ||
399 | put_super(sb); | |
400 | return false; | |
401 | } | |
402 | ||
914e2637 AV |
403 | /* |
404 | * Superblock locking. We really ought to get rid of these two. | |
405 | */ | |
406 | void lock_super(struct super_block * sb) | |
407 | { | |
914e2637 AV |
408 | mutex_lock(&sb->s_lock); |
409 | } | |
410 | ||
411 | void unlock_super(struct super_block * sb) | |
412 | { | |
914e2637 AV |
413 | mutex_unlock(&sb->s_lock); |
414 | } | |
415 | ||
416 | EXPORT_SYMBOL(lock_super); | |
417 | EXPORT_SYMBOL(unlock_super); | |
418 | ||
1da177e4 LT |
419 | /** |
420 | * generic_shutdown_super - common helper for ->kill_sb() | |
421 | * @sb: superblock to kill | |
422 | * | |
423 | * generic_shutdown_super() does all fs-independent work on superblock | |
424 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects | |
425 | * that need destruction out of superblock, call generic_shutdown_super() | |
426 | * and release aforementioned objects. Note: dentries and inodes _are_ | |
427 | * taken care of and do not need specific handling. | |
c636ebdb DH |
428 | * |
429 | * Upon calling this function, the filesystem may no longer alter or | |
430 | * rearrange the set of dentries belonging to this super_block, nor may it | |
431 | * change the attachments of dentries to inodes. | |
1da177e4 LT |
432 | */ |
433 | void generic_shutdown_super(struct super_block *sb) | |
434 | { | |
ee9b6d61 | 435 | const struct super_operations *sop = sb->s_op; |
1da177e4 | 436 | |
c636ebdb DH |
437 | if (sb->s_root) { |
438 | shrink_dcache_for_umount(sb); | |
60b0680f | 439 | sync_filesystem(sb); |
1da177e4 | 440 | sb->s_flags &= ~MS_ACTIVE; |
efaee192 | 441 | |
63997e98 AV |
442 | fsnotify_unmount_inodes(&sb->s_inodes); |
443 | ||
444 | evict_inodes(sb); | |
1da177e4 | 445 | |
1da177e4 LT |
446 | if (sop->put_super) |
447 | sop->put_super(sb); | |
448 | ||
63997e98 | 449 | if (!list_empty(&sb->s_inodes)) { |
7b4fe29e DJ |
450 | printk("VFS: Busy inodes after unmount of %s. " |
451 | "Self-destruct in 5 seconds. Have a nice day...\n", | |
452 | sb->s_id); | |
1da177e4 | 453 | } |
1da177e4 LT |
454 | } |
455 | spin_lock(&sb_lock); | |
456 | /* should be initialized for __put_super_and_need_restart() */ | |
a5166169 | 457 | hlist_del_init(&sb->s_instances); |
1da177e4 LT |
458 | spin_unlock(&sb_lock); |
459 | up_write(&sb->s_umount); | |
460 | } | |
461 | ||
462 | EXPORT_SYMBOL(generic_shutdown_super); | |
463 | ||
464 | /** | |
465 | * sget - find or create a superblock | |
466 | * @type: filesystem type superblock should belong to | |
467 | * @test: comparison callback | |
468 | * @set: setup callback | |
9249e17f | 469 | * @flags: mount flags |
1da177e4 LT |
470 | * @data: argument to each of them |
471 | */ | |
472 | struct super_block *sget(struct file_system_type *type, | |
473 | int (*test)(struct super_block *,void *), | |
474 | int (*set)(struct super_block *,void *), | |
9249e17f | 475 | int flags, |
1da177e4 LT |
476 | void *data) |
477 | { | |
478 | struct super_block *s = NULL; | |
a5166169 | 479 | struct hlist_node *node; |
d4730127 | 480 | struct super_block *old; |
1da177e4 LT |
481 | int err; |
482 | ||
483 | retry: | |
484 | spin_lock(&sb_lock); | |
d4730127 | 485 | if (test) { |
a5166169 | 486 | hlist_for_each_entry(old, node, &type->fs_supers, s_instances) { |
d4730127 MK |
487 | if (!test(old, data)) |
488 | continue; | |
489 | if (!grab_super(old)) | |
490 | goto retry; | |
a3cfbb53 LZ |
491 | if (s) { |
492 | up_write(&s->s_umount); | |
d4730127 | 493 | destroy_super(s); |
7a4dec53 | 494 | s = NULL; |
a3cfbb53 | 495 | } |
d3f21473 | 496 | down_write(&old->s_umount); |
7a4dec53 AV |
497 | if (unlikely(!(old->s_flags & MS_BORN))) { |
498 | deactivate_locked_super(old); | |
499 | goto retry; | |
500 | } | |
d4730127 MK |
501 | return old; |
502 | } | |
1da177e4 LT |
503 | } |
504 | if (!s) { | |
505 | spin_unlock(&sb_lock); | |
9249e17f | 506 | s = alloc_super(type, flags); |
1da177e4 LT |
507 | if (!s) |
508 | return ERR_PTR(-ENOMEM); | |
509 | goto retry; | |
510 | } | |
511 | ||
512 | err = set(s, data); | |
513 | if (err) { | |
514 | spin_unlock(&sb_lock); | |
a3cfbb53 | 515 | up_write(&s->s_umount); |
1da177e4 LT |
516 | destroy_super(s); |
517 | return ERR_PTR(err); | |
518 | } | |
519 | s->s_type = type; | |
520 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); | |
521 | list_add_tail(&s->s_list, &super_blocks); | |
a5166169 | 522 | hlist_add_head(&s->s_instances, &type->fs_supers); |
1da177e4 LT |
523 | spin_unlock(&sb_lock); |
524 | get_filesystem(type); | |
b0d40c92 | 525 | register_shrinker(&s->s_shrink); |
1da177e4 LT |
526 | return s; |
527 | } | |
528 | ||
529 | EXPORT_SYMBOL(sget); | |
530 | ||
531 | void drop_super(struct super_block *sb) | |
532 | { | |
533 | up_read(&sb->s_umount); | |
534 | put_super(sb); | |
535 | } | |
536 | ||
537 | EXPORT_SYMBOL(drop_super); | |
538 | ||
e5004753 CH |
539 | /** |
540 | * sync_supers - helper for periodic superblock writeback | |
541 | * | |
