]>
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 | /* |
eb6ef3df | 123 | * Don't call trylock_super as it is a potential |
d23da150 TC |
124 | * scalability bottleneck. The counts could get updated |
125 | * between super_cache_count and super_cache_scan anyway. | |
126 | * Call to super_cache_count with shrinker_rwsem held | |
503c358c | 127 | * ensures the safety of call to list_lru_shrink_count() and |
d23da150 TC |
128 | * s_op->nr_cached_objects(). |
129 | */ | |
0a234c6d | 130 | if (sb->s_op && sb->s_op->nr_cached_objects) |
4101b624 | 131 | total_objects = sb->s_op->nr_cached_objects(sb, sc); |
0a234c6d | 132 | |
503c358c VD |
133 | total_objects += list_lru_shrink_count(&sb->s_dentry_lru, sc); |
134 | total_objects += list_lru_shrink_count(&sb->s_inode_lru, sc); | |
0a234c6d | 135 | |
55f841ce | 136 | total_objects = vfs_pressure_ratio(total_objects); |
0e1fdafd | 137 | return total_objects; |
b0d40c92 DC |
138 | } |
139 | ||
853b39a7 ON |
140 | static void destroy_super_work(struct work_struct *work) |
141 | { | |
142 | struct super_block *s = container_of(work, struct super_block, | |
143 | destroy_work); | |
144 | int i; | |
145 | ||
146 | for (i = 0; i < SB_FREEZE_LEVELS; i++) | |
8129ed29 | 147 | percpu_free_rwsem(&s->s_writers.rw_sem[i]); |
853b39a7 ON |
148 | kfree(s); |
149 | } | |
150 | ||
151 | static void destroy_super_rcu(struct rcu_head *head) | |
152 | { | |
153 | struct super_block *s = container_of(head, struct super_block, rcu); | |
154 | INIT_WORK(&s->destroy_work, destroy_super_work); | |
155 | schedule_work(&s->destroy_work); | |
156 | } | |
157 | ||
0200894d AV |
158 | /* Free a superblock that has never been seen by anyone */ |
159 | static void destroy_unused_super(struct super_block *s) | |
5accdf82 | 160 | { |
0200894d AV |
161 | if (!s) |
162 | return; | |
163 | up_write(&s->s_umount); | |
7eb5e882 AV |
164 | list_lru_destroy(&s->s_dentry_lru); |
165 | list_lru_destroy(&s->s_inode_lru); | |
7eb5e882 | 166 | security_sb_free(s); |
6e4eab57 | 167 | put_user_ns(s->s_user_ns); |
7eb5e882 | 168 | kfree(s->s_subtype); |
7632554f | 169 | free_prealloced_shrinker(&s->s_shrink); |
0200894d AV |
170 | /* no delays needed */ |
171 | destroy_super_work(&s->destroy_work); | |
5accdf82 JK |
172 | } |
173 | ||
1da177e4 LT |
174 | /** |
175 | * alloc_super - create new superblock | |
fe2bbc48 | 176 | * @type: filesystem type superblock should belong to |
9249e17f | 177 | * @flags: the mount flags |
6e4eab57 | 178 | * @user_ns: User namespace for the super_block |
1da177e4 LT |
179 | * |
180 | * Allocates and initializes a new &struct super_block. alloc_super() | |
181 | * returns a pointer new superblock or %NULL if allocation had failed. | |
182 | */ | |
6e4eab57 EB |
183 | static struct super_block *alloc_super(struct file_system_type *type, int flags, |
184 | struct user_namespace *user_ns) | |
1da177e4 | 185 | { |
11b0b5ab | 186 | struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER); |
b87221de | 187 | static const struct super_operations default_op; |
7eb5e882 AV |
188 | int i; |
189 | ||
190 | if (!s) | |
191 | return NULL; | |
1da177e4 | 192 | |
b5bd856a | 193 | INIT_LIST_HEAD(&s->s_mounts); |
6e4eab57 | 194 | s->s_user_ns = get_user_ns(user_ns); |
ca0168e8 AV |
195 | init_rwsem(&s->s_umount); |
196 | lockdep_set_class(&s->s_umount, &type->s_umount_key); | |
197 | /* | |
198 | * sget() can have s_umount recursion. | |
199 | * | |
200 | * When it cannot find a suitable sb, it allocates a new | |
201 | * one (this one), and tries again to find a suitable old | |
202 | * one. | |
203 | * | |
204 | * In case that succeeds, it will acquire the s_umount | |
205 | * lock of the old one. Since these are clearly distrinct | |
206 | * locks, and this object isn't exposed yet, there's no | |
207 | * risk of deadlocks. | |
208 | * | |
209 | * Annotate this by putting this lock in a different | |
210 | * subclass. | |
211 | */ | |
212 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); | |
b5bd856a | 213 | |
7eb5e882 AV |
214 | if (security_sb_alloc(s)) |
215 | goto fail; | |
7b7a8665 | 216 | |
7eb5e882 | 217 | for (i = 0; i < SB_FREEZE_LEVELS; i++) { |
8129ed29 ON |
218 | if (__percpu_init_rwsem(&s->s_writers.rw_sem[i], |
219 | sb_writers_name[i], | |
220 | &type->s_writers_key[i])) | |
7eb5e882 | 221 | goto fail; |
1da177e4 | 222 | } |
7eb5e882 | 223 | init_waitqueue_head(&s->s_writers.wait_unfrozen); |
df0ce26c | 224 | s->s_bdi = &noop_backing_dev_info; |
7eb5e882 | 225 | s->s_flags = flags; |
cc50a07a | 226 | if (s->s_user_ns != &init_user_ns) |
67690f93 | 227 | s->s_iflags |= SB_I_NODEV; |
7eb5e882 AV |
228 | INIT_HLIST_NODE(&s->s_instances); |
229 | INIT_HLIST_BL_HEAD(&s->s_anon); | |
e97fedb9 | 230 | mutex_init(&s->s_sync_lock); |
7eb5e882 | 231 | INIT_LIST_HEAD(&s->s_inodes); |
74278da9 | 232 | spin_lock_init(&s->s_inode_list_lock); |
6c60d2b5 DC |
233 | INIT_LIST_HEAD(&s->s_inodes_wb); |
234 | spin_lock_init(&s->s_inode_wblist_lock); | |
7eb5e882 | 235 | |
2acb60a0 | 236 | if (list_lru_init_memcg(&s->s_dentry_lru)) |
7eb5e882 | 237 | goto fail; |
2acb60a0 | 238 | if (list_lru_init_memcg(&s->s_inode_lru)) |
7eb5e882 | 239 | goto fail; |
7eb5e882 AV |
240 | s->s_count = 1; |
241 | atomic_set(&s->s_active, 1); | |
242 | mutex_init(&s->s_vfs_rename_mutex); | |
243 | lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key); | |
bc8230ee | 244 | init_rwsem(&s->s_dquot.dqio_sem); |
7eb5e882 AV |
245 | s->s_maxbytes = MAX_NON_LFS; |
246 | s->s_op = &default_op; | |
247 | s->s_time_gran = 1000000000; | |
3cb29d11 | 248 | s->cleancache_poolid = CLEANCACHE_NO_POOL; |
7eb5e882 AV |
249 | |
250 | s->s_shrink.seeks = DEFAULT_SEEKS; | |
251 | s->s_shrink.scan_objects = super_cache_scan; | |
252 | s->s_shrink.count_objects = super_cache_count; | |
253 | s->s_shrink.batch = 1024; | |
2acb60a0 | 254 | s->s_shrink.flags = SHRINKER_NUMA_AWARE | SHRINKER_MEMCG_AWARE; |
7632554f TH |
255 | if (prealloc_shrinker(&s->s_shrink)) |
256 | goto fail; | |
1da177e4 | 257 | return s; |
5ca302c8 | 258 | |
7eb5e882 | 259 | fail: |
0200894d | 260 | destroy_unused_super(s); |
7eb5e882 | 261 | return NULL; |
1da177e4 LT |
262 | } |
263 | ||
264 | /* Superblock refcounting */ | |
265 | ||
266 | /* | |
35cf7ba0 | 267 | * Drop a superblock's refcount. The caller must hold sb_lock. |
1da177e4 | 268 | */ |
c645b930 | 269 | static void __put_super(struct super_block *s) |
1da177e4 | 270 | { |
c645b930 AV |
271 | if (!--s->s_count) { |
272 | list_del_init(&s->s_list); | |
273 | WARN_ON(s->s_dentry_lru.node); | |
274 | WARN_ON(s->s_inode_lru.node); | |
275 | WARN_ON(!list_empty(&s->s_mounts)); | |
276 | security_sb_free(s); | |
277 | put_user_ns(s->s_user_ns); | |
278 | kfree(s->s_subtype); | |
279 | call_rcu(&s->rcu, destroy_super_rcu); | |
1da177e4 | 280 | } |
1da177e4 LT |
281 | } |
282 | ||
283 | /** | |
284 | * put_super - drop a temporary reference to superblock | |
285 | * @sb: superblock in question | |
286 | * | |
287 | * Drops a temporary reference, frees superblock if there's no | |
288 | * references left. | |
