]>
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> |
c515e1fd | 34 | #include <linux/cleancache.h> |
40401530 | 35 | #include <linux/fsnotify.h> |
5accdf82 | 36 | #include <linux/lockdep.h> |
6e4eab57 | 37 | #include <linux/user_namespace.h> |
6d59e7f5 | 38 | #include "internal.h" |
1da177e4 LT |
39 | |
40 | ||
15d0f5ea AV |
41 | static LIST_HEAD(super_blocks); |
42 | static DEFINE_SPINLOCK(sb_lock); | |
1da177e4 | 43 | |
5accdf82 JK |
44 | static char *sb_writers_name[SB_FREEZE_LEVELS] = { |
45 | "sb_writers", | |
46 | "sb_pagefaults", | |
47 | "sb_internal", | |
48 | }; | |
49 | ||
b0d40c92 DC |
50 | /* |
51 | * One thing we have to be careful of with a per-sb shrinker is that we don't | |
52 | * drop the last active reference to the superblock from within the shrinker. | |
53 | * If that happens we could trigger unregistering the shrinker from within the | |
54 | * shrinker path and that leads to deadlock on the shrinker_rwsem. Hence we | |
55 | * take a passive reference to the superblock to avoid this from occurring. | |
56 | */ | |
0a234c6d DC |
57 | static unsigned long super_cache_scan(struct shrinker *shrink, |
58 | struct shrink_control *sc) | |
b0d40c92 DC |
59 | { |
60 | struct super_block *sb; | |
0a234c6d DC |
61 | long fs_objects = 0; |
62 | long total_objects; | |
63 | long freed = 0; | |
64 | long dentries; | |
65 | long inodes; | |
b0d40c92 DC |
66 | |
67 | sb = container_of(shrink, struct super_block, s_shrink); | |
68 | ||
69 | /* | |
70 | * Deadlock avoidance. We may hold various FS locks, and we don't want | |
71 | * to recurse into the FS that called us in clear_inode() and friends.. | |
72 | */ | |
0a234c6d DC |
73 | if (!(sc->gfp_mask & __GFP_FS)) |
74 | return SHRINK_STOP; | |
b0d40c92 | 75 | |
eb6ef3df | 76 | if (!trylock_super(sb)) |
0a234c6d | 77 | return SHRINK_STOP; |
b0d40c92 | 78 | |
d0407903 | 79 | if (sb->s_op->nr_cached_objects) |
4101b624 | 80 | fs_objects = sb->s_op->nr_cached_objects(sb, sc); |
0e1fdafd | 81 | |
503c358c VD |
82 | inodes = list_lru_shrink_count(&sb->s_inode_lru, sc); |
83 | dentries = list_lru_shrink_count(&sb->s_dentry_lru, sc); | |
f6041567 | 84 | total_objects = dentries + inodes + fs_objects + 1; |
475d0db7 TH |
85 | if (!total_objects) |
86 | total_objects = 1; | |
0e1fdafd | 87 | |
0a234c6d | 88 | /* proportion the scan between the caches */ |
f6041567 | 89 | dentries = mult_frac(sc->nr_to_scan, dentries, total_objects); |
bc3b14cb | 90 | inodes = mult_frac(sc->nr_to_scan, inodes, total_objects); |
503c358c | 91 | fs_objects = mult_frac(sc->nr_to_scan, fs_objects, total_objects); |
b0d40c92 | 92 | |
0a234c6d DC |
93 | /* |
94 | * prune the dcache first as the icache is pinned by it, then | |
95 | * prune the icache, followed by the filesystem specific caches | |
49e7e7ff VD |
96 | * |
97 | * Ensure that we always scan at least one object - memcg kmem | |
98 | * accounting uses this to fully empty the caches. | |
0a234c6d | 99 | */ |
49e7e7ff | 100 | sc->nr_to_scan = dentries + 1; |
503c358c | 101 | freed = prune_dcache_sb(sb, sc); |
49e7e7ff | 102 | sc->nr_to_scan = inodes + 1; |
503c358c | 103 | freed += prune_icache_sb(sb, sc); |
0a234c6d DC |
104 | |
105 | if (fs_objects) { | |
49e7e7ff | 106 | sc->nr_to_scan = fs_objects + 1; |
4101b624 | 107 | freed += sb->s_op->free_cached_objects(sb, sc); |
b0d40c92 DC |
108 | } |
109 | ||
eb6ef3df | 110 | up_read(&sb->s_umount); |
0a234c6d DC |
111 | return freed; |
112 | } | |
113 | ||
114 | static unsigned long super_cache_count(struct shrinker *shrink, | |
115 | struct shrink_control *sc) | |
116 | { | |
117 | struct super_block *sb; | |
118 | long total_objects = 0; | |
119 | ||
120 | sb = container_of(shrink, struct super_block, s_shrink); | |
121 | ||
d23da150 | 122 | /* |
b9659ff3 DC |
123 | * We don't call trylock_super() here as it is a scalability bottleneck, |
124 | * so we're exposed to partial setup state. The shrinker rwsem does not | |
125 | * protect filesystem operations backing list_lru_shrink_count() or | |
126 | * s_op->nr_cached_objects(). Counts can change between | |
127 | * super_cache_count and super_cache_scan, so we really don't need locks | |
128 | * here. | |
129 | * | |
130 | * However, if we are currently mounting the superblock, the underlying | |
131 | * filesystem might be in a state of partial construction and hence it | |
132 | * is dangerous to access it. trylock_super() uses a SB_BORN check to | |
133 | * avoid this situation, so do the same here. The memory barrier is | |
134 | * matched with the one in mount_fs() as we don't hold locks here. | |
d23da150 | 135 | */ |
b9659ff3 DC |
136 | if (!(sb->s_flags & SB_BORN)) |
137 | return 0; | |
138 | smp_rmb(); | |
139 | ||
0a234c6d | 140 | if (sb->s_op && sb->s_op->nr_cached_objects) |
4101b624 | 141 | total_objects = sb->s_op->nr_cached_objects(sb, sc); |
0a234c6d | 142 | |
503c358c VD |
143 | total_objects += list_lru_shrink_count(&sb->s_dentry_lru, sc); |
144 | total_objects += list_lru_shrink_count(&sb->s_inode_lru, sc); | |
0a234c6d | 145 | |
55f841ce | 146 | total_objects = vfs_pressure_ratio(total_objects); |
0e1fdafd | 147 | return total_objects; |
b0d40c92 DC |
148 | } |
149 | ||
853b39a7 ON |
150 | static void destroy_super_work(struct work_struct *work) |
151 | { | |
152 | struct super_block *s = container_of(work, struct super_block, | |
153 | destroy_work); | |
154 | int i; | |
155 | ||
156 | for (i = 0; i < SB_FREEZE_LEVELS; i++) | |
8129ed29 | 157 | percpu_free_rwsem(&s->s_writers.rw_sem[i]); |
853b39a7 ON |
158 | kfree(s); |
159 | } | |
160 | ||
161 | static void destroy_super_rcu(struct rcu_head *head) | |
162 | { | |
163 | struct super_block *s = container_of(head, struct super_block, rcu); | |
164 | INIT_WORK(&s->destroy_work, destroy_super_work); | |
165 | schedule_work(&s->destroy_work); | |
166 | } | |
167 | ||
7eb5e882 AV |
168 | /** |
169 | * destroy_super - frees a superblock | |
170 | * @s: superblock to free | |
171 | * | |
172 | * Frees a superblock. | |
173 | */ | |
174 | static void destroy_super(struct super_block *s) | |
5accdf82 | 175 | { |
7eb5e882 AV |
176 | list_lru_destroy(&s->s_dentry_lru); |
177 | list_lru_destroy(&s->s_inode_lru); | |
7eb5e882 AV |
178 | security_sb_free(s); |
179 | WARN_ON(!list_empty(&s->s_mounts)); | |
6e4eab57 | 180 | put_user_ns(s->s_user_ns); |
7eb5e882 | 181 | kfree(s->s_subtype); |
853b39a7 | 182 | call_rcu(&s->rcu, destroy_super_rcu); |
5accdf82 JK |
183 | } |
184 | ||
1da177e4 LT |
185 | /** |
186 | * alloc_super - create new superblock | |
fe2bbc48 | 187 | * @type: filesystem type superblock should belong to |
9249e17f | 188 | * @flags: the mount flags |
6e4eab57 | 189 | * @user_ns: User namespace for the super_block |
1da177e4 LT |
190 | * |
191 | * Allocates and initializes a new &struct super_block. alloc_super() | |
192 | * returns a pointer new superblock or %NULL if allocation had failed. | |
193 | */ | |
6e4eab57 EB |
194 | static struct super_block *alloc_super(struct file_system_type *type, int flags, |
195 | struct user_namespace *user_ns) | |
1da177e4 | 196 | { |
11b0b5ab | 197 | struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER); |
b87221de | 198 | static const struct super_operations default_op; |
7eb5e882 AV |
199 | int i; |
200 | ||
201 | if (!s) | |
202 | return NULL; | |
1da177e4 | 203 | |
b5bd856a | 204 | INIT_LIST_HEAD(&s->s_mounts); |
6e4eab57 | 205 | s->s_user_ns = get_user_ns(user_ns); |
b5bd856a | 206 | |
7eb5e882 AV |
