]>
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> |
9bc61ab1 | 38 | #include <linux/fs_context.h> |
e262e32d | 39 | #include <uapi/linux/mount.h> |
6d59e7f5 | 40 | #include "internal.h" |
1da177e4 | 41 | |
08fdc8a0 | 42 | static int thaw_super_locked(struct super_block *sb); |
1da177e4 | 43 | |
15d0f5ea AV |
44 | static LIST_HEAD(super_blocks); |
45 | static DEFINE_SPINLOCK(sb_lock); | |
1da177e4 | 46 | |
5accdf82 JK |
47 | static char *sb_writers_name[SB_FREEZE_LEVELS] = { |
48 | "sb_writers", | |
49 | "sb_pagefaults", | |
50 | "sb_internal", | |
51 | }; | |
52 | ||
b0d40c92 DC |
53 | /* |
54 | * One thing we have to be careful of with a per-sb shrinker is that we don't | |
55 | * drop the last active reference to the superblock from within the shrinker. | |
56 | * If that happens we could trigger unregistering the shrinker from within the | |
57 | * shrinker path and that leads to deadlock on the shrinker_rwsem. Hence we | |
58 | * take a passive reference to the superblock to avoid this from occurring. | |
59 | */ | |
0a234c6d DC |
60 | static unsigned long super_cache_scan(struct shrinker *shrink, |
61 | struct shrink_control *sc) | |
b0d40c92 DC |
62 | { |
63 | struct super_block *sb; | |
0a234c6d DC |
64 | long fs_objects = 0; |
65 | long total_objects; | |
66 | long freed = 0; | |
67 | long dentries; | |
68 | long inodes; | |
b0d40c92 DC |
69 | |
70 | sb = container_of(shrink, struct super_block, s_shrink); | |
71 | ||
72 | /* | |
73 | * Deadlock avoidance. We may hold various FS locks, and we don't want | |
74 | * to recurse into the FS that called us in clear_inode() and friends.. | |
75 | */ | |
0a234c6d DC |
76 | if (!(sc->gfp_mask & __GFP_FS)) |
77 | return SHRINK_STOP; | |
b0d40c92 | 78 | |
eb6ef3df | 79 | if (!trylock_super(sb)) |
0a234c6d | 80 | return SHRINK_STOP; |
b0d40c92 | 81 | |
d0407903 | 82 | if (sb->s_op->nr_cached_objects) |
4101b624 | 83 | fs_objects = sb->s_op->nr_cached_objects(sb, sc); |
0e1fdafd | 84 | |
503c358c VD |
85 | inodes = list_lru_shrink_count(&sb->s_inode_lru, sc); |
86 | dentries = list_lru_shrink_count(&sb->s_dentry_lru, sc); | |
f6041567 | 87 | total_objects = dentries + inodes + fs_objects + 1; |
475d0db7 TH |
88 | if (!total_objects) |
89 | total_objects = 1; | |
0e1fdafd | 90 | |
0a234c6d | 91 | /* proportion the scan between the caches */ |
f6041567 | 92 | dentries = mult_frac(sc->nr_to_scan, dentries, total_objects); |
bc3b14cb | 93 | inodes = mult_frac(sc->nr_to_scan, inodes, total_objects); |
503c358c | 94 | fs_objects = mult_frac(sc->nr_to_scan, fs_objects, total_objects); |
b0d40c92 | 95 | |
0a234c6d DC |
96 | /* |
97 | * prune the dcache first as the icache is pinned by it, then | |
98 | * prune the icache, followed by the filesystem specific caches | |
49e7e7ff VD |
99 | * |
100 | * Ensure that we always scan at least one object - memcg kmem | |
101 | * accounting uses this to fully empty the caches. | |
0a234c6d | 102 | */ |
49e7e7ff | 103 | sc->nr_to_scan = dentries + 1; |
503c358c | 104 | freed = prune_dcache_sb(sb, sc); |
49e7e7ff | 105 | sc->nr_to_scan = inodes + 1; |
503c358c | 106 | freed += prune_icache_sb(sb, sc); |
0a234c6d DC |
107 | |
108 | if (fs_objects) { | |
49e7e7ff | 109 | sc->nr_to_scan = fs_objects + 1; |
4101b624 | 110 | freed += sb->s_op->free_cached_objects(sb, sc); |
b0d40c92 DC |
111 | } |
112 | ||
eb6ef3df | 113 | up_read(&sb->s_umount); |
0a234c6d DC |
114 | return freed; |
115 | } | |
116 | ||
117 | static unsigned long super_cache_count(struct shrinker *shrink, | |
118 | struct shrink_control *sc) | |
119 | { | |
120 | struct super_block *sb; | |
121 | long total_objects = 0; | |
122 | ||
123 | sb = container_of(shrink, struct super_block, s_shrink); | |
124 | ||
d23da150 | 125 | /* |
79f546a6 DC |
126 | * We don't call trylock_super() here as it is a scalability bottleneck, |
127 | * so we're exposed to partial setup state. The shrinker rwsem does not | |
128 | * protect filesystem operations backing list_lru_shrink_count() or | |
129 | * s_op->nr_cached_objects(). Counts can change between | |
130 | * super_cache_count and super_cache_scan, so we really don't need locks | |
131 | * here. | |
132 | * | |
133 | * However, if we are currently mounting the superblock, the underlying | |
134 | * filesystem might be in a state of partial construction and hence it | |
135 | * is dangerous to access it. trylock_super() uses a SB_BORN check to | |
136 | * avoid this situation, so do the same here. The memory barrier is | |
137 | * matched with the one in mount_fs() as we don't hold locks here. | |
d23da150 | 138 | */ |
79f546a6 DC |
139 | if (!(sb->s_flags & SB_BORN)) |
140 | return 0; | |
141 | smp_rmb(); | |
142 | ||
0a234c6d | 143 | if (sb->s_op && sb->s_op->nr_cached_objects) |
4101b624 | 144 | total_objects = sb->s_op->nr_cached_objects(sb, sc); |
0a234c6d | 145 | |
503c358c VD |
146 | total_objects += list_lru_shrink_count(&sb->s_dentry_lru, sc); |
147 | total_objects += list_lru_shrink_count(&sb->s_inode_lru, sc); | |
0a234c6d | 148 | |
9b996468 KT |
149 | if (!total_objects) |
150 | return SHRINK_EMPTY; | |
151 | ||
55f841ce | 152 | total_objects = vfs_pressure_ratio(total_objects); |
0e1fdafd | 153 | return total_objects; |
b0d40c92 DC |
154 | } |
155 | ||
853b39a7 ON |
156 | static void destroy_super_work(struct work_struct *work) |
157 | { | |
158 | struct super_block *s = container_of(work, struct super_block, | |
159 | destroy_work); | |
160 | int i; | |
161 | ||
162 | for (i = 0; i < SB_FREEZE_LEVELS; i++) | |
8129ed29 | 163 | percpu_free_rwsem(&s->s_writers.rw_sem[i]); |
853b39a7 ON |
164 | kfree(s); |
165 | } | |
166 | ||
167 | static void destroy_super_rcu(struct rcu_head *head) | |
168 | { | |
169 | struct super_block *s = container_of(head, struct super_block, rcu); | |
170 | INIT_WORK(&s->destroy_work, destroy_super_work); | |
171 | schedule_work(&s->destroy_work); | |
172 | } | |
173 | ||
0200894d AV |
174 | /* Free a superblock that has never been seen by anyone */ |
175 | static void destroy_unused_super(struct super_block *s) | |
5accdf82 | 176 | { |
0200894d AV |
177 | if (!s) |
178 | return; | |
179 | up_write(&s->s_umount); | |
7eb5e882 AV |
180 | list_lru_destroy(&s->s_dentry_lru); |
181 | list_lru_destroy(&s->s_inode_lru); | |
7eb5e882 | 182 | security_sb_free(s); |
6e4eab57 | 183 | put_user_ns(s->s_user_ns); |
7eb5e882 | 184 | kfree(s->s_subtype); |
8e04944f | 185 | free_prealloced_shrinker(&s->s_shrink); |
0200894d AV |
186 | /* no delays needed */ |
187 | destroy_super_work(&s->destroy_work); | |
5accdf82 JK |
188 | } |
189 | ||
1da177e4 LT |
190 | /** |
191 | * alloc_super - create new superblock | |
fe2bbc48 | 192 | * @type: filesystem type superblock should belong to |
9249e17f | 193 | * @flags: the mount flags |
6e4eab57 | 194 | * @user_ns: User namespace for the super_block |
1da177e4 LT |
195 | * |
196 | * Allocates and initializes a new &struct super_block. alloc_super() | |
197 | * returns a pointer new superblock or %NULL if allocation had failed. | |
198 | */ | |
6e4eab57 EB |
199 | static struct super_block *alloc_super(struct file_system_type *type, int flags, |
200 | struct user_namespace *user_ns) | |
1da177e4 | 201 | { |
11b0b5ab | 202 | struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER); |
b87221de | 203 | static const struct super_operations default_op; |
7eb5e882 AV |
204 | int i; |
205 | ||
206 | if (!s) | |
207 | return NULL; | |
1da177e4 | 208 | |
b5bd856a | 209 | INIT_LIST_HEAD(&s->s_mounts); |
6e4eab57 | 210 | s->s_user_ns = get_user_ns(user_ns); |
ca0168e8 AV |
211 | init_rwsem(&s->s_umount); |
212 | lockdep_set_class(&s->s_umount, &type->s_umount_key); | |
213 | /* | |
214 | * sget() can have s_umount recursion. | |
215 | * | |
216 | * When it cannot find a suitable sb, it allocates a new | |
217 | * one (this one), and tries again to find a suitable old | |