289 | */ | |
f47ec3f2 | 290 | static void put_super(struct super_block *sb) |
1da177e4 LT |
291 | { |
292 | spin_lock(&sb_lock); | |
293 | __put_super(sb); | |
294 | spin_unlock(&sb_lock); | |
295 | } | |
296 | ||
297 | ||
298 | /** | |
1712ac8f | 299 | * deactivate_locked_super - drop an active reference to superblock |
1da177e4 LT |
300 | * @s: superblock to deactivate |
301 | * | |
bd7ced98 | 302 | * Drops an active reference to superblock, converting it into a temporary |
1712ac8f | 303 | * one if there is no other active references left. In that case we |
1da177e4 LT |
304 | * tell fs driver to shut it down and drop the temporary reference we |
305 | * had just acquired. | |
1712ac8f AV |
306 | * |
307 | * Caller holds exclusive lock on superblock; that lock is released. | |
1da177e4 | 308 | */ |
1712ac8f | 309 | void deactivate_locked_super(struct super_block *s) |
1da177e4 LT |
310 | { |
311 | struct file_system_type *fs = s->s_type; | |
b20bd1a5 | 312 | if (atomic_dec_and_test(&s->s_active)) { |
3167760f | 313 | cleancache_invalidate_fs(s); |
b0d40c92 | 314 | unregister_shrinker(&s->s_shrink); |
28f2cd4f | 315 | fs->kill_sb(s); |
f5e1dd34 | 316 | |
c0a5b560 VD |
317 | /* |
318 | * Since list_lru_destroy() may sleep, we cannot call it from | |
319 | * put_super(), where we hold the sb_lock. Therefore we destroy | |
320 | * the lru lists right now. | |
321 | */ | |
322 | list_lru_destroy(&s->s_dentry_lru); | |
323 | list_lru_destroy(&s->s_inode_lru); | |
324 | ||
1da177e4 LT |
325 | put_filesystem(fs); |
326 | put_super(s); | |
1712ac8f AV |
327 | } else { |
328 | up_write(&s->s_umount); | |
1da177e4 LT |
329 | } |
330 | } | |
331 | ||
1712ac8f | 332 | EXPORT_SYMBOL(deactivate_locked_super); |
1da177e4 | 333 | |
74dbbdd7 | 334 | /** |
1712ac8f | 335 | * deactivate_super - drop an active reference to superblock |
74dbbdd7 AV |
336 | * @s: superblock to deactivate |
337 | * | |
1712ac8f AV |
338 | * Variant of deactivate_locked_super(), except that superblock is *not* |
339 | * locked by caller. If we are going to drop the final active reference, | |
340 | * lock will be acquired prior to that. | |
74dbbdd7 | 341 | */ |
1712ac8f | 342 | void deactivate_super(struct super_block *s) |
74dbbdd7 | 343 | { |
1712ac8f AV |
344 | if (!atomic_add_unless(&s->s_active, -1, 1)) { |
345 | down_write(&s->s_umount); | |
346 | deactivate_locked_super(s); | |
74dbbdd7 AV |
347 | } |
348 | } | |
349 | ||
1712ac8f | 350 | EXPORT_SYMBOL(deactivate_super); |
74dbbdd7 | 351 | |
1da177e4 LT |
352 | /** |
353 | * grab_super - acquire an active reference | |
354 | * @s: reference we are trying to make active | |
355 | * | |
356 | * Tries to acquire an active reference. grab_super() is used when we | |
357 | * had just found a superblock in super_blocks or fs_type->fs_supers | |
358 | * and want to turn it into a full-blown active reference. grab_super() | |
359 | * is called with sb_lock held and drops it. Returns 1 in case of | |
360 | * success, 0 if we had failed (superblock contents was already dead or | |
acfec9a5 AV |
361 | * dying when grab_super() had been called). Note that this is only |
362 | * called for superblocks not in rundown mode (== ones still on ->fs_supers | |
363 | * of their type), so increment of ->s_count is OK here. | |
1da177e4 | 364 | */ |
9c4dbee7 | 365 | static int grab_super(struct super_block *s) __releases(sb_lock) |
1da177e4 LT |
366 | { |
367 | s->s_count++; | |
368 | spin_unlock(&sb_lock); | |
369 | down_write(&s->s_umount); | |
e462ec50 | 370 | if ((s->s_flags & SB_BORN) && atomic_inc_not_zero(&s->s_active)) { |
acfec9a5 AV |
371 | put_super(s); |
372 | return 1; | |
373 | } | |
1da177e4 LT |
374 | up_write(&s->s_umount); |
375 | put_super(s); | |
1da177e4 LT |
376 | return 0; |
377 | } | |
378 | ||
12ad3ab6 | 379 | /* |
eb6ef3df | 380 | * trylock_super - try to grab ->s_umount shared |
331cbdee | 381 | * @sb: reference we are trying to grab |
12ad3ab6 | 382 | * |
eb6ef3df | 383 | * Try to prevent fs shutdown. This is used in places where we |
12ad3ab6 | 384 | * cannot take an active reference but we need to ensure that the |
eb6ef3df KK |
385 | * filesystem is not shut down while we are working on it. It returns |
386 | * false if we cannot acquire s_umount or if we lose the race and | |
387 | * filesystem already got into shutdown, and returns true with the s_umount | |
388 | * lock held in read mode in case of success. On successful return, | |
389 | * the caller must drop the s_umount lock when done. | |
390 | * | |
391 | * Note that unlike get_super() et.al. this one does *not* bump ->s_count. | |
392 | * The reason why it's safe is that we are OK with doing trylock instead | |
393 | * of down_read(). There's a couple of places that are OK with that, but | |
394 | * it's very much not a general-purpose interface. | |
12ad3ab6 | 395 | */ |
eb6ef3df | 396 | bool trylock_super(struct super_block *sb) |
12ad3ab6 | 397 | { |
12ad3ab6 | 398 | if (down_read_trylock(&sb->s_umount)) { |
eb6ef3df | 399 | if (!hlist_unhashed(&sb->s_instances) && |
e462ec50 | 400 | sb->s_root && (sb->s_flags & SB_BORN)) |
12ad3ab6 DC |
401 | return true; |
402 | up_read(&sb->s_umount); | |
403 | } | |
404 | ||
12ad3ab6 DC |
405 | return false; |
406 | } | |
407 | ||
1da177e4 LT |
408 | /** |
409 | * generic_shutdown_super - common helper for ->kill_sb() | |
410 | * @sb: superblock to kill | |
411 | * | |
412 | * generic_shutdown_super() does all fs-independent work on superblock | |
413 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects | |
414 | * that need destruction out of superblock, call generic_shutdown_super() | |
415 | * and release aforementioned objects. Note: dentries and inodes _are_ | |
416 | * taken care of and do not need specific handling. | |
c636ebdb DH |
417 | * |
418 | * Upon calling this function, the filesystem may no longer alter or | |
419 | * rearrange the set of dentries belonging to this super_block, nor may it | |
420 | * change the attachments of dentries to inodes. | |
1da177e4 LT |
421 | */ |
422 | void generic_shutdown_super(struct super_block *sb) | |
423 | { | |
ee9b6d61 | 424 | const struct super_operations *sop = sb->s_op; |
1da177e4 | 425 | |
c636ebdb DH |
426 | if (sb->s_root) { |
427 | shrink_dcache_for_umount(sb); | |
60b0680f | 428 | sync_filesystem(sb); |
e462ec50 | 429 | sb->s_flags &= ~SB_ACTIVE; |
efaee192 | 430 | |
74278da9 | 431 | fsnotify_unmount_inodes(sb); |
a1a0e23e | 432 | cgroup_writeback_umount(); |
63997e98 AV |
433 | |
434 | evict_inodes(sb); | |
1da177e4 | 435 | |
7b7a8665 CH |
436 | if (sb->s_dio_done_wq) { |
437 | destroy_workqueue(sb->s_dio_done_wq); | |
438 | sb->s_dio_done_wq = NULL; | |
439 | } | |
440 | ||
1da177e4 LT |
441 | if (sop->put_super) |
442 | sop->put_super(sb); | |
443 | ||
63997e98 | 444 | if (!list_empty(&sb->s_inodes)) { |
7b4fe29e DJ |
445 | printk("VFS: Busy inodes after unmount of %s. " |
446 | "Self-destruct in 5 seconds. Have a nice day...\n", | |
447 | sb->s_id); | |
1da177e4 | 448 | } |
1da177e4 LT |
449 | } |