207 | if (security_sb_alloc(s)) |
208 | goto fail; | |
7b7a8665 | 209 | |
7eb5e882 | 210 | for (i = 0; i < SB_FREEZE_LEVELS; i++) { |
8129ed29 ON |
211 | if (__percpu_init_rwsem(&s->s_writers.rw_sem[i], |
212 | sb_writers_name[i], | |
213 | &type->s_writers_key[i])) | |
7eb5e882 | 214 | goto fail; |
1da177e4 | 215 | } |
7eb5e882 | 216 | init_waitqueue_head(&s->s_writers.wait_unfrozen); |
df0ce26c | 217 | s->s_bdi = &noop_backing_dev_info; |
7eb5e882 | 218 | s->s_flags = flags; |
cc50a07a | 219 | if (s->s_user_ns != &init_user_ns) |
67690f93 | 220 | s->s_iflags |= SB_I_NODEV; |
7eb5e882 AV |
221 | INIT_HLIST_NODE(&s->s_instances); |
222 | INIT_HLIST_BL_HEAD(&s->s_anon); | |
e97fedb9 | 223 | mutex_init(&s->s_sync_lock); |
7eb5e882 | 224 | INIT_LIST_HEAD(&s->s_inodes); |
74278da9 | 225 | spin_lock_init(&s->s_inode_list_lock); |
6c60d2b5 DC |
226 | INIT_LIST_HEAD(&s->s_inodes_wb); |
227 | spin_lock_init(&s->s_inode_wblist_lock); | |
7eb5e882 | 228 | |
2acb60a0 | 229 | if (list_lru_init_memcg(&s->s_dentry_lru)) |
7eb5e882 | 230 | goto fail; |
2acb60a0 | 231 | if (list_lru_init_memcg(&s->s_inode_lru)) |
7eb5e882 AV |
232 | goto fail; |
233 | ||
7eb5e882 AV |
234 | init_rwsem(&s->s_umount); |
235 | lockdep_set_class(&s->s_umount, &type->s_umount_key); | |
236 | /* | |
237 | * sget() can have s_umount recursion. | |
238 | * | |
239 | * When it cannot find a suitable sb, it allocates a new | |
240 | * one (this one), and tries again to find a suitable old | |
241 | * one. | |
242 | * | |
243 | * In case that succeeds, it will acquire the s_umount | |
244 | * lock of the old one. Since these are clearly distrinct | |
245 | * locks, and this object isn't exposed yet, there's no | |
246 | * risk of deadlocks. | |
247 | * | |
248 | * Annotate this by putting this lock in a different | |
249 | * subclass. | |
250 | */ | |
251 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); | |
252 | s->s_count = 1; | |
253 | atomic_set(&s->s_active, 1); | |
254 | mutex_init(&s->s_vfs_rename_mutex); | |
255 | lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key); | |
bc8230ee | 256 | init_rwsem(&s->s_dquot.dqio_sem); |
7eb5e882 AV |
257 | s->s_maxbytes = MAX_NON_LFS; |
258 | s->s_op = &default_op; | |
259 | s->s_time_gran = 1000000000; | |
3cb29d11 | 260 | s->cleancache_poolid = CLEANCACHE_NO_POOL; |
7eb5e882 AV |
261 | |
262 | s->s_shrink.seeks = DEFAULT_SEEKS; | |
263 | s->s_shrink.scan_objects = super_cache_scan; | |
264 | s->s_shrink.count_objects = super_cache_count; | |
265 | s->s_shrink.batch = 1024; | |
2acb60a0 | 266 | s->s_shrink.flags = SHRINKER_NUMA_AWARE | SHRINKER_MEMCG_AWARE; |
1da177e4 | 267 | return s; |
5ca302c8 | 268 | |
7eb5e882 AV |
269 | fail: |
270 | destroy_super(s); | |
271 | return NULL; | |
1da177e4 LT |
272 | } |
273 | ||
274 | /* Superblock refcounting */ | |
275 | ||
276 | /* | |
35cf7ba0 | 277 | * Drop a superblock's refcount. The caller must hold sb_lock. |
1da177e4 | 278 | */ |
f47ec3f2 | 279 | static void __put_super(struct super_block *sb) |
1da177e4 | 280 | { |
1da177e4 | 281 | if (!--sb->s_count) { |
551de6f3 | 282 | list_del_init(&sb->s_list); |
1da177e4 | 283 | destroy_super(sb); |
1da177e4 | 284 | } |
1da177e4 LT |
285 | } |
286 | ||
287 | /** | |
288 | * put_super - drop a temporary reference to superblock | |
289 | * @sb: superblock in question | |
290 | * | |
291 | * Drops a temporary reference, frees superblock if there's no | |
292 | * references left. | |
293 | */ | |
f47ec3f2 | 294 | static void put_super(struct super_block *sb) |
1da177e4 LT |
295 | { |
296 | spin_lock(&sb_lock); | |
297 | __put_super(sb); | |
298 | spin_unlock(&sb_lock); | |
299 | } | |
300 | ||
301 | ||
302 | /** | |
1712ac8f | 303 | * deactivate_locked_super - drop an active reference to superblock |
1da177e4 LT |
304 | * @s: superblock to deactivate |
305 | * | |
bd7ced98 | 306 | * Drops an active reference to superblock, converting it into a temporary |
1712ac8f | 307 | * one if there is no other active references left. In that case we |
1da177e4 LT |
308 | * tell fs driver to shut it down and drop the temporary reference we |
309 | * had just acquired. | |
1712ac8f AV |
310 | * |
311 | * Caller holds exclusive lock on superblock; that lock is released. | |
1da177e4 | 312 | */ |
1712ac8f | 313 | void deactivate_locked_super(struct super_block *s) |
1da177e4 LT |
314 | { |
315 | struct file_system_type *fs = s->s_type; | |
b20bd1a5 | 316 | if (atomic_dec_and_test(&s->s_active)) { |
3167760f | 317 | cleancache_invalidate_fs(s); |
b0d40c92 | 318 | unregister_shrinker(&s->s_shrink); |
28f2cd4f | 319 | fs->kill_sb(s); |
f5e1dd34 | 320 | |
c0a5b560 VD |
321 | /* |
322 | * Since list_lru_destroy() may sleep, we cannot call it from | |
323 | * put_super(), where we hold the sb_lock. Therefore we destroy | |
324 | * the lru lists right now. | |
325 | */ | |
326 | list_lru_destroy(&s->s_dentry_lru); | |
327 | list_lru_destroy(&s->s_inode_lru); | |
328 | ||
1da177e4 LT |
329 | put_filesystem(fs); |
330 | put_super(s); | |
1712ac8f AV |
331 | } else { |
332 | up_write(&s->s_umount); | |
1da177e4 LT |
333 | } |
334 | } | |
335 | ||
1712ac8f | 336 | EXPORT_SYMBOL(deactivate_locked_super); |
1da177e4 | 337 | |
74dbbdd7 | 338 | /** |
1712ac8f | 339 | * deactivate_super - drop an active reference to superblock |
74dbbdd7 AV |
340 | * @s: superblock to deactivate |
341 | * | |
1712ac8f AV |
342 | * Variant of deactivate_locked_super(), except that superblock is *not* |
343 | * locked by caller. If we are going to drop the final active reference, | |
344 | * lock will be acquired prior to that. | |
74dbbdd7 | 345 | */ |
1712ac8f | 346 | void deactivate_super(struct super_block *s) |
74dbbdd7 | 347 | { |
1712ac8f AV |
348 | if (!atomic_add_unless(&s->s_active, -1, 1)) { |
349 | down_write(&s->s_umount); | |
350 | deactivate_locked_super(s); | |
74dbbdd7 AV |
351 | } |
352 | } | |
353 | ||
1712ac8f | 354 | EXPORT_SYMBOL(deactivate_super); |
74dbbdd7 | 355 | |
1da177e4 LT |
356 | /** |
357 | * grab_super - acquire an active reference | |
358 | * @s: reference we are trying to make active | |
359 | * | |
360 | * Tries to acquire an active reference. grab_super() is used when we | |
361 | * had just found a superblock in super_blocks or fs_type->fs_supers | |
362 | * and want to turn it into a full-blown active reference. grab_super() | |
363 | * is called with sb_lock held and drops it. Returns 1 in case of | |
364 | * success, 0 if we had failed (superblock contents was already dead or | |
acfec9a5 AV |
365 | * dying when grab_super() had been called). Note that this is only |
366 | * called for superblocks not in rundown mode (== ones still on ->fs_supers | |
367 | * of their type), so increment of ->s_count is OK here. | |
1da177e4 | 368 | */ |
9c4dbee7 | 369 | static int grab_super(struct super_block *s) __releases(sb_lock) |
1da177e4 LT |
370 | { |
371 | s->s_count++; | |
372 | spin_unlock(&sb_lock); | |
373 | down_write(&s->s_umount); | |
e462ec50 | 374 | if ((s->s_flags & SB_BORN) && atomic_inc_not_zero(&s->s_active)) { |
acfec9a5 AV |
375 | put_super(s); |
376 | return 1; | |
377 | } | |
1da177e4 LT |
378 | up_write(&s->s_umount); |
379 | put_super(s); | |
1da177e4 LT |
380 | return 0; |
381 | } | |
382 | ||
12ad3ab6 | 383 | /* |
eb6ef3df | 384 | * trylock_super - try to grab ->s_umount shared |
331cbdee | 385 | * @sb: reference we are trying to grab |
12ad3ab6 | 386 | * |
eb6ef3df | 387 | * Try to prevent fs shutdown. This is used in places where we |