218 | * one. | |
219 | * | |
220 | * In case that succeeds, it will acquire the s_umount | |
221 | * lock of the old one. Since these are clearly distrinct | |
222 | * locks, and this object isn't exposed yet, there's no | |
223 | * risk of deadlocks. | |
224 | * | |
225 | * Annotate this by putting this lock in a different | |
226 | * subclass. | |
227 | */ | |
228 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); | |
b5bd856a | 229 | |
7eb5e882 AV |
230 | if (security_sb_alloc(s)) |
231 | goto fail; | |
7b7a8665 | 232 | |
7eb5e882 | 233 | for (i = 0; i < SB_FREEZE_LEVELS; i++) { |
8129ed29 ON |
234 | if (__percpu_init_rwsem(&s->s_writers.rw_sem[i], |
235 | sb_writers_name[i], | |
236 | &type->s_writers_key[i])) | |
7eb5e882 | 237 | goto fail; |
1da177e4 | 238 | } |
7eb5e882 | 239 | init_waitqueue_head(&s->s_writers.wait_unfrozen); |
df0ce26c | 240 | s->s_bdi = &noop_backing_dev_info; |
7eb5e882 | 241 | s->s_flags = flags; |
cc50a07a | 242 | if (s->s_user_ns != &init_user_ns) |
67690f93 | 243 | s->s_iflags |= SB_I_NODEV; |
7eb5e882 | 244 | INIT_HLIST_NODE(&s->s_instances); |
f1ee6162 | 245 | INIT_HLIST_BL_HEAD(&s->s_roots); |
e97fedb9 | 246 | mutex_init(&s->s_sync_lock); |
7eb5e882 | 247 | INIT_LIST_HEAD(&s->s_inodes); |
74278da9 | 248 | spin_lock_init(&s->s_inode_list_lock); |
6c60d2b5 DC |
249 | INIT_LIST_HEAD(&s->s_inodes_wb); |
250 | spin_lock_init(&s->s_inode_wblist_lock); | |
7eb5e882 | 251 | |
7eb5e882 AV |
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; |
8e04944f TH |
267 | if (prealloc_shrinker(&s->s_shrink)) |
268 | goto fail; | |
c92e8e10 | 269 | if (list_lru_init_memcg(&s->s_dentry_lru, &s->s_shrink)) |
2b3648a6 | 270 | goto fail; |
c92e8e10 | 271 | if (list_lru_init_memcg(&s->s_inode_lru, &s->s_shrink)) |
2b3648a6 | 272 | goto fail; |
1da177e4 | 273 | return s; |
5ca302c8 | 274 | |
7eb5e882 | 275 | fail: |
0200894d | 276 | destroy_unused_super(s); |
7eb5e882 | 277 | return NULL; |
1da177e4 LT |
278 | } |
279 | ||
280 | /* Superblock refcounting */ | |
281 | ||
282 | /* | |
35cf7ba0 | 283 | * Drop a superblock's refcount. The caller must hold sb_lock. |
1da177e4 | 284 | */ |
c645b930 | 285 | static void __put_super(struct super_block *s) |
1da177e4 | 286 | { |
c645b930 AV |
287 | if (!--s->s_count) { |
288 | list_del_init(&s->s_list); | |
289 | WARN_ON(s->s_dentry_lru.node); | |
290 | WARN_ON(s->s_inode_lru.node); | |
291 | WARN_ON(!list_empty(&s->s_mounts)); | |
292 | security_sb_free(s); | |
293 | put_user_ns(s->s_user_ns); | |
294 | kfree(s->s_subtype); | |
295 | call_rcu(&s->rcu, destroy_super_rcu); | |
1da177e4 | 296 | } |
1da177e4 LT |
297 | } |
298 | ||
299 | /** | |
300 | * put_super - drop a temporary reference to superblock | |
301 | * @sb: superblock in question | |
302 | * | |
303 | * Drops a temporary reference, frees superblock if there's no | |
304 | * references left. | |
305 | */ | |
f47ec3f2 | 306 | static void put_super(struct super_block *sb) |
1da177e4 LT |
307 | { |
308 | spin_lock(&sb_lock); | |
309 | __put_super(sb); | |
310 | spin_unlock(&sb_lock); | |
311 | } | |
312 | ||
313 | ||
314 | /** | |
1712ac8f | 315 | * deactivate_locked_super - drop an active reference to superblock |
1da177e4 LT |
316 | * @s: superblock to deactivate |
317 | * | |
bd7ced98 | 318 | * Drops an active reference to superblock, converting it into a temporary |
1712ac8f | 319 | * one if there is no other active references left. In that case we |
1da177e4 LT |
320 | * tell fs driver to shut it down and drop the temporary reference we |
321 | * had just acquired. | |
1712ac8f AV |
322 | * |
323 | * Caller holds exclusive lock on superblock; that lock is released. | |
1da177e4 | 324 | */ |
1712ac8f | 325 | void deactivate_locked_super(struct super_block *s) |
1da177e4 LT |
326 | { |
327 | struct file_system_type *fs = s->s_type; | |
b20bd1a5 | 328 | if (atomic_dec_and_test(&s->s_active)) { |
3167760f | 329 | cleancache_invalidate_fs(s); |
b0d40c92 | 330 | unregister_shrinker(&s->s_shrink); |
28f2cd4f | 331 | fs->kill_sb(s); |
f5e1dd34 | 332 | |
c0a5b560 VD |
333 | /* |
334 | * Since list_lru_destroy() may sleep, we cannot call it from | |
335 | * put_super(), where we hold the sb_lock. Therefore we destroy | |
336 | * the lru lists right now. | |
337 | */ | |
338 | list_lru_destroy(&s->s_dentry_lru); | |
339 | list_lru_destroy(&s->s_inode_lru); | |
340 | ||
1da177e4 LT |
341 | put_filesystem(fs); |
342 | put_super(s); | |
1712ac8f AV |
343 | } else { |
344 | up_write(&s->s_umount); | |
1da177e4 LT |
345 | } |
346 | } | |
347 | ||
1712ac8f | 348 | EXPORT_SYMBOL(deactivate_locked_super); |
1da177e4 | 349 | |
74dbbdd7 | 350 | /** |
1712ac8f | 351 | * deactivate_super - drop an active reference to superblock |
74dbbdd7 AV |
352 | * @s: superblock to deactivate |
353 | * | |
1712ac8f AV |
354 | * Variant of deactivate_locked_super(), except that superblock is *not* |
355 | * locked by caller. If we are going to drop the final active reference, | |
356 | * lock will be acquired prior to that. | |
74dbbdd7 | 357 | */ |
1712ac8f | 358 | void deactivate_super(struct super_block *s) |
74dbbdd7 | 359 | { |
1712ac8f AV |
360 | if (!atomic_add_unless(&s->s_active, -1, 1)) { |
361 | down_write(&s->s_umount); | |
362 | deactivate_locked_super(s); | |
74dbbdd7 AV |
363 | } |
364 | } | |
365 | ||
1712ac8f | 366 | EXPORT_SYMBOL(deactivate_super); |
74dbbdd7 | 367 | |
1da177e4 LT |
368 | /** |
369 | * grab_super - acquire an active reference | |
370 | * @s: reference we are trying to make active | |
371 | * | |
372 | * Tries to acquire an active reference. grab_super() is used when we | |
373 | * had just found a superblock in super_blocks or fs_type->fs_supers | |
374 | * and want to turn it into a full-blown active reference. grab_super() | |
375 | * is called with sb_lock held and drops it. Returns 1 in case of | |
376 | * success, 0 if we had failed (superblock contents was already dead or | |
acfec9a5 AV |
377 | * dying when grab_super() had been called). Note that this is only |
378 | * called for superblocks not in rundown mode (== ones still on ->fs_supers | |
379 | * of their type), so increment of ->s_count is OK here. | |
1da177e4 | 380 | */ |
9c4dbee7 | 381 | static int grab_super(struct super_block *s) __releases(sb_lock) |
1da177e4 LT |
382 | { |
383 | s->s_count++; | |
384 | spin_unlock(&sb_lock); | |
385 | down_write(&s->s_umount); | |
e462ec50 | 386 | if ((s->s_flags & SB_BORN) && atomic_inc_not_zero(&s->s_active)) { |
acfec9a5 AV |
387 | put_super(s); |
388 | return 1; | |
389 | } | |
1da177e4 LT |
390 | up_write(&s->s_umount); |
391 | put_super(s); | |
1da177e4 LT |
392 | return 0; |
393 | } | |
394 | ||
12ad3ab6 | 395 | /* |
eb6ef3df | 396 | * trylock_super - try to grab ->s_umount shared |
331cbdee | 397 | * @sb: reference we are trying to grab |
12ad3ab6 | 398 | * |
eb6ef3df | 399 | * Try to prevent fs shutdown. This is used in places where we |
12ad3ab6 | 400 | * cannot take an active reference but we need to ensure that the |
eb6ef3df KK |
401 | * filesystem is not shut down while we are working on it. It returns |
402 | * false if we cannot acquire s_umount or if we lose the race and | |
403 | * filesystem already got into shutdown, and returns true with the s_umount | |
404 | * lock held in read mode in case of success. On successful return, | |
405 | * the caller must drop the s_umount lock when done. | |
406 | * | |
407 | * Note that unlike get_super() et.al. this one does *not* bump ->s_count. | |
408 | * The reason why it's safe is that we are OK with doing trylock instead | |
409 | * of down_read(). There's a couple of places that are OK with that, but | |
410 | * it's very much not a general-purpose interface. | |
12ad3ab6 | 411 | */ |
eb6ef3df | 412 | bool trylock_super(struct super_block *sb) |
12ad3ab6 | 413 | { |