450 | spin_lock(&sb_lock); | |
451 | /* should be initialized for __put_super_and_need_restart() */ | |
a5166169 | 452 | hlist_del_init(&sb->s_instances); |
1da177e4 LT |
453 | spin_unlock(&sb_lock); |
454 | up_write(&sb->s_umount); | |
c1844d53 | 455 | if (sb->s_bdi != &noop_backing_dev_info) { |
fca39346 JK |
456 | bdi_put(sb->s_bdi); |
457 | sb->s_bdi = &noop_backing_dev_info; | |
fca39346 | 458 | } |
1da177e4 LT |
459 | } |
460 | ||
461 | EXPORT_SYMBOL(generic_shutdown_super); | |
462 | ||
463 | /** | |
6e4eab57 | 464 | * sget_userns - find or create a superblock |
1da177e4 LT |
465 | * @type: filesystem type superblock should belong to |
466 | * @test: comparison callback | |
467 | * @set: setup callback | |
9249e17f | 468 | * @flags: mount flags |
6e4eab57 | 469 | * @user_ns: User namespace for the super_block |
1da177e4 LT |
470 | * @data: argument to each of them |
471 | */ | |
6e4eab57 | 472 | struct super_block *sget_userns(struct file_system_type *type, |
1da177e4 LT |
473 | int (*test)(struct super_block *,void *), |
474 | int (*set)(struct super_block *,void *), | |
6e4eab57 | 475 | int flags, struct user_namespace *user_ns, |
1da177e4 LT |
476 | void *data) |
477 | { | |
478 | struct super_block *s = NULL; | |
d4730127 | 479 | struct super_block *old; |
1da177e4 LT |
480 | int err; |
481 | ||
e462ec50 | 482 | if (!(flags & (SB_KERNMOUNT|SB_SUBMOUNT)) && |
a001e74c EB |
483 | !(type->fs_flags & FS_USERNS_MOUNT) && |
484 | !capable(CAP_SYS_ADMIN)) | |
485 | return ERR_PTR(-EPERM); | |
1da177e4 LT |
486 | retry: |
487 | spin_lock(&sb_lock); | |
d4730127 | 488 | if (test) { |
b67bfe0d | 489 | hlist_for_each_entry(old, &type->fs_supers, s_instances) { |
d4730127 MK |
490 | if (!test(old, data)) |
491 | continue; | |
6e4eab57 EB |
492 | if (user_ns != old->s_user_ns) { |
493 | spin_unlock(&sb_lock); | |
0200894d | 494 | destroy_unused_super(s); |
6e4eab57 EB |
495 | return ERR_PTR(-EBUSY); |
496 | } | |
d4730127 MK |
497 | if (!grab_super(old)) |
498 | goto retry; | |
0200894d | 499 | destroy_unused_super(s); |
d4730127 MK |
500 | return old; |
501 | } | |
1da177e4 LT |
502 | } |
503 | if (!s) { | |
504 | spin_unlock(&sb_lock); | |
e462ec50 | 505 | s = alloc_super(type, (flags & ~SB_SUBMOUNT), user_ns); |
1da177e4 LT |
506 | if (!s) |
507 | return ERR_PTR(-ENOMEM); | |
508 | goto retry; | |
509 | } | |
dd111b31 | 510 | |
1da177e4 LT |
511 | err = set(s, data); |
512 | if (err) { | |
513 | spin_unlock(&sb_lock); | |
0200894d | 514 | destroy_unused_super(s); |
1da177e4 LT |
515 | return ERR_PTR(err); |
516 | } | |
517 | s->s_type = type; | |
518 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); | |
519 | list_add_tail(&s->s_list, &super_blocks); | |
a5166169 | 520 | hlist_add_head(&s->s_instances, &type->fs_supers); |
1da177e4 LT |
521 | spin_unlock(&sb_lock); |
522 | get_filesystem(type); | |
7632554f | 523 | register_shrinker_prepared(&s->s_shrink); |
1da177e4 LT |
524 | return s; |
525 | } | |
526 | ||
6e4eab57 EB |
527 | EXPORT_SYMBOL(sget_userns); |
528 | ||
529 | /** | |
530 | * sget - find or create a superblock | |
531 | * @type: filesystem type superblock should belong to | |
532 | * @test: comparison callback | |
533 | * @set: setup callback | |
534 | * @flags: mount flags | |
535 | * @data: argument to each of them | |
536 | */ | |
537 | struct super_block *sget(struct file_system_type *type, | |
538 | int (*test)(struct super_block *,void *), | |
539 | int (*set)(struct super_block *,void *), | |
540 | int flags, | |
541 | void *data) | |
542 | { | |
543 | struct user_namespace *user_ns = current_user_ns(); | |
544 | ||
93faccbb EB |
545 | /* We don't yet pass the user namespace of the parent |
546 | * mount through to here so always use &init_user_ns | |
547 | * until that changes. | |
548 | */ | |
e462ec50 | 549 | if (flags & SB_SUBMOUNT) |
93faccbb EB |
550 | user_ns = &init_user_ns; |
551 | ||
6e4eab57 | 552 | /* Ensure the requestor has permissions over the target filesystem */ |
e462ec50 | 553 | if (!(flags & (SB_KERNMOUNT|SB_SUBMOUNT)) && !ns_capable(user_ns, CAP_SYS_ADMIN)) |
6e4eab57 EB |
554 | return ERR_PTR(-EPERM); |
555 | ||
556 | return sget_userns(type, test, set, flags, user_ns, data); | |
557 | } | |
558 | ||
1da177e4 LT |
559 | EXPORT_SYMBOL(sget); |
560 | ||
561 | void drop_super(struct super_block *sb) | |
562 | { | |
563 | up_read(&sb->s_umount); | |
564 | put_super(sb); | |
565 | } | |
566 | ||
567 | EXPORT_SYMBOL(drop_super); | |
568 | ||
ba6379f7 JK |
569 | void drop_super_exclusive(struct super_block *sb) |
570 | { | |
571 | up_write(&sb->s_umount); | |
572 | put_super(sb); | |
573 | } | |
574 | EXPORT_SYMBOL(drop_super_exclusive); | |
575 | ||
01a05b33 AV |
576 | /** |
577 | * iterate_supers - call function for all active superblocks | |
578 | * @f: function to call | |
579 | * @arg: argument to pass to it | |
580 | * | |
581 | * Scans the superblock list and calls given function, passing it | |
582 | * locked superblock and given argument. | |
583 | */ | |
584 | void iterate_supers(void (*f)(struct super_block *, void *), void *arg) | |
585 | { | |
dca33252 | 586 | struct super_block *sb, *p = NULL; |
01a05b33 AV |
587 | |
588 | spin_lock(&sb_lock); | |
dca33252 | 589 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 590 | if (hlist_unhashed(&sb->s_instances)) |
01a05b33 AV |
591 | continue; |
592 | sb->s_count++; | |
593 | spin_unlock(&sb_lock); | |
594 | ||
595 | down_read(&sb->s_umount); | |
e462ec50 | 596 | if (sb->s_root && (sb->s_flags & SB_BORN)) |
01a05b33 AV |
597 | f(sb, arg); |
598 | up_read(&sb->s_umount); | |
599 | ||
600 | spin_lock(&sb_lock); | |
dca33252 AV |
601 | if (p) |
602 | __put_super(p); | |
603 | p = sb; | |
01a05b33 | 604 | } |
dca33252 AV |
605 | if (p) |
606 | __put_super(p); | |
01a05b33 AV |
607 | spin_unlock(&sb_lock); |
608 | } | |
609 | ||
43e15cdb AV |
610 | /** |
611 | * iterate_supers_type - call function for superblocks of given type | |
612 | * @type: fs type | |
613 | * @f: function to call | |
614 | * @arg: argument to pass to it | |
615 | * | |
616 | * Scans the superblock list and calls given function, passing it | |
617 | * locked superblock and given argument. | |
618 | */ | |
619 | void iterate_supers_type(struct file_system_type *type, | |
620 | void (*f)(struct super_block *, void *), void *arg) | |
621 | { | |
622 | struct super_block *sb, *p = NULL; | |
623 | ||
624 | spin_lock(&sb_lock); | |
b67bfe0d | 625 | hlist_for_each_entry(sb, &type->fs_supers, s_instances) { |
43e15cdb AV |
626 | sb->s_count++; |
627 | spin_unlock(&sb_lock); | |
628 | ||
629 | down_read(&sb->s_umount); | |
e462ec50 | 630 | if (sb->s_root && (sb->s_flags & SB_BORN)) |
43e15cdb AV |
631 | f(sb, arg); |
632 | up_read(&sb->s_umount); | |
633 | ||
634 | spin_lock(&sb_lock); | |
635 | if (p) | |
636 | __put_super(p); | |
637 | p = sb; | |
638 | } | |
639 | if (p) | |
640 | __put_super(p); | |
641 | spin_unlock(&sb_lock); | |
642 | } | |
643 | ||
644 | EXPORT_SYMBOL(iterate_supers_type); | |
645 | ||
ba6379f7 | 646 | static struct super_block *__get_super(struct block_device *bdev, bool excl) |