542 | * Call the write_super method if present on all dirty superblocks in | |
543 | * the system. This is for the periodic writeback used by most older | |
544 | * filesystems. For data integrity superblock writeback use | |
545 | * sync_filesystems() instead. | |
546 | * | |
1da177e4 LT |
547 | * Note: check the dirty flag before waiting, so we don't |
548 | * hold up the sync while mounting a device. (The newly | |
549 | * mounted device won't need syncing.) | |
550 | */ | |
551 | void sync_supers(void) | |
552 | { | |
dca33252 | 553 | struct super_block *sb, *p = NULL; |
618f0636 | 554 | |
1da177e4 | 555 | spin_lock(&sb_lock); |
dca33252 | 556 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 557 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 558 | continue; |
e5004753 | 559 | if (sb->s_op->write_super && sb->s_dirt) { |
1da177e4 LT |
560 | sb->s_count++; |
561 | spin_unlock(&sb_lock); | |
e5004753 | 562 | |
1da177e4 | 563 | down_read(&sb->s_umount); |
dabe0dc1 | 564 | if (sb->s_root && sb->s_dirt && (sb->s_flags & MS_BORN)) |
e5004753 | 565 | sb->s_op->write_super(sb); |
618f0636 | 566 | up_read(&sb->s_umount); |
e5004753 | 567 | |
618f0636 | 568 | spin_lock(&sb_lock); |
dca33252 AV |
569 | if (p) |
570 | __put_super(p); | |
571 | p = sb; | |
618f0636 KK |
572 | } |
573 | } | |
dca33252 AV |
574 | if (p) |
575 | __put_super(p); | |
1da177e4 LT |
576 | spin_unlock(&sb_lock); |
577 | } | |
578 | ||
01a05b33 AV |
579 | /** |
580 | * iterate_supers - call function for all active superblocks | |
581 | * @f: function to call | |
582 | * @arg: argument to pass to it | |
583 | * | |
584 | * Scans the superblock list and calls given function, passing it | |
585 | * locked superblock and given argument. | |
586 | */ | |
587 | void iterate_supers(void (*f)(struct super_block *, void *), void *arg) | |
588 | { | |
dca33252 | 589 | struct super_block *sb, *p = NULL; |
01a05b33 AV |
590 | |
591 | spin_lock(&sb_lock); | |
dca33252 | 592 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 593 | if (hlist_unhashed(&sb->s_instances)) |
01a05b33 AV |
594 | continue; |
595 | sb->s_count++; | |
596 | spin_unlock(&sb_lock); | |
597 | ||
598 | down_read(&sb->s_umount); | |
dabe0dc1 | 599 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
01a05b33 AV |
600 | f(sb, arg); |
601 | up_read(&sb->s_umount); | |
602 | ||
603 | spin_lock(&sb_lock); | |
dca33252 AV |
604 | if (p) |
605 | __put_super(p); | |
606 | p = sb; | |
01a05b33 | 607 | } |
dca33252 AV |
608 | if (p) |
609 | __put_super(p); | |
01a05b33 AV |
610 | spin_unlock(&sb_lock); |
611 | } | |
612 | ||
43e15cdb AV |
613 | /** |
614 | * iterate_supers_type - call function for superblocks of given type | |
615 | * @type: fs type | |
616 | * @f: function to call | |
617 | * @arg: argument to pass to it | |
618 | * | |
619 | * Scans the superblock list and calls given function, passing it | |
620 | * locked superblock and given argument. | |
621 | */ | |
622 | void iterate_supers_type(struct file_system_type *type, | |
623 | void (*f)(struct super_block *, void *), void *arg) | |
624 | { | |
625 | struct super_block *sb, *p = NULL; | |
a5166169 | 626 | struct hlist_node *node; |
43e15cdb AV |
627 | |
628 | spin_lock(&sb_lock); | |
a5166169 | 629 | hlist_for_each_entry(sb, node, &type->fs_supers, s_instances) { |
43e15cdb AV |
630 | sb->s_count++; |
631 | spin_unlock(&sb_lock); | |
632 | ||
633 | down_read(&sb->s_umount); | |
dabe0dc1 | 634 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
43e15cdb AV |
635 | f(sb, arg); |
636 | up_read(&sb->s_umount); | |
637 | ||
638 | spin_lock(&sb_lock); | |
639 | if (p) | |
640 | __put_super(p); | |
641 | p = sb; | |
642 | } | |
643 | if (p) | |
644 | __put_super(p); | |
645 | spin_unlock(&sb_lock); | |
646 | } | |
647 | ||
648 | EXPORT_SYMBOL(iterate_supers_type); | |
649 | ||
1da177e4 LT |
650 | /** |
651 | * get_super - get the superblock of a device | |
652 | * @bdev: device to get the superblock for | |
653 | * | |
654 | * Scans the superblock list and finds the superblock of the file system | |
655 | * mounted on the device given. %NULL is returned if no match is found. | |
656 | */ | |
657 | ||
df40c01a | 658 | struct super_block *get_super(struct block_device *bdev) |
1da177e4 | 659 | { |
618f0636 KK |
660 | struct super_block *sb; |
661 | ||
1da177e4 LT |
662 | if (!bdev) |
663 | return NULL; | |
618f0636 | 664 | |
1da177e4 | 665 | spin_lock(&sb_lock); |
618f0636 KK |
666 | rescan: |
667 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 668 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 669 | continue; |
618f0636 KK |
670 | if (sb->s_bdev == bdev) { |
671 | sb->s_count++; | |
1da177e4 | 672 | spin_unlock(&sb_lock); |
618f0636 | 673 | down_read(&sb->s_umount); |
df40c01a | 674 | /* still alive? */ |
dabe0dc1 | 675 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
618f0636 KK |
676 | return sb; |
677 | up_read(&sb->s_umount); | |
df40c01a | 678 | /* nope, got unmounted */ |
618f0636 | 679 | spin_lock(&sb_lock); |
