]>
Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 LT |
2 | /* |
3 | * linux/fs/super.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992 Linus Torvalds | |
6 | * | |
7 | * super.c contains code to handle: - mount structures | |
8 | * - super-block tables | |
9 | * - filesystem drivers list | |
10 | * - mount system call | |
11 | * - umount system call | |
12 | * - ustat system call | |
13 | * | |
14 | * GK 2/5/95 - Changed to support mounting the root fs via NFS | |
15 | * | |
16 | * Added kerneld support: Jacques Gelinas and Bjorn Ekwall | |
17 | * Added change_root: Werner Almesberger & Hans Lermen, Feb '96 | |
18 | * Added options to /proc/mounts: | |
96de0e25 | 19 | * Torbjörn Lindh (torbjorn.lindh@gopta.se), April 14, 1996. |
1da177e4 LT |
20 | * Added devfs support: Richard Gooch <rgooch@atnf.csiro.au>, 13-JAN-1998 |
21 | * Heavily rewritten for 'one fs - one tree' dcache architecture. AV, Mar 2000 | |
22 | */ | |
23 | ||
630d9c47 | 24 | #include <linux/export.h> |
1da177e4 | 25 | #include <linux/slab.h> |
1da177e4 | 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> |
22d94f49 | 34 | #include <linux/fscrypt.h> |
40401530 | 35 | #include <linux/fsnotify.h> |
5accdf82 | 36 | #include <linux/lockdep.h> |
6e4eab57 | 37 | #include <linux/user_namespace.h> |
9bc61ab1 | 38 | #include <linux/fs_context.h> |
e262e32d | 39 | #include <uapi/linux/mount.h> |
6d59e7f5 | 40 | #include "internal.h" |
1da177e4 | 41 | |
08fdc8a0 | 42 | static int thaw_super_locked(struct super_block *sb); |
1da177e4 | 43 | |
15d0f5ea AV |
44 | static LIST_HEAD(super_blocks); |
45 | static DEFINE_SPINLOCK(sb_lock); | |
1da177e4 | 46 | |
5accdf82 JK |
47 | static char *sb_writers_name[SB_FREEZE_LEVELS] = { |
48 | "sb_writers", | |
49 | "sb_pagefaults", | |
50 | "sb_internal", | |
51 | }; | |
52 | ||
5e874914 | 53 | static inline void __super_lock(struct super_block *sb, bool excl) |
0ed33598 CB |
54 | { |
55 | if (excl) | |
56 | down_write(&sb->s_umount); | |
57 | else | |
58 | down_read(&sb->s_umount); | |
59 | } | |
60 | ||
61 | static inline void super_unlock(struct super_block *sb, bool excl) | |
62 | { | |
63 | if (excl) | |
64 | up_write(&sb->s_umount); | |
65 | else | |
66 | up_read(&sb->s_umount); | |
67 | } | |
68 | ||
5e874914 | 69 | static inline void __super_lock_excl(struct super_block *sb) |
0ed33598 | 70 | { |
5e874914 | 71 | __super_lock(sb, true); |
0ed33598 CB |
72 | } |
73 | ||
74 | static inline void super_unlock_excl(struct super_block *sb) | |
75 | { | |
76 | super_unlock(sb, true); | |
77 | } | |
78 | ||
79 | static inline void super_unlock_shared(struct super_block *sb) | |
80 | { | |
81 | super_unlock(sb, false); | |
82 | } | |
83 | ||
5e874914 CB |
84 | static inline bool wait_born(struct super_block *sb) |
85 | { | |
86 | unsigned int flags; | |
87 | ||
88 | /* | |
89 | * Pairs with smp_store_release() in super_wake() and ensures | |
90 | * that we see SB_BORN or SB_DYING after we're woken. | |
91 | */ | |
92 | flags = smp_load_acquire(&sb->s_flags); | |
93 | return flags & (SB_BORN | SB_DYING); | |
94 | } | |
95 | ||
96 | /** | |
97 | * super_lock - wait for superblock to become ready and lock it | |
98 | * @sb: superblock to wait for | |
99 | * @excl: whether exclusive access is required | |
100 | * | |
101 | * If the superblock has neither passed through vfs_get_tree() or | |
102 | * generic_shutdown_super() yet wait for it to happen. Either superblock | |
103 | * creation will succeed and SB_BORN is set by vfs_get_tree() or we're | |
104 | * woken and we'll see SB_DYING. | |
105 | * | |
106 | * The caller must have acquired a temporary reference on @sb->s_count. | |
107 | * | |
108 | * Return: This returns true if SB_BORN was set, false if SB_DYING was | |
109 | * set. The function acquires s_umount and returns with it held. | |
110 | */ | |
111 | static __must_check bool super_lock(struct super_block *sb, bool excl) | |
112 | { | |
113 | ||
114 | lockdep_assert_not_held(&sb->s_umount); | |
115 | ||
116 | relock: | |
117 | __super_lock(sb, excl); | |
118 | ||
119 | /* | |
120 | * Has gone through generic_shutdown_super() in the meantime. | |
121 | * @sb->s_root is NULL and @sb->s_active is 0. No one needs to | |
122 | * grab a reference to this. Tell them so. | |
123 | */ | |
124 | if (sb->s_flags & SB_DYING) | |
125 | return false; | |
126 | ||
127 | /* Has called ->get_tree() successfully. */ | |
128 | if (sb->s_flags & SB_BORN) | |
129 | return true; | |
130 | ||
131 | super_unlock(sb, excl); | |
132 | ||
133 | /* wait until the superblock is ready or dying */ | |
134 | wait_var_event(&sb->s_flags, wait_born(sb)); | |
135 | ||
136 | /* | |
137 | * Neither SB_BORN nor SB_DYING are ever unset so we never loop. | |
138 | * Just reacquire @sb->s_umount for the caller. | |
139 | */ | |
140 | goto relock; | |
141 | } | |
142 | ||
143 | /* wait and acquire read-side of @sb->s_umount */ | |
144 | static inline bool super_lock_shared(struct super_block *sb) | |
145 | { | |
146 | return super_lock(sb, false); | |
147 | } | |
148 | ||
149 | /* wait and acquire write-side of @sb->s_umount */ | |
150 | static inline bool super_lock_excl(struct super_block *sb) | |
151 | { | |
152 | return super_lock(sb, true); | |
153 | } | |
154 | ||
155 | /* wake waiters */ | |
2c18a63b | 156 | #define SUPER_WAKE_FLAGS (SB_BORN | SB_DYING | SB_DEAD) |
5e874914 CB |
157 | static void super_wake(struct super_block *sb, unsigned int flag) |
158 | { | |
159 | WARN_ON_ONCE((flag & ~SUPER_WAKE_FLAGS)); | |
160 | WARN_ON_ONCE(hweight32(flag & SUPER_WAKE_FLAGS) > 1); | |
161 | ||
162 | /* | |
163 | * Pairs with smp_load_acquire() in super_lock() to make sure | |
164 | * all initializations in the superblock are seen by the user | |
165 | * seeing SB_BORN sent. | |
166 | */ | |
167 | smp_store_release(&sb->s_flags, sb->s_flags | flag); | |
168 | /* | |
169 | * Pairs with the barrier in prepare_to_wait_event() to make sure | |
170 | * ___wait_var_event() either sees SB_BORN set or | |
171 | * waitqueue_active() check in wake_up_var() sees the waiter. | |
172 | */ | |
173 | smp_mb(); | |
174 | wake_up_var(&sb->s_flags); | |
175 | } | |
176 | ||
b0d40c92 DC |
177 | /* |
178 | * One thing we have to be careful of with a per-sb shrinker is that we don't | |
179 | * drop the last active reference to the superblock from within the shrinker. | |
180 | * If that happens we could trigger unregistering the shrinker from within the | |
47a7c01c | 181 | * shrinker path and that leads to deadlock on the shrinker_rwsem. Hence we |
b0d40c92 DC |
182 | * take a passive reference to the superblock to avoid this from occurring. |
183 | */ | |
0a234c6d DC |
184 | static unsigned long super_cache_scan(struct shrinker *shrink, |
185 | struct shrink_control *sc) | |
b0d40c92 DC |
186 | { |
187 | struct super_block *sb; | |
0a234c6d DC |
188 | long fs_objects = 0; |
189 | long total_objects; | |
190 | long freed = 0; | |
191 | long dentries; | |
192 | long inodes; | |
b0d40c92 DC |
193 | |
194 | sb = container_of(shrink, struct super_block, s_shrink); | |
195 | ||
196 | /* | |
197 | * Deadlock avoidance. We may hold various FS locks, and we don't want | |
198 | * to recurse into the FS that called us in clear_inode() and friends.. | |
199 | */ | |
0a234c6d DC |
200 | if (!(sc->gfp_mask & __GFP_FS)) |
201 | return SHRINK_STOP; | |
b0d40c92 | 202 | |
d8ce82ef | 203 | if (!super_trylock_shared(sb)) |
0a234c6d | 204 | return SHRINK_STOP; |
b0d40c92 | 205 | |
d0407903 | 206 | if (sb->s_op->nr_cached_objects) |
4101b624 | 207 | fs_objects = sb->s_op->nr_cached_objects(sb, sc); |
0e1fdafd | 208 | |
503c358c VD |
209 | inodes = list_lru_shrink_count(&sb->s_inode_lru, sc); |
210 | dentries = list_lru_shrink_count(&sb->s_dentry_lru, sc); | |
f6041567 | 211 | total_objects = dentries + inodes + fs_objects + 1; |
475d0db7 TH |
212 | if (!total_objects) |
213 | total_objects = 1; | |
0e1fdafd | 214 | |
0a234c6d | 215 | /* proportion the scan between the caches */ |
f6041567 | 216 | dentries = mult_frac(sc->nr_to_scan, dentries, total_objects); |
bc3b14cb | 217 | inodes = mult_frac(sc->nr_to_scan, inodes, total_objects); |
503c358c | 218 | fs_objects = mult_frac(sc->nr_to_scan, fs_objects, total_objects); |
b0d40c92 | 219 | |
0a234c6d DC |
220 | /* |
221 | * prune the dcache first as the icache is pinned by it, then | |
222 | * prune the icache, followed by the filesystem specific caches | |
49e7e7ff VD |
223 | * |
224 | * Ensure that we always scan at least one object - memcg kmem | |
225 | * accounting uses this to fully empty the caches. | |
0a234c6d | 226 | */ |
49e7e7ff | 227 | sc->nr_to_scan = dentries + 1; |
503c358c | 228 | freed = prune_dcache_sb(sb, sc); |
49e7e7ff | 229 | sc->nr_to_scan = inodes + 1; |
503c358c | 230 | freed += prune_icache_sb(sb, sc); |
0a234c6d DC |
231 | |
232 | if (fs_objects) { | |
49e7e7ff | 233 | sc->nr_to_scan = fs_objects + 1; |
4101b624 | 234 | freed += sb->s_op->free_cached_objects(sb, sc); |
b0d40c92 DC |
235 | } |
236 | ||
0ed33598 | 237 | super_unlock_shared(sb); |
0a234c6d DC |
238 | return freed; |
239 | } | |
240 | ||
241 | static unsigned long super_cache_count(struct shrinker *shrink, | |
242 | struct shrink_control *sc) | |
243 | { | |
244 | struct super_block *sb; | |
245 | long total_objects = 0; | |
246 | ||
247 | sb = container_of(shrink, struct super_block, s_shrink); | |
248 | ||
d23da150 | 249 | /* |
d8ce82ef CB |
250 | * We don't call super_trylock_shared() here as it is a scalability |
251 | * bottleneck, so we're exposed to partial setup state. The shrinker | |
252 | * rwsem does not protect filesystem operations backing | |
253 | * list_lru_shrink_count() or s_op->nr_cached_objects(). Counts can | |
254 | * change between super_cache_count and super_cache_scan, so we really | |
255 | * don't need locks here. | |