12ad3ab6 | 388 | * cannot take an active reference but we need to ensure that the |
eb6ef3df KK |
389 | * filesystem is not shut down while we are working on it. It returns |
390 | * false if we cannot acquire s_umount or if we lose the race and | |
391 | * filesystem already got into shutdown, and returns true with the s_umount | |
392 | * lock held in read mode in case of success. On successful return, | |
393 | * the caller must drop the s_umount lock when done. | |
394 | * | |
395 | * Note that unlike get_super() et.al. this one does *not* bump ->s_count. | |
396 | * The reason why it's safe is that we are OK with doing trylock instead | |
397 | * of down_read(). There's a couple of places that are OK with that, but | |
398 | * it's very much not a general-purpose interface. | |
12ad3ab6 | 399 | */ |
eb6ef3df | 400 | bool trylock_super(struct super_block *sb) |
12ad3ab6 | 401 | { |
12ad3ab6 | 402 | if (down_read_trylock(&sb->s_umount)) { |
eb6ef3df | 403 | if (!hlist_unhashed(&sb->s_instances) && |
e462ec50 | 404 | sb->s_root && (sb->s_flags & SB_BORN)) |
12ad3ab6 DC |
405 | return true; |
406 | up_read(&sb->s_umount); | |
407 | } | |
408 | ||
12ad3ab6 DC |
409 | return false; |
410 | } | |
411 | ||
1da177e4 LT |
412 | /** |
413 | * generic_shutdown_super - common helper for ->kill_sb() | |
414 | * @sb: superblock to kill | |
415 | * | |
416 | * generic_shutdown_super() does all fs-independent work on superblock | |
417 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects | |
418 | * that need destruction out of superblock, call generic_shutdown_super() | |
419 | * and release aforementioned objects. Note: dentries and inodes _are_ | |
420 | * taken care of and do not need specific handling. | |
c636ebdb DH |
421 | * |
422 | * Upon calling this function, the filesystem may no longer alter or | |
423 | * rearrange the set of dentries belonging to this super_block, nor may it | |
424 | * change the attachments of dentries to inodes. | |
1da177e4 LT |
425 | */ |
426 | void generic_shutdown_super(struct super_block *sb) | |
427 | { | |
ee9b6d61 | 428 | const struct super_operations *sop = sb->s_op; |
1da177e4 | 429 | |
c636ebdb DH |
430 | if (sb->s_root) { |
431 | shrink_dcache_for_umount(sb); | |
60b0680f | 432 | sync_filesystem(sb); |
e462ec50 | 433 | sb->s_flags &= ~SB_ACTIVE; |
efaee192 | 434 | |
74278da9 | 435 | fsnotify_unmount_inodes(sb); |
a1a0e23e | 436 | cgroup_writeback_umount(); |
63997e98 AV |
437 | |
438 | evict_inodes(sb); | |
1da177e4 | 439 | |
7b7a8665 CH |
440 | if (sb->s_dio_done_wq) { |
441 | destroy_workqueue(sb->s_dio_done_wq); | |
442 | sb->s_dio_done_wq = NULL; | |
443 | } | |
444 | ||
1da177e4 LT |
445 | if (sop->put_super) |
446 | sop->put_super(sb); | |
447 | ||
63997e98 | 448 | if (!list_empty(&sb->s_inodes)) { |
7b4fe29e DJ |
449 | printk("VFS: Busy inodes after unmount of %s. " |
450 | "Self-destruct in 5 seconds. Have a nice day...\n", | |
451 | sb->s_id); | |
1da177e4 | 452 | } |
1da177e4 LT |
453 | } |
454 | spin_lock(&sb_lock); | |
455 | /* should be initialized for __put_super_and_need_restart() */ | |
a5166169 | 456 | hlist_del_init(&sb->s_instances); |
1da177e4 LT |
457 | spin_unlock(&sb_lock); |
458 | up_write(&sb->s_umount); | |
c1844d53 | 459 | if (sb->s_bdi != &noop_backing_dev_info) { |
fca39346 JK |
460 | bdi_put(sb->s_bdi); |
461 | sb->s_bdi = &noop_backing_dev_info; | |
fca39346 | 462 | } |
1da177e4 LT |
463 | } |
464 | ||
465 | EXPORT_SYMBOL(generic_shutdown_super); | |
466 | ||
467 | /** | |
6e4eab57 | 468 | * sget_userns - find or create a superblock |
1da177e4 LT |
469 | * @type: filesystem type superblock should belong to |
470 | * @test: comparison callback | |
471 | * @set: setup callback | |
9249e17f | 472 | * @flags: mount flags |
6e4eab57 | 473 | * @user_ns: User namespace for the super_block |
1da177e4 LT |
474 | * @data: argument to each of them |
475 | */ | |
6e4eab57 | 476 | struct super_block *sget_userns(struct file_system_type *type, |
1da177e4 LT |
477 | int (*test)(struct super_block *,void *), |
478 | int (*set)(struct super_block *,void *), | |
6e4eab57 | 479 | int flags, struct user_namespace *user_ns, |
1da177e4 LT |
480 | void *data) |
481 | { | |
482 | struct super_block *s = NULL; | |
d4730127 | 483 | struct super_block *old; |
1da177e4 LT |
484 | int err; |
485 | ||
e462ec50 | 486 | if (!(flags & (SB_KERNMOUNT|SB_SUBMOUNT)) && |
a001e74c EB |
487 | !(type->fs_flags & FS_USERNS_MOUNT) && |
488 | !capable(CAP_SYS_ADMIN)) | |
489 | return ERR_PTR(-EPERM); | |
1da177e4 LT |
490 | retry: |
491 | spin_lock(&sb_lock); | |
d4730127 | 492 | if (test) { |
b67bfe0d | 493 | hlist_for_each_entry(old, &type->fs_supers, s_instances) { |
d4730127 MK |
494 | if (!test(old, data)) |
495 | continue; | |
6e4eab57 EB |
496 | if (user_ns != old->s_user_ns) { |
497 | spin_unlock(&sb_lock); | |
498 | if (s) { | |
499 | up_write(&s->s_umount); | |
500 | destroy_super(s); | |
501 | } | |
502 | return ERR_PTR(-EBUSY); | |
503 | } | |
d4730127 MK |
504 | if (!grab_super(old)) |
505 | goto retry; | |
a3cfbb53 LZ |
506 | if (s) { |
507 | up_write(&s->s_umount); | |
d4730127 | 508 | destroy_super(s); |
7a4dec53 | 509 | s = NULL; |
a3cfbb53 | 510 | } |
d4730127 MK |
511 | return old; |
512 | } | |
1da177e4 LT |
513 | } |
514 | if (!s) { | |
515 | spin_unlock(&sb_lock); | |
e462ec50 | 516 | s = alloc_super(type, (flags & ~SB_SUBMOUNT), user_ns); |
1da177e4 LT |
517 | if (!s) |
518 | return ERR_PTR(-ENOMEM); | |
519 | goto retry; | |
520 | } | |
dd111b31 | 521 | |
1da177e4 LT |
522 | err = set(s, data); |
523 | if (err) { | |
524 | spin_unlock(&sb_lock); | |
a3cfbb53 | 525 | up_write(&s->s_umount); |
1da177e4 LT |
526 | destroy_super(s); |
527 | return ERR_PTR(err); | |
528 | } | |
529 | s->s_type = type; | |
530 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); | |
531 | list_add_tail(&s->s_list, &super_blocks); | |
a5166169 | 532 | hlist_add_head(&s->s_instances, &type->fs_supers); |
1da177e4 LT |
533 | spin_unlock(&sb_lock); |
534 | get_filesystem(type); | |
ac4dc9f1 AV |
535 | err = register_shrinker(&s->s_shrink); |
536 | if (err) { | |
537 | deactivate_locked_super(s); | |
538 | s = ERR_PTR(err); | |
539 | } | |
1da177e4 LT |
540 | return s; |
541 | } | |
542 | ||
6e4eab57 EB |
543 | EXPORT_SYMBOL(sget_userns); |
544 | ||
545 | /** | |
546 | * sget - find or create a superblock | |
547 | * @type: filesystem type superblock should belong to | |
548 | * @test: comparison callback | |
549 | * @set: setup callback | |
550 | * @flags: mount flags | |
551 | * @data: argument to each of them | |
552 | */ | |
553 | struct super_block *sget(struct file_system_type *type, | |
554 | int (*test)(struct super_block *,void *), | |
555 | int (*set)(struct super_block *,void *), | |
556 | int flags, | |
557 | void *data) | |
558 | { | |
559 | struct user_namespace *user_ns = current_user_ns(); | |
560 | ||
93faccbb EB |
561 | /* We don't yet pass the user namespace of the parent |
562 | * mount through to here so always use &init_user_ns | |
563 | * until that changes. | |
564 | */ | |
e462ec50 | 565 | if (flags & SB_SUBMOUNT) |
93faccbb EB |
566 | user_ns = &init_user_ns; |
567 | ||
6e4eab57 | 568 | /* Ensure the requestor has permissions over the target filesystem */ |
e462ec50 | 569 | if (!(flags & (SB_KERNMOUNT|SB_SUBMOUNT)) && !ns_capable(user_ns, CAP_SYS_ADMIN)) |
6e4eab57 EB |
570 | return ERR_PTR(-EPERM); |
571 | ||
572 | return sget_userns(type, test, set, flags, user_ns, data); | |