1da177e4 | 647 | { |
618f0636 KK |
648 | struct super_block *sb; |
649 | ||
1da177e4 LT |
650 | if (!bdev) |
651 | return NULL; | |
618f0636 | 652 | |
1da177e4 | 653 | spin_lock(&sb_lock); |
618f0636 KK |
654 | rescan: |
655 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 656 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 657 | continue; |
618f0636 KK |
658 | if (sb->s_bdev == bdev) { |
659 | sb->s_count++; | |
1da177e4 | 660 | spin_unlock(&sb_lock); |
ba6379f7 JK |
661 | if (!excl) |
662 | down_read(&sb->s_umount); | |
663 | else | |
664 | down_write(&sb->s_umount); | |
df40c01a | 665 | /* still alive? */ |
e462ec50 | 666 | if (sb->s_root && (sb->s_flags & SB_BORN)) |
618f0636 | 667 | return sb; |
ba6379f7 JK |
668 | if (!excl) |
669 | up_read(&sb->s_umount); | |
670 | else | |
671 | up_write(&sb->s_umount); | |
df40c01a | 672 | /* nope, got unmounted */ |
618f0636 | 673 | spin_lock(&sb_lock); |
df40c01a AV |
674 | __put_super(sb); |
675 | goto rescan; | |
1da177e4 LT |
676 | } |
677 | } | |
678 | spin_unlock(&sb_lock); | |
679 | return NULL; | |
680 | } | |
681 | ||
6b6dc836 | 682 | /** |
ba6379f7 | 683 | * get_super - get the superblock of a device |
6b6dc836 JK |
684 | * @bdev: device to get the superblock for |
685 | * | |
686 | * Scans the superblock list and finds the superblock of the file system | |
ba6379f7 | 687 | * mounted on the device given. %NULL is returned if no match is found. |
6b6dc836 | 688 | */ |
ba6379f7 JK |
689 | struct super_block *get_super(struct block_device *bdev) |
690 | { | |
691 | return __get_super(bdev, false); | |
692 | } | |
693 | EXPORT_SYMBOL(get_super); | |
694 | ||
695 | static struct super_block *__get_super_thawed(struct block_device *bdev, | |
696 | bool excl) | |
6b6dc836 JK |
697 | { |
698 | while (1) { | |
ba6379f7 | 699 | struct super_block *s = __get_super(bdev, excl); |
5accdf82 | 700 | if (!s || s->s_writers.frozen == SB_UNFROZEN) |
6b6dc836 | 701 | return s; |
ba6379f7 JK |
702 | if (!excl) |
703 | up_read(&s->s_umount); | |
704 | else | |
705 | up_write(&s->s_umount); | |
5accdf82 JK |
706 | wait_event(s->s_writers.wait_unfrozen, |
707 | s->s_writers.frozen == SB_UNFROZEN); | |
6b6dc836 JK |
708 | put_super(s); |
709 | } | |
710 | } | |
ba6379f7 JK |
711 | |
712 | /** | |
713 | * get_super_thawed - get thawed superblock of a device | |
714 | * @bdev: device to get the superblock for | |
715 | * | |
716 | * Scans the superblock list and finds the superblock of the file system | |
717 | * mounted on the device. The superblock is returned once it is thawed | |
718 | * (or immediately if it was not frozen). %NULL is returned if no match | |
719 | * is found. | |
720 | */ | |
721 | struct super_block *get_super_thawed(struct block_device *bdev) | |
722 | { | |
723 | return __get_super_thawed(bdev, false); | |
724 | } | |
6b6dc836 JK |
725 | EXPORT_SYMBOL(get_super_thawed); |
726 | ||
ba6379f7 JK |
727 | /** |
728 | * get_super_exclusive_thawed - get thawed superblock of a device | |
729 | * @bdev: device to get the superblock for | |
730 | * | |
731 | * Scans the superblock list and finds the superblock of the file system | |
732 | * mounted on the device. The superblock is returned once it is thawed | |
733 | * (or immediately if it was not frozen) and s_umount semaphore is held | |
734 | * in exclusive mode. %NULL is returned if no match is found. | |
735 | */ | |
736 | struct super_block *get_super_exclusive_thawed(struct block_device *bdev) | |
737 | { | |
738 | return __get_super_thawed(bdev, true); | |
739 | } | |
740 | EXPORT_SYMBOL(get_super_exclusive_thawed); | |
741 | ||
4504230a CH |
742 | /** |
743 | * get_active_super - get an active reference to the superblock of a device | |
744 | * @bdev: device to get the superblock for | |
745 | * | |
746 | * Scans the superblock list and finds the superblock of the file system | |
747 | * mounted on the device given. Returns the superblock with an active | |
d3f21473 | 748 | * reference or %NULL if none was found. |
4504230a CH |
749 | */ |
750 | struct super_block *get_active_super(struct block_device *bdev) | |
751 | { | |
752 | struct super_block *sb; | |
753 | ||
754 | if (!bdev) | |
755 | return NULL; | |
756 | ||
1494583d | 757 | restart: |
4504230a CH |
758 | spin_lock(&sb_lock); |
759 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 760 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 761 | continue; |
1494583d | 762 | if (sb->s_bdev == bdev) { |
acfec9a5 | 763 | if (!grab_super(sb)) |
1494583d | 764 | goto restart; |
acfec9a5 AV |
765 | up_write(&sb->s_umount); |
766 | return sb; | |
1494583d | 767 | } |
4504230a CH |
768 | } |
769 | spin_unlock(&sb_lock); | |
770 | return NULL; | |
771 | } | |
dd111b31 | 772 | |
df40c01a | 773 | struct super_block *user_get_super(dev_t dev) |
1da177e4 | 774 | { |
618f0636 | 775 | struct super_block *sb; |
1da177e4 | 776 | |
1da177e4 | 777 | spin_lock(&sb_lock); |
618f0636 KK |
778 | rescan: |
779 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 780 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 781 | continue; |
618f0636 KK |
782 | if (sb->s_dev == dev) { |
783 | sb->s_count++; | |
1da177e4 | 784 | spin_unlock(&sb_lock); |
618f0636 | 785 | down_read(&sb->s_umount); |
df40c01a | 786 | /* still alive? */ |
e462ec50 | 787 | if (sb->s_root && (sb->s_flags & SB_BORN)) |
618f0636 KK |
788 | return sb; |
789 | up_read(&sb->s_umount); | |
df40c01a | 790 | /* nope, got unmounted */ |
618f0636 | 791 | spin_lock(&sb_lock); |
df40c01a AV |
792 | __put_super(sb); |
793 | goto rescan; | |
1da177e4 LT |
794 | } |
795 | } | |
796 | spin_unlock(&sb_lock); | |
797 | return NULL; | |
798 | } | |
799 | ||
1da177e4 LT |
800 | /** |
801 | * do_remount_sb - asks filesystem to change mount options. | |
802 | * @sb: superblock in question | |
e462ec50 | 803 | * @sb_flags: revised superblock flags |
1da177e4 LT |
804 | * @data: the rest of options |
805 | * @force: whether or not to force the change | |
806 | * | |
807 | * Alters the mount options of a mounted file system. | |
808 | */ | |
e462ec50 | 809 | int do_remount_sb(struct super_block *sb, int sb_flags, void *data, int force) |
1da177e4 LT |
810 | { |
811 | int retval; | |
c79d967d | 812 | int remount_ro; |
4504230a | 813 | |
5accdf82 | 814 | if (sb->s_writers.frozen != SB_UNFROZEN) |
4504230a CH |
815 | return -EBUSY; |
816 | ||
9361401e | 817 | #ifdef CONFIG_BLOCK |
e462ec50 | 818 | if (!(sb_flags & SB_RDONLY) && bdev_read_only(sb->s_bdev)) |
1da177e4 | 819 | return -EACCES; |
9361401e | 820 | #endif |
4504230a | 821 | |
e462ec50 | 822 | remount_ro = (sb_flags & SB_RDONLY) && !sb_rdonly(sb); |
d208bbdd | 823 | |
0aec09d0 | 824 | if (remount_ro) { |
fdab684d | 825 | if (!hlist_empty(&sb->s_pins)) { |
0aec09d0 | 826 | up_write(&sb->s_umount); |
fdab684d | 827 | group_pin_kill(&sb->s_pins); |
0aec09d0 AV |
828 | down_write(&sb->s_umount); |
829 | if (!sb->s_root) | |
830 | return 0; | |
831 | if (sb->s_writers.frozen != SB_UNFROZEN) | |
832 | return -EBUSY; | |