df40c01a AV |
680 | __put_super(sb); |
681 | goto rescan; | |
1da177e4 LT |
682 | } |
683 | } | |
684 | spin_unlock(&sb_lock); | |
685 | return NULL; | |
686 | } | |
687 | ||
688 | EXPORT_SYMBOL(get_super); | |
4504230a | 689 | |
6b6dc836 JK |
690 | /** |
691 | * get_super_thawed - get thawed superblock of a device | |
692 | * @bdev: device to get the superblock for | |
693 | * | |
694 | * Scans the superblock list and finds the superblock of the file system | |
695 | * mounted on the device. The superblock is returned once it is thawed | |
696 | * (or immediately if it was not frozen). %NULL is returned if no match | |
697 | * is found. | |
698 | */ | |
699 | struct super_block *get_super_thawed(struct block_device *bdev) | |
700 | { | |
701 | while (1) { | |
702 | struct super_block *s = get_super(bdev); | |
5accdf82 | 703 | if (!s || s->s_writers.frozen == SB_UNFROZEN) |
6b6dc836 JK |
704 | return s; |
705 | up_read(&s->s_umount); | |
5accdf82 JK |
706 | wait_event(s->s_writers.wait_unfrozen, |
707 | s->s_writers.frozen == SB_UNFROZEN); | |
6b6dc836 JK |
708 | put_super(s); |
709 | } | |
710 | } | |
711 | EXPORT_SYMBOL(get_super_thawed); | |
712 | ||
4504230a CH |
713 | /** |
714 | * get_active_super - get an active reference to the superblock of a device | |
715 | * @bdev: device to get the superblock for | |
716 | * | |
717 | * Scans the superblock list and finds the superblock of the file system | |
718 | * mounted on the device given. Returns the superblock with an active | |
d3f21473 | 719 | * reference or %NULL if none was found. |
4504230a CH |
720 | */ |
721 | struct super_block *get_active_super(struct block_device *bdev) | |
722 | { | |
723 | struct super_block *sb; | |
724 | ||
725 | if (!bdev) | |
726 | return NULL; | |
727 | ||
1494583d | 728 | restart: |
4504230a CH |
729 | spin_lock(&sb_lock); |
730 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 731 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 732 | continue; |
1494583d AV |
733 | if (sb->s_bdev == bdev) { |
734 | if (grab_super(sb)) /* drops sb_lock */ | |
735 | return sb; | |
736 | else | |
737 | goto restart; | |
738 | } | |
4504230a CH |
739 | } |
740 | spin_unlock(&sb_lock); | |
741 | return NULL; | |
742 | } | |
1da177e4 | 743 | |
df40c01a | 744 | struct super_block *user_get_super(dev_t dev) |
1da177e4 | 745 | { |
618f0636 | 746 | struct super_block *sb; |
1da177e4 | 747 | |
1da177e4 | 748 | spin_lock(&sb_lock); |
618f0636 KK |
749 | rescan: |
750 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 751 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 752 | continue; |
618f0636 KK |
753 | if (sb->s_dev == dev) { |
754 | sb->s_count++; | |
1da177e4 | 755 | spin_unlock(&sb_lock); |
618f0636 | 756 | down_read(&sb->s_umount); |
df40c01a | 757 | /* still alive? */ |
dabe0dc1 | 758 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
618f0636 KK |
759 | return sb; |
760 | up_read(&sb->s_umount); | |
df40c01a | 761 | /* nope, got unmounted */ |
618f0636 | 762 | spin_lock(&sb_lock); |
df40c01a AV |
763 | __put_super(sb); |
764 | goto rescan; | |
1da177e4 LT |
765 | } |
766 | } | |
767 | spin_unlock(&sb_lock); | |
768 | return NULL; | |
769 | } | |
770 | ||
1da177e4 LT |
771 | /** |
772 | * do_remount_sb - asks filesystem to change mount options. | |
773 | * @sb: superblock in question | |
774 | * @flags: numeric part of options | |
775 | * @data: the rest of options | |
776 | * @force: whether or not to force the change | |
777 | * | |
778 | * Alters the mount options of a mounted file system. | |
779 | */ | |
780 | int do_remount_sb(struct super_block *sb, int flags, void *data, int force) | |
781 | { | |
782 | int retval; | |
c79d967d | 783 | int remount_ro; |
4504230a | 784 | |
5accdf82 | 785 | if (sb->s_writers.frozen != SB_UNFROZEN) |
4504230a CH |
786 | return -EBUSY; |
787 | ||
9361401e | 788 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
789 | if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev)) |
790 | return -EACCES; | |
9361401e | 791 | #endif |
4504230a | 792 | |
1da177e4 LT |
793 | if (flags & MS_RDONLY) |
794 | acct_auto_close(sb); | |
795 | shrink_dcache_sb(sb); | |
60b0680f | 796 | sync_filesystem(sb); |
1da177e4 | 797 | |
d208bbdd | 798 | remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY); |
d208bbdd | 799 | |
1da177e4 LT |
800 | /* If we are remounting RDONLY and current sb is read/write, |
801 | make sure there are no rw files opened */ | |
d208bbdd | 802 | if (remount_ro) { |
4ed5e82f | 803 | if (force) { |
1da177e4 | 804 | mark_files_ro(sb); |
4ed5e82f MS |
805 | } else { |
806 | retval = sb_prepare_remount_readonly(sb); | |
807 | if (retval) | |
808 | return retval; | |
4ed5e82f | 809 | } |
1da177e4 LT |
810 | } |
811 | ||
812 | if (sb->s_op->remount_fs) { | |
1da177e4 | 813 | retval = sb->s_op->remount_fs(sb, &flags, data); |
2833eb2b MS |
814 | if (retval) { |
815 | if (!force) | |
4ed5e82f | 816 | goto cancel_readonly; |
2833eb2b MS |