79f546a6 DC |
256 | * |
257 | * However, if we are currently mounting the superblock, the underlying | |
258 | * filesystem might be in a state of partial construction and hence it | |
d8ce82ef CB |
259 | * is dangerous to access it. super_trylock_shared() uses a SB_BORN check |
260 | * to avoid this situation, so do the same here. The memory barrier is | |
79f546a6 | 261 | * matched with the one in mount_fs() as we don't hold locks here. |
d23da150 | 262 | */ |
79f546a6 DC |
263 | if (!(sb->s_flags & SB_BORN)) |
264 | return 0; | |
265 | smp_rmb(); | |
266 | ||
0a234c6d | 267 | if (sb->s_op && sb->s_op->nr_cached_objects) |
4101b624 | 268 | total_objects = sb->s_op->nr_cached_objects(sb, sc); |
0a234c6d | 269 | |
503c358c VD |
270 | total_objects += list_lru_shrink_count(&sb->s_dentry_lru, sc); |
271 | total_objects += list_lru_shrink_count(&sb->s_inode_lru, sc); | |
0a234c6d | 272 | |
9b996468 KT |
273 | if (!total_objects) |
274 | return SHRINK_EMPTY; | |
275 | ||
55f841ce | 276 | total_objects = vfs_pressure_ratio(total_objects); |
0e1fdafd | 277 | return total_objects; |
b0d40c92 DC |
278 | } |
279 | ||
853b39a7 ON |
280 | static void destroy_super_work(struct work_struct *work) |
281 | { | |
282 | struct super_block *s = container_of(work, struct super_block, | |
283 | destroy_work); | |
284 | int i; | |
285 | ||
286 | for (i = 0; i < SB_FREEZE_LEVELS; i++) | |
8129ed29 | 287 | percpu_free_rwsem(&s->s_writers.rw_sem[i]); |
853b39a7 ON |
288 | kfree(s); |
289 | } | |
290 | ||
291 | static void destroy_super_rcu(struct rcu_head *head) | |
292 | { | |
293 | struct super_block *s = container_of(head, struct super_block, rcu); | |
294 | INIT_WORK(&s->destroy_work, destroy_super_work); | |
295 | schedule_work(&s->destroy_work); | |
296 | } | |
297 | ||
0200894d AV |
298 | /* Free a superblock that has never been seen by anyone */ |
299 | static void destroy_unused_super(struct super_block *s) | |
5accdf82 | 300 | { |
0200894d AV |
301 | if (!s) |
302 | return; | |
0ed33598 | 303 | super_unlock_excl(s); |
7eb5e882 AV |
304 | list_lru_destroy(&s->s_dentry_lru); |
305 | list_lru_destroy(&s->s_inode_lru); | |
7eb5e882 | 306 | security_sb_free(s); |
6e4eab57 | 307 | put_user_ns(s->s_user_ns); |
7eb5e882 | 308 | kfree(s->s_subtype); |
8e04944f | 309 | free_prealloced_shrinker(&s->s_shrink); |
0200894d AV |
310 | /* no delays needed */ |
311 | destroy_super_work(&s->destroy_work); | |
5accdf82 JK |
312 | } |
313 | ||
1da177e4 LT |
314 | /** |
315 | * alloc_super - create new superblock | |
fe2bbc48 | 316 | * @type: filesystem type superblock should belong to |
9249e17f | 317 | * @flags: the mount flags |
6e4eab57 | 318 | * @user_ns: User namespace for the super_block |
1da177e4 LT |
319 | * |
320 | * Allocates and initializes a new &struct super_block. alloc_super() | |
321 | * returns a pointer new superblock or %NULL if allocation had failed. | |
322 | */ | |
6e4eab57 EB |
323 | static struct super_block *alloc_super(struct file_system_type *type, int flags, |
324 | struct user_namespace *user_ns) | |
1da177e4 | 325 | { |
11b0b5ab | 326 | struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER); |
b87221de | 327 | static const struct super_operations default_op; |
7eb5e882 AV |
328 | int i; |
329 | ||
330 | if (!s) | |
331 | return NULL; | |
1da177e4 | 332 | |
b5bd856a | 333 | INIT_LIST_HEAD(&s->s_mounts); |
6e4eab57 | 334 | s->s_user_ns = get_user_ns(user_ns); |
ca0168e8 AV |
335 | init_rwsem(&s->s_umount); |
336 | lockdep_set_class(&s->s_umount, &type->s_umount_key); | |
337 | /* | |
338 | * sget() can have s_umount recursion. | |
339 | * | |
340 | * When it cannot find a suitable sb, it allocates a new | |
341 | * one (this one), and tries again to find a suitable old | |
342 | * one. | |
343 | * | |
344 | * In case that succeeds, it will acquire the s_umount | |
345 | * lock of the old one. Since these are clearly distrinct | |
346 | * locks, and this object isn't exposed yet, there's no | |
347 | * risk of deadlocks. | |
348 | * | |
349 | * Annotate this by putting this lock in a different | |
350 | * subclass. | |
351 | */ | |
352 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); | |
b5bd856a | 353 | |
7eb5e882 AV |
354 | if (security_sb_alloc(s)) |
355 | goto fail; | |
7b7a8665 | 356 | |
7eb5e882 | 357 | for (i = 0; i < SB_FREEZE_LEVELS; i++) { |
8129ed29 ON |
358 | if (__percpu_init_rwsem(&s->s_writers.rw_sem[i], |
359 | sb_writers_name[i], | |
360 | &type->s_writers_key[i])) | |
7eb5e882 | 361 | goto fail; |
1da177e4 | 362 | } |
df0ce26c | 363 | s->s_bdi = &noop_backing_dev_info; |
7eb5e882 | 364 | s->s_flags = flags; |
cc50a07a | 365 | if (s->s_user_ns != &init_user_ns) |
67690f93 | 366 | s->s_iflags |= SB_I_NODEV; |
7eb5e882 | 367 | INIT_HLIST_NODE(&s->s_instances); |
f1ee6162 | 368 | INIT_HLIST_BL_HEAD(&s->s_roots); |
e97fedb9 | 369 | mutex_init(&s->s_sync_lock); |
7eb5e882 | 370 | INIT_LIST_HEAD(&s->s_inodes); |
74278da9 | 371 | spin_lock_init(&s->s_inode_list_lock); |
6c60d2b5 DC |
372 | INIT_LIST_HEAD(&s->s_inodes_wb); |
373 | spin_lock_init(&s->s_inode_wblist_lock); | |
7eb5e882 | 374 | |
7eb5e882 AV |
375 | s->s_count = 1; |
376 | atomic_set(&s->s_active, 1); | |
377 | mutex_init(&s->s_vfs_rename_mutex); | |
378 | lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key); | |
bc8230ee | 379 | init_rwsem(&s->s_dquot.dqio_sem); |
7eb5e882 AV |
380 | s->s_maxbytes = MAX_NON_LFS; |
381 | s->s_op = &default_op; | |
382 | s->s_time_gran = 1000000000; | |
188d20bc DD |
383 | s->s_time_min = TIME64_MIN; |
384 | s->s_time_max = TIME64_MAX; | |
7eb5e882 AV |
385 | |
386 | s->s_shrink.seeks = DEFAULT_SEEKS; | |
387 | s->s_shrink.scan_objects = super_cache_scan; | |
388 | s->s_shrink.count_objects = super_cache_count; | |
389 | s->s_shrink.batch = 1024; | |
2acb60a0 | 390 | s->s_shrink.flags = SHRINKER_NUMA_AWARE | SHRINKER_MEMCG_AWARE; |
e33c267a | 391 | if (prealloc_shrinker(&s->s_shrink, "sb-%s", type->name)) |
8e04944f | 392 | goto fail; |
c92e8e10 | 393 | if (list_lru_init_memcg(&s->s_dentry_lru, &s->s_shrink)) |
2b3648a6 | 394 | goto fail; |
c92e8e10 | 395 | if (list_lru_init_memcg(&s->s_inode_lru, &s->s_shrink)) |
2b3648a6 | 396 | goto fail; |
1da177e4 | 397 | return s; |
5ca302c8 | 398 | |
7eb5e882 | 399 | fail: |
0200894d | 400 | destroy_unused_super(s); |
7eb5e882 | 401 | return NULL; |
1da177e4 LT |
402 | } |
403 | ||
404 | /* Superblock refcounting */ | |
405 | ||
406 | /* | |
35cf7ba0 | 407 | * Drop a superblock's refcount. The caller must hold sb_lock. |
1da177e4 | 408 | */ |
c645b930 | 409 | static void __put_super(struct super_block *s) |
1da177e4 | 410 | { |
c645b930 AV |
411 | if (!--s->s_count) { |
412 | list_del_init(&s->s_list); | |
413 | WARN_ON(s->s_dentry_lru.node); | |
414 | WARN_ON(s->s_inode_lru.node); | |
415 | WARN_ON(!list_empty(&s->s_mounts)); | |
416 | security_sb_free(s); | |
417 | put_user_ns(s->s_user_ns); | |
418 | kfree(s->s_subtype); | |
419 | call_rcu(&s->rcu, destroy_super_rcu); | |
1da177e4 | 420 | } |
1da177e4 LT |
421 | } |
422 | ||
423 | /** | |
424 | * put_super - drop a temporary reference to superblock | |
425 | * @sb: superblock in question | |
426 | * | |
427 | * Drops a temporary reference, frees superblock if there's no | |
428 | * references left. | |
429 | */ | |
60b49885 | 430 | void put_super(struct super_block *sb) |
1da177e4 LT |
431 | { |
432 | spin_lock(&sb_lock); | |
433 | __put_super(sb); | |
434 | spin_unlock(&sb_lock); | |
435 | } | |
436 | ||
437 | ||
438 | /** | |
1712ac8f | 439 | * deactivate_locked_super - drop an active reference to superblock |
1da177e4 LT |
440 | * @s: superblock to deactivate |
441 | * | |
bd7ced98 | 442 | * Drops an active reference to superblock, converting it into a temporary |
1712ac8f | 443 | * one if there is no other active references left. In that case we |
1da177e4 LT |
444 | * tell fs driver to shut it down and drop the temporary reference we |
445 | * had just acquired. | |
1712ac8f AV |
446 | * |
447 | * Caller holds exclusive lock on superblock; that lock is released. | |
1da177e4 | 448 | */ |
1712ac8f | 449 | void deactivate_locked_super(struct super_block *s) |
1da177e4 LT |
450 | { |
451 | struct file_system_type *fs = s->s_type; | |
b20bd1a5 | 452 | if (atomic_dec_and_test(&s->s_active)) { |
b0d40c92 | 453 | unregister_shrinker(&s->s_shrink); |
28f2cd4f | 454 | fs->kill_sb(s); |
f5e1dd34 | 455 | |
c0a5b560 VD |
456 | /* |
457 | * Since list_lru_destroy() may sleep, we cannot call it from | |
458 | * put_super(), where we hold the sb_lock. Therefore we destroy | |
459 | * the lru lists right now. | |
460 | */ | |
461 | list_lru_destroy(&s->s_dentry_lru); | |
462 | list_lru_destroy(&s->s_inode_lru); | |
463 | ||
2c18a63b CB |
464 | /* |
465 | * Remove it from @fs_supers so it isn't found by new | |
466 | * sget{_fc}() walkers anymore. Any concurrent mounter still | |
467 | * managing to grab a temporary reference is guaranteed to | |
468 | * already see SB_DYING and will wait until we notify them about | |
469 | * SB_DEAD. | |
470 | */ | |
471 | spin_lock(&sb_lock); | |
472 | hlist_del_init(&s->s_instances); | |
473 | spin_unlock(&sb_lock); | |
474 | ||
475 | /* | |
476 | * Let concurrent mounts know that this thing is really dead. | |
477 | * We don't need @sb->s_umount here as every concurrent caller | |
478 | * will see SB_DYING and either discard the superblock or wait | |
479 | * for SB_DEAD. | |
480 | */ | |
481 | super_wake(s, SB_DEAD); | |
482 | ||
1da177e4 LT |
483 | put_filesystem(fs); |
484 | put_super(s); | |
1712ac8f | 485 | } else { |
0ed33598 | 486 | super_unlock_excl(s); |
1da177e4 LT |
487 | } |
488 | } | |
489 | ||
1712ac8f | 490 | EXPORT_SYMBOL(deactivate_locked_super); |
1da177e4 | 491 | |
74dbbdd7 | 492 | /** |
1712ac8f | 493 | * deactivate_super - drop an active reference to superblock |