12ad3ab6 | 414 | if (down_read_trylock(&sb->s_umount)) { |
eb6ef3df | 415 | if (!hlist_unhashed(&sb->s_instances) && |
e462ec50 | 416 | sb->s_root && (sb->s_flags & SB_BORN)) |
12ad3ab6 DC |
417 | return true; |
418 | up_read(&sb->s_umount); | |
419 | } | |
420 | ||
12ad3ab6 DC |
421 | return false; |
422 | } | |
423 | ||
1da177e4 LT |
424 | /** |
425 | * generic_shutdown_super - common helper for ->kill_sb() | |
426 | * @sb: superblock to kill | |
427 | * | |
428 | * generic_shutdown_super() does all fs-independent work on superblock | |
429 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects | |
430 | * that need destruction out of superblock, call generic_shutdown_super() | |
431 | * and release aforementioned objects. Note: dentries and inodes _are_ | |
432 | * taken care of and do not need specific handling. | |
c636ebdb DH |
433 | * |
434 | * Upon calling this function, the filesystem may no longer alter or | |
435 | * rearrange the set of dentries belonging to this super_block, nor may it | |
436 | * change the attachments of dentries to inodes. | |
1da177e4 LT |
437 | */ |
438 | void generic_shutdown_super(struct super_block *sb) | |
439 | { | |
ee9b6d61 | 440 | const struct super_operations *sop = sb->s_op; |
1da177e4 | 441 | |
c636ebdb DH |
442 | if (sb->s_root) { |
443 | shrink_dcache_for_umount(sb); | |
60b0680f | 444 | sync_filesystem(sb); |
e462ec50 | 445 | sb->s_flags &= ~SB_ACTIVE; |
efaee192 | 446 | |
1e6cb723 | 447 | fsnotify_sb_delete(sb); |
a1a0e23e | 448 | cgroup_writeback_umount(); |
63997e98 AV |
449 | |
450 | evict_inodes(sb); | |
1da177e4 | 451 | |
7b7a8665 CH |
452 | if (sb->s_dio_done_wq) { |
453 | destroy_workqueue(sb->s_dio_done_wq); | |
454 | sb->s_dio_done_wq = NULL; | |
455 | } | |
456 | ||
1da177e4 LT |
457 | if (sop->put_super) |
458 | sop->put_super(sb); | |
459 | ||
63997e98 | 460 | if (!list_empty(&sb->s_inodes)) { |
7b4fe29e DJ |
461 | printk("VFS: Busy inodes after unmount of %s. " |
462 | "Self-destruct in 5 seconds. Have a nice day...\n", | |
463 | sb->s_id); | |
1da177e4 | 464 | } |
1da177e4 LT |
465 | } |
466 | spin_lock(&sb_lock); | |
467 | /* should be initialized for __put_super_and_need_restart() */ | |
a5166169 | 468 | hlist_del_init(&sb->s_instances); |
1da177e4 LT |
469 | spin_unlock(&sb_lock); |
470 | up_write(&sb->s_umount); | |
c1844d53 | 471 | if (sb->s_bdi != &noop_backing_dev_info) { |
fca39346 JK |
472 | bdi_put(sb->s_bdi); |
473 | sb->s_bdi = &noop_backing_dev_info; | |
fca39346 | 474 | } |
1da177e4 LT |
475 | } |
476 | ||
477 | EXPORT_SYMBOL(generic_shutdown_super); | |
478 | ||
20284ab7 | 479 | bool mount_capable(struct fs_context *fc) |
0ce0cf12 | 480 | { |
20284ab7 AV |
481 | struct user_namespace *user_ns = fc->global ? &init_user_ns |
482 | : fc->user_ns; | |
483 | ||
484 | if (!(fc->fs_type->fs_flags & FS_USERNS_MOUNT)) | |
0ce0cf12 AV |
485 | return capable(CAP_SYS_ADMIN); |
486 | else | |
20284ab7 | 487 | return ns_capable(user_ns, CAP_SYS_ADMIN); |
0ce0cf12 AV |
488 | } |
489 | ||
cb50b348 AV |
490 | /** |
491 | * sget_fc - Find or create a superblock | |
492 | * @fc: Filesystem context. | |
493 | * @test: Comparison callback | |
494 | * @set: Setup callback | |
495 | * | |
496 | * Find or create a superblock using the parameters stored in the filesystem | |
497 | * context and the two callback functions. | |
498 | * | |
499 | * If an extant superblock is matched, then that will be returned with an | |
500 | * elevated reference count that the caller must transfer or discard. | |
501 | * | |
502 | * If no match is made, a new superblock will be allocated and basic | |
503 | * initialisation will be performed (s_type, s_fs_info and s_id will be set and | |
504 | * the set() callback will be invoked), the superblock will be published and it | |
505 | * will be returned in a partially constructed state with SB_BORN and SB_ACTIVE | |
506 | * as yet unset. | |
507 | */ | |
508 | struct super_block *sget_fc(struct fs_context *fc, | |
509 | int (*test)(struct super_block *, struct fs_context *), | |
510 | int (*set)(struct super_block *, struct fs_context *)) | |
511 | { | |
512 | struct super_block *s = NULL; | |
513 | struct super_block *old; | |
514 | struct user_namespace *user_ns = fc->global ? &init_user_ns : fc->user_ns; | |
515 | int err; | |
516 | ||
cb50b348 AV |
517 | retry: |
518 | spin_lock(&sb_lock); | |
519 | if (test) { | |
520 | hlist_for_each_entry(old, &fc->fs_type->fs_supers, s_instances) { | |
521 | if (test(old, fc)) | |
522 | goto share_extant_sb; | |
523 | } | |
524 | } | |
525 | if (!s) { | |
526 | spin_unlock(&sb_lock); | |
527 | s = alloc_super(fc->fs_type, fc->sb_flags, user_ns); | |
528 | if (!s) | |
529 | return ERR_PTR(-ENOMEM); | |
530 | goto retry; | |
531 | } | |
532 | ||
533 | s->s_fs_info = fc->s_fs_info; | |
534 | err = set(s, fc); | |
535 | if (err) { | |
536 | s->s_fs_info = NULL; | |
537 | spin_unlock(&sb_lock); | |
538 | destroy_unused_super(s); | |
539 | return ERR_PTR(err); | |
540 | } | |
541 | fc->s_fs_info = NULL; | |
542 | s->s_type = fc->fs_type; | |
c80fa7c8 | 543 | s->s_iflags |= fc->s_iflags; |
cb50b348 AV |
544 | strlcpy(s->s_id, s->s_type->name, sizeof(s->s_id)); |
545 | list_add_tail(&s->s_list, &super_blocks); | |
546 | hlist_add_head(&s->s_instances, &s->s_type->fs_supers); | |
547 | spin_unlock(&sb_lock); | |
548 | get_filesystem(s->s_type); | |
549 | register_shrinker_prepared(&s->s_shrink); | |
550 | return s; | |
551 | ||
552 | share_extant_sb: | |
553 | if (user_ns != old->s_user_ns) { | |
554 | spin_unlock(&sb_lock); | |
555 | destroy_unused_super(s); | |
556 | return ERR_PTR(-EBUSY); | |
557 | } | |
558 | if (!grab_super(old)) | |
559 | goto retry; | |
560 | destroy_unused_super(s); | |
561 | return old; | |
562 | } | |
563 | EXPORT_SYMBOL(sget_fc); | |
564 | ||
1da177e4 | 565 | /** |
023d066a DH |
566 | * sget - find or create a superblock |
567 | * @type: filesystem type superblock should belong to | |
568 | * @test: comparison callback | |
569 | * @set: setup callback | |
570 | * @flags: mount flags | |
571 | * @data: argument to each of them | |
1da177e4 | 572 | */ |
023d066a | 573 | struct super_block *sget(struct file_system_type *type, |
1da177e4 LT |
574 | int (*test)(struct super_block *,void *), |
575 | int (*set)(struct super_block *,void *), | |
023d066a | 576 | int flags, |
1da177e4 LT |
577 | void *data) |
578 | { | |
023d066a | 579 | struct user_namespace *user_ns = current_user_ns(); |
1da177e4 | 580 | struct super_block *s = NULL; |
d4730127 | 581 | struct super_block *old; |
1da177e4 LT |
582 | int err; |
583 | ||
023d066a DH |
584 | /* We don't yet pass the user namespace of the parent |
585 | * mount through to here so always use &init_user_ns | |
586 | * until that changes. | |
587 | */ | |
588 | if (flags & SB_SUBMOUNT) | |
589 | user_ns = &init_user_ns; | |
590 | ||
1da177e4 LT |
591 | retry: |
592 | spin_lock(&sb_lock); | |
d4730127 | 593 | if (test) { |
b67bfe0d | 594 | hlist_for_each_entry(old, &type->fs_supers, s_instances) { |
d4730127 MK |
595 | if (!test(old, data)) |
596 | continue; | |
6e4eab57 EB |
597 | if (user_ns != old->s_user_ns) { |
598 | spin_unlock(&sb_lock); | |
0200894d | 599 | destroy_unused_super(s); |
6e4eab57 EB |
600 | return ERR_PTR(-EBUSY); |
601 | } | |
d4730127 MK |
602 | if (!grab_super(old)) |
603 | goto retry; | |
0200894d | 604 | destroy_unused_super(s); |
d4730127 MK |
605 | return old; |
606 | } | |
1da177e4 LT |
607 | } |
608 | if (!s) { | |
609 | spin_unlock(&sb_lock); | |
e462ec50 | 610 | s = alloc_super(type, (flags & ~SB_SUBMOUNT), user_ns); |
1da177e4 LT |
611 | if (!s) |
612 | return ERR_PTR(-ENOMEM); | |
613 | goto retry; | |
614 | } | |
dd111b31 | 615 | |
1da177e4 LT |
616 | err = set(s, data); |
617 | if (err) { | |
618 | spin_unlock(&sb_lock); | |
0200894d | 619 | destroy_unused_super(s); |
1da177e4 LT |