817 | /* If forced remount, go ahead despite any errors */ |
818 | WARN(1, "forced remount of a %s fs returned %i\n", | |
819 | sb->s_type->name, retval); | |
820 | } | |
1da177e4 LT |
821 | } |
822 | sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK); | |
4ed5e82f MS |
823 | /* Needs to be ordered wrt mnt_is_readonly() */ |
824 | smp_wmb(); | |
825 | sb->s_readonly_remount = 0; | |
c79d967d | 826 | |
d208bbdd NP |
827 | /* |
828 | * Some filesystems modify their metadata via some other path than the | |
829 | * bdev buffer cache (eg. use a private mapping, or directories in | |
830 | * pagecache, etc). Also file data modifications go via their own | |
831 | * mappings. So If we try to mount readonly then copy the filesystem | |
832 | * from bdev, we could get stale data, so invalidate it to give a best | |
833 | * effort at coherency. | |
834 | */ | |
835 | if (remount_ro && sb->s_bdev) | |
836 | invalidate_bdev(sb->s_bdev); | |
1da177e4 | 837 | return 0; |
4ed5e82f MS |
838 | |
839 | cancel_readonly: | |
840 | sb->s_readonly_remount = 0; | |
841 | return retval; | |
1da177e4 LT |
842 | } |
843 | ||
a2a9537a | 844 | static void do_emergency_remount(struct work_struct *work) |
1da177e4 | 845 | { |
dca33252 | 846 | struct super_block *sb, *p = NULL; |
1da177e4 LT |
847 | |
848 | spin_lock(&sb_lock); | |
dca33252 | 849 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 850 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 851 | continue; |
1da177e4 LT |
852 | sb->s_count++; |
853 | spin_unlock(&sb_lock); | |
443b94ba | 854 | down_write(&sb->s_umount); |
dabe0dc1 AV |
855 | if (sb->s_root && sb->s_bdev && (sb->s_flags & MS_BORN) && |
856 | !(sb->s_flags & MS_RDONLY)) { | |
1da177e4 | 857 | /* |
1da177e4 LT |
858 | * What lock protects sb->s_flags?? |
859 | */ | |
1da177e4 | 860 | do_remount_sb(sb, MS_RDONLY, NULL, 1); |
1da177e4 | 861 | } |
443b94ba | 862 | up_write(&sb->s_umount); |
1da177e4 | 863 | spin_lock(&sb_lock); |
dca33252 AV |
864 | if (p) |
865 | __put_super(p); | |
866 | p = sb; | |
1da177e4 | 867 | } |
dca33252 AV |
868 | if (p) |
869 | __put_super(p); | |
1da177e4 | 870 | spin_unlock(&sb_lock); |
a2a9537a | 871 | kfree(work); |
1da177e4 LT |
872 | printk("Emergency Remount complete\n"); |
873 | } | |
874 | ||
875 | void emergency_remount(void) | |
876 | { | |
a2a9537a JA |
877 | struct work_struct *work; |
878 | ||
879 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | |
880 | if (work) { | |
881 | INIT_WORK(work, do_emergency_remount); | |
882 | schedule_work(work); | |
883 | } | |
1da177e4 LT |
884 | } |
885 | ||
886 | /* | |
887 | * Unnamed block devices are dummy devices used by virtual | |
888 | * filesystems which don't use real block-devices. -- jrs | |
889 | */ | |
890 | ||
ad76cbc6 | 891 | static DEFINE_IDA(unnamed_dev_ida); |
1da177e4 | 892 | static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */ |
c63e09ec | 893 | static int unnamed_dev_start = 0; /* don't bother trying below it */ |
1da177e4 | 894 | |
0ee5dc67 | 895 | int get_anon_bdev(dev_t *p) |
1da177e4 LT |
896 | { |
897 | int dev; | |
898 | int error; | |
899 | ||
900 | retry: | |
ad76cbc6 | 901 | if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0) |
1da177e4 LT |
902 | return -ENOMEM; |
903 | spin_lock(&unnamed_dev_lock); | |
c63e09ec | 904 | error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev); |
f21f6220 AV |
905 | if (!error) |
906 | unnamed_dev_start = dev + 1; | |
1da177e4 LT |
907 | spin_unlock(&unnamed_dev_lock); |
908 | if (error == -EAGAIN) | |
909 | /* We raced and lost with another CPU. */ | |
910 | goto retry; | |
911 | else if (error) | |
912 | return -EAGAIN; | |
913 | ||
914 | if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) { | |
915 | spin_lock(&unnamed_dev_lock); | |
ad76cbc6 | 916 | ida_remove(&unnamed_dev_ida, dev); |
f21f6220 AV |
917 | if (unnamed_dev_start > dev) |
918 | unnamed_dev_start = dev; | |
1da177e4 LT |
919 | spin_unlock(&unnamed_dev_lock); |
920 | return -EMFILE; | |
921 | } | |
0ee5dc67 | 922 | *p = MKDEV(0, dev & MINORMASK); |
1da177e4 LT |
923 | return 0; |
924 | } | |
0ee5dc67 | 925 | EXPORT_SYMBOL(get_anon_bdev); |
1da177e4 | 926 | |
0ee5dc67 | 927 | void free_anon_bdev(dev_t dev) |
1da177e4 | 928 | { |
0ee5dc67 | 929 | int slot = MINOR(dev); |
1da177e4 | 930 | spin_lock(&unnamed_dev_lock); |
ad76cbc6 | 931 | ida_remove(&unnamed_dev_ida, slot); |
c63e09ec AV |
932 | if (slot < unnamed_dev_start) |
933 | unnamed_dev_start = slot; | |
1da177e4 LT |
934 | spin_unlock(&unnamed_dev_lock); |
935 | } | |
0ee5dc67 AV |
936 | EXPORT_SYMBOL(free_anon_bdev); |
937 | ||
938 | int set_anon_super(struct super_block *s, void *data) | |
939 | { | |
940 | int error = get_anon_bdev(&s->s_dev); | |
941 | if (!error) | |
942 | s->s_bdi = &noop_backing_dev_info; | |
943 | return error; | |
944 | } | |
945 | ||
946 | EXPORT_SYMBOL(set_anon_super); | |
947 | ||
948 | void kill_anon_super(struct super_block *sb) | |
949 | { | |
950 | dev_t dev = sb->s_dev; | |