74dbbdd7 AV |
494 | * @s: superblock to deactivate |
495 | * | |
1712ac8f AV |
496 | * Variant of deactivate_locked_super(), except that superblock is *not* |
497 | * locked by caller. If we are going to drop the final active reference, | |
498 | * lock will be acquired prior to that. | |
74dbbdd7 | 499 | */ |
1712ac8f | 500 | void deactivate_super(struct super_block *s) |
74dbbdd7 | 501 | { |
cc23402c | 502 | if (!atomic_add_unless(&s->s_active, -1, 1)) { |
5e874914 | 503 | __super_lock_excl(s); |
1712ac8f | 504 | deactivate_locked_super(s); |
74dbbdd7 AV |
505 | } |
506 | } | |
507 | ||
1712ac8f | 508 | EXPORT_SYMBOL(deactivate_super); |
74dbbdd7 | 509 | |
1da177e4 LT |
510 | /** |
511 | * grab_super - acquire an active reference | |
512 | * @s: reference we are trying to make active | |
513 | * | |
514 | * Tries to acquire an active reference. grab_super() is used when we | |
515 | * had just found a superblock in super_blocks or fs_type->fs_supers | |
516 | * and want to turn it into a full-blown active reference. grab_super() | |
517 | * is called with sb_lock held and drops it. Returns 1 in case of | |
518 | * success, 0 if we had failed (superblock contents was already dead or | |
acfec9a5 AV |
519 | * dying when grab_super() had been called). Note that this is only |
520 | * called for superblocks not in rundown mode (== ones still on ->fs_supers | |
521 | * of their type), so increment of ->s_count is OK here. | |
1da177e4 | 522 | */ |
9c4dbee7 | 523 | static int grab_super(struct super_block *s) __releases(sb_lock) |
1da177e4 | 524 | { |
5e874914 CB |
525 | bool born; |
526 | ||
1da177e4 LT |
527 | s->s_count++; |
528 | spin_unlock(&sb_lock); | |
5e874914 CB |
529 | born = super_lock_excl(s); |
530 | if (born && atomic_inc_not_zero(&s->s_active)) { | |
acfec9a5 AV |
531 | put_super(s); |
532 | return 1; | |
533 | } | |
0ed33598 | 534 | super_unlock_excl(s); |
1da177e4 | 535 | put_super(s); |
1da177e4 LT |
536 | return 0; |
537 | } | |
538 | ||
2c18a63b CB |
539 | static inline bool wait_dead(struct super_block *sb) |
540 | { | |
541 | unsigned int flags; | |
542 | ||
543 | /* | |
544 | * Pairs with memory barrier in super_wake() and ensures | |
545 | * that we see SB_DEAD after we're woken. | |
546 | */ | |
547 | flags = smp_load_acquire(&sb->s_flags); | |
548 | return flags & SB_DEAD; | |
549 | } | |
550 | ||
551 | /** | |
552 | * grab_super_dead - acquire an active reference to a superblock | |
553 | * @sb: superblock to acquire | |
554 | * | |
555 | * Acquire a temporary reference on a superblock and try to trade it for | |
556 | * an active reference. This is used in sget{_fc}() to wait for a | |
557 | * superblock to either become SB_BORN or for it to pass through | |
558 | * sb->kill() and be marked as SB_DEAD. | |
559 | * | |
560 | * Return: This returns true if an active reference could be acquired, | |
561 | * false if not. | |
562 | */ | |
563 | static bool grab_super_dead(struct super_block *sb) | |
564 | { | |
565 | ||
566 | sb->s_count++; | |
567 | if (grab_super(sb)) { | |
568 | put_super(sb); | |
569 | lockdep_assert_held(&sb->s_umount); | |
570 | return true; | |
571 | } | |
572 | wait_var_event(&sb->s_flags, wait_dead(sb)); | |
573 | put_super(sb); | |
574 | lockdep_assert_not_held(&sb->s_umount); | |
575 | return false; | |
576 | } | |
577 | ||
12ad3ab6 | 578 | /* |
d8ce82ef | 579 | * super_trylock_shared - try to grab ->s_umount shared |
331cbdee | 580 | * @sb: reference we are trying to grab |
12ad3ab6 | 581 | * |
eb6ef3df | 582 | * Try to prevent fs shutdown. This is used in places where we |
12ad3ab6 | 583 | * cannot take an active reference but we need to ensure that the |
eb6ef3df KK |
584 | * filesystem is not shut down while we are working on it. It returns |
585 | * false if we cannot acquire s_umount or if we lose the race and | |
586 | * filesystem already got into shutdown, and returns true with the s_umount | |
587 | * lock held in read mode in case of success. On successful return, | |
588 | * the caller must drop the s_umount lock when done. | |
589 | * | |
590 | * Note that unlike get_super() et.al. this one does *not* bump ->s_count. | |
591 | * The reason why it's safe is that we are OK with doing trylock instead | |
592 | * of down_read(). There's a couple of places that are OK with that, but | |
593 | * it's very much not a general-purpose interface. | |
12ad3ab6 | 594 | */ |
d8ce82ef | 595 | bool super_trylock_shared(struct super_block *sb) |
12ad3ab6 | 596 | { |
12ad3ab6 | 597 | if (down_read_trylock(&sb->s_umount)) { |
5e874914 CB |
598 | if (!(sb->s_flags & SB_DYING) && sb->s_root && |
599 | (sb->s_flags & SB_BORN)) | |
12ad3ab6 | 600 | return true; |
0ed33598 | 601 | super_unlock_shared(sb); |
12ad3ab6 DC |
602 | } |
603 | ||
12ad3ab6 DC |
604 | return false; |
605 | } | |
606 | ||
04b94071 DL |
607 | /** |
608 | * retire_super - prevents superblock from being reused | |
609 | * @sb: superblock to retire | |
610 | * | |
611 | * The function marks superblock to be ignored in superblock test, which | |
612 | * prevents it from being reused for any new mounts. If the superblock has | |
613 | * a private bdi, it also unregisters it, but doesn't reduce the refcount | |
614 | * of the superblock to prevent potential races. The refcount is reduced | |
615 | * by generic_shutdown_super(). The function can not be called | |
616 | * concurrently with generic_shutdown_super(). It is safe to call the | |
617 | * function multiple times, subsequent calls have no effect. | |
618 | * | |
619 | * The marker will affect the re-use only for block-device-based | |
620 | * superblocks. Other superblocks will still get marked if this function | |
621 | * is used, but that will not affect their reusability. | |
622 | */ | |
623 | void retire_super(struct super_block *sb) | |
624 | { | |
625 | WARN_ON(!sb->s_bdev); | |
5e874914 | 626 | __super_lock_excl(sb); |
04b94071 DL |
627 | if (sb->s_iflags & SB_I_PERSB_BDI) { |
628 | bdi_unregister(sb->s_bdi); | |
629 | sb->s_iflags &= ~SB_I_PERSB_BDI; | |
630 | } | |
631 | sb->s_iflags |= SB_I_RETIRED; | |
0ed33598 | 632 | super_unlock_excl(sb); |
04b94071 DL |
633 | } |
634 | EXPORT_SYMBOL(retire_super); | |
635 | ||
1da177e4 LT |
636 | /** |
637 | * generic_shutdown_super - common helper for ->kill_sb() | |
638 | * @sb: superblock to kill | |
639 | * | |
640 | * generic_shutdown_super() does all fs-independent work on superblock | |
641 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects | |
642 | * that need destruction out of superblock, call generic_shutdown_super() | |
643 | * and release aforementioned objects. Note: dentries and inodes _are_ | |
644 | * taken care of and do not need specific handling. | |
c636ebdb DH |
645 | * |
646 | * Upon calling this function, the filesystem may no longer alter or | |
647 | * rearrange the set of dentries belonging to this super_block, nor may it | |
648 | * change the attachments of dentries to inodes. | |
1da177e4 LT |
649 | */ |
650 | void generic_shutdown_super(struct super_block *sb) | |
651 | { | |
ee9b6d61 | 652 | const struct super_operations *sop = sb->s_op; |
1da177e4 | 653 | |
c636ebdb DH |
654 | if (sb->s_root) { |
655 | shrink_dcache_for_umount(sb); | |
60b0680f | 656 | sync_filesystem(sb); |
e462ec50 | 657 | sb->s_flags &= ~SB_ACTIVE; |
efaee192 | 658 | |
a1a0e23e | 659 | cgroup_writeback_umount(); |
63997e98 | 660 | |
ccb820dc | 661 | /* Evict all inodes with zero refcount. */ |
63997e98 | 662 | evict_inodes(sb); |
ccb820dc EB |
663 | |
664 | /* | |
665 | * Clean up and evict any inodes that still have references due | |
666 | * to fsnotify or the security policy. | |
667 | */ | |
1edc8eb2 | 668 | fsnotify_sb_delete(sb); |
83e804f0 | 669 | security_sb_delete(sb); |
1da177e4 | 670 | |
ccb820dc EB |
671 | /* |
672 | * Now that all potentially-encrypted inodes have been evicted, | |
673 | * the fscrypt keyring can be destroyed. | |
674 | */ | |
675 | fscrypt_destroy_keyring(sb); | |
676 | ||
7b7a8665 CH |
677 | if (sb->s_dio_done_wq) { |
678 | destroy_workqueue(sb->s_dio_done_wq); | |
679 | sb->s_dio_done_wq = NULL; | |
680 | } | |
681 | ||
1da177e4 LT |
682 | if (sop->put_super) |
683 | sop->put_super(sb); | |
684 | ||
47d58691 JH |
685 | if (CHECK_DATA_CORRUPTION(!list_empty(&sb->s_inodes), |
686 | "VFS: Busy inodes after unmount of %s (%s)", | |
687 | sb->s_id, sb->s_type->name)) { | |
688 | /* | |
689 | * Adding a proper bailout path here would be hard, but | |
690 | * we can at least make it more likely that a later | |
691 | * iput_final() or such crashes cleanly. | |
692 | */ | |
693 | struct inode *inode; | |
694 | ||
695 | spin_lock(&sb->s_inode_list_lock); | |
696 | list_for_each_entry(inode, &sb->s_inodes, i_sb_list) { | |
697 | inode->i_op = VFS_PTR_POISON; | |
698 | inode->i_sb = VFS_PTR_POISON; | |
699 | inode->i_mapping = VFS_PTR_POISON; | |
700 | } | |
701 | spin_unlock(&sb->s_inode_list_lock); | |
1da177e4 | 702 | } |
1da177e4 | 703 | } |
5e874914 CB |
704 | /* |
705 | * Broadcast to everyone that grabbed a temporary reference to this | |
706 | * superblock before we removed it from @fs_supers that the superblock | |
707 | * is dying. Every walker of @fs_supers outside of sget{_fc}() will now | |
708 | * discard this superblock and treat it as dead. | |
2c18a63b CB |
709 | * |
710 | * We leave the superblock on @fs_supers so it can be found by | |
711 | * sget{_fc}() until we passed sb->kill_sb(). | |
5e874914 CB |
712 | */ |
713 | super_wake(sb, SB_DYING); | |
0ed33598 | 714 | super_unlock_excl(sb); |
c1844d53 | 715 | if (sb->s_bdi != &noop_backing_dev_info) { |
0b3ea092 CH |
716 | if (sb->s_iflags & SB_I_PERSB_BDI) |
717 | bdi_unregister(sb->s_bdi); | |
fca39346 JK |
718 | bdi_put(sb->s_bdi); |
719 | sb->s_bdi = &noop_backing_dev_info; | |
fca39346 | 720 | } |