e462ec50 | 833 | remount_ro = (sb_flags & SB_RDONLY) && !sb_rdonly(sb); |
0aec09d0 AV |
834 | } |
835 | } | |
836 | shrink_dcache_sb(sb); | |
837 | ||
1da177e4 LT |
838 | /* If we are remounting RDONLY and current sb is read/write, |
839 | make sure there are no rw files opened */ | |
d208bbdd | 840 | if (remount_ro) { |
4ed5e82f | 841 | if (force) { |
eee5cc27 AV |
842 | sb->s_readonly_remount = 1; |
843 | smp_wmb(); | |
4ed5e82f MS |
844 | } else { |
845 | retval = sb_prepare_remount_readonly(sb); | |
846 | if (retval) | |
847 | return retval; | |
4ed5e82f | 848 | } |
1da177e4 LT |
849 | } |
850 | ||
851 | if (sb->s_op->remount_fs) { | |
e462ec50 | 852 | retval = sb->s_op->remount_fs(sb, &sb_flags, data); |
2833eb2b MS |
853 | if (retval) { |
854 | if (!force) | |
4ed5e82f | 855 | goto cancel_readonly; |
2833eb2b MS |
856 | /* If forced remount, go ahead despite any errors */ |
857 | WARN(1, "forced remount of a %s fs returned %i\n", | |
858 | sb->s_type->name, retval); | |
859 | } | |
1da177e4 | 860 | } |
e462ec50 | 861 | sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (sb_flags & MS_RMT_MASK); |
4ed5e82f MS |
862 | /* Needs to be ordered wrt mnt_is_readonly() */ |
863 | smp_wmb(); | |
864 | sb->s_readonly_remount = 0; | |
c79d967d | 865 | |
d208bbdd NP |
866 | /* |
867 | * Some filesystems modify their metadata via some other path than the | |
868 | * bdev buffer cache (eg. use a private mapping, or directories in | |
869 | * pagecache, etc). Also file data modifications go via their own | |
870 | * mappings. So If we try to mount readonly then copy the filesystem | |
871 | * from bdev, we could get stale data, so invalidate it to give a best | |
872 | * effort at coherency. | |
873 | */ | |
874 | if (remount_ro && sb->s_bdev) | |
875 | invalidate_bdev(sb->s_bdev); | |
1da177e4 | 876 | return 0; |
4ed5e82f MS |
877 | |
878 | cancel_readonly: | |
879 | sb->s_readonly_remount = 0; | |
880 | return retval; | |
1da177e4 LT |
881 | } |
882 | ||
a2a9537a | 883 | static void do_emergency_remount(struct work_struct *work) |
1da177e4 | 884 | { |
dca33252 | 885 | struct super_block *sb, *p = NULL; |
1da177e4 LT |
886 | |
887 | spin_lock(&sb_lock); | |
dca33252 | 888 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 889 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 890 | continue; |
1da177e4 LT |
891 | sb->s_count++; |
892 | spin_unlock(&sb_lock); | |
443b94ba | 893 | down_write(&sb->s_umount); |
e462ec50 | 894 | if (sb->s_root && sb->s_bdev && (sb->s_flags & SB_BORN) && |
bc98a42c | 895 | !sb_rdonly(sb)) { |
1da177e4 | 896 | /* |
1da177e4 LT |
897 | * What lock protects sb->s_flags?? |
898 | */ | |
e462ec50 | 899 | do_remount_sb(sb, SB_RDONLY, NULL, 1); |
1da177e4 | 900 | } |
443b94ba | 901 | up_write(&sb->s_umount); |
1da177e4 | 902 | spin_lock(&sb_lock); |
dca33252 AV |
903 | if (p) |
904 | __put_super(p); | |
905 | p = sb; | |
1da177e4 | 906 | } |
dca33252 AV |
907 | if (p) |
908 | __put_super(p); | |
1da177e4 | 909 | spin_unlock(&sb_lock); |
a2a9537a | 910 | kfree(work); |
1da177e4 LT |
911 | printk("Emergency Remount complete\n"); |
912 | } | |
913 | ||
914 | void emergency_remount(void) | |
915 | { | |
a2a9537a JA |
916 | struct work_struct *work; |
917 | ||
918 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | |
919 | if (work) { | |
920 | INIT_WORK(work, do_emergency_remount); | |
921 | schedule_work(work); | |
922 | } | |
1da177e4 LT |
923 | } |
924 | ||
925 | /* | |
926 | * Unnamed block devices are dummy devices used by virtual | |
927 | * filesystems which don't use real block-devices. -- jrs | |
928 | */ | |
929 | ||
ad76cbc6 | 930 | static DEFINE_IDA(unnamed_dev_ida); |
1da177e4 | 931 | static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */ |
a2a4dc49 TB |
932 | /* Many userspace utilities consider an FSID of 0 invalid. |
933 | * Always return at least 1 from get_anon_bdev. | |
934 | */ | |
935 | static int unnamed_dev_start = 1; | |
1da177e4 | 936 | |
0ee5dc67 | 937 | int get_anon_bdev(dev_t *p) |
1da177e4 LT |
938 | { |
939 | int dev; | |
940 | int error; | |
941 | ||
942 | retry: | |
ad76cbc6 | 943 | if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0) |
1da177e4 LT |
944 | return -ENOMEM; |
945 | spin_lock(&unnamed_dev_lock); | |
c63e09ec | 946 | error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev); |
f21f6220 AV |
947 | if (!error) |
948 | unnamed_dev_start = dev + 1; | |
1da177e4 LT |
949 | spin_unlock(&unnamed_dev_lock); |
950 | if (error == -EAGAIN) | |
951 | /* We raced and lost with another CPU. */ | |
952 | goto retry; | |
953 | else if (error) | |
954 | return -EAGAIN; | |
955 | ||
1af95de6 | 956 | if (dev >= (1 << MINORBITS)) { |
1da177e4 | 957 | spin_lock(&unnamed_dev_lock); |
ad76cbc6 | 958 | ida_remove(&unnamed_dev_ida, dev); |
f21f6220 AV |
959 | if (unnamed_dev_start > dev) |
960 | unnamed_dev_start = dev; | |
1da177e4 LT |
961 | spin_unlock(&unnamed_dev_lock); |
962 | return -EMFILE; | |
963 | } | |
0ee5dc67 | 964 | *p = MKDEV(0, dev & MINORMASK); |
1da177e4 LT |
965 | return 0; |
966 | } | |
0ee5dc67 | 967 | EXPORT_SYMBOL(get_anon_bdev); |
1da177e4 | 968 | |
0ee5dc67 | 969 | void free_anon_bdev(dev_t dev) |
1da177e4 | 970 | { |
0ee5dc67 | 971 | int slot = MINOR(dev); |
1da177e4 | 972 | spin_lock(&unnamed_dev_lock); |
ad76cbc6 | 973 | ida_remove(&unnamed_dev_ida, slot); |
c63e09ec AV |
974 | if (slot < unnamed_dev_start) |
975 | unnamed_dev_start = slot; | |
1da177e4 LT |
976 | spin_unlock(&unnamed_dev_lock); |
977 | } | |
0ee5dc67 AV |
978 | EXPORT_SYMBOL(free_anon_bdev); |
979 | ||
980 | int set_anon_super(struct super_block *s, void *data) | |
981 | { | |
df0ce26c | 982 | return get_anon_bdev(&s->s_dev); |
0ee5dc67 AV |
983 | } |
984 | ||
985 | EXPORT_SYMBOL(set_anon_super); | |
986 | ||
987 | void kill_anon_super(struct super_block *sb) | |
988 | { | |
989 | dev_t dev = sb->s_dev; | |
990 | generic_shutdown_super(sb); | |
991 | free_anon_bdev(dev); | |
992 | } | |
1da177e4 LT |
993 | |
994 | EXPORT_SYMBOL(kill_anon_super); | |
995 | ||
1da177e4 LT |
996 | void kill_litter_super(struct super_block *sb) |
997 | { | |
998 | if (sb->s_root) | |
999 | d_genocide(sb->s_root); | |
1000 | kill_anon_super(sb); | |
1001 | } | |
1002 | ||
1003 | EXPORT_SYMBOL(kill_litter_super); | |
1004 | ||
909e6d94 SH |
1005 | static int ns_test_super(struct super_block *sb, void *data) |
1006 | { | |
1007 | return sb->s_fs_info == data; | |
1008 | } | |
1009 | ||
1010 | static int ns_set_super(struct super_block *sb, void *data) | |
1011 | { | |
1012 | sb->s_fs_info = data; | |
1013 | return set_anon_super(sb, NULL); | |
1014 | } | |
1015 | ||
d91ee87d EB |
1016 | struct dentry *mount_ns(struct file_system_type *fs_type, |
1017 | int flags, void *data, void *ns, struct user_namespace *user_ns, | |
1018 | int (*fill_super)(struct super_block *, void *, int)) | |
909e6d94 SH |
1019 | { |
1020 | struct super_block *sb; | |
1021 | ||
d91ee87d EB |
1022 | /* Don't allow mounting unless the caller has CAP_SYS_ADMIN |
1023 | * over the namespace. | |
1024 | */ | |