573 | } | |
574 | ||
1da177e4 LT |
575 | EXPORT_SYMBOL(sget); |
576 | ||
577 | void drop_super(struct super_block *sb) | |
578 | { | |
579 | up_read(&sb->s_umount); | |
580 | put_super(sb); | |
581 | } | |
582 | ||
583 | EXPORT_SYMBOL(drop_super); | |
584 | ||
ba6379f7 JK |
585 | void drop_super_exclusive(struct super_block *sb) |
586 | { | |
587 | up_write(&sb->s_umount); | |
588 | put_super(sb); | |
589 | } | |
590 | EXPORT_SYMBOL(drop_super_exclusive); | |
591 | ||
01a05b33 AV |
592 | /** |
593 | * iterate_supers - call function for all active superblocks | |
594 | * @f: function to call | |
595 | * @arg: argument to pass to it | |
596 | * | |
597 | * Scans the superblock list and calls given function, passing it | |
598 | * locked superblock and given argument. | |
599 | */ | |
600 | void iterate_supers(void (*f)(struct super_block *, void *), void *arg) | |
601 | { | |
dca33252 | 602 | struct super_block *sb, *p = NULL; |
01a05b33 AV |
603 | |
604 | spin_lock(&sb_lock); | |
dca33252 | 605 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 606 | if (hlist_unhashed(&sb->s_instances)) |
01a05b33 AV |
607 | continue; |
608 | sb->s_count++; | |
609 | spin_unlock(&sb_lock); | |
610 | ||
611 | down_read(&sb->s_umount); | |
e462ec50 | 612 | if (sb->s_root && (sb->s_flags & SB_BORN)) |
01a05b33 AV |
613 | f(sb, arg); |
614 | up_read(&sb->s_umount); | |
615 | ||
616 | spin_lock(&sb_lock); | |
dca33252 AV |
617 | if (p) |
618 | __put_super(p); | |
619 | p = sb; | |
01a05b33 | 620 | } |
dca33252 AV |
621 | if (p) |
622 | __put_super(p); | |
01a05b33 AV |
623 | spin_unlock(&sb_lock); |
624 | } | |
625 | ||
43e15cdb AV |
626 | /** |
627 | * iterate_supers_type - call function for superblocks of given type | |
628 | * @type: fs type | |
629 | * @f: function to call | |
630 | * @arg: argument to pass to it | |
631 | * | |
632 | * Scans the superblock list and calls given function, passing it | |
633 | * locked superblock and given argument. | |
634 | */ | |
635 | void iterate_supers_type(struct file_system_type *type, | |
636 | void (*f)(struct super_block *, void *), void *arg) | |
637 | { | |
638 | struct super_block *sb, *p = NULL; | |
639 | ||
640 | spin_lock(&sb_lock); | |
b67bfe0d | 641 | hlist_for_each_entry(sb, &type->fs_supers, s_instances) { |
43e15cdb AV |
642 | sb->s_count++; |
643 | spin_unlock(&sb_lock); | |
644 | ||
645 | down_read(&sb->s_umount); | |
e462ec50 | 646 | if (sb->s_root && (sb->s_flags & SB_BORN)) |
43e15cdb AV |
647 | f(sb, arg); |
648 | up_read(&sb->s_umount); | |
649 | ||
650 | spin_lock(&sb_lock); | |
651 | if (p) | |
652 | __put_super(p); | |
653 | p = sb; | |
654 | } | |
655 | if (p) | |
656 | __put_super(p); | |
657 | spin_unlock(&sb_lock); | |
658 | } | |
659 | ||
660 | EXPORT_SYMBOL(iterate_supers_type); | |
661 | ||
ba6379f7 | 662 | static struct super_block *__get_super(struct block_device *bdev, bool excl) |
1da177e4 | 663 | { |
618f0636 KK |
664 | struct super_block *sb; |
665 | ||
1da177e4 LT |
666 | if (!bdev) |
667 | return NULL; | |
618f0636 | 668 | |
1da177e4 | 669 | spin_lock(&sb_lock); |
618f0636 KK |
670 | rescan: |
671 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 672 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 673 | continue; |
618f0636 KK |
674 | if (sb->s_bdev == bdev) { |
675 | sb->s_count++; | |
1da177e4 | 676 | spin_unlock(&sb_lock); |
ba6379f7 JK |
677 | if (!excl) |
678 | down_read(&sb->s_umount); | |
679 | else | |
680 | down_write(&sb->s_umount); | |
df40c01a | 681 | /* still alive? */ |
e462ec50 | 682 | if (sb->s_root && (sb->s_flags & SB_BORN)) |
618f0636 | 683 | return sb; |
ba6379f7 JK |
684 | if (!excl) |
685 | up_read(&sb->s_umount); | |
686 | else | |
687 | up_write(&sb->s_umount); | |
df40c01a | 688 | /* nope, got unmounted */ |
618f0636 | 689 | spin_lock(&sb_lock); |
df40c01a AV |
690 | __put_super(sb); |
691 | goto rescan; | |
1da177e4 LT |
692 | } |
693 | } | |
694 | spin_unlock(&sb_lock); | |
695 | return NULL; | |
696 | } | |
697 | ||
6b6dc836 | 698 | /** |
ba6379f7 | 699 | * get_super - get the superblock of a device |
6b6dc836 JK |
700 | * @bdev: device to get the superblock for |
701 | * | |
702 | * Scans the superblock list and finds the superblock of the file system | |
ba6379f7 | 703 | * mounted on the device given. %NULL is returned if no match is found. |
6b6dc836 | 704 | */ |
ba6379f7 JK |
705 | struct super_block *get_super(struct block_device *bdev) |
706 | { | |
707 | return __get_super(bdev, false); | |
708 | } | |
709 | EXPORT_SYMBOL(get_super); | |
710 | ||
711 | static struct super_block *__get_super_thawed(struct block_device *bdev, | |
712 | bool excl) | |
6b6dc836 JK |
713 | { |
714 | while (1) { | |
ba6379f7 | 715 | struct super_block *s = __get_super(bdev, excl); |
5accdf82 | 716 | if (!s || s->s_writers.frozen == SB_UNFROZEN) |
6b6dc836 | 717 | return s; |
ba6379f7 JK |
718 | if (!excl) |
719 | up_read(&s->s_umount); | |
720 | else | |
721 | up_write(&s->s_umount); | |
5accdf82 JK |
722 | wait_event(s->s_writers.wait_unfrozen, |
723 | s->s_writers.frozen == SB_UNFROZEN); | |
6b6dc836 JK |
724 | put_super(s); |
725 | } | |
726 | } | |
ba6379f7 JK |
727 | |
728 | /** | |
729 | * get_super_thawed - get thawed superblock of a device | |
730 | * @bdev: device to get the superblock for | |
731 | * | |
732 | * Scans the superblock list and finds the superblock of the file system | |
733 | * mounted on the device. The superblock is returned once it is thawed | |
734 | * (or immediately if it was not frozen). %NULL is returned if no match | |
735 | * is found. | |
736 | */ | |
737 | struct super_block *get_super_thawed(struct block_device *bdev) | |
738 | { | |
739 | return __get_super_thawed(bdev, false); | |
740 | } | |
6b6dc836 JK |
741 | EXPORT_SYMBOL(get_super_thawed); |
742 | ||
ba6379f7 JK |
743 | /** |
744 | * get_super_exclusive_thawed - get thawed superblock of a device | |
745 | * @bdev: device to get the superblock for | |
746 | * | |
747 | * Scans the superblock list and finds the superblock of the file system | |
748 | * mounted on the device. The superblock is returned once it is thawed | |
749 | * (or immediately if it was not frozen) and s_umount semaphore is held | |
750 | * in exclusive mode. %NULL is returned if no match is found. | |
751 | */ | |
752 | struct super_block *get_super_exclusive_thawed(struct block_device *bdev) | |
753 | { | |
754 | return __get_super_thawed(bdev, true); | |
755 | } | |
756 | EXPORT_SYMBOL(get_super_exclusive_thawed); | |
757 | ||
4504230a CH |
758 | /** |
759 | * get_active_super - get an active reference to the superblock of a device | |
760 | * @bdev: device to get the superblock for | |
761 | * | |
762 | * Scans the superblock list and finds the superblock of the file system | |
763 | * mounted on the device given. Returns the superblock with an active | |
d3f21473 | 764 | * reference or %NULL if none was found. |
4504230a CH |
765 | */ |
766 | struct super_block *get_active_super(struct block_device *bdev) | |
767 | { | |
768 | struct super_block *sb; | |
769 | ||
770 | if (!bdev) | |
771 | return NULL; | |
772 | ||
1494583d | 773 | restart: |
4504230a CH |
774 | spin_lock(&sb_lock); |
775 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 776 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 777 | continue; |
1494583d | 778 | if (sb->s_bdev == bdev) { |
acfec9a5 | 779 | if (!grab_super(sb)) |
1494583d | 780 | goto restart; |
acfec9a5 AV |