620 | return ERR_PTR(err); |
621 | } | |
622 | s->s_type = type; | |
623 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); | |
624 | list_add_tail(&s->s_list, &super_blocks); | |
a5166169 | 625 | hlist_add_head(&s->s_instances, &type->fs_supers); |
1da177e4 LT |
626 | spin_unlock(&sb_lock); |
627 | get_filesystem(type); | |
8e04944f | 628 | register_shrinker_prepared(&s->s_shrink); |
1da177e4 LT |
629 | return s; |
630 | } | |
1da177e4 LT |
631 | EXPORT_SYMBOL(sget); |
632 | ||
633 | void drop_super(struct super_block *sb) | |
634 | { | |
635 | up_read(&sb->s_umount); | |
636 | put_super(sb); | |
637 | } | |
638 | ||
639 | EXPORT_SYMBOL(drop_super); | |
640 | ||
ba6379f7 JK |
641 | void drop_super_exclusive(struct super_block *sb) |
642 | { | |
643 | up_write(&sb->s_umount); | |
644 | put_super(sb); | |
645 | } | |
646 | EXPORT_SYMBOL(drop_super_exclusive); | |
647 | ||
fa7c1d50 MG |
648 | static void __iterate_supers(void (*f)(struct super_block *)) |
649 | { | |
650 | struct super_block *sb, *p = NULL; | |
651 | ||
652 | spin_lock(&sb_lock); | |
653 | list_for_each_entry(sb, &super_blocks, s_list) { | |
654 | if (hlist_unhashed(&sb->s_instances)) | |
655 | continue; | |
656 | sb->s_count++; | |
657 | spin_unlock(&sb_lock); | |
658 | ||
659 | f(sb); | |
660 | ||
661 | spin_lock(&sb_lock); | |
662 | if (p) | |
663 | __put_super(p); | |
664 | p = sb; | |
665 | } | |
666 | if (p) | |
667 | __put_super(p); | |
668 | spin_unlock(&sb_lock); | |
669 | } | |
01a05b33 AV |
670 | /** |
671 | * iterate_supers - call function for all active superblocks | |
672 | * @f: function to call | |
673 | * @arg: argument to pass to it | |
674 | * | |
675 | * Scans the superblock list and calls given function, passing it | |
676 | * locked superblock and given argument. | |
677 | */ | |
678 | void iterate_supers(void (*f)(struct super_block *, void *), void *arg) | |
679 | { | |
dca33252 | 680 | struct super_block *sb, *p = NULL; |
01a05b33 AV |
681 | |
682 | spin_lock(&sb_lock); | |
dca33252 | 683 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 684 | if (hlist_unhashed(&sb->s_instances)) |
01a05b33 AV |
685 | continue; |
686 | sb->s_count++; | |
687 | spin_unlock(&sb_lock); | |
688 | ||
689 | down_read(&sb->s_umount); | |
e462ec50 | 690 | if (sb->s_root && (sb->s_flags & SB_BORN)) |
01a05b33 AV |
691 | f(sb, arg); |
692 | up_read(&sb->s_umount); | |
693 | ||
694 | spin_lock(&sb_lock); | |
dca33252 AV |
695 | if (p) |
696 | __put_super(p); | |
697 | p = sb; | |
01a05b33 | 698 | } |
dca33252 AV |
699 | if (p) |
700 | __put_super(p); | |
01a05b33 AV |
701 | spin_unlock(&sb_lock); |
702 | } | |
703 | ||
43e15cdb AV |
704 | /** |
705 | * iterate_supers_type - call function for superblocks of given type | |
706 | * @type: fs type | |
707 | * @f: function to call | |
708 | * @arg: argument to pass to it | |
709 | * | |
710 | * Scans the superblock list and calls given function, passing it | |
711 | * locked superblock and given argument. | |
712 | */ | |
713 | void iterate_supers_type(struct file_system_type *type, | |
714 | void (*f)(struct super_block *, void *), void *arg) | |
715 | { | |
716 | struct super_block *sb, *p = NULL; | |
717 | ||
718 | spin_lock(&sb_lock); | |
b67bfe0d | 719 | hlist_for_each_entry(sb, &type->fs_supers, s_instances) { |
43e15cdb AV |
720 | sb->s_count++; |
721 | spin_unlock(&sb_lock); | |
722 | ||
723 | down_read(&sb->s_umount); | |
e462ec50 | 724 | if (sb->s_root && (sb->s_flags & SB_BORN)) |
43e15cdb AV |
725 | f(sb, arg); |
726 | up_read(&sb->s_umount); | |
727 | ||
728 | spin_lock(&sb_lock); | |
729 | if (p) | |
730 | __put_super(p); | |
731 | p = sb; | |
732 | } | |
733 | if (p) | |
734 | __put_super(p); | |
735 | spin_unlock(&sb_lock); | |
736 | } | |
737 | ||
738 | EXPORT_SYMBOL(iterate_supers_type); | |
739 | ||
ba6379f7 | 740 | static struct super_block *__get_super(struct block_device *bdev, bool excl) |
1da177e4 | 741 | { |
618f0636 KK |
742 | struct super_block *sb; |
743 | ||
1da177e4 LT |
744 | if (!bdev) |
745 | return NULL; | |
618f0636 | 746 | |
1da177e4 | 747 | spin_lock(&sb_lock); |
618f0636 KK |
748 | rescan: |
749 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 750 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 751 | continue; |
618f0636 KK |
752 | if (sb->s_bdev == bdev) { |
753 | sb->s_count++; | |
1da177e4 | 754 | spin_unlock(&sb_lock); |
ba6379f7 JK |
755 | if (!excl) |
756 | down_read(&sb->s_umount); | |
757 | else | |
758 | down_write(&sb->s_umount); | |
df40c01a | 759 | /* still alive? */ |
e462ec50 | 760 | if (sb->s_root && (sb->s_flags & SB_BORN)) |
618f0636 | 761 | return sb; |
ba6379f7 JK |
762 | if (!excl) |
763 | up_read(&sb->s_umount); | |
764 | else | |
765 | up_write(&sb->s_umount); | |
df40c01a | 766 | /* nope, got unmounted */ |
618f0636 | 767 | spin_lock(&sb_lock); |
df40c01a AV |
768 | __put_super(sb); |
769 | goto rescan; | |
1da177e4 LT |
770 | } |
771 | } | |
772 | spin_unlock(&sb_lock); | |
773 | return NULL; | |
774 | } | |
775 | ||
6b6dc836 | 776 | /** |
ba6379f7 | 777 | * get_super - get the superblock of a device |
6b6dc836 JK |
778 | * @bdev: device to get the superblock for |
779 | * | |
780 | * Scans the superblock list and finds the superblock of the file system | |
ba6379f7 | 781 | * mounted on the device given. %NULL is returned if no match is found. |
6b6dc836 | 782 | */ |
ba6379f7 JK |
783 | struct super_block *get_super(struct block_device *bdev) |
784 | { | |
785 | return __get_super(bdev, false); | |
786 | } | |
787 | EXPORT_SYMBOL(get_super); | |
788 | ||
789 | static struct super_block *__get_super_thawed(struct block_device *bdev, | |
790 | bool excl) | |
6b6dc836 JK |
791 | { |
792 | while (1) { | |
ba6379f7 | 793 | struct super_block *s = __get_super(bdev, excl); |
5accdf82 | 794 | if (!s || s->s_writers.frozen == SB_UNFROZEN) |
6b6dc836 | 795 | return s; |
ba6379f7 JK |
796 | if (!excl) |
797 | up_read(&s->s_umount); | |
798 | else | |
799 | up_write(&s->s_umount); | |
5accdf82 JK |
800 | wait_event(s->s_writers.wait_unfrozen, |
801 | s->s_writers.frozen == SB_UNFROZEN); | |
6b6dc836 JK |
802 | put_super(s); |
803 | } | |
804 | } | |
ba6379f7 JK |
805 | |
806 | /** | |
807 | * get_super_thawed - get thawed superblock of a device | |
808 | * @bdev: device to get the superblock for | |
809 | * | |
810 | * Scans the superblock list and finds the superblock of the file system | |
811 | * mounted on the device. The superblock is returned once it is thawed | |
812 | * (or immediately if it was not frozen). %NULL is returned if no match | |
813 | * is found. | |
814 | */ | |
815 | struct super_block *get_super_thawed(struct block_device *bdev) | |
816 | { | |
817 | return __get_super_thawed(bdev, false); | |
818 | } | |
6b6dc836 JK |
819 | EXPORT_SYMBOL(get_super_thawed); |
820 | ||
ba6379f7 JK |
821 | /** |
822 | * get_super_exclusive_thawed - get thawed superblock of a device | |
823 | * @bdev: device to get the superblock for | |
824 | * | |
825 | * Scans the superblock list and finds the superblock of the file system | |
826 | * mounted on the device. The superblock is returned once it is thawed | |
827 | * (or immediately if it was not frozen) and s_umount semaphore is held | |
828 | * in exclusive mode. %NULL is returned if no match is found. | |
829 | */ | |
830 | struct super_block *get_super_exclusive_thawed(struct block_device *bdev) | |
831 | { | |
832 | return __get_super_thawed(bdev, true); | |
833 | } | |
834 | EXPORT_SYMBOL(get_super_exclusive_thawed); | |
835 | ||
4504230a CH |
836 | /** |
837 | * get_active_super - get an active reference to the superblock of a device | |
838 | * @bdev: device to get the superblock for | |
839 | * | |
840 | * Scans the superblock list and finds the superblock of the file system | |
841 | * mounted on the device given. Returns the superblock with an active | |