1da177e4 LT |
721 | } |
722 | ||
723 | EXPORT_SYMBOL(generic_shutdown_super); | |
724 | ||
20284ab7 | 725 | bool mount_capable(struct fs_context *fc) |
0ce0cf12 | 726 | { |
20284ab7 | 727 | if (!(fc->fs_type->fs_flags & FS_USERNS_MOUNT)) |
0ce0cf12 AV |
728 | return capable(CAP_SYS_ADMIN); |
729 | else | |
c2c44ec2 | 730 | return ns_capable(fc->user_ns, CAP_SYS_ADMIN); |
0ce0cf12 AV |
731 | } |
732 | ||
cb50b348 AV |
733 | /** |
734 | * sget_fc - Find or create a superblock | |
735 | * @fc: Filesystem context. | |
736 | * @test: Comparison callback | |
737 | * @set: Setup callback | |
738 | * | |
739 | * Find or create a superblock using the parameters stored in the filesystem | |
740 | * context and the two callback functions. | |
741 | * | |
742 | * If an extant superblock is matched, then that will be returned with an | |
743 | * elevated reference count that the caller must transfer or discard. | |
744 | * | |
745 | * If no match is made, a new superblock will be allocated and basic | |
746 | * initialisation will be performed (s_type, s_fs_info and s_id will be set and | |
747 | * the set() callback will be invoked), the superblock will be published and it | |
748 | * will be returned in a partially constructed state with SB_BORN and SB_ACTIVE | |
749 | * as yet unset. | |
750 | */ | |
751 | struct super_block *sget_fc(struct fs_context *fc, | |
752 | int (*test)(struct super_block *, struct fs_context *), | |
753 | int (*set)(struct super_block *, struct fs_context *)) | |
754 | { | |
755 | struct super_block *s = NULL; | |
756 | struct super_block *old; | |
757 | struct user_namespace *user_ns = fc->global ? &init_user_ns : fc->user_ns; | |
758 | int err; | |
759 | ||
cb50b348 AV |
760 | retry: |
761 | spin_lock(&sb_lock); | |
762 | if (test) { | |
763 | hlist_for_each_entry(old, &fc->fs_type->fs_supers, s_instances) { | |
764 | if (test(old, fc)) | |
765 | goto share_extant_sb; | |
766 | } | |
767 | } | |
768 | if (!s) { | |
769 | spin_unlock(&sb_lock); | |
770 | s = alloc_super(fc->fs_type, fc->sb_flags, user_ns); | |
771 | if (!s) | |
772 | return ERR_PTR(-ENOMEM); | |
773 | goto retry; | |
774 | } | |
775 | ||
776 | s->s_fs_info = fc->s_fs_info; | |
777 | err = set(s, fc); | |
778 | if (err) { | |
779 | s->s_fs_info = NULL; | |
780 | spin_unlock(&sb_lock); | |
781 | destroy_unused_super(s); | |
782 | return ERR_PTR(err); | |
783 | } | |
784 | fc->s_fs_info = NULL; | |
785 | s->s_type = fc->fs_type; | |
c80fa7c8 | 786 | s->s_iflags |= fc->s_iflags; |
c642256b | 787 | strscpy(s->s_id, s->s_type->name, sizeof(s->s_id)); |
5e874914 CB |
788 | /* |
789 | * Make the superblock visible on @super_blocks and @fs_supers. | |
790 | * It's in a nascent state and users should wait on SB_BORN or | |
791 | * SB_DYING to be set. | |
792 | */ | |
cb50b348 AV |
793 | list_add_tail(&s->s_list, &super_blocks); |
794 | hlist_add_head(&s->s_instances, &s->s_type->fs_supers); | |
795 | spin_unlock(&sb_lock); | |
796 | get_filesystem(s->s_type); | |
797 | register_shrinker_prepared(&s->s_shrink); | |
798 | return s; | |
799 | ||
800 | share_extant_sb: | |
801 | if (user_ns != old->s_user_ns) { | |
802 | spin_unlock(&sb_lock); | |
803 | destroy_unused_super(s); | |
804 | return ERR_PTR(-EBUSY); | |
805 | } | |
2c18a63b | 806 | if (!grab_super_dead(old)) |
cb50b348 AV |
807 | goto retry; |
808 | destroy_unused_super(s); | |
809 | return old; | |
810 | } | |
811 | EXPORT_SYMBOL(sget_fc); | |
812 | ||
1da177e4 | 813 | /** |
023d066a DH |
814 | * sget - find or create a superblock |
815 | * @type: filesystem type superblock should belong to | |
816 | * @test: comparison callback | |
817 | * @set: setup callback | |
818 | * @flags: mount flags | |
819 | * @data: argument to each of them | |
1da177e4 | 820 | */ |
023d066a | 821 | struct super_block *sget(struct file_system_type *type, |
1da177e4 LT |
822 | int (*test)(struct super_block *,void *), |
823 | int (*set)(struct super_block *,void *), | |
023d066a | 824 | int flags, |
1da177e4 LT |
825 | void *data) |
826 | { | |
023d066a | 827 | struct user_namespace *user_ns = current_user_ns(); |
1da177e4 | 828 | struct super_block *s = NULL; |
d4730127 | 829 | struct super_block *old; |
1da177e4 LT |
830 | int err; |
831 | ||
023d066a DH |
832 | /* We don't yet pass the user namespace of the parent |
833 | * mount through to here so always use &init_user_ns | |
834 | * until that changes. | |
835 | */ | |
836 | if (flags & SB_SUBMOUNT) | |
837 | user_ns = &init_user_ns; | |
838 | ||
1da177e4 LT |
839 | retry: |
840 | spin_lock(&sb_lock); | |
d4730127 | 841 | if (test) { |
b67bfe0d | 842 | hlist_for_each_entry(old, &type->fs_supers, s_instances) { |
d4730127 MK |
843 | if (!test(old, data)) |
844 | continue; | |
6e4eab57 EB |
845 | if (user_ns != old->s_user_ns) { |
846 | spin_unlock(&sb_lock); | |
0200894d | 847 | destroy_unused_super(s); |
6e4eab57 EB |
848 | return ERR_PTR(-EBUSY); |
849 | } | |
2c18a63b | 850 | if (!grab_super_dead(old)) |
d4730127 | 851 | goto retry; |
0200894d | 852 | destroy_unused_super(s); |
d4730127 MK |
853 | return old; |
854 | } | |
1da177e4 LT |
855 | } |
856 | if (!s) { | |
857 | spin_unlock(&sb_lock); | |
e462ec50 | 858 | s = alloc_super(type, (flags & ~SB_SUBMOUNT), user_ns); |
1da177e4 LT |
859 | if (!s) |
860 | return ERR_PTR(-ENOMEM); | |
861 | goto retry; | |
862 | } | |
dd111b31 | 863 | |
1da177e4 LT |
864 | err = set(s, data); |
865 | if (err) { | |
866 | spin_unlock(&sb_lock); | |
0200894d | 867 | destroy_unused_super(s); |
1da177e4 LT |
868 | return ERR_PTR(err); |
869 | } | |
870 | s->s_type = type; | |
c642256b | 871 | strscpy(s->s_id, type->name, sizeof(s->s_id)); |
1da177e4 | 872 | list_add_tail(&s->s_list, &super_blocks); |
a5166169 | 873 | hlist_add_head(&s->s_instances, &type->fs_supers); |
1da177e4 LT |
874 | spin_unlock(&sb_lock); |
875 | get_filesystem(type); | |
8e04944f | 876 | register_shrinker_prepared(&s->s_shrink); |
1da177e4 LT |
877 | return s; |
878 | } | |
1da177e4 LT |
879 | EXPORT_SYMBOL(sget); |
880 | ||
881 | void drop_super(struct super_block *sb) | |
882 | { | |
0ed33598 | 883 | super_unlock_shared(sb); |
1da177e4 LT |
884 | put_super(sb); |
885 | } | |
886 | ||
887 | EXPORT_SYMBOL(drop_super); | |
888 | ||
ba6379f7 JK |
889 | void drop_super_exclusive(struct super_block *sb) |
890 | { | |
0ed33598 | 891 | super_unlock_excl(sb); |
ba6379f7 JK |
892 | put_super(sb); |
893 | } | |
894 | EXPORT_SYMBOL(drop_super_exclusive); | |
895 | ||
fa7c1d50 MG |
896 | static void __iterate_supers(void (*f)(struct super_block *)) |
897 | { | |
898 | struct super_block *sb, *p = NULL; | |
899 | ||
900 | spin_lock(&sb_lock); | |
901 | list_for_each_entry(sb, &super_blocks, s_list) { | |
5e874914 CB |
902 | /* Pairs with memory marrier in super_wake(). */ |
903 | if (smp_load_acquire(&sb->s_flags) & SB_DYING) | |
fa7c1d50 MG |
904 | continue; |
905 | sb->s_count++; | |
906 | spin_unlock(&sb_lock); | |
907 | ||
908 | f(sb); | |
909 | ||
910 | spin_lock(&sb_lock); | |
911 | if (p) | |
912 | __put_super(p); | |
913 | p = sb; | |
914 | } | |
915 | if (p) | |
916 | __put_super(p); | |
917 | spin_unlock(&sb_lock); | |
918 | } | |
01a05b33 AV |
919 | /** |
920 | * iterate_supers - call function for all active superblocks | |
921 | * @f: function to call | |
922 | * @arg: argument to pass to it | |
923 | * | |
924 | * Scans the superblock list and calls given function, passing it | |
925 | * locked superblock and given argument. | |
926 | */ | |
927 | void iterate_supers(void (*f)(struct super_block *, void *), void *arg) | |
928 | { | |
dca33252 | 929 | struct super_block *sb, *p = NULL; |
01a05b33 AV |
930 | |
931 | spin_lock(&sb_lock); | |
dca33252 | 932 | list_for_each_entry(sb, &super_blocks, s_list) { |
5e874914 CB |
933 | bool born; |
934 | ||
01a05b33 AV |
935 | sb->s_count++; |
936 | spin_unlock(&sb_lock); | |
937 | ||
5e874914 CB |
938 | born = super_lock_shared(sb); |
939 | if (born && sb->s_root) | |
01a05b33 | 940 | f(sb, arg); |
0ed33598 | 941 | super_unlock_shared(sb); |
01a05b33 AV |
942 | |
943 | spin_lock(&sb_lock); | |
dca33252 AV |
944 | if (p) |
945 | __put_super(p); | |
946 | p = sb; | |
01a05b33 | 947 | } |
dca33252 AV |
948 | if (p) |
949 | __put_super(p); | |
01a05b33 AV |
950 | spin_unlock(&sb_lock); |
951 | } | |
952 | ||
43e15cdb AV |
953 | /** |
954 | * iterate_supers_type - call function for superblocks of given type | |
955 | * @type: fs type | |
956 | * @f: function to call | |
957 | * @arg: argument to pass to it | |
958 | * | |
959 | * Scans the superblock list and calls given function, passing it | |
960 | * locked superblock and given argument. | |
961 | */ | |
962 | void iterate_supers_type(struct file_system_type *type, | |
963 | void (*f)(struct super_block *, void *), void *arg) | |
964 | { | |
965 | struct super_block *sb, *p = NULL; | |
966 | ||
967 | spin_lock(&sb_lock); | |
b67bfe0d | 968 | hlist_for_each_entry(sb, &type->fs_supers, s_instances) { |
5e874914 CB |
969 | bool born; |
970 | ||
43e15cdb AV |
971 | sb->s_count++; |
972 | spin_unlock(&sb_lock); | |
973 | ||
5e874914 CB |
974 | born = super_lock_shared(sb); |
975 | if (born && sb->s_root) | |
43e15cdb | 976 | f(sb, arg); |
0ed33598 | 977 | super_unlock_shared(sb); |
43e15cdb AV |
978 | |
979 | spin_lock(&sb_lock); | |
980 | if (p) | |
981 | __put_super(p); | |
982 | p = sb; | |
983 | } | |
984 | if (p) | |
985 | __put_super(p); | |
986 | spin_unlock(&sb_lock); | |
987 | } | |
988 | ||
989 | EXPORT_SYMBOL(iterate_supers_type); | |
990 | ||
4504230a CH |
991 | /** |
992 | * get_active_super - get an active reference to the superblock of a device | |
993 | * @bdev: device to get the superblock for | |
994 | * | |
995 | * Scans the superblock list and finds the superblock of the file system | |
996 | * mounted on the device given. Returns the superblock with an active | |