781 | up_write(&sb->s_umount); |
782 | return sb; | |
1494583d | 783 | } |
4504230a CH |
784 | } |
785 | spin_unlock(&sb_lock); | |
786 | return NULL; | |
787 | } | |
dd111b31 | 788 | |
df40c01a | 789 | struct super_block *user_get_super(dev_t dev) |
1da177e4 | 790 | { |
618f0636 | 791 | struct super_block *sb; |
1da177e4 | 792 | |
1da177e4 | 793 | spin_lock(&sb_lock); |
618f0636 KK |
794 | rescan: |
795 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 796 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 797 | continue; |
618f0636 KK |
798 | if (sb->s_dev == dev) { |
799 | sb->s_count++; | |
1da177e4 | 800 | spin_unlock(&sb_lock); |
618f0636 | 801 | down_read(&sb->s_umount); |
df40c01a | 802 | /* still alive? */ |
e462ec50 | 803 | if (sb->s_root && (sb->s_flags & SB_BORN)) |
618f0636 KK |
804 | return sb; |
805 | up_read(&sb->s_umount); | |
df40c01a | 806 | /* nope, got unmounted */ |
618f0636 | 807 | spin_lock(&sb_lock); |
df40c01a AV |
808 | __put_super(sb); |
809 | goto rescan; | |
1da177e4 LT |
810 | } |
811 | } | |
812 | spin_unlock(&sb_lock); | |
813 | return NULL; | |
814 | } | |
815 | ||
1da177e4 LT |
816 | /** |
817 | * do_remount_sb - asks filesystem to change mount options. | |
818 | * @sb: superblock in question | |
e462ec50 | 819 | * @sb_flags: revised superblock flags |
1da177e4 LT |
820 | * @data: the rest of options |
821 | * @force: whether or not to force the change | |
822 | * | |
823 | * Alters the mount options of a mounted file system. | |
824 | */ | |
e462ec50 | 825 | int do_remount_sb(struct super_block *sb, int sb_flags, void *data, int force) |
1da177e4 LT |
826 | { |
827 | int retval; | |
c79d967d | 828 | int remount_ro; |
4504230a | 829 | |
5accdf82 | 830 | if (sb->s_writers.frozen != SB_UNFROZEN) |
4504230a CH |
831 | return -EBUSY; |
832 | ||
9361401e | 833 | #ifdef CONFIG_BLOCK |
e462ec50 | 834 | if (!(sb_flags & SB_RDONLY) && bdev_read_only(sb->s_bdev)) |
1da177e4 | 835 | return -EACCES; |
9361401e | 836 | #endif |
4504230a | 837 | |
e462ec50 | 838 | remount_ro = (sb_flags & SB_RDONLY) && !sb_rdonly(sb); |
d208bbdd | 839 | |
0aec09d0 | 840 | if (remount_ro) { |
fdab684d | 841 | if (!hlist_empty(&sb->s_pins)) { |
0aec09d0 | 842 | up_write(&sb->s_umount); |
fdab684d | 843 | group_pin_kill(&sb->s_pins); |
0aec09d0 AV |
844 | down_write(&sb->s_umount); |
845 | if (!sb->s_root) | |
846 | return 0; | |
847 | if (sb->s_writers.frozen != SB_UNFROZEN) | |
848 | return -EBUSY; | |
e462ec50 | 849 | remount_ro = (sb_flags & SB_RDONLY) && !sb_rdonly(sb); |
0aec09d0 AV |
850 | } |
851 | } | |
852 | shrink_dcache_sb(sb); | |
853 | ||
1da177e4 LT |
854 | /* If we are remounting RDONLY and current sb is read/write, |
855 | make sure there are no rw files opened */ | |
d208bbdd | 856 | if (remount_ro) { |
4ed5e82f | 857 | if (force) { |
eee5cc27 AV |
858 | sb->s_readonly_remount = 1; |
859 | smp_wmb(); | |
4ed5e82f MS |
860 | } else { |
861 | retval = sb_prepare_remount_readonly(sb); | |
862 | if (retval) | |
863 | return retval; | |
4ed5e82f | 864 | } |
1da177e4 LT |
865 | } |
866 | ||
867 | if (sb->s_op->remount_fs) { | |
e462ec50 | 868 | retval = sb->s_op->remount_fs(sb, &sb_flags, data); |
2833eb2b MS |
869 | if (retval) { |
870 | if (!force) | |
4ed5e82f | 871 | goto cancel_readonly; |
2833eb2b MS |
872 | /* If forced remount, go ahead despite any errors */ |
873 | WARN(1, "forced remount of a %s fs returned %i\n", | |
874 | sb->s_type->name, retval); | |
875 | } | |
1da177e4 | 876 | } |
e462ec50 | 877 | sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (sb_flags & MS_RMT_MASK); |
4ed5e82f MS |
878 | /* Needs to be ordered wrt mnt_is_readonly() */ |
879 | smp_wmb(); | |
880 | sb->s_readonly_remount = 0; | |
c79d967d | 881 | |
d208bbdd NP |
882 | /* |
883 | * Some filesystems modify their metadata via some other path than the | |
884 | * bdev buffer cache (eg. use a private mapping, or directories in | |
885 | * pagecache, etc). Also file data modifications go via their own | |
886 | * mappings. So If we try to mount readonly then copy the filesystem | |
887 | * from bdev, we could get stale data, so invalidate it to give a best | |
888 | * effort at coherency. | |
889 | */ | |
890 | if (remount_ro && sb->s_bdev) | |
891 | invalidate_bdev(sb->s_bdev); | |
1da177e4 | 892 | return 0; |
4ed5e82f MS |
893 | |
894 | cancel_readonly: | |
895 | sb->s_readonly_remount = 0; | |
896 | return retval; | |
1da177e4 LT |
897 | } |
898 | ||
a2a9537a | 899 | static void do_emergency_remount(struct work_struct *work) |
1da177e4 | 900 | { |
dca33252 | 901 | struct super_block *sb, *p = NULL; |
1da177e4 LT |
902 | |
903 | spin_lock(&sb_lock); | |
dca33252 | 904 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 905 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 906 | continue; |
1da177e4 LT |
907 | sb->s_count++; |
908 | spin_unlock(&sb_lock); | |
443b94ba | 909 | down_write(&sb->s_umount); |
e462ec50 | 910 | if (sb->s_root && sb->s_bdev && (sb->s_flags & SB_BORN) && |
bc98a42c | 911 | !sb_rdonly(sb)) { |
1da177e4 | 912 | /* |
1da177e4 LT |
913 | * What lock protects sb->s_flags?? |
914 | */ | |
e462ec50 | 915 | do_remount_sb(sb, SB_RDONLY, NULL, 1); |
1da177e4 | 916 | } |
443b94ba | 917 | up_write(&sb->s_umount); |
1da177e4 | 918 | spin_lock(&sb_lock); |
dca33252 AV |
919 | if (p) |
920 | __put_super(p); | |
921 | p = sb; | |
1da177e4 | 922 | } |
dca33252 AV |
923 | if (p) |
924 | __put_super(p); | |
1da177e4 | 925 | spin_unlock(&sb_lock); |
a2a9537a | 926 | kfree(work); |
1da177e4 LT |
927 | printk("Emergency Remount complete\n"); |
928 | } | |
929 | ||
930 | void emergency_remount(void) | |
931 | { | |
a2a9537a JA |
932 | struct work_struct *work; |
933 | ||
934 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | |
935 | if (work) { | |
936 | INIT_WORK(work, do_emergency_remount); | |
937 | schedule_work(work); | |
938 | } | |
1da177e4 LT |
939 | } |
940 | ||
941 | /* | |
942 | * Unnamed block devices are dummy devices used by virtual | |
943 | * filesystems which don't use real block-devices. -- jrs | |
944 | */ | |
945 | ||
ad76cbc6 | 946 | static DEFINE_IDA(unnamed_dev_ida); |
1da177e4 | 947 | static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */ |
a2a4dc49 TB |
948 | /* Many userspace utilities consider an FSID of 0 invalid. |
949 | * Always return at least 1 from get_anon_bdev. | |
950 | */ | |
951 | static int unnamed_dev_start = 1; | |
1da177e4 | 952 | |
0ee5dc67 | 953 | int get_anon_bdev(dev_t *p) |
1da177e4 LT |
954 | { |
955 | int dev; | |
956 | int error; | |
957 | ||
958 | retry: | |
ad76cbc6 | 959 | if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0) |
1da177e4 LT |
960 | return -ENOMEM; |
961 | spin_lock(&unnamed_dev_lock); | |
c63e09ec | 962 | error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev); |
f21f6220 AV |
963 | if (!error) |
964 | unnamed_dev_start = dev + 1; | |
1da177e4 LT |
965 | spin_unlock(&unnamed_dev_lock); |
966 | if (error == -EAGAIN) | |
967 | /* We raced and lost with another CPU. */ | |
968 | goto retry; | |
969 | else if (error) | |
970 | return -EAGAIN; | |
971 | ||
1af95de6 | 972 | if (dev >= (1 << MINORBITS)) { |
1da177e4 | 973 | spin_lock(&unnamed_dev_lock); |
ad76cbc6 | 974 | ida_remove(&unnamed_dev_ida, dev); |
f21f6220 AV |
975 | if (unnamed_dev_start > dev) |
976 | unnamed_dev_start = dev; | |
1da177e4 LT |
977 | spin_unlock(&unnamed_dev_lock); |
978 | return -EMFILE; | |
979 | } | |
0ee5dc67 | 980 | *p = MKDEV(0, dev & MINORMASK); |