951 | generic_shutdown_super(sb); | |
952 | free_anon_bdev(dev); | |
953 | } | |
1da177e4 LT |
954 | |
955 | EXPORT_SYMBOL(kill_anon_super); | |
956 | ||
1da177e4 LT |
957 | void kill_litter_super(struct super_block *sb) |
958 | { | |
959 | if (sb->s_root) | |
960 | d_genocide(sb->s_root); | |
961 | kill_anon_super(sb); | |
962 | } | |
963 | ||
964 | EXPORT_SYMBOL(kill_litter_super); | |
965 | ||
909e6d94 SH |
966 | static int ns_test_super(struct super_block *sb, void *data) |
967 | { | |
968 | return sb->s_fs_info == data; | |
969 | } | |
970 | ||
971 | static int ns_set_super(struct super_block *sb, void *data) | |
972 | { | |
973 | sb->s_fs_info = data; | |
974 | return set_anon_super(sb, NULL); | |
975 | } | |
976 | ||
ceefda69 AV |
977 | struct dentry *mount_ns(struct file_system_type *fs_type, int flags, |
978 | void *data, int (*fill_super)(struct super_block *, void *, int)) | |
909e6d94 SH |
979 | { |
980 | struct super_block *sb; | |
981 | ||
9249e17f | 982 | sb = sget(fs_type, ns_test_super, ns_set_super, flags, data); |
909e6d94 | 983 | if (IS_ERR(sb)) |
ceefda69 | 984 | return ERR_CAST(sb); |
909e6d94 SH |
985 | |
986 | if (!sb->s_root) { | |
987 | int err; | |
909e6d94 SH |
988 | err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0); |
989 | if (err) { | |
74dbbdd7 | 990 | deactivate_locked_super(sb); |
ceefda69 | 991 | return ERR_PTR(err); |
909e6d94 SH |
992 | } |
993 | ||
994 | sb->s_flags |= MS_ACTIVE; | |
995 | } | |
996 | ||
ceefda69 | 997 | return dget(sb->s_root); |
909e6d94 SH |
998 | } |
999 | ||
ceefda69 | 1000 | EXPORT_SYMBOL(mount_ns); |
909e6d94 | 1001 | |
9361401e | 1002 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
1003 | static int set_bdev_super(struct super_block *s, void *data) |
1004 | { | |
1005 | s->s_bdev = data; | |
1006 | s->s_dev = s->s_bdev->bd_dev; | |
32a88aa1 JA |
1007 | |
1008 | /* | |
1009 | * We set the bdi here to the queue backing, file systems can | |
1010 | * overwrite this in ->fill_super() | |
1011 | */ | |
1012 | s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info; | |
1da177e4 LT |
1013 | return 0; |
1014 | } | |
1015 | ||
1016 | static int test_bdev_super(struct super_block *s, void *data) | |
1017 | { | |
1018 | return (void *)s->s_bdev == data; | |
1019 | } | |
1020 | ||
152a0836 | 1021 | struct dentry *mount_bdev(struct file_system_type *fs_type, |
1da177e4 | 1022 | int flags, const char *dev_name, void *data, |
152a0836 | 1023 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1024 | { |
1025 | struct block_device *bdev; | |
1026 | struct super_block *s; | |
d4d77629 | 1027 | fmode_t mode = FMODE_READ | FMODE_EXCL; |
1da177e4 LT |
1028 | int error = 0; |
1029 | ||
30c40d2c AV |
1030 | if (!(flags & MS_RDONLY)) |
1031 | mode |= FMODE_WRITE; | |
1032 | ||
d4d77629 | 1033 | bdev = blkdev_get_by_path(dev_name, mode, fs_type); |
1da177e4 | 1034 | if (IS_ERR(bdev)) |
152a0836 | 1035 | return ERR_CAST(bdev); |
1da177e4 LT |
1036 | |
1037 | /* | |
1038 | * once the super is inserted into the list by sget, s_umount | |
1039 | * will protect the lockfs code from trying to start a snapshot | |
1040 | * while we are mounting | |
1041 | */ | |
4fadd7bb CH |
1042 | mutex_lock(&bdev->bd_fsfreeze_mutex); |
1043 | if (bdev->bd_fsfreeze_count > 0) { | |
1044 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
1045 | error = -EBUSY; | |
1046 | goto error_bdev; | |
1047 | } | |
9249e17f DH |
1048 | s = sget(fs_type, test_bdev_super, set_bdev_super, flags | MS_NOSEC, |
1049 | bdev); | |
4fadd7bb | 1050 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
1da177e4 | 1051 | if (IS_ERR(s)) |
454e2398 | 1052 | goto error_s; |
1da177e4 LT |
1053 | |
1054 | if (s->s_root) { | |
1055 | if ((flags ^ s->s_flags) & MS_RDONLY) { | |
74dbbdd7 | 1056 | deactivate_locked_super(s); |
454e2398 DH |
1057 | error = -EBUSY; |
1058 | goto error_bdev; | |
1da177e4 | 1059 | } |
454e2398 | 1060 | |
4f331f01 TH |
1061 | /* |
1062 | * s_umount nests inside bd_mutex during | |
e525fd89 TH |
1063 | * __invalidate_device(). blkdev_put() acquires |
1064 | * bd_mutex and can't be called under s_umount. Drop | |
1065 | * s_umount temporarily. This is safe as we're | |
1066 | * holding an active reference. | |
4f331f01 TH |
1067 | */ |
1068 | up_write(&s->s_umount); | |
d4d77629 | 1069 | blkdev_put(bdev, mode); |
4f331f01 | 1070 | down_write(&s->s_umount); |
1da177e4 LT |
1071 | } else { |
1072 | char b[BDEVNAME_SIZE]; | |
1073 | ||
30c40d2c | 1074 | s->s_mode = mode; |
1da177e4 | 1075 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); |
e78c9a00 | 1076 | sb_set_blocksize(s, block_size(bdev)); |
9b04c997 | 1077 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1078 | if (error) { |
74dbbdd7 | 1079 | deactivate_locked_super(s); |
454e2398 | 1080 | goto error; |
fa675765 | 1081 | } |
454e2398 DH |
1082 | |
1083 | s->s_flags |= MS_ACTIVE; | |
87d8fe1e | 1084 | bdev->bd_super = s; |
1da177e4 LT |
1085 | } |
1086 | ||
152a0836 | 1087 | return dget(s->s_root); |