d3f21473 | 997 | * reference or %NULL if none was found. |
4504230a CH |
998 | */ |
999 | struct super_block *get_active_super(struct block_device *bdev) | |
1000 | { | |
1001 | struct super_block *sb; | |
1002 | ||
1003 | if (!bdev) | |
1004 | return NULL; | |
1005 | ||
1006 | spin_lock(&sb_lock); | |
1007 | list_for_each_entry(sb, &super_blocks, s_list) { | |
1494583d | 1008 | if (sb->s_bdev == bdev) { |
acfec9a5 | 1009 | if (!grab_super(sb)) |
5e874914 | 1010 | return NULL; |
0ed33598 | 1011 | super_unlock_excl(sb); |
acfec9a5 | 1012 | return sb; |
1494583d | 1013 | } |
4504230a CH |
1014 | } |
1015 | spin_unlock(&sb_lock); | |
1016 | return NULL; | |
1017 | } | |
dd111b31 | 1018 | |
4e7b5671 | 1019 | struct super_block *user_get_super(dev_t dev, bool excl) |
1da177e4 | 1020 | { |
618f0636 | 1021 | struct super_block *sb; |
1da177e4 | 1022 | |
1da177e4 | 1023 | spin_lock(&sb_lock); |
618f0636 KK |
1024 | list_for_each_entry(sb, &super_blocks, s_list) { |
1025 | if (sb->s_dev == dev) { | |
5e874914 CB |
1026 | bool born; |
1027 | ||
618f0636 | 1028 | sb->s_count++; |
1da177e4 | 1029 | spin_unlock(&sb_lock); |
df40c01a | 1030 | /* still alive? */ |
5e874914 CB |
1031 | born = super_lock(sb, excl); |
1032 | if (born && sb->s_root) | |
618f0636 | 1033 | return sb; |
0ed33598 | 1034 | super_unlock(sb, excl); |
df40c01a | 1035 | /* nope, got unmounted */ |
618f0636 | 1036 | spin_lock(&sb_lock); |
df40c01a | 1037 | __put_super(sb); |
5e874914 | 1038 | break; |
1da177e4 LT |
1039 | } |
1040 | } | |
1041 | spin_unlock(&sb_lock); | |
1042 | return NULL; | |
1043 | } | |
1044 | ||
1da177e4 | 1045 | /** |
8d0347f6 DH |
1046 | * reconfigure_super - asks filesystem to change superblock parameters |
1047 | * @fc: The superblock and configuration | |
1da177e4 | 1048 | * |
8d0347f6 | 1049 | * Alters the configuration parameters of a live superblock. |
1da177e4 | 1050 | */ |
8d0347f6 | 1051 | int reconfigure_super(struct fs_context *fc) |
1da177e4 | 1052 | { |
8d0347f6 | 1053 | struct super_block *sb = fc->root->d_sb; |
1da177e4 | 1054 | int retval; |
8d0347f6 | 1055 | bool remount_ro = false; |
c541dce8 | 1056 | bool remount_rw = false; |
8d0347f6 | 1057 | bool force = fc->sb_flags & SB_FORCE; |
4504230a | 1058 | |
8d0347f6 DH |
1059 | if (fc->sb_flags_mask & ~MS_RMT_MASK) |
1060 | return -EINVAL; | |
5accdf82 | 1061 | if (sb->s_writers.frozen != SB_UNFROZEN) |
4504230a CH |
1062 | return -EBUSY; |
1063 | ||
8d0347f6 DH |
1064 | retval = security_sb_remount(sb, fc->security); |
1065 | if (retval) | |
1066 | return retval; | |
1067 | ||
1068 | if (fc->sb_flags_mask & SB_RDONLY) { | |
9361401e | 1069 | #ifdef CONFIG_BLOCK |
6f0d9689 CH |
1070 | if (!(fc->sb_flags & SB_RDONLY) && sb->s_bdev && |
1071 | bdev_read_only(sb->s_bdev)) | |
8d0347f6 | 1072 | return -EACCES; |
9361401e | 1073 | #endif |
c541dce8 | 1074 | remount_rw = !(fc->sb_flags & SB_RDONLY) && sb_rdonly(sb); |
8d0347f6 DH |
1075 | remount_ro = (fc->sb_flags & SB_RDONLY) && !sb_rdonly(sb); |
1076 | } | |
d208bbdd | 1077 | |
0aec09d0 | 1078 | if (remount_ro) { |
fdab684d | 1079 | if (!hlist_empty(&sb->s_pins)) { |
0ed33598 | 1080 | super_unlock_excl(sb); |
fdab684d | 1081 | group_pin_kill(&sb->s_pins); |
5e874914 | 1082 | __super_lock_excl(sb); |
0aec09d0 AV |
1083 | if (!sb->s_root) |
1084 | return 0; | |
1085 | if (sb->s_writers.frozen != SB_UNFROZEN) | |
1086 | return -EBUSY; | |
8d0347f6 | 1087 | remount_ro = !sb_rdonly(sb); |
0aec09d0 AV |
1088 | } |
1089 | } | |
1090 | shrink_dcache_sb(sb); | |
1091 | ||
8d0347f6 DH |
1092 | /* If we are reconfiguring to RDONLY and current sb is read/write, |
1093 | * make sure there are no files open for writing. | |
1094 | */ | |
d208bbdd | 1095 | if (remount_ro) { |
4ed5e82f | 1096 | if (force) { |
d7439fb1 | 1097 | sb_start_ro_state_change(sb); |
4ed5e82f MS |
1098 | } else { |
1099 | retval = sb_prepare_remount_readonly(sb); | |
1100 | if (retval) | |
1101 | return retval; | |
4ed5e82f | 1102 | } |
c541dce8 JK |
1103 | } else if (remount_rw) { |
1104 | /* | |
d7439fb1 JK |
1105 | * Protect filesystem's reconfigure code from writes from |
1106 | * userspace until reconfigure finishes. | |
c541dce8 | 1107 | */ |
d7439fb1 | 1108 | sb_start_ro_state_change(sb); |
1da177e4 LT |
1109 | } |
1110 | ||
f3a09c92 AV |
1111 | if (fc->ops->reconfigure) { |
1112 | retval = fc->ops->reconfigure(fc); | |
1113 | if (retval) { | |
1114 | if (!force) | |
1115 | goto cancel_readonly; | |
1116 | /* If forced remount, go ahead despite any errors */ | |
1117 | WARN(1, "forced remount of a %s fs returned %i\n", | |
1118 | sb->s_type->name, retval); | |
1119 | } | |
1da177e4 | 1120 | } |
8d0347f6 DH |
1121 | |
1122 | WRITE_ONCE(sb->s_flags, ((sb->s_flags & ~fc->sb_flags_mask) | | |
1123 | (fc->sb_flags & fc->sb_flags_mask))); | |
d7439fb1 | 1124 | sb_end_ro_state_change(sb); |
c79d967d | 1125 | |
d208bbdd NP |
1126 | /* |
1127 | * Some filesystems modify their metadata via some other path than the | |
1128 | * bdev buffer cache (eg. use a private mapping, or directories in | |
1129 | * pagecache, etc). Also file data modifications go via their own | |
1130 | * mappings. So If we try to mount readonly then copy the filesystem | |
1131 | * from bdev, we could get stale data, so invalidate it to give a best | |
1132 | * effort at coherency. | |
1133 | */ | |
1134 | if (remount_ro && sb->s_bdev) | |
1135 | invalidate_bdev(sb->s_bdev); | |
1da177e4 | 1136 | return 0; |
4ed5e82f MS |
1137 | |
1138 | cancel_readonly: | |
d7439fb1 | 1139 | sb_end_ro_state_change(sb); |
4ed5e82f | 1140 | return retval; |
1da177e4 LT |
1141 | } |
1142 | ||
fa7c1d50 | 1143 | static void do_emergency_remount_callback(struct super_block *sb) |
1da177e4 | 1144 | { |
5e874914 CB |
1145 | bool born = super_lock_excl(sb); |
1146 | ||
1147 | if (born && sb->s_root && sb->s_bdev && !sb_rdonly(sb)) { | |
8d0347f6 DH |
1148 | struct fs_context *fc; |
1149 | ||
1150 | fc = fs_context_for_reconfigure(sb->s_root, | |
1151 | SB_RDONLY | SB_FORCE, SB_RDONLY); | |
1152 | if (!IS_ERR(fc)) { | |
1153 | if (parse_monolithic_mount_data(fc, NULL) == 0) | |
1154 | (void)reconfigure_super(fc); | |
1155 | put_fs_context(fc); | |
1156 | } | |
1da177e4 | 1157 | } |
0ed33598 | 1158 | super_unlock_excl(sb); |
fa7c1d50 MG |
1159 | } |
1160 | ||
1161 | static void do_emergency_remount(struct work_struct *work) | |
1162 | { | |
1163 | __iterate_supers(do_emergency_remount_callback); | |
a2a9537a | 1164 | kfree(work); |
1da177e4 LT |
1165 | printk("Emergency Remount complete\n"); |
1166 | } | |
1167 | ||
1168 | void emergency_remount(void) | |
1169 | { | |
a2a9537a JA |
1170 | struct work_struct *work; |
1171 | ||
1172 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | |
1173 | if (work) { | |
1174 | INIT_WORK(work, do_emergency_remount); | |
1175 | schedule_work(work); | |
1176 | } | |
1da177e4 LT |
1177 | } |
1178 | ||
08fdc8a0 MG |
1179 | static void do_thaw_all_callback(struct super_block *sb) |
1180 | { | |
5e874914 CB |
1181 | bool born = super_lock_excl(sb); |
1182 | ||
1183 | if (born && sb->s_root) { | |
08fdc8a0 MG |
1184 | emergency_thaw_bdev(sb); |
1185 | thaw_super_locked(sb); | |
1186 | } else { | |
0ed33598 | 1187 | super_unlock_excl(sb); |
08fdc8a0 MG |
1188 | } |
1189 | } | |
1190 | ||
1191 | static void do_thaw_all(struct work_struct *work) | |
1192 | { | |
1193 | __iterate_supers(do_thaw_all_callback); | |
1194 | kfree(work); | |
1195 | printk(KERN_WARNING "Emergency Thaw complete\n"); | |
1196 | } | |
1197 | ||
1198 | /** | |
1199 | * emergency_thaw_all -- forcibly thaw every frozen filesystem | |
1200 | * | |
1201 | * Used for emergency unfreeze of all filesystems via SysRq | |
1202 | */ | |
1203 | void emergency_thaw_all(void) | |
1204 | { | |
1205 | struct work_struct *work; | |
1206 | ||
1207 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | |
1208 | if (work) { | |
1209 | INIT_WORK(work, do_thaw_all); | |
1210 | schedule_work(work); | |
1211 | } | |
1212 | } | |
1213 | ||
ad76cbc6 | 1214 | static DEFINE_IDA(unnamed_dev_ida); |
1da177e4 | 1215 | |
5a66847e MW |
1216 | /** |
1217 | * get_anon_bdev - Allocate a block device for filesystems which don't have one. | |
1218 | * @p: Pointer to a dev_t. | |
1219 | * | |
1220 | * Filesystems which don't use real block devices can call this function | |
1221 | * to allocate a virtual block device. | |
1222 | * | |
1223 | * Context: Any context. Frequently called while holding sb_lock. | |
1224 | * Return: 0 on success, -EMFILE if there are no anonymous bdevs left | |
1225 | * or -ENOMEM if memory allocation failed. | |
1226 | */ | |
0ee5dc67 | 1227 | int get_anon_bdev(dev_t *p) |
1da177e4 LT |
1228 | { |
1229 | int dev; | |
5a66847e MW |
1230 | |
1231 | /* | |
1232 | * Many userspace utilities consider an FSID of 0 invalid. | |
1233 | * Always return at least 1 from get_anon_bdev. | |
1234 | */ | |
1235 | dev = ida_alloc_range(&unnamed_dev_ida, 1, (1 << MINORBITS) - 1, | |
1236 | GFP_ATOMIC); | |
1237 | if (dev == -ENOSPC) | |
1238 | dev = -EMFILE; | |
1239 | if (dev < 0) | |
1240 | return dev; | |
1241 | ||
1242 | *p = MKDEV(0, dev); | |
1da177e4 LT |
1243 | return 0; |
1244 | } | |
0ee5dc67 | 1245 | EXPORT_SYMBOL(get_anon_bdev); |
1da177e4 | 1246 | |
0ee5dc67 | 1247 | void free_anon_bdev(dev_t dev) |
1da177e4 | 1248 | { |
5a66847e | 1249 | ida_free(&unnamed_dev_ida, MINOR(dev)); |
1da177e4 | 1250 | } |
0ee5dc67 AV |
1251 | EXPORT_SYMBOL(free_anon_bdev); |
1252 | ||
1253 | int set_anon_super(struct super_block *s, void *data) | |
1254 | { | |
df0ce26c | 1255 | return get_anon_bdev(&s->s_dev); |
0ee5dc67 | 1256 | } |
0ee5dc67 AV |
1257 | EXPORT_SYMBOL(set_anon_super); |
1258 | ||
1259 | void kill_anon_super(struct super_block *sb) | |
1260 | { | |
1261 | dev_t dev = sb->s_dev; | |
1262 | generic_shutdown_super(sb); | |