1da177e4 LT |
981 | return 0; |
982 | } | |
0ee5dc67 | 983 | EXPORT_SYMBOL(get_anon_bdev); |
1da177e4 | 984 | |
0ee5dc67 | 985 | void free_anon_bdev(dev_t dev) |
1da177e4 | 986 | { |
0ee5dc67 | 987 | int slot = MINOR(dev); |
1da177e4 | 988 | spin_lock(&unnamed_dev_lock); |
ad76cbc6 | 989 | ida_remove(&unnamed_dev_ida, slot); |
c63e09ec AV |
990 | if (slot < unnamed_dev_start) |
991 | unnamed_dev_start = slot; | |
1da177e4 LT |
992 | spin_unlock(&unnamed_dev_lock); |
993 | } | |
0ee5dc67 AV |
994 | EXPORT_SYMBOL(free_anon_bdev); |
995 | ||
996 | int set_anon_super(struct super_block *s, void *data) | |
997 | { | |
df0ce26c | 998 | return get_anon_bdev(&s->s_dev); |
0ee5dc67 AV |
999 | } |
1000 | ||
1001 | EXPORT_SYMBOL(set_anon_super); | |
1002 | ||
1003 | void kill_anon_super(struct super_block *sb) | |
1004 | { | |
1005 | dev_t dev = sb->s_dev; | |
1006 | generic_shutdown_super(sb); | |
1007 | free_anon_bdev(dev); | |
1008 | } | |
1da177e4 LT |
1009 | |
1010 | EXPORT_SYMBOL(kill_anon_super); | |
1011 | ||
1da177e4 LT |
1012 | void kill_litter_super(struct super_block *sb) |
1013 | { | |
1014 | if (sb->s_root) | |
1015 | d_genocide(sb->s_root); | |
1016 | kill_anon_super(sb); | |
1017 | } | |
1018 | ||
1019 | EXPORT_SYMBOL(kill_litter_super); | |
1020 | ||
909e6d94 SH |
1021 | static int ns_test_super(struct super_block *sb, void *data) |
1022 | { | |
1023 | return sb->s_fs_info == data; | |
1024 | } | |
1025 | ||
1026 | static int ns_set_super(struct super_block *sb, void *data) | |
1027 | { | |
1028 | sb->s_fs_info = data; | |
1029 | return set_anon_super(sb, NULL); | |
1030 | } | |
1031 | ||
d91ee87d EB |
1032 | struct dentry *mount_ns(struct file_system_type *fs_type, |
1033 | int flags, void *data, void *ns, struct user_namespace *user_ns, | |
1034 | int (*fill_super)(struct super_block *, void *, int)) | |
909e6d94 SH |
1035 | { |
1036 | struct super_block *sb; | |
1037 | ||
d91ee87d EB |
1038 | /* Don't allow mounting unless the caller has CAP_SYS_ADMIN |
1039 | * over the namespace. | |
1040 | */ | |
e462ec50 | 1041 | if (!(flags & SB_KERNMOUNT) && !ns_capable(user_ns, CAP_SYS_ADMIN)) |
d91ee87d EB |
1042 | return ERR_PTR(-EPERM); |
1043 | ||
6e4eab57 EB |
1044 | sb = sget_userns(fs_type, ns_test_super, ns_set_super, flags, |
1045 | user_ns, ns); | |
909e6d94 | 1046 | if (IS_ERR(sb)) |
ceefda69 | 1047 | return ERR_CAST(sb); |
909e6d94 SH |
1048 | |
1049 | if (!sb->s_root) { | |
1050 | int err; | |
e462ec50 | 1051 | err = fill_super(sb, data, flags & SB_SILENT ? 1 : 0); |
909e6d94 | 1052 | if (err) { |
74dbbdd7 | 1053 | deactivate_locked_super(sb); |
ceefda69 | 1054 | return ERR_PTR(err); |
909e6d94 SH |
1055 | } |
1056 | ||
e462ec50 | 1057 | sb->s_flags |= SB_ACTIVE; |
909e6d94 SH |
1058 | } |
1059 | ||
ceefda69 | 1060 | return dget(sb->s_root); |
909e6d94 SH |
1061 | } |
1062 | ||
ceefda69 | 1063 | EXPORT_SYMBOL(mount_ns); |
909e6d94 | 1064 | |
9361401e | 1065 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
1066 | static int set_bdev_super(struct super_block *s, void *data) |
1067 | { | |
1068 | s->s_bdev = data; | |
1069 | s->s_dev = s->s_bdev->bd_dev; | |
13eec236 | 1070 | s->s_bdi = bdi_get(s->s_bdev->bd_bdi); |
32a88aa1 | 1071 | |
1da177e4 LT |
1072 | return 0; |
1073 | } | |
1074 | ||
1075 | static int test_bdev_super(struct super_block *s, void *data) | |
1076 | { | |
1077 | return (void *)s->s_bdev == data; | |
1078 | } | |
1079 | ||
152a0836 | 1080 | struct dentry *mount_bdev(struct file_system_type *fs_type, |
1da177e4 | 1081 | int flags, const char *dev_name, void *data, |
152a0836 | 1082 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1083 | { |
1084 | struct block_device *bdev; | |
1085 | struct super_block *s; | |
d4d77629 | 1086 | fmode_t mode = FMODE_READ | FMODE_EXCL; |
1da177e4 LT |
1087 | int error = 0; |
1088 | ||
e462ec50 | 1089 | if (!(flags & SB_RDONLY)) |
30c40d2c AV |
1090 | mode |= FMODE_WRITE; |
1091 | ||
d4d77629 | 1092 | bdev = blkdev_get_by_path(dev_name, mode, fs_type); |
1da177e4 | 1093 | if (IS_ERR(bdev)) |
152a0836 | 1094 | return ERR_CAST(bdev); |
1da177e4 LT |
1095 | |
1096 | /* | |
1097 | * once the super is inserted into the list by sget, s_umount | |
1098 | * will protect the lockfs code from trying to start a snapshot | |
1099 | * while we are mounting | |
1100 | */ | |
4fadd7bb CH |
1101 | mutex_lock(&bdev->bd_fsfreeze_mutex); |
1102 | if (bdev->bd_fsfreeze_count > 0) { | |
1103 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
1104 | error = -EBUSY; | |
1105 | goto error_bdev; | |
1106 | } | |
e462ec50 | 1107 | s = sget(fs_type, test_bdev_super, set_bdev_super, flags | SB_NOSEC, |
9249e17f | 1108 | bdev); |
4fadd7bb | 1109 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
1da177e4 | 1110 | if (IS_ERR(s)) |
454e2398 | 1111 | goto error_s; |
1da177e4 LT |
1112 | |
1113 | if (s->s_root) { | |
e462ec50 | 1114 | if ((flags ^ s->s_flags) & SB_RDONLY) { |
74dbbdd7 | 1115 | deactivate_locked_super(s); |
454e2398 DH |
1116 | error = -EBUSY; |
1117 | goto error_bdev; | |
1da177e4 | 1118 | } |
454e2398 | 1119 | |
4f331f01 TH |
1120 | /* |
1121 | * s_umount nests inside bd_mutex during | |
e525fd89 TH |
1122 | * __invalidate_device(). blkdev_put() acquires |
1123 | * bd_mutex and can't be called under s_umount. Drop | |
1124 | * s_umount temporarily. This is safe as we're | |
1125 | * holding an active reference. | |
4f331f01 TH |
1126 | */ |
1127 | up_write(&s->s_umount); | |
d4d77629 | 1128 | blkdev_put(bdev, mode); |
4f331f01 | 1129 | down_write(&s->s_umount); |
1da177e4 | 1130 | } else { |
30c40d2c | 1131 | s->s_mode = mode; |
a1c6f057 | 1132 | snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev); |
e78c9a00 | 1133 | sb_set_blocksize(s, block_size(bdev)); |
e462ec50 | 1134 | error = fill_super(s, data, flags & SB_SILENT ? 1 : 0); |
1da177e4 | 1135 | if (error) { |
74dbbdd7 | 1136 | deactivate_locked_super(s); |
454e2398 | 1137 | goto error; |
fa675765 | 1138 | } |
454e2398 | 1139 | |
e462ec50 | 1140 | s->s_flags |= SB_ACTIVE; |
87d8fe1e | 1141 | bdev->bd_super = s; |
1da177e4 LT |
1142 | } |
1143 | ||
152a0836 | 1144 | return dget(s->s_root); |
1da177e4 | 1145 | |
454e2398 DH |
1146 | error_s: |
1147 | error = PTR_ERR(s); | |
1148 | error_bdev: | |
d4d77629 | 1149 | blkdev_put(bdev, mode); |
454e2398 | 1150 | error: |
152a0836 AV |
1151 | return ERR_PTR(error); |
1152 | } | |
1153 | EXPORT_SYMBOL(mount_bdev); | |
1154 | ||
1da177e4 LT |
1155 | void kill_block_super(struct super_block *sb) |
1156 | { | |
1157 | struct block_device *bdev = sb->s_bdev; | |
30c40d2c | 1158 | fmode_t mode = sb->s_mode; |
1da177e4 | 1159 | |
ddbaaf30 | 1160 | bdev->bd_super = NULL; |
1da177e4 LT |
1161 | generic_shutdown_super(sb); |
1162 | sync_blockdev(bdev); | |
d4d77629 | 1163 | WARN_ON_ONCE(!(mode & FMODE_EXCL)); |
e525fd89 | 1164 | blkdev_put(bdev, mode | FMODE_EXCL); |
1da177e4 LT |
1165 | } |
1166 | ||
1167 | EXPORT_SYMBOL(kill_block_super); | |
9361401e | 1168 | #endif |
1da177e4 | 1169 | |
3c26ff6e | 1170 | struct dentry *mount_nodev(struct file_system_type *fs_type, |
1da177e4 | 1171 | int flags, void *data, |
3c26ff6e | 1172 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1173 | { |
1174 | int error; | |
9249e17f | 1175 | struct super_block *s = sget(fs_type, NULL, set_anon_super, flags, NULL); |
1da177e4 LT |
1176 | |
1177 | if (IS_ERR(s)) | |
3c26ff6e | 1178 | return ERR_CAST(s); |
1da177e4 | 1179 | |
e462ec50 | 1180 | error = fill_super(s, data, flags & SB_SILENT ? 1 : 0); |