1da177e4 | 1088 | |
454e2398 DH |
1089 | error_s: |
1090 | error = PTR_ERR(s); | |
1091 | error_bdev: | |
d4d77629 | 1092 | blkdev_put(bdev, mode); |
454e2398 | 1093 | error: |
152a0836 AV |
1094 | return ERR_PTR(error); |
1095 | } | |
1096 | EXPORT_SYMBOL(mount_bdev); | |
1097 | ||
1da177e4 LT |
1098 | void kill_block_super(struct super_block *sb) |
1099 | { | |
1100 | struct block_device *bdev = sb->s_bdev; | |
30c40d2c | 1101 | fmode_t mode = sb->s_mode; |
1da177e4 | 1102 | |
ddbaaf30 | 1103 | bdev->bd_super = NULL; |
1da177e4 LT |
1104 | generic_shutdown_super(sb); |
1105 | sync_blockdev(bdev); | |
d4d77629 | 1106 | WARN_ON_ONCE(!(mode & FMODE_EXCL)); |
e525fd89 | 1107 | blkdev_put(bdev, mode | FMODE_EXCL); |
1da177e4 LT |
1108 | } |
1109 | ||
1110 | EXPORT_SYMBOL(kill_block_super); | |
9361401e | 1111 | #endif |
1da177e4 | 1112 | |
3c26ff6e | 1113 | struct dentry *mount_nodev(struct file_system_type *fs_type, |
1da177e4 | 1114 | int flags, void *data, |
3c26ff6e | 1115 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1116 | { |
1117 | int error; | |
9249e17f | 1118 | struct super_block *s = sget(fs_type, NULL, set_anon_super, flags, NULL); |
1da177e4 LT |
1119 | |
1120 | if (IS_ERR(s)) | |
3c26ff6e | 1121 | return ERR_CAST(s); |
1da177e4 | 1122 | |
9b04c997 | 1123 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1124 | if (error) { |
74dbbdd7 | 1125 | deactivate_locked_super(s); |
3c26ff6e | 1126 | return ERR_PTR(error); |
1da177e4 LT |
1127 | } |
1128 | s->s_flags |= MS_ACTIVE; | |
3c26ff6e | 1129 | return dget(s->s_root); |
1da177e4 | 1130 | } |
3c26ff6e AV |
1131 | EXPORT_SYMBOL(mount_nodev); |
1132 | ||
1da177e4 LT |
1133 | static int compare_single(struct super_block *s, void *p) |
1134 | { | |
1135 | return 1; | |
1136 | } | |
1137 | ||
fc14f2fe | 1138 | struct dentry *mount_single(struct file_system_type *fs_type, |
1da177e4 | 1139 | int flags, void *data, |
fc14f2fe | 1140 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1141 | { |
1142 | struct super_block *s; | |
1143 | int error; | |
1144 | ||
9249e17f | 1145 | s = sget(fs_type, compare_single, set_anon_super, flags, NULL); |
1da177e4 | 1146 | if (IS_ERR(s)) |
fc14f2fe | 1147 | return ERR_CAST(s); |
1da177e4 | 1148 | if (!s->s_root) { |
9b04c997 | 1149 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1150 | if (error) { |
74dbbdd7 | 1151 | deactivate_locked_super(s); |
fc14f2fe | 1152 | return ERR_PTR(error); |
1da177e4 LT |
1153 | } |
1154 | s->s_flags |= MS_ACTIVE; | |
9329d1be KS |
1155 | } else { |
1156 | do_remount_sb(s, flags, data, 0); | |
1da177e4 | 1157 | } |
fc14f2fe AV |
1158 | return dget(s->s_root); |
1159 | } | |
1160 | EXPORT_SYMBOL(mount_single); | |
1161 | ||
9d412a43 AV |
1162 | struct dentry * |
1163 | mount_fs(struct file_system_type *type, int flags, const char *name, void *data) | |
1da177e4 | 1164 | { |
c96e41e9 | 1165 | struct dentry *root; |
9d412a43 | 1166 | struct super_block *sb; |
1da177e4 | 1167 | char *secdata = NULL; |
9d412a43 | 1168 | int error = -ENOMEM; |
8089352a | 1169 | |
e0007529 | 1170 | if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) { |
1da177e4 | 1171 | secdata = alloc_secdata(); |
454e2398 | 1172 | if (!secdata) |
9d412a43 | 1173 | goto out; |
1da177e4 | 1174 | |
e0007529 | 1175 | error = security_sb_copy_data(data, secdata); |
454e2398 | 1176 | if (error) |
1da177e4 | 1177 | goto out_free_secdata; |
1da177e4 LT |
1178 | } |
1179 | ||
1a102ff9 AV |
1180 | root = type->mount(type, flags, name, data); |
1181 | if (IS_ERR(root)) { | |
1182 | error = PTR_ERR(root); | |
1183 | goto out_free_secdata; | |
c96e41e9 | 1184 | } |
9d412a43 AV |
1185 | sb = root->d_sb; |
1186 | BUG_ON(!sb); | |
1187 | WARN_ON(!sb->s_bdi); | |
6c510389 | 1188 | WARN_ON(sb->s_bdi == &default_backing_dev_info); |
9d412a43 | 1189 | sb->s_flags |= MS_BORN; |
454e2398 | 1190 | |
9d412a43 | 1191 | error = security_sb_kern_mount(sb, flags, secdata); |
5129a469 JE |
1192 | if (error) |
1193 | goto out_sb; | |
454e2398 | 1194 | |
42cb56ae JL |
1195 | /* |
1196 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE | |
1197 | * but s_maxbytes was an unsigned long long for many releases. Throw | |
1198 | * this warning for a little while to try and catch filesystems that | |
4358b567 | 1199 | * violate this rule. |
42cb56ae | 1200 | */ |
9d412a43 AV |
1201 | WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " |
1202 | "negative value (%lld)\n", type->name, sb->s_maxbytes); | |
42cb56ae | 1203 | |
9d412a43 | 1204 | up_write(&sb->s_umount); |
8680e22f | 1205 | free_secdata(secdata); |
9d412a43 | 1206 | return root; |
1da177e4 | 1207 | out_sb: |
9d412a43 AV |
1208 | dput(root); |
1209 | deactivate_locked_super(sb); | |
1da177e4 LT |
1210 | out_free_secdata: |
1211 | free_secdata(secdata); | |
1da177e4 | 1212 | out: |
454e2398 | 1213 | return ERR_PTR(error); |
1da177e4 LT |
1214 | } |
1215 | ||
5accdf82 JK |
1216 | /* |
1217 | * This is an internal function, please use sb_end_{write,pagefault,intwrite} | |