1263 | free_anon_bdev(dev); | |
1264 | } | |
1da177e4 LT |
1265 | EXPORT_SYMBOL(kill_anon_super); |
1266 | ||
1da177e4 LT |
1267 | void kill_litter_super(struct super_block *sb) |
1268 | { | |
1269 | if (sb->s_root) | |
1270 | d_genocide(sb->s_root); | |
1271 | kill_anon_super(sb); | |
1272 | } | |
1da177e4 LT |
1273 | EXPORT_SYMBOL(kill_litter_super); |
1274 | ||
cb50b348 AV |
1275 | int set_anon_super_fc(struct super_block *sb, struct fs_context *fc) |
1276 | { | |
1277 | return set_anon_super(sb, NULL); | |
1278 | } | |
1279 | EXPORT_SYMBOL(set_anon_super_fc); | |
1280 | ||
1281 | static int test_keyed_super(struct super_block *sb, struct fs_context *fc) | |
1282 | { | |
1283 | return sb->s_fs_info == fc->s_fs_info; | |
1284 | } | |
1285 | ||
1286 | static int test_single_super(struct super_block *s, struct fs_context *fc) | |
1287 | { | |
1288 | return 1; | |
1289 | } | |
1290 | ||
cda2ed05 CH |
1291 | static int vfs_get_super(struct fs_context *fc, bool reconf, |
1292 | int (*test)(struct super_block *, struct fs_context *), | |
1293 | int (*fill_super)(struct super_block *sb, | |
1294 | struct fs_context *fc)) | |
cb50b348 | 1295 | { |
cb50b348 | 1296 | struct super_block *sb; |
43ce4c1f | 1297 | int err; |
cb50b348 | 1298 | |
cb50b348 AV |
1299 | sb = sget_fc(fc, test, set_anon_super_fc); |
1300 | if (IS_ERR(sb)) | |
1301 | return PTR_ERR(sb); | |
1302 | ||
1303 | if (!sb->s_root) { | |
43ce4c1f DH |
1304 | err = fill_super(sb, fc); |
1305 | if (err) | |
1306 | goto error; | |
cb50b348 AV |
1307 | |
1308 | sb->s_flags |= SB_ACTIVE; | |
43ce4c1f DH |
1309 | fc->root = dget(sb->s_root); |
1310 | } else { | |
1311 | fc->root = dget(sb->s_root); | |
cda2ed05 | 1312 | if (reconf) { |
43ce4c1f DH |
1313 | err = reconfigure_super(fc); |
1314 | if (err < 0) { | |
1315 | dput(fc->root); | |
1316 | fc->root = NULL; | |
1317 | goto error; | |
1318 | } | |
1319 | } | |
cb50b348 AV |
1320 | } |
1321 | ||
cb50b348 | 1322 | return 0; |
43ce4c1f DH |
1323 | |
1324 | error: | |
1325 | deactivate_locked_super(sb); | |
1326 | return err; | |
cb50b348 | 1327 | } |
cb50b348 | 1328 | |
2ac295d4 AV |
1329 | int get_tree_nodev(struct fs_context *fc, |
1330 | int (*fill_super)(struct super_block *sb, | |
1331 | struct fs_context *fc)) | |
1332 | { | |
cda2ed05 | 1333 | return vfs_get_super(fc, false, NULL, fill_super); |
2ac295d4 AV |
1334 | } |
1335 | EXPORT_SYMBOL(get_tree_nodev); | |
1336 | ||
c23a0bba AV |
1337 | int get_tree_single(struct fs_context *fc, |
1338 | int (*fill_super)(struct super_block *sb, | |
1339 | struct fs_context *fc)) | |
1340 | { | |
cda2ed05 | 1341 | return vfs_get_super(fc, false, test_single_super, fill_super); |
c23a0bba AV |
1342 | } |
1343 | EXPORT_SYMBOL(get_tree_single); | |
1344 | ||
43ce4c1f DH |
1345 | int get_tree_single_reconf(struct fs_context *fc, |
1346 | int (*fill_super)(struct super_block *sb, | |
1347 | struct fs_context *fc)) | |
1348 | { | |
cda2ed05 | 1349 | return vfs_get_super(fc, true, test_single_super, fill_super); |
43ce4c1f DH |
1350 | } |
1351 | EXPORT_SYMBOL(get_tree_single_reconf); | |
1352 | ||
533770cc AV |
1353 | int get_tree_keyed(struct fs_context *fc, |
1354 | int (*fill_super)(struct super_block *sb, | |
1355 | struct fs_context *fc), | |
1356 | void *key) | |
1357 | { | |
1358 | fc->s_fs_info = key; | |
cda2ed05 | 1359 | return vfs_get_super(fc, false, test_keyed_super, fill_super); |
533770cc AV |
1360 | } |
1361 | EXPORT_SYMBOL(get_tree_keyed); | |
1362 | ||
9361401e | 1363 | #ifdef CONFIG_BLOCK |
9c09a7cf CH |
1364 | /* |
1365 | * Lock a super block that the callers holds a reference to. | |
1366 | * | |
1367 | * The caller needs to ensure that the super_block isn't being freed while | |
1368 | * calling this function, e.g. by holding a lock over the call to this function | |
1369 | * and the place that clears the pointer to the superblock used by this function | |
1370 | * before freeing the superblock. | |
1371 | */ | |
d8ce82ef | 1372 | static bool super_lock_shared_active(struct super_block *sb) |
9c09a7cf | 1373 | { |
5e874914 CB |
1374 | bool born = super_lock_shared(sb); |
1375 | ||
1376 | if (!born || !sb->s_root || !(sb->s_flags & SB_ACTIVE)) { | |
0ed33598 | 1377 | super_unlock_shared(sb); |
9c09a7cf CH |
1378 | return false; |
1379 | } | |
1380 | return true; | |
1381 | } | |
1382 | ||
d8530de5 | 1383 | static void fs_bdev_mark_dead(struct block_device *bdev, bool surprise) |
87efb390 | 1384 | { |
9c09a7cf | 1385 | struct super_block *sb = bdev->bd_holder; |
87efb390 | 1386 | |
9c09a7cf CH |
1387 | /* bd_holder_lock ensures that the sb isn't freed */ |
1388 | lockdep_assert_held(&bdev->bd_holder_lock); | |
1389 | ||
d8ce82ef | 1390 | if (!super_lock_shared_active(sb)) |
87efb390 CH |
1391 | return; |
1392 | ||
d8530de5 CH |
1393 | if (!surprise) |
1394 | sync_filesystem(sb); | |
1395 | shrink_dcache_sb(sb); | |
e127b9bc | 1396 | invalidate_inodes(sb); |
87efb390 CH |
1397 | if (sb->s_op->shutdown) |
1398 | sb->s_op->shutdown(sb); | |
9c09a7cf | 1399 | |
0ed33598 | 1400 | super_unlock_shared(sb); |
87efb390 CH |
1401 | } |
1402 | ||
2142b88c CH |
1403 | static void fs_bdev_sync(struct block_device *bdev) |
1404 | { | |
1405 | struct super_block *sb = bdev->bd_holder; | |
1406 | ||
1407 | lockdep_assert_held(&bdev->bd_holder_lock); | |
1408 | ||
d8ce82ef | 1409 | if (!super_lock_shared_active(sb)) |
2142b88c CH |
1410 | return; |
1411 | sync_filesystem(sb); | |
0ed33598 | 1412 | super_unlock_shared(sb); |
2142b88c CH |
1413 | } |
1414 | ||
7ecd0b6f | 1415 | const struct blk_holder_ops fs_holder_ops = { |
d8530de5 | 1416 | .mark_dead = fs_bdev_mark_dead, |
2142b88c | 1417 | .sync = fs_bdev_sync, |
87efb390 | 1418 | }; |
7ecd0b6f | 1419 | EXPORT_SYMBOL_GPL(fs_holder_ops); |
fe62c3a4 | 1420 | |
1da177e4 LT |
1421 | static int set_bdev_super(struct super_block *s, void *data) |
1422 | { | |
aca740ce | 1423 | s->s_dev = *(dev_t *)data; |
1da177e4 LT |
1424 | return 0; |
1425 | } | |
1426 | ||
fe62c3a4 DH |
1427 | static int set_bdev_super_fc(struct super_block *s, struct fs_context *fc) |
1428 | { | |
1429 | return set_bdev_super(s, fc->sget_key); | |
1430 | } | |
1431 | ||
1432 | static int test_bdev_super_fc(struct super_block *s, struct fs_context *fc) | |
1433 | { | |
aca740ce JK |
1434 | return !(s->s_iflags & SB_I_RETIRED) && |
1435 | s->s_dev == *(dev_t *)fc->sget_key; | |
1436 | } | |
1437 | ||
cf6da236 | 1438 | int setup_bdev_super(struct super_block *sb, int sb_flags, |
aca740ce JK |
1439 | struct fs_context *fc) |
1440 | { | |
1441 | blk_mode_t mode = sb_open_mode(sb_flags); | |
1442 | struct block_device *bdev; | |
1443 | ||
2ea6f689 | 1444 | bdev = blkdev_get_by_dev(sb->s_dev, mode, sb, &fs_holder_ops); |
aca740ce JK |
1445 | if (IS_ERR(bdev)) { |
1446 | if (fc) | |
1447 | errorf(fc, "%s: Can't open blockdev", fc->source); | |
1448 | return PTR_ERR(bdev); | |
1449 | } | |
1450 | ||
1451 | /* | |
1452 | * This really should be in blkdev_get_by_dev, but right now can't due | |
1453 | * to legacy issues that require us to allow opening a block device node | |
1454 | * writable from userspace even for a read-only block device. | |
1455 | */ | |
1456 | if ((mode & BLK_OPEN_WRITE) && bdev_read_only(bdev)) { | |
2ea6f689 | 1457 | blkdev_put(bdev, sb); |
aca740ce JK |
1458 | return -EACCES; |
1459 | } | |
1460 | ||
1461 | /* | |
1462 | * Until SB_BORN flag is set, there can be no active superblock | |
1463 | * references and thus no filesystem freezing. get_active_super() will | |
1464 | * just loop waiting for SB_BORN so even freeze_bdev() cannot proceed. | |
1465 | * | |
1466 | * It is enough to check bdev was not frozen before we set s_bdev. | |
1467 | */ | |
1468 | mutex_lock(&bdev->bd_fsfreeze_mutex); | |
1469 | if (bdev->bd_fsfreeze_count > 0) { | |
1470 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
1471 | if (fc) | |
1472 | warnf(fc, "%pg: Can't mount, blockdev is frozen", bdev); | |
2ea6f689 | 1473 | blkdev_put(bdev, sb); |
aca740ce JK |
1474 | return -EBUSY; |
1475 | } | |
1476 | spin_lock(&sb_lock); | |
1477 | sb->s_bdev = bdev; | |
1478 | sb->s_bdi = bdi_get(bdev->bd_disk->bdi); | |
1479 | if (bdev_stable_writes(bdev)) | |
1480 | sb->s_iflags |= SB_I_STABLE_WRITES; | |
1481 | spin_unlock(&sb_lock); | |
1482 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
1483 | ||
1484 | snprintf(sb->s_id, sizeof(sb->s_id), "%pg", bdev); | |
1485 | shrinker_debugfs_rename(&sb->s_shrink, "sb-%s:%s", sb->s_type->name, | |
1486 | sb->s_id); | |
1487 | sb_set_blocksize(sb, block_size(bdev)); | |
1488 | return 0; | |
fe62c3a4 | 1489 | } |
cf6da236 | 1490 | EXPORT_SYMBOL_GPL(setup_bdev_super); |
fe62c3a4 DH |
1491 | |
1492 | /** | |
1493 | * get_tree_bdev - Get a superblock based on a single block device | |
1494 | * @fc: The filesystem context holding the parameters | |
1495 | * @fill_super: Helper to initialise a new superblock | |
1496 | */ | |
1497 | int get_tree_bdev(struct fs_context *fc, | |
1498 | int (*fill_super)(struct super_block *, | |
1499 | struct fs_context *)) | |
1500 | { | |
fe62c3a4 | 1501 | struct super_block *s; |
fe62c3a4 | 1502 | int error = 0; |
aca740ce | 1503 | dev_t dev; |
fe62c3a4 | 1504 | |
fe62c3a4 DH |
1505 | if (!fc->source) |
1506 | return invalf(fc, "No source specified"); | |
1507 | ||
aca740ce JK |
1508 | error = lookup_bdev(fc->source, &dev); |
1509 | if (error) { | |
1510 | errorf(fc, "%s: Can't lookup blockdev", fc->source); | |
1511 | return error; | |
fe62c3a4 DH |
1512 | } |
1513 | ||
1514 | fc->sb_flags |= SB_NOSEC; | |
aca740ce | 1515 | fc->sget_key = &dev; |
fe62c3a4 | 1516 | s = sget_fc(fc, test_bdev_super_fc, set_bdev_super_fc); |
aca740ce | 1517 | if (IS_ERR(s)) |
fe62c3a4 DH |
1518 | return PTR_ERR(s); |
1519 | ||
1520 | if (s->s_root) { | |
1521 | /* Don't summarily change the RO/RW state. */ | |