1da177e4 | 1181 | if (error) { |
74dbbdd7 | 1182 | deactivate_locked_super(s); |
3c26ff6e | 1183 | return ERR_PTR(error); |
1da177e4 | 1184 | } |
e462ec50 | 1185 | s->s_flags |= SB_ACTIVE; |
3c26ff6e | 1186 | return dget(s->s_root); |
1da177e4 | 1187 | } |
3c26ff6e AV |
1188 | EXPORT_SYMBOL(mount_nodev); |
1189 | ||
1da177e4 LT |
1190 | static int compare_single(struct super_block *s, void *p) |
1191 | { | |
1192 | return 1; | |
1193 | } | |
1194 | ||
fc14f2fe | 1195 | struct dentry *mount_single(struct file_system_type *fs_type, |
1da177e4 | 1196 | int flags, void *data, |
fc14f2fe | 1197 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1198 | { |
1199 | struct super_block *s; | |
1200 | int error; | |
1201 | ||
9249e17f | 1202 | s = sget(fs_type, compare_single, set_anon_super, flags, NULL); |
1da177e4 | 1203 | if (IS_ERR(s)) |
fc14f2fe | 1204 | return ERR_CAST(s); |
1da177e4 | 1205 | if (!s->s_root) { |
e462ec50 | 1206 | error = fill_super(s, data, flags & SB_SILENT ? 1 : 0); |
1da177e4 | 1207 | if (error) { |
74dbbdd7 | 1208 | deactivate_locked_super(s); |
fc14f2fe | 1209 | return ERR_PTR(error); |
1da177e4 | 1210 | } |
e462ec50 | 1211 | s->s_flags |= SB_ACTIVE; |
9329d1be KS |
1212 | } else { |
1213 | do_remount_sb(s, flags, data, 0); | |
1da177e4 | 1214 | } |
fc14f2fe AV |
1215 | return dget(s->s_root); |
1216 | } | |
1217 | EXPORT_SYMBOL(mount_single); | |
1218 | ||
9d412a43 AV |
1219 | struct dentry * |
1220 | mount_fs(struct file_system_type *type, int flags, const char *name, void *data) | |
1da177e4 | 1221 | { |
c96e41e9 | 1222 | struct dentry *root; |
9d412a43 | 1223 | struct super_block *sb; |
1da177e4 | 1224 | char *secdata = NULL; |
9d412a43 | 1225 | int error = -ENOMEM; |
8089352a | 1226 | |
e0007529 | 1227 | if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) { |
1da177e4 | 1228 | secdata = alloc_secdata(); |
454e2398 | 1229 | if (!secdata) |
9d412a43 | 1230 | goto out; |
1da177e4 | 1231 | |
e0007529 | 1232 | error = security_sb_copy_data(data, secdata); |
454e2398 | 1233 | if (error) |
1da177e4 | 1234 | goto out_free_secdata; |
1da177e4 LT |
1235 | } |
1236 | ||
1a102ff9 AV |
1237 | root = type->mount(type, flags, name, data); |
1238 | if (IS_ERR(root)) { | |
1239 | error = PTR_ERR(root); | |
1240 | goto out_free_secdata; | |
c96e41e9 | 1241 | } |
9d412a43 AV |
1242 | sb = root->d_sb; |
1243 | BUG_ON(!sb); | |
1244 | WARN_ON(!sb->s_bdi); | |
b9659ff3 DC |
1245 | |
1246 | /* | |
1247 | * Write barrier is for super_cache_count(). We place it before setting | |
1248 | * SB_BORN as the data dependency between the two functions is the | |
1249 | * superblock structure contents that we just set up, not the SB_BORN | |
1250 | * flag. | |
1251 | */ | |
1252 | smp_wmb(); | |
e462ec50 | 1253 | sb->s_flags |= SB_BORN; |
454e2398 | 1254 | |
9d412a43 | 1255 | error = security_sb_kern_mount(sb, flags, secdata); |
5129a469 JE |
1256 | if (error) |
1257 | goto out_sb; | |
454e2398 | 1258 | |
42cb56ae JL |
1259 | /* |
1260 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE | |
1261 | * but s_maxbytes was an unsigned long long for many releases. Throw | |
1262 | * this warning for a little while to try and catch filesystems that | |
4358b567 | 1263 | * violate this rule. |
42cb56ae | 1264 | */ |
9d412a43 AV |
1265 | WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " |
1266 | "negative value (%lld)\n", type->name, sb->s_maxbytes); | |
42cb56ae | 1267 | |
9d412a43 | 1268 | up_write(&sb->s_umount); |
8680e22f | 1269 | free_secdata(secdata); |
9d412a43 | 1270 | return root; |
1da177e4 | 1271 | out_sb: |
9d412a43 AV |
1272 | dput(root); |
1273 | deactivate_locked_super(sb); | |
1da177e4 LT |
1274 | out_free_secdata: |
1275 | free_secdata(secdata); | |
1da177e4 | 1276 | out: |
454e2398 | 1277 | return ERR_PTR(error); |
1da177e4 LT |
1278 | } |
1279 | ||
fca39346 JK |
1280 | /* |
1281 | * Setup private BDI for given superblock. It gets automatically cleaned up | |
1282 | * in generic_shutdown_super(). | |
1283 | */ | |
1284 | int super_setup_bdi_name(struct super_block *sb, char *fmt, ...) | |
1285 | { | |
1286 | struct backing_dev_info *bdi; | |
1287 | int err; | |
1288 | va_list args; | |
1289 | ||
1290 | bdi = bdi_alloc(GFP_KERNEL); | |
1291 | if (!bdi) | |
1292 | return -ENOMEM; | |
1293 | ||
1294 | bdi->name = sb->s_type->name; | |
1295 | ||
1296 | va_start(args, fmt); | |
7c4cc300 | 1297 | err = bdi_register_va(bdi, fmt, args); |
fca39346 JK |
1298 | va_end(args); |
1299 | if (err) { | |
1300 | bdi_put(bdi); | |
1301 | return err; | |
1302 | } | |
1303 | WARN_ON(sb->s_bdi != &noop_backing_dev_info); | |
1304 | sb->s_bdi = bdi; | |
fca39346 JK |
1305 | |
1306 | return 0; | |
1307 | } | |
1308 | EXPORT_SYMBOL(super_setup_bdi_name); | |
1309 | ||
1310 | /* | |
1311 | * Setup private BDI for given superblock. I gets automatically cleaned up | |
1312 | * in generic_shutdown_super(). | |
1313 | */ | |
1314 | int super_setup_bdi(struct super_block *sb) | |
1315 | { | |
1316 | static atomic_long_t bdi_seq = ATOMIC_LONG_INIT(0); | |
1317 | ||
1318 | return super_setup_bdi_name(sb, "%.28s-%ld", sb->s_type->name, | |
1319 | atomic_long_inc_return(&bdi_seq)); | |
1320 | } | |
1321 | EXPORT_SYMBOL(super_setup_bdi); | |
1322 | ||
5accdf82 JK |
1323 | /* |
1324 | * This is an internal function, please use sb_end_{write,pagefault,intwrite} | |
1325 | * instead. | |
1326 | */ | |
1327 | void __sb_end_write(struct super_block *sb, int level) | |
1328 | { | |
8129ed29 | 1329 | percpu_up_read(sb->s_writers.rw_sem + level-1); |
5accdf82 JK |
1330 | } |
1331 | EXPORT_SYMBOL(__sb_end_write); | |
1332 | ||
f4b554af ON |
1333 | /* |
1334 | * This is an internal function, please use sb_start_{write,pagefault,intwrite} | |
1335 | * instead. | |
1336 | */ | |
1337 | int __sb_start_write(struct super_block *sb, int level, bool wait) | |
1338 | { | |
1339 | bool force_trylock = false; | |
8129ed29 | 1340 | int ret = 1; |
f4b554af ON |
1341 | |
1342 | #ifdef CONFIG_LOCKDEP | |
1343 | /* | |
1344 | * We want lockdep to tell us about possible deadlocks with freezing | |
1345 | * but it's it bit tricky to properly instrument it. Getting a freeze | |
1346 | * protection works as getting a read lock but there are subtle | |
1347 | * problems. XFS for example gets freeze protection on internal level | |
1348 | * twice in some cases, which is OK only because we already hold a | |
1349 | * freeze protection also on higher level. Due to these cases we have | |
1350 | * to use wait == F (trylock mode) which must not fail. | |
1351 | */ | |
1352 | if (wait) { | |
1353 | int i; | |
1354 | ||
1355 | for (i = 0; i < level - 1; i++) | |
8129ed29 | 1356 | if (percpu_rwsem_is_held(sb->s_writers.rw_sem + i)) { |
f4b554af ON |
1357 | force_trylock = true; |
1358 | break; | |
1359 | } | |
1360 | } | |
1361 | #endif | |
8129ed29 ON |
1362 | if (wait && !force_trylock) |
1363 | percpu_down_read(sb->s_writers.rw_sem + level-1); | |
1364 | else | |
1365 | ret = percpu_down_read_trylock(sb->s_writers.rw_sem + level-1); | |
1366 | ||
22224a17 | 1367 | WARN_ON(force_trylock && !ret); |
f4b554af ON |
1368 | return ret; |
1369 | } | |
5accdf82 JK |
1370 | EXPORT_SYMBOL(__sb_start_write); |
1371 | ||
1372 | /** | |
1373 | * sb_wait_write - wait until all writers to given file system finish | |
1374 | * @sb: the super for which we wait | |
1375 | * @level: type of writers we wait for (normal vs page fault) | |
1376 | * | |