1218 | * instead. | |
1219 | */ | |
1220 | void __sb_end_write(struct super_block *sb, int level) | |
1221 | { | |
1222 | percpu_counter_dec(&sb->s_writers.counter[level-1]); | |
1223 | /* | |
1224 | * Make sure s_writers are updated before we wake up waiters in | |
1225 | * freeze_super(). | |
1226 | */ | |
1227 | smp_mb(); | |
1228 | if (waitqueue_active(&sb->s_writers.wait)) | |
1229 | wake_up(&sb->s_writers.wait); | |
1230 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _RET_IP_); | |
1231 | } | |
1232 | EXPORT_SYMBOL(__sb_end_write); | |
1233 | ||
1234 | #ifdef CONFIG_LOCKDEP | |
1235 | /* | |
1236 | * We want lockdep to tell us about possible deadlocks with freezing but | |
1237 | * it's it bit tricky to properly instrument it. Getting a freeze protection | |
1238 | * works as getting a read lock but there are subtle problems. XFS for example | |
1239 | * gets freeze protection on internal level twice in some cases, which is OK | |
1240 | * only because we already hold a freeze protection also on higher level. Due | |
1241 | * to these cases we have to tell lockdep we are doing trylock when we | |
1242 | * already hold a freeze protection for a higher freeze level. | |
1243 | */ | |
1244 | static void acquire_freeze_lock(struct super_block *sb, int level, bool trylock, | |
1245 | unsigned long ip) | |
1246 | { | |
1247 | int i; | |
1248 | ||
1249 | if (!trylock) { | |
1250 | for (i = 0; i < level - 1; i++) | |
1251 | if (lock_is_held(&sb->s_writers.lock_map[i])) { | |
1252 | trylock = true; | |
1253 | break; | |
1254 | } | |
1255 | } | |
1256 | rwsem_acquire_read(&sb->s_writers.lock_map[level-1], 0, trylock, ip); | |
1257 | } | |
1258 | #endif | |
1259 | ||
1260 | /* | |
1261 | * This is an internal function, please use sb_start_{write,pagefault,intwrite} | |
1262 | * instead. | |
1263 | */ | |
1264 | int __sb_start_write(struct super_block *sb, int level, bool wait) | |
1265 | { | |
1266 | retry: | |
1267 | if (unlikely(sb->s_writers.frozen >= level)) { | |
1268 | if (!wait) | |
1269 | return 0; | |
1270 | wait_event(sb->s_writers.wait_unfrozen, | |
1271 | sb->s_writers.frozen < level); | |
1272 | } | |
1273 | ||
1274 | #ifdef CONFIG_LOCKDEP | |
1275 | acquire_freeze_lock(sb, level, !wait, _RET_IP_); | |
1276 | #endif | |
1277 | percpu_counter_inc(&sb->s_writers.counter[level-1]); | |
1278 | /* | |
1279 | * Make sure counter is updated before we check for frozen. | |
1280 | * freeze_super() first sets frozen and then checks the counter. | |
1281 | */ | |
1282 | smp_mb(); | |
1283 | if (unlikely(sb->s_writers.frozen >= level)) { | |
1284 | __sb_end_write(sb, level); | |
1285 | goto retry; | |
1286 | } | |
1287 | return 1; | |
1288 | } | |
1289 | EXPORT_SYMBOL(__sb_start_write); | |
1290 | ||
1291 | /** | |
1292 | * sb_wait_write - wait until all writers to given file system finish | |
1293 | * @sb: the super for which we wait | |
1294 | * @level: type of writers we wait for (normal vs page fault) | |
1295 | * | |
1296 | * This function waits until there are no writers of given type to given file | |
1297 | * system. Caller of this function should make sure there can be no new writers | |
1298 | * of type @level before calling this function. Otherwise this function can | |
1299 | * livelock. | |
1300 | */ | |
1301 | static void sb_wait_write(struct super_block *sb, int level) | |
1302 | { | |
1303 | s64 writers; | |
1304 | ||
1305 | /* | |
1306 | * We just cycle-through lockdep here so that it does not complain | |
1307 | * about returning with lock to userspace | |
1308 | */ | |
1309 | rwsem_acquire(&sb->s_writers.lock_map[level-1], 0, 0, _THIS_IP_); | |
1310 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _THIS_IP_); | |
1311 | ||
1312 | do { | |
1313 | DEFINE_WAIT(wait); | |
1314 | ||
1315 | /* | |
1316 | * We use a barrier in prepare_to_wait() to separate setting | |
1317 | * of frozen and checking of the counter | |
1318 | */ | |
1319 | prepare_to_wait(&sb->s_writers.wait, &wait, | |
1320 | TASK_UNINTERRUPTIBLE); | |
1321 | ||
1322 | writers = percpu_counter_sum(&sb->s_writers.counter[level-1]); | |
1323 | if (writers) | |
1324 | schedule(); | |
1325 | ||
1326 | finish_wait(&sb->s_writers.wait, &wait); | |
1327 | } while (writers); | |
1328 | } | |
1329 | ||
18e9e510 | 1330 | /** |
7000d3c4 RD |
1331 | * freeze_super - lock the filesystem and force it into a consistent state |
1332 | * @sb: the super to lock | |
18e9e510 JB |
1333 | * |
1334 | * Syncs the super to make sure the filesystem is consistent and calls the fs's | |
1335 | * freeze_fs. Subsequent calls to this without first thawing the fs will return | |
1336 | * -EBUSY. | |
5accdf82 JK |
1337 | * |
1338 | * During this function, sb->s_writers.frozen goes through these values: | |
1339 | * | |
1340 | * SB_UNFROZEN: File system is normal, all writes progress as usual. | |
1341 | * | |
1342 | * SB_FREEZE_WRITE: The file system is in the process of being frozen. New | |
1343 | * writes should be blocked, though page faults are still allowed. We wait for | |