d3f21473 | 842 | * reference or %NULL if none was found. |
4504230a CH |
843 | */ |
844 | struct super_block *get_active_super(struct block_device *bdev) | |
845 | { | |
846 | struct super_block *sb; | |
847 | ||
848 | if (!bdev) | |
849 | return NULL; | |
850 | ||
1494583d | 851 | restart: |
4504230a CH |
852 | spin_lock(&sb_lock); |
853 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 854 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 855 | continue; |
1494583d | 856 | if (sb->s_bdev == bdev) { |
acfec9a5 | 857 | if (!grab_super(sb)) |
1494583d | 858 | goto restart; |
acfec9a5 AV |
859 | up_write(&sb->s_umount); |
860 | return sb; | |
1494583d | 861 | } |
4504230a CH |
862 | } |
863 | spin_unlock(&sb_lock); | |
864 | return NULL; | |
865 | } | |
dd111b31 | 866 | |
df40c01a | 867 | struct super_block *user_get_super(dev_t dev) |
1da177e4 | 868 | { |
618f0636 | 869 | struct super_block *sb; |
1da177e4 | 870 | |
1da177e4 | 871 | spin_lock(&sb_lock); |
618f0636 KK |
872 | rescan: |
873 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 874 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 875 | continue; |
618f0636 KK |
876 | if (sb->s_dev == dev) { |
877 | sb->s_count++; | |
1da177e4 | 878 | spin_unlock(&sb_lock); |
618f0636 | 879 | down_read(&sb->s_umount); |
df40c01a | 880 | /* still alive? */ |
e462ec50 | 881 | if (sb->s_root && (sb->s_flags & SB_BORN)) |
618f0636 KK |
882 | return sb; |
883 | up_read(&sb->s_umount); | |
df40c01a | 884 | /* nope, got unmounted */ |
618f0636 | 885 | spin_lock(&sb_lock); |
df40c01a AV |
886 | __put_super(sb); |
887 | goto rescan; | |
1da177e4 LT |
888 | } |
889 | } | |
890 | spin_unlock(&sb_lock); | |
891 | return NULL; | |
892 | } | |
893 | ||
1da177e4 | 894 | /** |
8d0347f6 DH |
895 | * reconfigure_super - asks filesystem to change superblock parameters |
896 | * @fc: The superblock and configuration | |
1da177e4 | 897 | * |
8d0347f6 | 898 | * Alters the configuration parameters of a live superblock. |
1da177e4 | 899 | */ |
8d0347f6 | 900 | int reconfigure_super(struct fs_context *fc) |
1da177e4 | 901 | { |
8d0347f6 | 902 | struct super_block *sb = fc->root->d_sb; |
1da177e4 | 903 | int retval; |
8d0347f6 DH |
904 | bool remount_ro = false; |
905 | bool force = fc->sb_flags & SB_FORCE; | |
4504230a | 906 | |
8d0347f6 DH |
907 | if (fc->sb_flags_mask & ~MS_RMT_MASK) |
908 | return -EINVAL; | |
5accdf82 | 909 | if (sb->s_writers.frozen != SB_UNFROZEN) |
4504230a CH |
910 | return -EBUSY; |
911 | ||
8d0347f6 DH |
912 | retval = security_sb_remount(sb, fc->security); |
913 | if (retval) | |
914 | return retval; | |
915 | ||
916 | if (fc->sb_flags_mask & SB_RDONLY) { | |
9361401e | 917 | #ifdef CONFIG_BLOCK |
8d0347f6 DH |
918 | if (!(fc->sb_flags & SB_RDONLY) && bdev_read_only(sb->s_bdev)) |
919 | return -EACCES; | |
9361401e | 920 | #endif |
4504230a | 921 | |
8d0347f6 DH |
922 | remount_ro = (fc->sb_flags & SB_RDONLY) && !sb_rdonly(sb); |
923 | } | |
d208bbdd | 924 | |
0aec09d0 | 925 | if (remount_ro) { |
fdab684d | 926 | if (!hlist_empty(&sb->s_pins)) { |
0aec09d0 | 927 | up_write(&sb->s_umount); |
fdab684d | 928 | group_pin_kill(&sb->s_pins); |
0aec09d0 AV |
929 | down_write(&sb->s_umount); |
930 | if (!sb->s_root) | |
931 | return 0; | |
932 | if (sb->s_writers.frozen != SB_UNFROZEN) | |
933 | return -EBUSY; | |
8d0347f6 | 934 | remount_ro = !sb_rdonly(sb); |
0aec09d0 AV |
935 | } |
936 | } | |
937 | shrink_dcache_sb(sb); | |
938 | ||
8d0347f6 DH |
939 | /* If we are reconfiguring to RDONLY and current sb is read/write, |
940 | * make sure there are no files open for writing. | |
941 | */ | |
d208bbdd | 942 | if (remount_ro) { |
4ed5e82f | 943 | if (force) { |
eee5cc27 AV |
944 | sb->s_readonly_remount = 1; |
945 | smp_wmb(); | |
4ed5e82f MS |
946 | } else { |
947 | retval = sb_prepare_remount_readonly(sb); | |
948 | if (retval) | |
949 | return retval; | |
4ed5e82f | 950 | } |
1da177e4 LT |
951 | } |
952 | ||
f3a09c92 AV |
953 | if (fc->ops->reconfigure) { |
954 | retval = fc->ops->reconfigure(fc); | |
955 | if (retval) { | |
956 | if (!force) | |
957 | goto cancel_readonly; | |
958 | /* If forced remount, go ahead despite any errors */ | |
959 | WARN(1, "forced remount of a %s fs returned %i\n", | |
960 | sb->s_type->name, retval); | |
961 | } | |
1da177e4 | 962 | } |
8d0347f6 DH |
963 | |
964 | WRITE_ONCE(sb->s_flags, ((sb->s_flags & ~fc->sb_flags_mask) | | |
965 | (fc->sb_flags & fc->sb_flags_mask))); | |
4ed5e82f MS |
966 | /* Needs to be ordered wrt mnt_is_readonly() */ |
967 | smp_wmb(); | |
968 | sb->s_readonly_remount = 0; | |
c79d967d | 969 | |
d208bbdd NP |
970 | /* |
971 | * Some filesystems modify their metadata via some other path than the | |
972 | * bdev buffer cache (eg. use a private mapping, or directories in | |
973 | * pagecache, etc). Also file data modifications go via their own | |
974 | * mappings. So If we try to mount readonly then copy the filesystem | |
975 | * from bdev, we could get stale data, so invalidate it to give a best | |
976 | * effort at coherency. | |
977 | */ | |
978 | if (remount_ro && sb->s_bdev) | |
979 | invalidate_bdev(sb->s_bdev); | |
1da177e4 | 980 | return 0; |
4ed5e82f MS |
981 | |
982 | cancel_readonly: | |
983 | sb->s_readonly_remount = 0; | |
984 | return retval; | |
1da177e4 LT |
985 | } |
986 | ||
fa7c1d50 | 987 | static void do_emergency_remount_callback(struct super_block *sb) |
1da177e4 | 988 | { |
fa7c1d50 MG |
989 | down_write(&sb->s_umount); |
990 | if (sb->s_root && sb->s_bdev && (sb->s_flags & SB_BORN) && | |
991 | !sb_rdonly(sb)) { | |
8d0347f6 DH |
992 | struct fs_context *fc; |
993 | ||
994 | fc = fs_context_for_reconfigure(sb->s_root, | |
995 | SB_RDONLY | SB_FORCE, SB_RDONLY); | |
996 | if (!IS_ERR(fc)) { | |
997 | if (parse_monolithic_mount_data(fc, NULL) == 0) | |
998 | (void)reconfigure_super(fc); | |
999 | put_fs_context(fc); | |
1000 | } | |
1da177e4 | 1001 | } |
fa7c1d50 MG |
1002 | up_write(&sb->s_umount); |
1003 | } | |
1004 | ||
1005 | static void do_emergency_remount(struct work_struct *work) | |
1006 | { | |
1007 | __iterate_supers(do_emergency_remount_callback); | |
a2a9537a | 1008 | kfree(work); |
1da177e4 LT |
1009 | printk("Emergency Remount complete\n"); |
1010 | } | |
1011 | ||
1012 | void emergency_remount(void) | |
1013 | { | |
a2a9537a JA |
1014 | struct work_struct *work; |
1015 | ||
1016 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | |
1017 | if (work) { | |
1018 | INIT_WORK(work, do_emergency_remount); | |
1019 | schedule_work(work); | |
1020 | } | |
1da177e4 LT |
1021 | } |
1022 | ||
08fdc8a0 MG |
1023 | static void do_thaw_all_callback(struct super_block *sb) |
1024 | { | |
1025 | down_write(&sb->s_umount); | |
1c18d2a1 | 1026 | if (sb->s_root && sb->s_flags & SB_BORN) { |
08fdc8a0 MG |
1027 | emergency_thaw_bdev(sb); |
1028 | thaw_super_locked(sb); | |
1029 | } else { | |
1030 | up_write(&sb->s_umount); | |
1031 | } | |
1032 | } | |
1033 | ||
1034 | static void do_thaw_all(struct work_struct *work) | |
1035 | { | |
1036 | __iterate_supers(do_thaw_all_callback); | |
1037 | kfree(work); | |
1038 | printk(KERN_WARNING "Emergency Thaw complete\n"); | |
1039 | } | |
1040 | ||
1041 | /** | |
1042 | * emergency_thaw_all -- forcibly thaw every frozen filesystem | |
1043 | * | |
1044 | * Used for emergency unfreeze of all filesystems via SysRq | |
1045 | */ | |
1046 | void emergency_thaw_all(void) | |
1047 | { | |
1048 | struct work_struct *work; | |
1049 | ||
1050 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | |
1051 | if (work) { | |
1052 | INIT_WORK(work, do_thaw_all); | |
1053 | schedule_work(work); | |
1054 | } | |
1055 | } | |
1056 | ||
ad76cbc6 | 1057 | static DEFINE_IDA(unnamed_dev_ida); |
1da177e4 | 1058 | |
5a66847e MW |
1059 | /** |
1060 | * get_anon_bdev - Allocate a block device for filesystems which don't have one. | |