e462ec50 | 1025 | if (!(flags & SB_KERNMOUNT) && !ns_capable(user_ns, CAP_SYS_ADMIN)) |
d91ee87d EB |
1026 | return ERR_PTR(-EPERM); |
1027 | ||
6e4eab57 EB |
1028 | sb = sget_userns(fs_type, ns_test_super, ns_set_super, flags, |
1029 | user_ns, ns); | |
909e6d94 | 1030 | if (IS_ERR(sb)) |
ceefda69 | 1031 | return ERR_CAST(sb); |
909e6d94 SH |
1032 | |
1033 | if (!sb->s_root) { | |
1034 | int err; | |
e462ec50 | 1035 | err = fill_super(sb, data, flags & SB_SILENT ? 1 : 0); |
909e6d94 | 1036 | if (err) { |
74dbbdd7 | 1037 | deactivate_locked_super(sb); |
ceefda69 | 1038 | return ERR_PTR(err); |
909e6d94 SH |
1039 | } |
1040 | ||
e462ec50 | 1041 | sb->s_flags |= SB_ACTIVE; |
909e6d94 SH |
1042 | } |
1043 | ||
ceefda69 | 1044 | return dget(sb->s_root); |
909e6d94 SH |
1045 | } |
1046 | ||
ceefda69 | 1047 | EXPORT_SYMBOL(mount_ns); |
909e6d94 | 1048 | |
9361401e | 1049 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
1050 | static int set_bdev_super(struct super_block *s, void *data) |
1051 | { | |
1052 | s->s_bdev = data; | |
1053 | s->s_dev = s->s_bdev->bd_dev; | |
13eec236 | 1054 | s->s_bdi = bdi_get(s->s_bdev->bd_bdi); |
32a88aa1 | 1055 | |
1da177e4 LT |
1056 | return 0; |
1057 | } | |
1058 | ||
1059 | static int test_bdev_super(struct super_block *s, void *data) | |
1060 | { | |
1061 | return (void *)s->s_bdev == data; | |
1062 | } | |
1063 | ||
152a0836 | 1064 | struct dentry *mount_bdev(struct file_system_type *fs_type, |
1da177e4 | 1065 | int flags, const char *dev_name, void *data, |
152a0836 | 1066 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1067 | { |
1068 | struct block_device *bdev; | |
1069 | struct super_block *s; | |
d4d77629 | 1070 | fmode_t mode = FMODE_READ | FMODE_EXCL; |
1da177e4 LT |
1071 | int error = 0; |
1072 | ||
e462ec50 | 1073 | if (!(flags & SB_RDONLY)) |
30c40d2c AV |
1074 | mode |= FMODE_WRITE; |
1075 | ||
d4d77629 | 1076 | bdev = blkdev_get_by_path(dev_name, mode, fs_type); |
1da177e4 | 1077 | if (IS_ERR(bdev)) |
152a0836 | 1078 | return ERR_CAST(bdev); |
1da177e4 | 1079 | |
dbf2d0a4 SF |
1080 | if (current_user_ns() != &init_user_ns) { |
1081 | /* | |
1082 | * For userns mounts, disallow mounting if bdev is open for | |
1083 | * writing | |
1084 | */ | |
1085 | if (!atomic_dec_unless_positive(&bdev->bd_inode->i_writecount)) { | |
1086 | error = -EBUSY; | |
1087 | goto error_bdev; | |
1088 | } | |
1089 | if (bdev->bd_contains != bdev && | |
1090 | !atomic_dec_unless_positive(&bdev->bd_contains->bd_inode->i_writecount)) { | |
1091 | atomic_inc(&bdev->bd_inode->i_writecount); | |
1092 | error = -EBUSY; | |
1093 | goto error_bdev; | |
1094 | } | |
1095 | } | |
1096 | ||
1da177e4 LT |
1097 | /* |
1098 | * once the super is inserted into the list by sget, s_umount | |
1099 | * will protect the lockfs code from trying to start a snapshot | |
1100 | * while we are mounting | |
1101 | */ | |
4fadd7bb CH |
1102 | mutex_lock(&bdev->bd_fsfreeze_mutex); |
1103 | if (bdev->bd_fsfreeze_count > 0) { | |
1104 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
1105 | error = -EBUSY; | |
dbf2d0a4 | 1106 | goto error_inc; |
4fadd7bb | 1107 | } |
e462ec50 | 1108 | s = sget(fs_type, test_bdev_super, set_bdev_super, flags | SB_NOSEC, |
9249e17f | 1109 | bdev); |
4fadd7bb | 1110 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
1da177e4 | 1111 | if (IS_ERR(s)) |
454e2398 | 1112 | goto error_s; |
1da177e4 LT |
1113 | |
1114 | if (s->s_root) { | |
e462ec50 | 1115 | if ((flags ^ s->s_flags) & SB_RDONLY) { |
74dbbdd7 | 1116 | deactivate_locked_super(s); |
454e2398 | 1117 | error = -EBUSY; |
dbf2d0a4 | 1118 | goto error_inc; |
1da177e4 | 1119 | } |
454e2398 | 1120 | |
4f331f01 TH |
1121 | /* |
1122 | * s_umount nests inside bd_mutex during | |
e525fd89 TH |
1123 | * __invalidate_device(). blkdev_put() acquires |
1124 | * bd_mutex and can't be called under s_umount. Drop | |
1125 | * s_umount temporarily. This is safe as we're | |
1126 | * holding an active reference. | |
4f331f01 TH |
1127 | */ |
1128 | up_write(&s->s_umount); | |
d4d77629 | 1129 | blkdev_put(bdev, mode); |
4f331f01 | 1130 | down_write(&s->s_umount); |
1da177e4 | 1131 | } else { |
30c40d2c | 1132 | s->s_mode = mode; |
a1c6f057 | 1133 | snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev); |
e78c9a00 | 1134 | sb_set_blocksize(s, block_size(bdev)); |
e462ec50 | 1135 | error = fill_super(s, data, flags & SB_SILENT ? 1 : 0); |
1da177e4 | 1136 | if (error) { |
74dbbdd7 | 1137 | deactivate_locked_super(s); |
454e2398 | 1138 | goto error; |
fa675765 | 1139 | } |
454e2398 | 1140 | |
e462ec50 | 1141 | s->s_flags |= SB_ACTIVE; |
87d8fe1e | 1142 | bdev->bd_super = s; |
1da177e4 LT |
1143 | } |
1144 | ||
152a0836 | 1145 | return dget(s->s_root); |
1da177e4 | 1146 | |
454e2398 DH |
1147 | error_s: |
1148 | error = PTR_ERR(s); | |
dbf2d0a4 SF |
1149 | error_inc: |
1150 | if (current_user_ns() != &init_user_ns) { | |
1151 | atomic_inc(&bdev->bd_inode->i_writecount); | |
1152 | if (bdev->bd_contains != bdev) | |
1153 | atomic_inc(&bdev->bd_contains->bd_inode->i_writecount); | |
1154 | } | |
454e2398 | 1155 | error_bdev: |
d4d77629 | 1156 | blkdev_put(bdev, mode); |
454e2398 | 1157 | error: |
152a0836 AV |
1158 | return ERR_PTR(error); |
1159 | } | |
1160 | EXPORT_SYMBOL(mount_bdev); | |
1161 | ||
1da177e4 LT |
1162 | void kill_block_super(struct super_block *sb) |
1163 | { | |
1164 | struct block_device *bdev = sb->s_bdev; | |
30c40d2c | 1165 | fmode_t mode = sb->s_mode; |
1da177e4 | 1166 | |
ddbaaf30 | 1167 | bdev->bd_super = NULL; |
1da177e4 LT |
1168 | generic_shutdown_super(sb); |
1169 | sync_blockdev(bdev); | |
d4d77629 | 1170 | WARN_ON_ONCE(!(mode & FMODE_EXCL)); |
dbf2d0a4 SF |
1171 | if (sb->s_user_ns != &init_user_ns) { |
1172 | atomic_inc(&bdev->bd_inode->i_writecount); | |
1173 | if (bdev->bd_contains != bdev) | |
1174 | atomic_inc(&bdev->bd_contains->bd_inode->i_writecount); | |
1175 | } | |
e525fd89 | 1176 | blkdev_put(bdev, mode | FMODE_EXCL); |
1da177e4 LT |
1177 | } |
1178 | ||
1179 | EXPORT_SYMBOL(kill_block_super); | |
9361401e | 1180 | #endif |
1da177e4 | 1181 | |
3c26ff6e | 1182 | struct dentry *mount_nodev(struct file_system_type *fs_type, |
1da177e4 | 1183 | int flags, void *data, |
3c26ff6e | 1184 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1185 | { |
1186 | int error; | |
9249e17f | 1187 | struct super_block *s = sget(fs_type, NULL, set_anon_super, flags, NULL); |
1da177e4 LT |
1188 | |
1189 | if (IS_ERR(s)) | |
3c26ff6e | 1190 | return ERR_CAST(s); |
1da177e4 | 1191 | |
e462ec50 | 1192 | error = fill_super(s, data, flags & SB_SILENT ? 1 : 0); |
1da177e4 | 1193 | if (error) { |
74dbbdd7 | 1194 | deactivate_locked_super(s); |
3c26ff6e | 1195 | return ERR_PTR(error); |
1da177e4 | 1196 | } |
e462ec50 | 1197 | s->s_flags |= SB_ACTIVE; |
3c26ff6e | 1198 | return dget(s->s_root); |
1da177e4 | 1199 | } |
3c26ff6e AV |
1200 | EXPORT_SYMBOL(mount_nodev); |
1201 | ||
1da177e4 LT |
1202 | static int compare_single(struct super_block *s, void *p) |
1203 | { | |
1204 | return 1; | |
1205 | } | |
1206 | ||