1522 | if ((fc->sb_flags ^ s->s_flags) & SB_RDONLY) { | |
aca740ce | 1523 | warnf(fc, "%pg: Can't mount, would change RO state", s->s_bdev); |
fe62c3a4 | 1524 | deactivate_locked_super(s); |
fe62c3a4 DH |
1525 | return -EBUSY; |
1526 | } | |
aca740ce | 1527 | } else { |
fe62c3a4 | 1528 | /* |
aca740ce JK |
1529 | * We drop s_umount here because we need to open the bdev and |
1530 | * bdev->open_mutex ranks above s_umount (blkdev_put() -> | |
560e20e4 | 1531 | * bdev_mark_dead()). It is safe because we have active sb |
aca740ce | 1532 | * reference and SB_BORN is not set yet. |
fe62c3a4 | 1533 | */ |
0ed33598 | 1534 | super_unlock_excl(s); |
aca740ce | 1535 | error = setup_bdev_super(s, fc->sb_flags, fc); |
5e874914 | 1536 | __super_lock_excl(s); |
aca740ce JK |
1537 | if (!error) |
1538 | error = fill_super(s, fc); | |
fe62c3a4 DH |
1539 | if (error) { |
1540 | deactivate_locked_super(s); | |
1541 | return error; | |
1542 | } | |
fe62c3a4 | 1543 | s->s_flags |= SB_ACTIVE; |
fe62c3a4 DH |
1544 | } |
1545 | ||
1546 | BUG_ON(fc->root); | |
1547 | fc->root = dget(s->s_root); | |
1548 | return 0; | |
1549 | } | |
1550 | EXPORT_SYMBOL(get_tree_bdev); | |
1551 | ||
1da177e4 LT |
1552 | static int test_bdev_super(struct super_block *s, void *data) |
1553 | { | |
aca740ce | 1554 | return !(s->s_iflags & SB_I_RETIRED) && s->s_dev == *(dev_t *)data; |
1da177e4 LT |
1555 | } |
1556 | ||
152a0836 | 1557 | struct dentry *mount_bdev(struct file_system_type *fs_type, |
1da177e4 | 1558 | int flags, const char *dev_name, void *data, |
152a0836 | 1559 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 | 1560 | { |
1da177e4 | 1561 | struct super_block *s; |
aca740ce JK |
1562 | int error; |
1563 | dev_t dev; | |
1da177e4 | 1564 | |
aca740ce JK |
1565 | error = lookup_bdev(dev_name, &dev); |
1566 | if (error) | |
1567 | return ERR_PTR(error); | |
1da177e4 | 1568 | |
aca740ce JK |
1569 | flags |= SB_NOSEC; |
1570 | s = sget(fs_type, test_bdev_super, set_bdev_super, flags, &dev); | |
1da177e4 | 1571 | if (IS_ERR(s)) |
aca740ce | 1572 | return ERR_CAST(s); |
1da177e4 LT |
1573 | |
1574 | if (s->s_root) { | |
e462ec50 | 1575 | if ((flags ^ s->s_flags) & SB_RDONLY) { |
74dbbdd7 | 1576 | deactivate_locked_super(s); |
aca740ce | 1577 | return ERR_PTR(-EBUSY); |
1da177e4 | 1578 | } |
aca740ce | 1579 | } else { |
4f331f01 | 1580 | /* |
aca740ce JK |
1581 | * We drop s_umount here because we need to open the bdev and |
1582 | * bdev->open_mutex ranks above s_umount (blkdev_put() -> | |
560e20e4 | 1583 | * bdev_mark_dead()). It is safe because we have active sb |
aca740ce | 1584 | * reference and SB_BORN is not set yet. |
4f331f01 | 1585 | */ |
0ed33598 | 1586 | super_unlock_excl(s); |
aca740ce | 1587 | error = setup_bdev_super(s, flags, NULL); |
5e874914 | 1588 | __super_lock_excl(s); |
aca740ce JK |
1589 | if (!error) |
1590 | error = fill_super(s, data, flags & SB_SILENT ? 1 : 0); | |
1da177e4 | 1591 | if (error) { |
74dbbdd7 | 1592 | deactivate_locked_super(s); |
aca740ce | 1593 | return ERR_PTR(error); |
fa675765 | 1594 | } |
454e2398 | 1595 | |
e462ec50 | 1596 | s->s_flags |= SB_ACTIVE; |
1da177e4 LT |
1597 | } |
1598 | ||
152a0836 | 1599 | return dget(s->s_root); |
152a0836 AV |
1600 | } |
1601 | EXPORT_SYMBOL(mount_bdev); | |
1602 | ||
1da177e4 LT |
1603 | void kill_block_super(struct super_block *sb) |
1604 | { | |
1605 | struct block_device *bdev = sb->s_bdev; | |
1606 | ||
1da177e4 | 1607 | generic_shutdown_super(sb); |
aca740ce JK |
1608 | if (bdev) { |
1609 | sync_blockdev(bdev); | |
2ea6f689 | 1610 | blkdev_put(bdev, sb); |
aca740ce | 1611 | } |
1da177e4 LT |
1612 | } |
1613 | ||
1614 | EXPORT_SYMBOL(kill_block_super); | |
9361401e | 1615 | #endif |
1da177e4 | 1616 | |
3c26ff6e | 1617 | struct dentry *mount_nodev(struct file_system_type *fs_type, |
1da177e4 | 1618 | int flags, void *data, |
3c26ff6e | 1619 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1620 | { |
1621 | int error; | |
9249e17f | 1622 | struct super_block *s = sget(fs_type, NULL, set_anon_super, flags, NULL); |
1da177e4 LT |
1623 | |
1624 | if (IS_ERR(s)) | |
3c26ff6e | 1625 | return ERR_CAST(s); |
1da177e4 | 1626 | |
e462ec50 | 1627 | error = fill_super(s, data, flags & SB_SILENT ? 1 : 0); |
1da177e4 | 1628 | if (error) { |
74dbbdd7 | 1629 | deactivate_locked_super(s); |
3c26ff6e | 1630 | return ERR_PTR(error); |
1da177e4 | 1631 | } |
e462ec50 | 1632 | s->s_flags |= SB_ACTIVE; |
3c26ff6e | 1633 | return dget(s->s_root); |
1da177e4 | 1634 | } |
3c26ff6e AV |
1635 | EXPORT_SYMBOL(mount_nodev); |
1636 | ||
a6097180 N |
1637 | int reconfigure_single(struct super_block *s, |
1638 | int flags, void *data) | |
8d0347f6 DH |
1639 | { |
1640 | struct fs_context *fc; | |
1641 | int ret; | |
1642 | ||
1643 | /* The caller really need to be passing fc down into mount_single(), | |
1644 | * then a chunk of this can be removed. [Bollocks -- AV] | |
1645 | * Better yet, reconfiguration shouldn't happen, but rather the second | |
1646 | * mount should be rejected if the parameters are not compatible. | |
1647 | */ | |
1648 | fc = fs_context_for_reconfigure(s->s_root, flags, MS_RMT_MASK); | |
1649 | if (IS_ERR(fc)) | |
1650 | return PTR_ERR(fc); | |
1651 | ||
1652 | ret = parse_monolithic_mount_data(fc, data); | |
1653 | if (ret < 0) | |
1654 | goto out; | |
1655 | ||
1656 | ret = reconfigure_super(fc); | |
1657 | out: | |
1658 | put_fs_context(fc); | |
1659 | return ret; | |
1660 | } | |
1661 | ||
1da177e4 LT |
1662 | static int compare_single(struct super_block *s, void *p) |
1663 | { | |
1664 | return 1; | |
1665 | } | |
1666 | ||
fc14f2fe | 1667 | struct dentry *mount_single(struct file_system_type *fs_type, |
1da177e4 | 1668 | int flags, void *data, |
fc14f2fe | 1669 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1670 | { |
1671 | struct super_block *s; | |
1672 | int error; | |
1673 | ||
9249e17f | 1674 | s = sget(fs_type, compare_single, set_anon_super, flags, NULL); |
1da177e4 | 1675 | if (IS_ERR(s)) |
fc14f2fe | 1676 | return ERR_CAST(s); |
1da177e4 | 1677 | if (!s->s_root) { |
e462ec50 | 1678 | error = fill_super(s, data, flags & SB_SILENT ? 1 : 0); |
8d0347f6 DH |
1679 | if (!error) |
1680 | s->s_flags |= SB_ACTIVE; | |
9329d1be | 1681 | } else { |
8d0347f6 DH |
1682 | error = reconfigure_single(s, flags, data); |
1683 | } | |
1684 | if (unlikely(error)) { | |
1685 | deactivate_locked_super(s); | |
1686 | return ERR_PTR(error); | |
1da177e4 | 1687 | } |
fc14f2fe AV |
1688 | return dget(s->s_root); |
1689 | } | |
1690 | EXPORT_SYMBOL(mount_single); | |
1691 | ||
9bc61ab1 DH |
1692 | /** |
1693 | * vfs_get_tree - Get the mountable root | |
1694 | * @fc: The superblock configuration context. | |
1695 | * | |
1696 | * The filesystem is invoked to get or create a superblock which can then later | |
1697 | * be used for mounting. The filesystem places a pointer to the root to be | |
1698 | * used for mounting in @fc->root. | |
1699 | */ | |
1700 | int vfs_get_tree(struct fs_context *fc) | |
1da177e4 | 1701 | { |
9d412a43 | 1702 | struct super_block *sb; |
9bc61ab1 | 1703 | int error; |
8089352a | 1704 | |
f3a09c92 AV |
1705 | if (fc->root) |
1706 | return -EBUSY; | |
1707 | ||
1708 | /* Get the mountable root in fc->root, with a ref on the root and a ref | |
1709 | * on the superblock. | |
1710 | */ | |
1711 | error = fc->ops->get_tree(fc); | |
9bc61ab1 DH |
1712 | if (error < 0) |
1713 | return error; | |
1da177e4 | 1714 | |
f3a09c92 AV |
1715 | if (!fc->root) { |
1716 | pr_err("Filesystem %s get_tree() didn't set fc->root\n", | |
1717 | fc->fs_type->name); | |
1718 | /* We don't know what the locking state of the superblock is - | |
1719 | * if there is a superblock. | |
1720 | */ | |
1721 | BUG(); | |
1722 | } | |
1723 | ||
9bc61ab1 | 1724 | sb = fc->root->d_sb; |
9d412a43 | 1725 | WARN_ON(!sb->s_bdi); |
79f546a6 DC |
1726 | |
1727 | /* | |
5e874914 CB |
1728 | * super_wake() contains a memory barrier which also care of |
1729 | * ordering for super_cache_count(). We place it before setting | |
1730 | * SB_BORN as the data dependency between the two functions is | |
1731 | * the superblock structure contents that we just set up, not | |
1732 | * the SB_BORN flag. | |
79f546a6 | 1733 | */ |
5e874914 | 1734 | super_wake(sb, SB_BORN); |
454e2398 | 1735 | |
9bc61ab1 | 1736 | error = security_sb_set_mnt_opts(sb, fc->security, 0, NULL); |
c9ce29ed AV |
1737 | if (unlikely(error)) { |
1738 | fc_drop_locked(fc); | |
1739 | return error; | |
a10d7c22 AV |
1740 | } |
1741 | ||
42cb56ae JL |
1742 | /* |
1743 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE | |
1744 | * but s_maxbytes was an unsigned long long for many releases. Throw | |
1745 | * this warning for a little while to try and catch filesystems that | |
4358b567 | 1746 | * violate this rule. |
42cb56ae | 1747 | */ |
9d412a43 | 1748 | WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " |
9bc61ab1 | 1749 | "negative value (%lld)\n", fc->fs_type->name, sb->s_maxbytes); |
42cb56ae | 1750 | |
9bc61ab1 | 1751 | return 0; |
1da177e4 | 1752 | } |
9bc61ab1 | 1753 | EXPORT_SYMBOL(vfs_get_tree); |
1da177e4 | 1754 | |
fca39346 JK |
1755 | /* |
1756 | * Setup private BDI for given superblock. It gets automatically cleaned up | |
1757 | * in generic_shutdown_super(). | |
1758 | */ | |
1759 | int super_setup_bdi_name(struct super_block *sb, char *fmt, ...) | |
1760 | { | |
1761 | struct backing_dev_info *bdi; | |
1762 | int err; | |
1763 | va_list args; | |
1764 | ||
aef33c2f | 1765 | bdi = bdi_alloc(NUMA_NO_NODE); |
fca39346 JK |
1766 | if (!bdi) |
1767 | return -ENOMEM; | |
1768 | ||
fca39346 | 1769 | va_start(args, fmt); |
7c4cc300 | 1770 | err = bdi_register_va(bdi, fmt, args); |
fca39346 JK |
1771 | va_end(args); |
1772 | if (err) { | |
1773 | bdi_put(bdi); | |
1774 | return err; | |
1775 | } | |