1377 | * This function waits until there are no writers of given type to given file | |
8129ed29 | 1378 | * system. |
5accdf82 JK |
1379 | */ |
1380 | static void sb_wait_write(struct super_block *sb, int level) | |
1381 | { | |
8129ed29 | 1382 | percpu_down_write(sb->s_writers.rw_sem + level-1); |
8129ed29 | 1383 | } |
5accdf82 | 1384 | |
f1a96220 ON |
1385 | /* |
1386 | * We are going to return to userspace and forget about these locks, the | |
1387 | * ownership goes to the caller of thaw_super() which does unlock(). | |
1388 | */ | |
1389 | static void lockdep_sb_freeze_release(struct super_block *sb) | |
1390 | { | |
1391 | int level; | |
1392 | ||
1393 | for (level = SB_FREEZE_LEVELS - 1; level >= 0; level--) | |
1394 | percpu_rwsem_release(sb->s_writers.rw_sem + level, 0, _THIS_IP_); | |
1395 | } | |
1396 | ||
1397 | /* | |
1398 | * Tell lockdep we are holding these locks before we call ->unfreeze_fs(sb). | |
1399 | */ | |
1400 | static void lockdep_sb_freeze_acquire(struct super_block *sb) | |
8129ed29 ON |
1401 | { |
1402 | int level; | |
5accdf82 | 1403 | |
8129ed29 ON |
1404 | for (level = 0; level < SB_FREEZE_LEVELS; ++level) |
1405 | percpu_rwsem_acquire(sb->s_writers.rw_sem + level, 0, _THIS_IP_); | |
f1a96220 ON |
1406 | } |
1407 | ||
1408 | static void sb_freeze_unlock(struct super_block *sb) | |
1409 | { | |
1410 | int level; | |
5accdf82 | 1411 | |
8129ed29 ON |
1412 | for (level = SB_FREEZE_LEVELS - 1; level >= 0; level--) |
1413 | percpu_up_write(sb->s_writers.rw_sem + level); | |
5accdf82 JK |
1414 | } |
1415 | ||
18e9e510 | 1416 | /** |
7000d3c4 RD |
1417 | * freeze_super - lock the filesystem and force it into a consistent state |
1418 | * @sb: the super to lock | |
18e9e510 JB |
1419 | * |
1420 | * Syncs the super to make sure the filesystem is consistent and calls the fs's | |
1421 | * freeze_fs. Subsequent calls to this without first thawing the fs will return | |
1422 | * -EBUSY. | |
5accdf82 JK |
1423 | * |
1424 | * During this function, sb->s_writers.frozen goes through these values: | |
1425 | * | |
1426 | * SB_UNFROZEN: File system is normal, all writes progress as usual. | |
1427 | * | |
1428 | * SB_FREEZE_WRITE: The file system is in the process of being frozen. New | |
1429 | * writes should be blocked, though page faults are still allowed. We wait for | |
1430 | * all writes to complete and then proceed to the next stage. | |
1431 | * | |
1432 | * SB_FREEZE_PAGEFAULT: Freezing continues. Now also page faults are blocked | |
1433 | * but internal fs threads can still modify the filesystem (although they | |
1434 | * should not dirty new pages or inodes), writeback can run etc. After waiting | |
1435 | * for all running page faults we sync the filesystem which will clean all | |
1436 | * dirty pages and inodes (no new dirty pages or inodes can be created when | |
1437 | * sync is running). | |
1438 | * | |
1439 | * SB_FREEZE_FS: The file system is frozen. Now all internal sources of fs | |
1440 | * modification are blocked (e.g. XFS preallocation truncation on inode | |
1441 | * reclaim). This is usually implemented by blocking new transactions for | |
1442 | * filesystems that have them and need this additional guard. After all | |
1443 | * internal writers are finished we call ->freeze_fs() to finish filesystem | |
1444 | * freezing. Then we transition to SB_FREEZE_COMPLETE state. This state is | |
1445 | * mostly auxiliary for filesystems to verify they do not modify frozen fs. | |
1446 | * | |
1447 | * sb->s_writers.frozen is protected by sb->s_umount. | |
18e9e510 JB |
1448 | */ |
1449 | int freeze_super(struct super_block *sb) | |
1450 | { | |
1451 | int ret; | |
1452 | ||
1453 | atomic_inc(&sb->s_active); | |
1454 | down_write(&sb->s_umount); | |
5accdf82 | 1455 | if (sb->s_writers.frozen != SB_UNFROZEN) { |
18e9e510 JB |
1456 | deactivate_locked_super(sb); |
1457 | return -EBUSY; | |
1458 | } | |
1459 | ||
e462ec50 | 1460 | if (!(sb->s_flags & SB_BORN)) { |
dabe0dc1 AV |
1461 | up_write(&sb->s_umount); |
1462 | return 0; /* sic - it's "nothing to do" */ | |
1463 | } | |
1464 | ||
bc98a42c | 1465 | if (sb_rdonly(sb)) { |
5accdf82 JK |
1466 | /* Nothing to do really... */ |
1467 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
18e9e510 JB |
1468 | up_write(&sb->s_umount); |
1469 | return 0; | |
1470 | } | |
1471 | ||
5accdf82 | 1472 | sb->s_writers.frozen = SB_FREEZE_WRITE; |
5accdf82 JK |
1473 | /* Release s_umount to preserve sb_start_write -> s_umount ordering */ |
1474 | up_write(&sb->s_umount); | |
5accdf82 | 1475 | sb_wait_write(sb, SB_FREEZE_WRITE); |
8129ed29 | 1476 | down_write(&sb->s_umount); |
5accdf82 JK |
1477 | |
1478 | /* Now we go and block page faults... */ | |
5accdf82 | 1479 | sb->s_writers.frozen = SB_FREEZE_PAGEFAULT; |
5accdf82 JK |
1480 | sb_wait_write(sb, SB_FREEZE_PAGEFAULT); |
1481 | ||
1482 | /* All writers are done so after syncing there won't be dirty data */ | |
18e9e510 JB |
1483 | sync_filesystem(sb); |
1484 | ||
5accdf82 JK |
1485 | /* Now wait for internal filesystem counter */ |
1486 | sb->s_writers.frozen = SB_FREEZE_FS; | |
5accdf82 | 1487 | sb_wait_write(sb, SB_FREEZE_FS); |
18e9e510 | 1488 | |
18e9e510 JB |
1489 | if (sb->s_op->freeze_fs) { |
1490 | ret = sb->s_op->freeze_fs(sb); | |
1491 | if (ret) { | |
1492 | printk(KERN_ERR | |
1493 | "VFS:Filesystem freeze failed\n"); | |
5accdf82 | 1494 | sb->s_writers.frozen = SB_UNFROZEN; |
8129ed29 | 1495 | sb_freeze_unlock(sb); |
5accdf82 | 1496 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 JB |
1497 | deactivate_locked_super(sb); |
1498 | return ret; | |
1499 | } | |
1500 | } | |
5accdf82 | 1501 | /* |
89f39af1 ON |
1502 | * For debugging purposes so that fs can warn if it sees write activity |
1503 | * when frozen is set to SB_FREEZE_COMPLETE, and for thaw_super(). | |
5accdf82 JK |
1504 | */ |
1505 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
f1a96220 | 1506 | lockdep_sb_freeze_release(sb); |
18e9e510 JB |
1507 | up_write(&sb->s_umount); |
1508 | return 0; | |
1509 | } | |
1510 | EXPORT_SYMBOL(freeze_super); | |
1511 | ||
1512 | /** | |
1513 | * thaw_super -- unlock filesystem | |
1514 | * @sb: the super to thaw | |
1515 | * | |
1516 | * Unlocks the filesystem and marks it writeable again after freeze_super(). | |
1517 | */ | |
1518 | int thaw_super(struct super_block *sb) | |
1519 | { | |
1520 | int error; | |
1521 | ||
1522 | down_write(&sb->s_umount); | |
89f39af1 | 1523 | if (sb->s_writers.frozen != SB_FREEZE_COMPLETE) { |
18e9e510 JB |
1524 | up_write(&sb->s_umount); |
1525 | return -EINVAL; | |
1526 | } | |
1527 | ||
bc98a42c | 1528 | if (sb_rdonly(sb)) { |
8129ed29 | 1529 | sb->s_writers.frozen = SB_UNFROZEN; |
18e9e510 | 1530 | goto out; |
8129ed29 | 1531 | } |
18e9e510 | 1532 | |
f1a96220 ON |
1533 | lockdep_sb_freeze_acquire(sb); |
1534 | ||
18e9e510 JB |
1535 | if (sb->s_op->unfreeze_fs) { |
1536 | error = sb->s_op->unfreeze_fs(sb); | |
1537 | if (error) { | |
1538 | printk(KERN_ERR | |
1539 | "VFS:Filesystem thaw failed\n"); | |
f1a96220 | 1540 | lockdep_sb_freeze_release(sb); |
18e9e510 JB |
1541 | up_write(&sb->s_umount); |
1542 | return error; | |
1543 | } | |
1544 | } | |
1545 | ||
5accdf82 | 1546 | sb->s_writers.frozen = SB_UNFROZEN; |
8129ed29 ON |
1547 | sb_freeze_unlock(sb); |
1548 | out: | |
5accdf82 | 1549 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 | 1550 | deactivate_locked_super(sb); |
18e9e510 JB |
1551 | return 0; |
1552 | } | |
1553 | EXPORT_SYMBOL(thaw_super); |