1061 | * @p: Pointer to a dev_t. | |
1062 | * | |
1063 | * Filesystems which don't use real block devices can call this function | |
1064 | * to allocate a virtual block device. | |
1065 | * | |
1066 | * Context: Any context. Frequently called while holding sb_lock. | |
1067 | * Return: 0 on success, -EMFILE if there are no anonymous bdevs left | |
1068 | * or -ENOMEM if memory allocation failed. | |
1069 | */ | |
0ee5dc67 | 1070 | int get_anon_bdev(dev_t *p) |
1da177e4 LT |
1071 | { |
1072 | int dev; | |
5a66847e MW |
1073 | |
1074 | /* | |
1075 | * Many userspace utilities consider an FSID of 0 invalid. | |
1076 | * Always return at least 1 from get_anon_bdev. | |
1077 | */ | |
1078 | dev = ida_alloc_range(&unnamed_dev_ida, 1, (1 << MINORBITS) - 1, | |
1079 | GFP_ATOMIC); | |
1080 | if (dev == -ENOSPC) | |
1081 | dev = -EMFILE; | |
1082 | if (dev < 0) | |
1083 | return dev; | |
1084 | ||
1085 | *p = MKDEV(0, dev); | |
1da177e4 LT |
1086 | return 0; |
1087 | } | |
0ee5dc67 | 1088 | EXPORT_SYMBOL(get_anon_bdev); |
1da177e4 | 1089 | |
0ee5dc67 | 1090 | void free_anon_bdev(dev_t dev) |
1da177e4 | 1091 | { |
5a66847e | 1092 | ida_free(&unnamed_dev_ida, MINOR(dev)); |
1da177e4 | 1093 | } |
0ee5dc67 AV |
1094 | EXPORT_SYMBOL(free_anon_bdev); |
1095 | ||
1096 | int set_anon_super(struct super_block *s, void *data) | |
1097 | { | |
df0ce26c | 1098 | return get_anon_bdev(&s->s_dev); |
0ee5dc67 | 1099 | } |
0ee5dc67 AV |
1100 | EXPORT_SYMBOL(set_anon_super); |
1101 | ||
1102 | void kill_anon_super(struct super_block *sb) | |
1103 | { | |
1104 | dev_t dev = sb->s_dev; | |
1105 | generic_shutdown_super(sb); | |
1106 | free_anon_bdev(dev); | |
1107 | } | |
1da177e4 LT |
1108 | EXPORT_SYMBOL(kill_anon_super); |
1109 | ||
1da177e4 LT |
1110 | void kill_litter_super(struct super_block *sb) |
1111 | { | |
1112 | if (sb->s_root) | |
1113 | d_genocide(sb->s_root); | |
1114 | kill_anon_super(sb); | |
1115 | } | |
1da177e4 LT |
1116 | EXPORT_SYMBOL(kill_litter_super); |
1117 | ||
cb50b348 AV |
1118 | int set_anon_super_fc(struct super_block *sb, struct fs_context *fc) |
1119 | { | |
1120 | return set_anon_super(sb, NULL); | |
1121 | } | |
1122 | EXPORT_SYMBOL(set_anon_super_fc); | |
1123 | ||
1124 | static int test_keyed_super(struct super_block *sb, struct fs_context *fc) | |
1125 | { | |
1126 | return sb->s_fs_info == fc->s_fs_info; | |
1127 | } | |
1128 | ||
1129 | static int test_single_super(struct super_block *s, struct fs_context *fc) | |
1130 | { | |
1131 | return 1; | |
1132 | } | |
1133 | ||
1134 | /** | |
1135 | * vfs_get_super - Get a superblock with a search key set in s_fs_info. | |
1136 | * @fc: The filesystem context holding the parameters | |
1137 | * @keying: How to distinguish superblocks | |
1138 | * @fill_super: Helper to initialise a new superblock | |
1139 | * | |
1140 | * Search for a superblock and create a new one if not found. The search | |
1141 | * criterion is controlled by @keying. If the search fails, a new superblock | |
1142 | * is created and @fill_super() is called to initialise it. | |
1143 | * | |
1144 | * @keying can take one of a number of values: | |
1145 | * | |
1146 | * (1) vfs_get_single_super - Only one superblock of this type may exist on the | |
1147 | * system. This is typically used for special system filesystems. | |
1148 | * | |
1149 | * (2) vfs_get_keyed_super - Multiple superblocks may exist, but they must have | |
1150 | * distinct keys (where the key is in s_fs_info). Searching for the same | |
1151 | * key again will turn up the superblock for that key. | |
1152 | * | |
1153 | * (3) vfs_get_independent_super - Multiple superblocks may exist and are | |
1154 | * unkeyed. Each call will get a new superblock. | |
1155 | * | |
1156 | * A permissions check is made by sget_fc() unless we're getting a superblock | |
1157 | * for a kernel-internal mount or a submount. | |
1158 | */ | |
1159 | int vfs_get_super(struct fs_context *fc, | |
1160 | enum vfs_get_super_keying keying, | |
1161 | int (*fill_super)(struct super_block *sb, | |
1162 | struct fs_context *fc)) | |
1163 | { | |
1164 | int (*test)(struct super_block *, struct fs_context *); | |
1165 | struct super_block *sb; | |
1166 | ||
1167 | switch (keying) { | |
1168 | case vfs_get_single_super: | |
1169 | test = test_single_super; | |
1170 | break; | |
1171 | case vfs_get_keyed_super: | |
1172 | test = test_keyed_super; | |
1173 | break; | |
1174 | case vfs_get_independent_super: | |
1175 | test = NULL; | |
1176 | break; | |
1177 | default: | |
1178 | BUG(); | |
1179 | } | |
1180 | ||
1181 | sb = sget_fc(fc, test, set_anon_super_fc); | |
1182 | if (IS_ERR(sb)) | |
1183 | return PTR_ERR(sb); | |
1184 | ||
1185 | if (!sb->s_root) { | |
1186 | int err = fill_super(sb, fc); | |
1187 | if (err) { | |
1188 | deactivate_locked_super(sb); | |
1189 | return err; | |
1190 | } | |
1191 | ||
1192 | sb->s_flags |= SB_ACTIVE; | |
1193 | } | |
1194 | ||
1195 | BUG_ON(fc->root); | |
1196 | fc->root = dget(sb->s_root); | |
1197 | return 0; | |
1198 | } | |
1199 | EXPORT_SYMBOL(vfs_get_super); | |
1200 | ||
2ac295d4 AV |
1201 | int get_tree_nodev(struct fs_context *fc, |
1202 | int (*fill_super)(struct super_block *sb, | |
1203 | struct fs_context *fc)) | |
1204 | { | |
1205 | return vfs_get_super(fc, vfs_get_independent_super, fill_super); | |
1206 | } | |
1207 | EXPORT_SYMBOL(get_tree_nodev); | |
1208 | ||
c23a0bba AV |
1209 | int get_tree_single(struct fs_context *fc, |
1210 | int (*fill_super)(struct super_block *sb, | |
1211 | struct fs_context *fc)) | |
1212 | { | |
1213 | return vfs_get_super(fc, vfs_get_single_super, fill_super); | |
1214 | } | |
1215 | EXPORT_SYMBOL(get_tree_single); | |
1216 | ||
9361401e | 1217 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
1218 | static int set_bdev_super(struct super_block *s, void *data) |
1219 | { | |
1220 | s->s_bdev = data; | |
1221 | s->s_dev = s->s_bdev->bd_dev; | |
13eec236 | 1222 | s->s_bdi = bdi_get(s->s_bdev->bd_bdi); |
32a88aa1 | 1223 | |
1da177e4 LT |
1224 | return 0; |
1225 | } | |
1226 | ||
1227 | static int test_bdev_super(struct super_block *s, void *data) | |
1228 | { | |
1229 | return (void *)s->s_bdev == data; | |
1230 | } | |
1231 | ||
152a0836 | 1232 | struct dentry *mount_bdev(struct file_system_type *fs_type, |
1da177e4 | 1233 | int flags, const char *dev_name, void *data, |
152a0836 | 1234 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1235 | { |
1236 | struct block_device *bdev; | |
1237 | struct super_block *s; | |
d4d77629 | 1238 | fmode_t mode = FMODE_READ | FMODE_EXCL; |
1da177e4 LT |
1239 | int error = 0; |
1240 | ||
e462ec50 | 1241 | if (!(flags & SB_RDONLY)) |
30c40d2c AV |
1242 | mode |= FMODE_WRITE; |
1243 | ||
d4d77629 | 1244 | bdev = blkdev_get_by_path(dev_name, mode, fs_type); |
1da177e4 | 1245 | if (IS_ERR(bdev)) |
152a0836 | 1246 | return ERR_CAST(bdev); |
1da177e4 LT |
1247 | |
1248 | /* | |
1249 | * once the super is inserted into the list by sget, s_umount | |
1250 | * will protect the lockfs code from trying to start a snapshot | |
1251 | * while we are mounting | |
1252 | */ | |
4fadd7bb CH |
1253 | mutex_lock(&bdev->bd_fsfreeze_mutex); |
1254 | if (bdev->bd_fsfreeze_count > 0) { | |
1255 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
1256 | error = -EBUSY; | |
1257 | goto error_bdev; | |
1258 | } | |
e462ec50 | 1259 | s = sget(fs_type, test_bdev_super, set_bdev_super, flags | SB_NOSEC, |
9249e17f | 1260 | bdev); |
4fadd7bb | 1261 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
1da177e4 | 1262 | if (IS_ERR(s)) |
454e2398 | 1263 | goto error_s; |
1da177e4 LT |
1264 | |
1265 | if (s->s_root) { | |
e462ec50 | 1266 | if ((flags ^ s->s_flags) & SB_RDONLY) { |
74dbbdd7 | 1267 | deactivate_locked_super(s); |
454e2398 DH |
1268 | error = -EBUSY; |
1269 | goto error_bdev; | |
1da177e4 | 1270 | } |