1344 | * all writes to complete and then proceed to the next stage. | |
1345 | * | |
1346 | * SB_FREEZE_PAGEFAULT: Freezing continues. Now also page faults are blocked | |
1347 | * but internal fs threads can still modify the filesystem (although they | |
1348 | * should not dirty new pages or inodes), writeback can run etc. After waiting | |
1349 | * for all running page faults we sync the filesystem which will clean all | |
1350 | * dirty pages and inodes (no new dirty pages or inodes can be created when | |
1351 | * sync is running). | |
1352 | * | |
1353 | * SB_FREEZE_FS: The file system is frozen. Now all internal sources of fs | |
1354 | * modification are blocked (e.g. XFS preallocation truncation on inode | |
1355 | * reclaim). This is usually implemented by blocking new transactions for | |
1356 | * filesystems that have them and need this additional guard. After all | |
1357 | * internal writers are finished we call ->freeze_fs() to finish filesystem | |
1358 | * freezing. Then we transition to SB_FREEZE_COMPLETE state. This state is | |
1359 | * mostly auxiliary for filesystems to verify they do not modify frozen fs. | |
1360 | * | |
1361 | * sb->s_writers.frozen is protected by sb->s_umount. | |
18e9e510 JB |
1362 | */ |
1363 | int freeze_super(struct super_block *sb) | |
1364 | { | |
1365 | int ret; | |
1366 | ||
1367 | atomic_inc(&sb->s_active); | |
1368 | down_write(&sb->s_umount); | |
5accdf82 | 1369 | if (sb->s_writers.frozen != SB_UNFROZEN) { |
18e9e510 JB |
1370 | deactivate_locked_super(sb); |
1371 | return -EBUSY; | |
1372 | } | |
1373 | ||
dabe0dc1 AV |
1374 | if (!(sb->s_flags & MS_BORN)) { |
1375 | up_write(&sb->s_umount); | |
1376 | return 0; /* sic - it's "nothing to do" */ | |
1377 | } | |
1378 | ||
18e9e510 | 1379 | if (sb->s_flags & MS_RDONLY) { |
5accdf82 JK |
1380 | /* Nothing to do really... */ |
1381 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
18e9e510 JB |
1382 | up_write(&sb->s_umount); |
1383 | return 0; | |
1384 | } | |
1385 | ||
5accdf82 JK |
1386 | /* From now on, no new normal writers can start */ |
1387 | sb->s_writers.frozen = SB_FREEZE_WRITE; | |
1388 | smp_wmb(); | |
1389 | ||
1390 | /* Release s_umount to preserve sb_start_write -> s_umount ordering */ | |
1391 | up_write(&sb->s_umount); | |
1392 | ||
1393 | sb_wait_write(sb, SB_FREEZE_WRITE); | |
1394 | ||
1395 | /* Now we go and block page faults... */ | |
1396 | down_write(&sb->s_umount); | |
1397 | sb->s_writers.frozen = SB_FREEZE_PAGEFAULT; | |
18e9e510 JB |
1398 | smp_wmb(); |
1399 | ||
5accdf82 JK |
1400 | sb_wait_write(sb, SB_FREEZE_PAGEFAULT); |
1401 | ||
1402 | /* All writers are done so after syncing there won't be dirty data */ | |
18e9e510 JB |
1403 | sync_filesystem(sb); |
1404 | ||
5accdf82 JK |
1405 | /* Now wait for internal filesystem counter */ |
1406 | sb->s_writers.frozen = SB_FREEZE_FS; | |
18e9e510 | 1407 | smp_wmb(); |
5accdf82 | 1408 | sb_wait_write(sb, SB_FREEZE_FS); |
18e9e510 | 1409 | |
18e9e510 JB |
1410 | if (sb->s_op->freeze_fs) { |
1411 | ret = sb->s_op->freeze_fs(sb); | |
1412 | if (ret) { | |
1413 | printk(KERN_ERR | |
1414 | "VFS:Filesystem freeze failed\n"); | |
5accdf82 | 1415 | sb->s_writers.frozen = SB_UNFROZEN; |
e1616300 | 1416 | smp_wmb(); |
5accdf82 | 1417 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 JB |
1418 | deactivate_locked_super(sb); |
1419 | return ret; | |
1420 | } | |
1421 | } | |
5accdf82 JK |
1422 | /* |
1423 | * This is just for debugging purposes so that fs can warn if it | |
1424 | * sees write activity when frozen is set to SB_FREEZE_COMPLETE. | |
1425 | */ | |
1426 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
18e9e510 JB |
1427 | up_write(&sb->s_umount); |
1428 | return 0; | |
1429 | } | |
1430 | EXPORT_SYMBOL(freeze_super); | |
1431 | ||
1432 | /** | |
1433 | * thaw_super -- unlock filesystem | |
1434 | * @sb: the super to thaw | |
1435 | * | |
1436 | * Unlocks the filesystem and marks it writeable again after freeze_super(). | |
1437 | */ | |
1438 | int thaw_super(struct super_block *sb) | |
1439 | { | |
1440 | int error; | |
1441 | ||
1442 | down_write(&sb->s_umount); | |
5accdf82 | 1443 | if (sb->s_writers.frozen == SB_UNFROZEN) { |
18e9e510 JB |
1444 | up_write(&sb->s_umount); |
1445 | return -EINVAL; | |
1446 | } | |
1447 | ||
1448 | if (sb->s_flags & MS_RDONLY) | |
1449 | goto out; | |
1450 | ||
1451 | if (sb->s_op->unfreeze_fs) { | |
1452 | error = sb->s_op->unfreeze_fs(sb); | |
1453 | if (error) { | |
1454 | printk(KERN_ERR | |
1455 | "VFS:Filesystem thaw failed\n"); | |
18e9e510 JB |
1456 | up_write(&sb->s_umount); |
1457 | return error; | |
1458 | } | |
1459 | } | |
1460 | ||
1461 | out: | |
5accdf82 | 1462 | sb->s_writers.frozen = SB_UNFROZEN; |
18e9e510 | 1463 | smp_wmb(); |
5accdf82 | 1464 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 JB |
1465 | deactivate_locked_super(sb); |
1466 | ||
1467 | return 0; | |
1468 | } | |
1469 | EXPORT_SYMBOL(thaw_super); |