1776 | WARN_ON(sb->s_bdi != &noop_backing_dev_info); | |
1777 | sb->s_bdi = bdi; | |
0b3ea092 | 1778 | sb->s_iflags |= SB_I_PERSB_BDI; |
fca39346 JK |
1779 | |
1780 | return 0; | |
1781 | } | |
1782 | EXPORT_SYMBOL(super_setup_bdi_name); | |
1783 | ||
1784 | /* | |
1785 | * Setup private BDI for given superblock. I gets automatically cleaned up | |
1786 | * in generic_shutdown_super(). | |
1787 | */ | |
1788 | int super_setup_bdi(struct super_block *sb) | |
1789 | { | |
1790 | static atomic_long_t bdi_seq = ATOMIC_LONG_INIT(0); | |
1791 | ||
1792 | return super_setup_bdi_name(sb, "%.28s-%ld", sb->s_type->name, | |
1793 | atomic_long_inc_return(&bdi_seq)); | |
1794 | } | |
1795 | EXPORT_SYMBOL(super_setup_bdi); | |
1796 | ||
5accdf82 JK |
1797 | /** |
1798 | * sb_wait_write - wait until all writers to given file system finish | |
1799 | * @sb: the super for which we wait | |
1800 | * @level: type of writers we wait for (normal vs page fault) | |
1801 | * | |
1802 | * This function waits until there are no writers of given type to given file | |
8129ed29 | 1803 | * system. |
5accdf82 JK |
1804 | */ |
1805 | static void sb_wait_write(struct super_block *sb, int level) | |
1806 | { | |
8129ed29 | 1807 | percpu_down_write(sb->s_writers.rw_sem + level-1); |
8129ed29 | 1808 | } |
5accdf82 | 1809 | |
f1a96220 ON |
1810 | /* |
1811 | * We are going to return to userspace and forget about these locks, the | |
1812 | * ownership goes to the caller of thaw_super() which does unlock(). | |
1813 | */ | |
1814 | static void lockdep_sb_freeze_release(struct super_block *sb) | |
1815 | { | |
1816 | int level; | |
1817 | ||
1818 | for (level = SB_FREEZE_LEVELS - 1; level >= 0; level--) | |
1819 | percpu_rwsem_release(sb->s_writers.rw_sem + level, 0, _THIS_IP_); | |
1820 | } | |
1821 | ||
1822 | /* | |
1823 | * Tell lockdep we are holding these locks before we call ->unfreeze_fs(sb). | |
1824 | */ | |
1825 | static void lockdep_sb_freeze_acquire(struct super_block *sb) | |
8129ed29 ON |
1826 | { |
1827 | int level; | |
5accdf82 | 1828 | |
8129ed29 ON |
1829 | for (level = 0; level < SB_FREEZE_LEVELS; ++level) |
1830 | percpu_rwsem_acquire(sb->s_writers.rw_sem + level, 0, _THIS_IP_); | |
f1a96220 ON |
1831 | } |
1832 | ||
2719c716 | 1833 | static void sb_freeze_unlock(struct super_block *sb, int level) |
f1a96220 | 1834 | { |
2719c716 | 1835 | for (level--; level >= 0; level--) |
8129ed29 | 1836 | percpu_up_write(sb->s_writers.rw_sem + level); |
5accdf82 JK |
1837 | } |
1838 | ||
18e9e510 | 1839 | /** |
7000d3c4 RD |
1840 | * freeze_super - lock the filesystem and force it into a consistent state |
1841 | * @sb: the super to lock | |
18e9e510 JB |
1842 | * |
1843 | * Syncs the super to make sure the filesystem is consistent and calls the fs's | |
1844 | * freeze_fs. Subsequent calls to this without first thawing the fs will return | |
1845 | * -EBUSY. | |
5accdf82 JK |
1846 | * |
1847 | * During this function, sb->s_writers.frozen goes through these values: | |
1848 | * | |
1849 | * SB_UNFROZEN: File system is normal, all writes progress as usual. | |
1850 | * | |
1851 | * SB_FREEZE_WRITE: The file system is in the process of being frozen. New | |
1852 | * writes should be blocked, though page faults are still allowed. We wait for | |
1853 | * all writes to complete and then proceed to the next stage. | |
1854 | * | |
1855 | * SB_FREEZE_PAGEFAULT: Freezing continues. Now also page faults are blocked | |
1856 | * but internal fs threads can still modify the filesystem (although they | |
1857 | * should not dirty new pages or inodes), writeback can run etc. After waiting | |
1858 | * for all running page faults we sync the filesystem which will clean all | |
1859 | * dirty pages and inodes (no new dirty pages or inodes can be created when | |
1860 | * sync is running). | |
1861 | * | |
1862 | * SB_FREEZE_FS: The file system is frozen. Now all internal sources of fs | |
1863 | * modification are blocked (e.g. XFS preallocation truncation on inode | |
1864 | * reclaim). This is usually implemented by blocking new transactions for | |
1865 | * filesystems that have them and need this additional guard. After all | |
1866 | * internal writers are finished we call ->freeze_fs() to finish filesystem | |
1867 | * freezing. Then we transition to SB_FREEZE_COMPLETE state. This state is | |
1868 | * mostly auxiliary for filesystems to verify they do not modify frozen fs. | |
1869 | * | |
1870 | * sb->s_writers.frozen is protected by sb->s_umount. | |
18e9e510 JB |
1871 | */ |
1872 | int freeze_super(struct super_block *sb) | |
1873 | { | |
1874 | int ret; | |
1875 | ||
1876 | atomic_inc(&sb->s_active); | |
5e874914 | 1877 | __super_lock_excl(sb); |
5accdf82 | 1878 | if (sb->s_writers.frozen != SB_UNFROZEN) { |
18e9e510 JB |
1879 | deactivate_locked_super(sb); |
1880 | return -EBUSY; | |
1881 | } | |
1882 | ||
e462ec50 | 1883 | if (!(sb->s_flags & SB_BORN)) { |
0ed33598 | 1884 | super_unlock_excl(sb); |
dabe0dc1 AV |
1885 | return 0; /* sic - it's "nothing to do" */ |
1886 | } | |
1887 | ||
bc98a42c | 1888 | if (sb_rdonly(sb)) { |
5accdf82 JK |
1889 | /* Nothing to do really... */ |
1890 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
0ed33598 | 1891 | super_unlock_excl(sb); |
18e9e510 JB |
1892 | return 0; |
1893 | } | |
1894 | ||
5accdf82 | 1895 | sb->s_writers.frozen = SB_FREEZE_WRITE; |
5accdf82 | 1896 | /* Release s_umount to preserve sb_start_write -> s_umount ordering */ |
0ed33598 | 1897 | super_unlock_excl(sb); |
5accdf82 | 1898 | sb_wait_write(sb, SB_FREEZE_WRITE); |
5e874914 | 1899 | __super_lock_excl(sb); |
5accdf82 JK |
1900 | |
1901 | /* Now we go and block page faults... */ | |
5accdf82 | 1902 | sb->s_writers.frozen = SB_FREEZE_PAGEFAULT; |
5accdf82 JK |
1903 | sb_wait_write(sb, SB_FREEZE_PAGEFAULT); |
1904 | ||
1905 | /* All writers are done so after syncing there won't be dirty data */ | |
2719c716 DW |
1906 | ret = sync_filesystem(sb); |
1907 | if (ret) { | |
1908 | sb->s_writers.frozen = SB_UNFROZEN; | |
1909 | sb_freeze_unlock(sb, SB_FREEZE_PAGEFAULT); | |
2719c716 DW |
1910 | deactivate_locked_super(sb); |
1911 | return ret; | |
1912 | } | |
18e9e510 | 1913 | |
5accdf82 JK |
1914 | /* Now wait for internal filesystem counter */ |
1915 | sb->s_writers.frozen = SB_FREEZE_FS; | |
5accdf82 | 1916 | sb_wait_write(sb, SB_FREEZE_FS); |
18e9e510 | 1917 | |
18e9e510 JB |
1918 | if (sb->s_op->freeze_fs) { |
1919 | ret = sb->s_op->freeze_fs(sb); | |
1920 | if (ret) { | |
1921 | printk(KERN_ERR | |
1922 | "VFS:Filesystem freeze failed\n"); | |
5accdf82 | 1923 | sb->s_writers.frozen = SB_UNFROZEN; |
2719c716 | 1924 | sb_freeze_unlock(sb, SB_FREEZE_FS); |
18e9e510 JB |
1925 | deactivate_locked_super(sb); |
1926 | return ret; | |
1927 | } | |
1928 | } | |
5accdf82 | 1929 | /* |
89f39af1 ON |
1930 | * For debugging purposes so that fs can warn if it sees write activity |
1931 | * when frozen is set to SB_FREEZE_COMPLETE, and for thaw_super(). | |
5accdf82 JK |
1932 | */ |
1933 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
f1a96220 | 1934 | lockdep_sb_freeze_release(sb); |
0ed33598 | 1935 | super_unlock_excl(sb); |
18e9e510 JB |
1936 | return 0; |
1937 | } | |
1938 | EXPORT_SYMBOL(freeze_super); | |
1939 | ||
08fdc8a0 | 1940 | static int thaw_super_locked(struct super_block *sb) |
18e9e510 JB |
1941 | { |
1942 | int error; | |
1943 | ||
89f39af1 | 1944 | if (sb->s_writers.frozen != SB_FREEZE_COMPLETE) { |
0ed33598 | 1945 | super_unlock_excl(sb); |
18e9e510 JB |
1946 | return -EINVAL; |
1947 | } | |
1948 | ||
bc98a42c | 1949 | if (sb_rdonly(sb)) { |
8129ed29 | 1950 | sb->s_writers.frozen = SB_UNFROZEN; |
18e9e510 | 1951 | goto out; |
8129ed29 | 1952 | } |
18e9e510 | 1953 | |
f1a96220 ON |
1954 | lockdep_sb_freeze_acquire(sb); |
1955 | ||
18e9e510 JB |
1956 | if (sb->s_op->unfreeze_fs) { |
1957 | error = sb->s_op->unfreeze_fs(sb); | |
1958 | if (error) { | |
1959 | printk(KERN_ERR | |
1960 | "VFS:Filesystem thaw failed\n"); | |
f1a96220 | 1961 | lockdep_sb_freeze_release(sb); |
0ed33598 | 1962 | super_unlock_excl(sb); |
18e9e510 JB |
1963 | return error; |
1964 | } | |
1965 | } | |
1966 | ||
5accdf82 | 1967 | sb->s_writers.frozen = SB_UNFROZEN; |
2719c716 | 1968 | sb_freeze_unlock(sb, SB_FREEZE_FS); |
8129ed29 | 1969 | out: |
18e9e510 | 1970 | deactivate_locked_super(sb); |
18e9e510 JB |
1971 | return 0; |
1972 | } | |
08fdc8a0 | 1973 | |
961f3c89 MCC |
1974 | /** |
1975 | * thaw_super -- unlock filesystem | |
1976 | * @sb: the super to thaw | |
1977 | * | |
1978 | * Unlocks the filesystem and marks it writeable again after freeze_super(). | |
1979 | */ | |
08fdc8a0 MG |
1980 | int thaw_super(struct super_block *sb) |
1981 | { | |
5e874914 | 1982 | __super_lock_excl(sb); |
08fdc8a0 MG |
1983 | return thaw_super_locked(sb); |
1984 | } | |
18e9e510 | 1985 | EXPORT_SYMBOL(thaw_super); |
439bc39b CH |
1986 | |
1987 | /* | |
1988 | * Create workqueue for deferred direct IO completions. We allocate the | |
1989 | * workqueue when it's first needed. This avoids creating workqueue for | |
1990 | * filesystems that don't need it and also allows us to create the workqueue | |
1991 | * late enough so the we can include s_id in the name of the workqueue. | |
1992 | */ | |
1993 | int sb_init_dio_done_wq(struct super_block *sb) | |
1994 | { | |
1995 | struct workqueue_struct *old; | |
1996 | struct workqueue_struct *wq = alloc_workqueue("dio/%s", | |
1997 | WQ_MEM_RECLAIM, 0, | |
1998 | sb->s_id); | |
1999 | if (!wq) | |
2000 | return -ENOMEM; | |
2001 | /* | |
2002 | * This has to be atomic as more DIOs can race to create the workqueue | |
2003 | */ | |
2004 | old = cmpxchg(&sb->s_dio_done_wq, NULL, wq); | |
2005 | /* Someone created workqueue before us? Free ours... */ | |
2006 | if (old) | |
2007 | destroy_workqueue(wq); | |
2008 | return 0; | |
2009 | } |