454e2398 | 1271 | |
4f331f01 TH |
1272 | /* |
1273 | * s_umount nests inside bd_mutex during | |
e525fd89 TH |
1274 | * __invalidate_device(). blkdev_put() acquires |
1275 | * bd_mutex and can't be called under s_umount. Drop | |
1276 | * s_umount temporarily. This is safe as we're | |
1277 | * holding an active reference. | |
4f331f01 TH |
1278 | */ |
1279 | up_write(&s->s_umount); | |
d4d77629 | 1280 | blkdev_put(bdev, mode); |
4f331f01 | 1281 | down_write(&s->s_umount); |
1da177e4 | 1282 | } else { |
30c40d2c | 1283 | s->s_mode = mode; |
a1c6f057 | 1284 | snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev); |
e78c9a00 | 1285 | sb_set_blocksize(s, block_size(bdev)); |
e462ec50 | 1286 | error = fill_super(s, data, flags & SB_SILENT ? 1 : 0); |
1da177e4 | 1287 | if (error) { |
74dbbdd7 | 1288 | deactivate_locked_super(s); |
454e2398 | 1289 | goto error; |
fa675765 | 1290 | } |
454e2398 | 1291 | |
e462ec50 | 1292 | s->s_flags |= SB_ACTIVE; |
87d8fe1e | 1293 | bdev->bd_super = s; |
1da177e4 LT |
1294 | } |
1295 | ||
152a0836 | 1296 | return dget(s->s_root); |
1da177e4 | 1297 | |
454e2398 DH |
1298 | error_s: |
1299 | error = PTR_ERR(s); | |
1300 | error_bdev: | |
d4d77629 | 1301 | blkdev_put(bdev, mode); |
454e2398 | 1302 | error: |
152a0836 AV |
1303 | return ERR_PTR(error); |
1304 | } | |
1305 | EXPORT_SYMBOL(mount_bdev); | |
1306 | ||
1da177e4 LT |
1307 | void kill_block_super(struct super_block *sb) |
1308 | { | |
1309 | struct block_device *bdev = sb->s_bdev; | |
30c40d2c | 1310 | fmode_t mode = sb->s_mode; |
1da177e4 | 1311 | |
ddbaaf30 | 1312 | bdev->bd_super = NULL; |
1da177e4 LT |
1313 | generic_shutdown_super(sb); |
1314 | sync_blockdev(bdev); | |
d4d77629 | 1315 | WARN_ON_ONCE(!(mode & FMODE_EXCL)); |
e525fd89 | 1316 | blkdev_put(bdev, mode | FMODE_EXCL); |
1da177e4 LT |
1317 | } |
1318 | ||
1319 | EXPORT_SYMBOL(kill_block_super); | |
9361401e | 1320 | #endif |
1da177e4 | 1321 | |
3c26ff6e | 1322 | struct dentry *mount_nodev(struct file_system_type *fs_type, |
1da177e4 | 1323 | int flags, void *data, |
3c26ff6e | 1324 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1325 | { |
1326 | int error; | |
9249e17f | 1327 | struct super_block *s = sget(fs_type, NULL, set_anon_super, flags, NULL); |
1da177e4 LT |
1328 | |
1329 | if (IS_ERR(s)) | |
3c26ff6e | 1330 | return ERR_CAST(s); |
1da177e4 | 1331 | |
e462ec50 | 1332 | error = fill_super(s, data, flags & SB_SILENT ? 1 : 0); |
1da177e4 | 1333 | if (error) { |
74dbbdd7 | 1334 | deactivate_locked_super(s); |
3c26ff6e | 1335 | return ERR_PTR(error); |
1da177e4 | 1336 | } |
e462ec50 | 1337 | s->s_flags |= SB_ACTIVE; |
3c26ff6e | 1338 | return dget(s->s_root); |
1da177e4 | 1339 | } |
3c26ff6e AV |
1340 | EXPORT_SYMBOL(mount_nodev); |
1341 | ||
8d0347f6 DH |
1342 | static int reconfigure_single(struct super_block *s, |
1343 | int flags, void *data) | |
1344 | { | |
1345 | struct fs_context *fc; | |
1346 | int ret; | |
1347 | ||
1348 | /* The caller really need to be passing fc down into mount_single(), | |
1349 | * then a chunk of this can be removed. [Bollocks -- AV] | |
1350 | * Better yet, reconfiguration shouldn't happen, but rather the second | |
1351 | * mount should be rejected if the parameters are not compatible. | |
1352 | */ | |
1353 | fc = fs_context_for_reconfigure(s->s_root, flags, MS_RMT_MASK); | |
1354 | if (IS_ERR(fc)) | |
1355 | return PTR_ERR(fc); | |
1356 | ||
1357 | ret = parse_monolithic_mount_data(fc, data); | |
1358 | if (ret < 0) | |
1359 | goto out; | |
1360 | ||
1361 | ret = reconfigure_super(fc); | |
1362 | out: | |
1363 | put_fs_context(fc); | |
1364 | return ret; | |
1365 | } | |
1366 | ||
1da177e4 LT |
1367 | static int compare_single(struct super_block *s, void *p) |
1368 | { | |
1369 | return 1; | |
1370 | } | |
1371 | ||
fc14f2fe | 1372 | struct dentry *mount_single(struct file_system_type *fs_type, |
1da177e4 | 1373 | int flags, void *data, |
fc14f2fe | 1374 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1375 | { |
1376 | struct super_block *s; | |
1377 | int error; | |
1378 | ||
9249e17f | 1379 | s = sget(fs_type, compare_single, set_anon_super, flags, NULL); |
1da177e4 | 1380 | if (IS_ERR(s)) |
fc14f2fe | 1381 | return ERR_CAST(s); |
1da177e4 | 1382 | if (!s->s_root) { |
e462ec50 | 1383 | error = fill_super(s, data, flags & SB_SILENT ? 1 : 0); |
8d0347f6 DH |
1384 | if (!error) |
1385 | s->s_flags |= SB_ACTIVE; | |
9329d1be | 1386 | } else { |
8d0347f6 DH |
1387 | error = reconfigure_single(s, flags, data); |
1388 | } | |
1389 | if (unlikely(error)) { | |
1390 | deactivate_locked_super(s); | |
1391 | return ERR_PTR(error); | |
1da177e4 | 1392 | } |
fc14f2fe AV |
1393 | return dget(s->s_root); |
1394 | } | |
1395 | EXPORT_SYMBOL(mount_single); | |
1396 | ||
9bc61ab1 DH |
1397 | /** |
1398 | * vfs_get_tree - Get the mountable root | |
1399 | * @fc: The superblock configuration context. | |
1400 | * | |
1401 | * The filesystem is invoked to get or create a superblock which can then later | |
1402 | * be used for mounting. The filesystem places a pointer to the root to be | |
1403 | * used for mounting in @fc->root. | |
1404 | */ | |
1405 | int vfs_get_tree(struct fs_context *fc) | |
1da177e4 | 1406 | { |
9d412a43 | 1407 | struct super_block *sb; |
9bc61ab1 | 1408 | int error; |
8089352a | 1409 | |
f3a09c92 AV |
1410 | if (fc->root) |
1411 | return -EBUSY; | |
1412 | ||
1413 | /* Get the mountable root in fc->root, with a ref on the root and a ref | |
1414 | * on the superblock. | |
1415 | */ | |
1416 | error = fc->ops->get_tree(fc); | |
9bc61ab1 DH |
1417 | if (error < 0) |
1418 | return error; | |
1da177e4 | 1419 | |
f3a09c92 AV |
1420 | if (!fc->root) { |
1421 | pr_err("Filesystem %s get_tree() didn't set fc->root\n", | |
1422 | fc->fs_type->name); | |
1423 | /* We don't know what the locking state of the superblock is - | |
1424 | * if there is a superblock. | |
1425 | */ | |
1426 | BUG(); | |
1427 | } | |
1428 | ||
9bc61ab1 | 1429 | sb = fc->root->d_sb; |
9d412a43 | 1430 | WARN_ON(!sb->s_bdi); |
79f546a6 | 1431 | |
a0c9a8b8 AV |
1432 | if (fc->subtype && !sb->s_subtype) { |
1433 | sb->s_subtype = fc->subtype; | |
1434 | fc->subtype = NULL; | |
1435 | } | |
1436 | ||
79f546a6 DC |
1437 | /* |
1438 | * Write barrier is for super_cache_count(). We place it before setting | |
1439 | * SB_BORN as the data dependency between the two functions is the | |
1440 | * superblock structure contents that we just set up, not the SB_BORN | |
1441 | * flag. | |
1442 | */ | |
1443 | smp_wmb(); | |
e462ec50 | 1444 | sb->s_flags |= SB_BORN; |
454e2398 | 1445 | |
9bc61ab1 | 1446 | error = security_sb_set_mnt_opts(sb, fc->security, 0, NULL); |
c9ce29ed AV |
1447 | if (unlikely(error)) { |
1448 | fc_drop_locked(fc); | |
1449 | return error; | |
a10d7c22 AV |
1450 | } |
1451 | ||
42cb56ae JL |
1452 | /* |
1453 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE | |
1454 | * but s_maxbytes was an unsigned long long for many releases. Throw | |
1455 | * this warning for a little while to try and catch filesystems that | |
4358b567 | 1456 | * violate this rule. |
42cb56ae | 1457 | */ |
9d412a43 | 1458 | WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " |
9bc61ab1 | 1459 | "negative value (%lld)\n", fc->fs_type->name, sb->s_maxbytes); |
42cb56ae | 1460 | |
9bc61ab1 | 1461 | return 0; |
1da177e4 | 1462 | } |
9bc61ab1 | 1463 | EXPORT_SYMBOL(vfs_get_tree); |
1da177e4 | 1464 | |
fca39346 JK |
1465 | /* |
1466 | * Setup private BDI for given superblock. It gets automatically cleaned up | |
1467 | * in generic_shutdown_super(). | |
1468 | */ | |
1469 | int super_setup_bdi_name(struct super_block *sb, char *fmt, ...) | |
1470 | { | |
1471 | struct backing_dev_info *bdi; | |
1472 | int err; | |
1473 | va_list args; | |
1474 | ||
1475 | bdi = bdi_alloc(GFP_KERNEL); | |
1476 | if (!bdi) | |
1477 | return -ENOMEM; | |