fc14f2fe | 1207 | struct dentry *mount_single(struct file_system_type *fs_type, |
1da177e4 | 1208 | int flags, void *data, |
fc14f2fe | 1209 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1210 | { |
1211 | struct super_block *s; | |
1212 | int error; | |
1213 | ||
9249e17f | 1214 | s = sget(fs_type, compare_single, set_anon_super, flags, NULL); |
1da177e4 | 1215 | if (IS_ERR(s)) |
fc14f2fe | 1216 | return ERR_CAST(s); |
1da177e4 | 1217 | if (!s->s_root) { |
e462ec50 | 1218 | error = fill_super(s, data, flags & SB_SILENT ? 1 : 0); |
1da177e4 | 1219 | if (error) { |
74dbbdd7 | 1220 | deactivate_locked_super(s); |
fc14f2fe | 1221 | return ERR_PTR(error); |
1da177e4 | 1222 | } |
e462ec50 | 1223 | s->s_flags |= SB_ACTIVE; |
9329d1be KS |
1224 | } else { |
1225 | do_remount_sb(s, flags, data, 0); | |
1da177e4 | 1226 | } |
fc14f2fe AV |
1227 | return dget(s->s_root); |
1228 | } | |
1229 | EXPORT_SYMBOL(mount_single); | |
1230 | ||
9d412a43 AV |
1231 | struct dentry * |
1232 | mount_fs(struct file_system_type *type, int flags, const char *name, void *data) | |
1da177e4 | 1233 | { |
c96e41e9 | 1234 | struct dentry *root; |
9d412a43 | 1235 | struct super_block *sb; |
1da177e4 | 1236 | char *secdata = NULL; |
9d412a43 | 1237 | int error = -ENOMEM; |
8089352a | 1238 | |
e0007529 | 1239 | if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) { |
1da177e4 | 1240 | secdata = alloc_secdata(); |
454e2398 | 1241 | if (!secdata) |
9d412a43 | 1242 | goto out; |
1da177e4 | 1243 | |
e0007529 | 1244 | error = security_sb_copy_data(data, secdata); |
454e2398 | 1245 | if (error) |
1da177e4 | 1246 | goto out_free_secdata; |
1da177e4 LT |
1247 | } |
1248 | ||
1a102ff9 AV |
1249 | root = type->mount(type, flags, name, data); |
1250 | if (IS_ERR(root)) { | |
1251 | error = PTR_ERR(root); | |
1252 | goto out_free_secdata; | |
c96e41e9 | 1253 | } |
9d412a43 AV |
1254 | sb = root->d_sb; |
1255 | BUG_ON(!sb); | |
1256 | WARN_ON(!sb->s_bdi); | |
e462ec50 | 1257 | sb->s_flags |= SB_BORN; |
454e2398 | 1258 | |
9d412a43 | 1259 | error = security_sb_kern_mount(sb, flags, secdata); |
5129a469 JE |
1260 | if (error) |
1261 | goto out_sb; | |
454e2398 | 1262 | |
42cb56ae JL |
1263 | /* |
1264 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE | |
1265 | * but s_maxbytes was an unsigned long long for many releases. Throw | |
1266 | * this warning for a little while to try and catch filesystems that | |
4358b567 | 1267 | * violate this rule. |
42cb56ae | 1268 | */ |
9d412a43 AV |
1269 | WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " |
1270 | "negative value (%lld)\n", type->name, sb->s_maxbytes); | |
42cb56ae | 1271 | |
9d412a43 | 1272 | up_write(&sb->s_umount); |
8680e22f | 1273 | free_secdata(secdata); |
9d412a43 | 1274 | return root; |
1da177e4 | 1275 | out_sb: |
9d412a43 AV |
1276 | dput(root); |
1277 | deactivate_locked_super(sb); | |
1da177e4 LT |
1278 | out_free_secdata: |
1279 | free_secdata(secdata); | |
1da177e4 | 1280 | out: |
454e2398 | 1281 | return ERR_PTR(error); |
1da177e4 LT |
1282 | } |
1283 | ||
fca39346 JK |
1284 | /* |
1285 | * Setup private BDI for given superblock. It gets automatically cleaned up | |
1286 | * in generic_shutdown_super(). | |
1287 | */ | |
1288 | int super_setup_bdi_name(struct super_block *sb, char *fmt, ...) | |
1289 | { | |
1290 | struct backing_dev_info *bdi; | |
1291 | int err; | |
1292 | va_list args; | |
1293 | ||
1294 | bdi = bdi_alloc(GFP_KERNEL); | |
1295 | if (!bdi) | |
1296 | return -ENOMEM; | |
1297 | ||
1298 | bdi->name = sb->s_type->name; | |
1299 | ||
1300 | va_start(args, fmt); | |
7c4cc300 | 1301 | err = bdi_register_va(bdi, fmt, args); |
fca39346 JK |
1302 | va_end(args); |
1303 | if (err) { | |
1304 | bdi_put(bdi); | |
1305 | return err; | |
1306 | } | |
1307 | WARN_ON(sb->s_bdi != &noop_backing_dev_info); | |
1308 | sb->s_bdi = bdi; | |
fca39346 JK |
1309 | |
1310 | return 0; | |
1311 | } | |
1312 | EXPORT_SYMBOL(super_setup_bdi_name); | |
1313 | ||
1314 | /* | |
1315 | * Setup private BDI for given superblock. I gets automatically cleaned up | |
1316 | * in generic_shutdown_super(). | |
1317 | */ | |
1318 | int super_setup_bdi(struct super_block *sb) | |
1319 | { | |
1320 | static atomic_long_t bdi_seq = ATOMIC_LONG_INIT(0); | |
1321 | ||
1322 | return super_setup_bdi_name(sb, "%.28s-%ld", sb->s_type->name, | |
1323 | atomic_long_inc_return(&bdi_seq)); | |
1324 | } | |
1325 | EXPORT_SYMBOL(super_setup_bdi); | |
1326 | ||
5accdf82 JK |
1327 | /* |
1328 | * This is an internal function, please use sb_end_{write,pagefault,intwrite} | |
1329 | * instead. | |
1330 | */ | |
1331 | void __sb_end_write(struct super_block *sb, int level) | |
1332 | { | |
8129ed29 | 1333 | percpu_up_read(sb->s_writers.rw_sem + level-1); |
5accdf82 JK |
1334 | } |
1335 | EXPORT_SYMBOL(__sb_end_write); | |
1336 | ||
f4b554af ON |
1337 | /* |
1338 | * This is an internal function, please use sb_start_{write,pagefault,intwrite} | |
1339 | * instead. | |
1340 | */ | |
1341 | int __sb_start_write(struct super_block *sb, int level, bool wait) | |
1342 | { | |
1343 | bool force_trylock = false; | |
8129ed29 | 1344 | int ret = 1; |
f4b554af ON |
1345 | |
1346 | #ifdef CONFIG_LOCKDEP | |
1347 | /* | |
1348 | * We want lockdep to tell us about possible deadlocks with freezing | |
1349 | * but it's it bit tricky to properly instrument it. Getting a freeze | |
1350 | * protection works as getting a read lock but there are subtle | |
1351 | * problems. XFS for example gets freeze protection on internal level | |
1352 | * twice in some cases, which is OK only because we already hold a | |
1353 | * freeze protection also on higher level. Due to these cases we have | |
1354 | * to use wait == F (trylock mode) which must not fail. | |
1355 | */ | |
1356 | if (wait) { | |
1357 | int i; | |
1358 | ||
1359 | for (i = 0; i < level - 1; i++) | |
8129ed29 | 1360 | if (percpu_rwsem_is_held(sb->s_writers.rw_sem + i)) { |
f4b554af ON |
1361 | force_trylock = true; |
1362 | break; | |
1363 | } | |
1364 | } | |
1365 | #endif | |
8129ed29 ON |
1366 | if (wait && !force_trylock) |
1367 | percpu_down_read(sb->s_writers.rw_sem + level-1); | |
1368 | else | |
1369 | ret = percpu_down_read_trylock(sb->s_writers.rw_sem + level-1); | |
1370 | ||
22224a17 | 1371 | WARN_ON(force_trylock && !ret); |
f4b554af ON |
1372 | return ret; |
1373 | } | |
5accdf82 JK |
1374 | EXPORT_SYMBOL(__sb_start_write); |
1375 | ||
1376 | /** | |
1377 | * sb_wait_write - wait until all writers to given file system finish | |
1378 | * @sb: the super for which we wait | |
1379 | * @level: type of writers we wait for (normal vs page fault) | |
1380 | * | |
1381 | * This function waits until there are no writers of given type to given file | |
8129ed29 | 1382 | * system. |
5accdf82 JK |
1383 | */ |
1384 | static void sb_wait_write(struct super_block *sb, int level) | |
1385 | { | |
8129ed29 | 1386 | percpu_down_write(sb->s_writers.rw_sem + level-1); |
8129ed29 | 1387 | } |
5accdf82 | 1388 | |
f1a96220 ON |
1389 | /* |
1390 | * We are going to return to userspace and forget about these locks, the | |
1391 | * ownership goes to the caller of thaw_super() which does unlock(). | |
1392 | */ | |
1393 | static void lockdep_sb_freeze_release(struct super_block *sb) | |
1394 | { | |
1395 | int level; | |
1396 | ||
1397 | for (level = SB_FREEZE_LEVELS - 1; level >= 0; level--) | |
1398 | percpu_rwsem_release(sb->s_writers.rw_sem + level, 0, _THIS_IP_); | |
1399 | } | |
1400 | ||
1401 | /* | |
1402 | * Tell lockdep we are holding these locks before we call ->unfreeze_fs(sb). | |
1403 | */ | |
1404 | static void lockdep_sb_freeze_acquire(struct super_block *sb) | |
8129ed29 ON |
1405 | { |
1406 | int level; | |
5accdf82 | 1407 | |
8129ed29 ON |
1408 | for (level = 0; level < SB_FREEZE_LEVELS; ++level) |
1409 | percpu_rwsem_acquire(sb->s_writers.rw_sem + level, 0, _THIS_IP_); | |
f1a96220 ON |
1410 | } |
1411 | ||
1412 | static void sb_freeze_unlock(struct super_block *sb) | |
1413 | { | |
1414 | int level; | |
5accdf82 | 1415 | |
8129ed29 ON |
1416 | for (level = SB_FREEZE_LEVELS - 1; level >= 0; level--) |
1417 | percpu_up_write(sb->s_writers.rw_sem + level); | |
5accdf82 JK |
1418 | } |
1419 | ||
18e9e510 | 1420 | /** |
7000d3c4 RD |
1421 | * freeze_super - lock the filesystem and force it into a consistent state |
1422 | * @sb: the super to lock | |
18e9e510 JB |
1423 | * |
1424 | * Syncs the super to make sure the filesystem is consistent and calls the fs's | |
1425 | * freeze_fs. Subsequent calls to this without first thawing the fs will return | |
1426 | * -EBUSY. | |
5accdf82 JK |
1427 | * |
1428 | * During this function, sb->s_writers.frozen goes through these values: | |
1429 | * | |
1430 | * SB_UNFROZEN: File system is normal, all writes progress as usual. | |
1431 | * | |
1432 | * SB_FREEZE_WRITE: The file system is in the process of being frozen. New | |
1433 | * writes should be blocked, though page faults are still allowed. We wait for | |
1434 | * all writes to complete and then proceed to the next stage. | |
1435 | * | |
1436 | * SB_FREEZE_PAGEFAULT: Freezing continues. Now also page faults are blocked | |
1437 | * but internal fs threads can still modify the filesystem (although they | |
1438 | * should not dirty new pages or inodes), writeback can run etc. After waiting | |
1439 | * for all running page faults we sync the filesystem which will clean all | |
1440 | * dirty pages and inodes (no new dirty pages or inodes can be created when | |
1441 | * sync is running). | |
1442 | * | |
1443 | * SB_FREEZE_FS: The file system is frozen. Now all internal sources of fs | |
1444 | * modification are blocked (e.g. XFS preallocation truncation on inode | |
1445 | * reclaim). This is usually implemented by blocking new transactions for | |
1446 | * filesystems that have them and need this additional guard. After all | |
1447 | * internal writers are finished we call ->freeze_fs() to finish filesystem | |
1448 | * freezing. Then we transition to SB_FREEZE_COMPLETE state. This state is | |
1449 | * mostly auxiliary for filesystems to verify they do not modify frozen fs. | |
1450 | * | |
1451 | * sb->s_writers.frozen is protected by sb->s_umount. | |
18e9e510 JB |
1452 | */ |
1453 | int freeze_super(struct super_block *sb) | |
1454 | { | |
1455 | int ret; | |
1456 | ||
1457 | atomic_inc(&sb->s_active); | |
1458 | down_write(&sb->s_umount); | |
5accdf82 | 1459 | if (sb->s_writers.frozen != SB_UNFROZEN) { |
18e9e510 JB |
1460 | deactivate_locked_super(sb); |
1461 | return -EBUSY; | |
1462 | } | |
1463 | ||
e462ec50 | 1464 | if (!(sb->s_flags & SB_BORN)) { |
dabe0dc1 AV |
1465 | up_write(&sb->s_umount); |
1466 | return 0; /* sic - it's "nothing to do" */ | |
1467 | } | |
1468 | ||
bc98a42c | 1469 | if (sb_rdonly(sb)) { |
5accdf82 JK |
1470 | /* Nothing to do really... */ |
1471 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
18e9e510 JB |
1472 | up_write(&sb->s_umount); |
1473 | return 0; | |
1474 | } | |
1475 | ||
5accdf82 | 1476 | sb->s_writers.frozen = SB_FREEZE_WRITE; |
5accdf82 JK |
1477 | /* Release s_umount to preserve sb_start_write -> s_umount ordering */ |
1478 | up_write(&sb->s_umount); | |
5accdf82 | 1479 | sb_wait_write(sb, SB_FREEZE_WRITE); |
8129ed29 | 1480 | down_write(&sb->s_umount); |
5accdf82 JK |
1481 | |
1482 | /* Now we go and block page faults... */ | |
5accdf82 | 1483 | sb->s_writers.frozen = SB_FREEZE_PAGEFAULT; |
5accdf82 JK |
1484 | sb_wait_write(sb, SB_FREEZE_PAGEFAULT); |
1485 | ||
1486 | /* All writers are done so after syncing there won't be dirty data */ | |
18e9e510 JB |
1487 | sync_filesystem(sb); |
1488 | ||
5accdf82 JK |
1489 | /* Now wait for internal filesystem counter */ |
1490 | sb->s_writers.frozen = SB_FREEZE_FS; | |
5accdf82 | 1491 | sb_wait_write(sb, SB_FREEZE_FS); |
18e9e510 | 1492 | |
18e9e510 JB |
1493 | if (sb->s_op->freeze_fs) { |
1494 | ret = sb->s_op->freeze_fs(sb); | |
1495 | if (ret) { | |
1496 | printk(KERN_ERR | |
1497 | "VFS:Filesystem freeze failed\n"); | |
5accdf82 | 1498 | sb->s_writers.frozen = SB_UNFROZEN; |
8129ed29 | 1499 | sb_freeze_unlock(sb); |
5accdf82 | 1500 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 JB |
1501 | deactivate_locked_super(sb); |
1502 | return ret; | |
1503 | } | |
1504 | } | |
5accdf82 | 1505 | /* |
89f39af1 ON |
1506 | * For debugging purposes so that fs can warn if it sees write activity |
1507 | * when frozen is set to SB_FREEZE_COMPLETE, and for thaw_super(). | |
5accdf82 JK |
1508 | */ |
1509 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
f1a96220 | 1510 | lockdep_sb_freeze_release(sb); |
18e9e510 JB |
1511 | up_write(&sb->s_umount); |
1512 | return 0; | |
1513 | } | |
1514 | EXPORT_SYMBOL(freeze_super); | |
1515 | ||
1516 | /** | |
1517 | * thaw_super -- unlock filesystem | |
1518 | * @sb: the super to thaw | |
1519 | * | |
1520 | * Unlocks the filesystem and marks it writeable again after freeze_super(). | |
1521 | */ | |
1522 | int thaw_super(struct super_block *sb) | |
1523 | { | |
1524 | int error; | |
1525 | ||
1526 | down_write(&sb->s_umount); | |
89f39af1 | 1527 | if (sb->s_writers.frozen != SB_FREEZE_COMPLETE) { |
18e9e510 JB |
1528 | up_write(&sb->s_umount); |
1529 | return -EINVAL; | |
1530 | } | |
1531 | ||
bc98a42c | 1532 | if (sb_rdonly(sb)) { |
8129ed29 | 1533 | sb->s_writers.frozen = SB_UNFROZEN; |
18e9e510 | 1534 | goto out; |
8129ed29 | 1535 | } |
18e9e510 | 1536 | |
f1a96220 ON |
1537 | lockdep_sb_freeze_acquire(sb); |
1538 | ||
18e9e510 JB |
1539 | if (sb->s_op->unfreeze_fs) { |
1540 | error = sb->s_op->unfreeze_fs(sb); | |
1541 | if (error) { | |
1542 | printk(KERN_ERR | |
1543 | "VFS:Filesystem thaw failed\n"); | |
f1a96220 | 1544 | lockdep_sb_freeze_release(sb); |
18e9e510 JB |
1545 | up_write(&sb->s_umount); |
1546 | return error; | |
1547 | } | |
1548 | } | |
1549 | ||
5accdf82 | 1550 | sb->s_writers.frozen = SB_UNFROZEN; |
8129ed29 ON |
1551 | sb_freeze_unlock(sb); |
1552 | out: | |
5accdf82 | 1553 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 | 1554 | deactivate_locked_super(sb); |
18e9e510 JB |
1555 | return 0; |
1556 | } | |
1557 | EXPORT_SYMBOL(thaw_super); |