1478 | ||
1479 | bdi->name = sb->s_type->name; | |
1480 | ||
1481 | va_start(args, fmt); | |
7c4cc300 | 1482 | err = bdi_register_va(bdi, fmt, args); |
fca39346 JK |
1483 | va_end(args); |
1484 | if (err) { | |
1485 | bdi_put(bdi); | |
1486 | return err; | |
1487 | } | |
1488 | WARN_ON(sb->s_bdi != &noop_backing_dev_info); | |
1489 | sb->s_bdi = bdi; | |
fca39346 JK |
1490 | |
1491 | return 0; | |
1492 | } | |
1493 | EXPORT_SYMBOL(super_setup_bdi_name); | |
1494 | ||
1495 | /* | |
1496 | * Setup private BDI for given superblock. I gets automatically cleaned up | |
1497 | * in generic_shutdown_super(). | |
1498 | */ | |
1499 | int super_setup_bdi(struct super_block *sb) | |
1500 | { | |
1501 | static atomic_long_t bdi_seq = ATOMIC_LONG_INIT(0); | |
1502 | ||
1503 | return super_setup_bdi_name(sb, "%.28s-%ld", sb->s_type->name, | |
1504 | atomic_long_inc_return(&bdi_seq)); | |
1505 | } | |
1506 | EXPORT_SYMBOL(super_setup_bdi); | |
1507 | ||
5accdf82 JK |
1508 | /* |
1509 | * This is an internal function, please use sb_end_{write,pagefault,intwrite} | |
1510 | * instead. | |
1511 | */ | |
1512 | void __sb_end_write(struct super_block *sb, int level) | |
1513 | { | |
8129ed29 | 1514 | percpu_up_read(sb->s_writers.rw_sem + level-1); |
5accdf82 JK |
1515 | } |
1516 | EXPORT_SYMBOL(__sb_end_write); | |
1517 | ||
f4b554af ON |
1518 | /* |
1519 | * This is an internal function, please use sb_start_{write,pagefault,intwrite} | |
1520 | * instead. | |
1521 | */ | |
1522 | int __sb_start_write(struct super_block *sb, int level, bool wait) | |
1523 | { | |
1524 | bool force_trylock = false; | |
8129ed29 | 1525 | int ret = 1; |
f4b554af ON |
1526 | |
1527 | #ifdef CONFIG_LOCKDEP | |
1528 | /* | |
1529 | * We want lockdep to tell us about possible deadlocks with freezing | |
1530 | * but it's it bit tricky to properly instrument it. Getting a freeze | |
1531 | * protection works as getting a read lock but there are subtle | |
1532 | * problems. XFS for example gets freeze protection on internal level | |
1533 | * twice in some cases, which is OK only because we already hold a | |
1534 | * freeze protection also on higher level. Due to these cases we have | |
1535 | * to use wait == F (trylock mode) which must not fail. | |
1536 | */ | |
1537 | if (wait) { | |
1538 | int i; | |
1539 | ||
1540 | for (i = 0; i < level - 1; i++) | |
8129ed29 | 1541 | if (percpu_rwsem_is_held(sb->s_writers.rw_sem + i)) { |
f4b554af ON |
1542 | force_trylock = true; |
1543 | break; | |
1544 | } | |
1545 | } | |
1546 | #endif | |
8129ed29 ON |
1547 | if (wait && !force_trylock) |
1548 | percpu_down_read(sb->s_writers.rw_sem + level-1); | |
1549 | else | |
1550 | ret = percpu_down_read_trylock(sb->s_writers.rw_sem + level-1); | |
1551 | ||
22224a17 | 1552 | WARN_ON(force_trylock && !ret); |
f4b554af ON |
1553 | return ret; |
1554 | } | |
5accdf82 JK |
1555 | EXPORT_SYMBOL(__sb_start_write); |
1556 | ||
1557 | /** | |
1558 | * sb_wait_write - wait until all writers to given file system finish | |
1559 | * @sb: the super for which we wait | |
1560 | * @level: type of writers we wait for (normal vs page fault) | |
1561 | * | |
1562 | * This function waits until there are no writers of given type to given file | |
8129ed29 | 1563 | * system. |
5accdf82 JK |
1564 | */ |
1565 | static void sb_wait_write(struct super_block *sb, int level) | |
1566 | { | |
8129ed29 | 1567 | percpu_down_write(sb->s_writers.rw_sem + level-1); |
8129ed29 | 1568 | } |
5accdf82 | 1569 | |
f1a96220 ON |
1570 | /* |
1571 | * We are going to return to userspace and forget about these locks, the | |
1572 | * ownership goes to the caller of thaw_super() which does unlock(). | |
1573 | */ | |
1574 | static void lockdep_sb_freeze_release(struct super_block *sb) | |
1575 | { | |
1576 | int level; | |
1577 | ||
1578 | for (level = SB_FREEZE_LEVELS - 1; level >= 0; level--) | |
1579 | percpu_rwsem_release(sb->s_writers.rw_sem + level, 0, _THIS_IP_); | |
1580 | } | |
1581 | ||
1582 | /* | |
1583 | * Tell lockdep we are holding these locks before we call ->unfreeze_fs(sb). | |
1584 | */ | |
1585 | static void lockdep_sb_freeze_acquire(struct super_block *sb) | |
8129ed29 ON |
1586 | { |
1587 | int level; | |
5accdf82 | 1588 | |
8129ed29 ON |
1589 | for (level = 0; level < SB_FREEZE_LEVELS; ++level) |
1590 | percpu_rwsem_acquire(sb->s_writers.rw_sem + level, 0, _THIS_IP_); | |
f1a96220 ON |
1591 | } |
1592 | ||
1593 | static void sb_freeze_unlock(struct super_block *sb) | |
1594 | { | |
1595 | int level; | |
5accdf82 | 1596 | |
8129ed29 ON |
1597 | for (level = SB_FREEZE_LEVELS - 1; level >= 0; level--) |
1598 | percpu_up_write(sb->s_writers.rw_sem + level); | |
5accdf82 JK |
1599 | } |
1600 | ||
18e9e510 | 1601 | /** |
7000d3c4 RD |
1602 | * freeze_super - lock the filesystem and force it into a consistent state |
1603 | * @sb: the super to lock | |
18e9e510 JB |
1604 | * |
1605 | * Syncs the super to make sure the filesystem is consistent and calls the fs's | |
1606 | * freeze_fs. Subsequent calls to this without first thawing the fs will return | |
1607 | * -EBUSY. | |
5accdf82 JK |
1608 | * |
1609 | * During this function, sb->s_writers.frozen goes through these values: | |
1610 | * | |
1611 | * SB_UNFROZEN: File system is normal, all writes progress as usual. | |
1612 | * | |
1613 | * SB_FREEZE_WRITE: The file system is in the process of being frozen. New | |
1614 | * writes should be blocked, though page faults are still allowed. We wait for | |
1615 | * all writes to complete and then proceed to the next stage. | |
1616 | * | |
1617 | * SB_FREEZE_PAGEFAULT: Freezing continues. Now also page faults are blocked | |
1618 | * but internal fs threads can still modify the filesystem (although they | |
1619 | * should not dirty new pages or inodes), writeback can run etc. After waiting | |
1620 | * for all running page faults we sync the filesystem which will clean all | |
1621 | * dirty pages and inodes (no new dirty pages or inodes can be created when | |
1622 | * sync is running). | |
1623 | * | |
1624 | * SB_FREEZE_FS: The file system is frozen. Now all internal sources of fs | |
1625 | * modification are blocked (e.g. XFS preallocation truncation on inode | |
1626 | * reclaim). This is usually implemented by blocking new transactions for | |
1627 | * filesystems that have them and need this additional guard. After all | |
1628 | * internal writers are finished we call ->freeze_fs() to finish filesystem | |
1629 | * freezing. Then we transition to SB_FREEZE_COMPLETE state. This state is | |
1630 | * mostly auxiliary for filesystems to verify they do not modify frozen fs. | |
1631 | * | |
1632 | * sb->s_writers.frozen is protected by sb->s_umount. | |
18e9e510 JB |
1633 | */ |
1634 | int freeze_super(struct super_block *sb) | |
1635 | { | |
1636 | int ret; | |
1637 | ||
1638 | atomic_inc(&sb->s_active); | |
1639 | down_write(&sb->s_umount); | |
5accdf82 | 1640 | if (sb->s_writers.frozen != SB_UNFROZEN) { |
18e9e510 JB |
1641 | deactivate_locked_super(sb); |
1642 | return -EBUSY; | |
1643 | } | |
1644 | ||
e462ec50 | 1645 | if (!(sb->s_flags & SB_BORN)) { |
dabe0dc1 AV |
1646 | up_write(&sb->s_umount); |
1647 | return 0; /* sic - it's "nothing to do" */ | |
1648 | } | |
1649 | ||
bc98a42c | 1650 | if (sb_rdonly(sb)) { |
5accdf82 JK |
1651 | /* Nothing to do really... */ |
1652 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
18e9e510 JB |
1653 | up_write(&sb->s_umount); |
1654 | return 0; | |
1655 | } | |
1656 | ||
5accdf82 | 1657 | sb->s_writers.frozen = SB_FREEZE_WRITE; |
5accdf82 JK |
1658 | /* Release s_umount to preserve sb_start_write -> s_umount ordering */ |
1659 | up_write(&sb->s_umount); | |
5accdf82 | 1660 | sb_wait_write(sb, SB_FREEZE_WRITE); |
8129ed29 | 1661 | down_write(&sb->s_umount); |
5accdf82 JK |
1662 | |
1663 | /* Now we go and block page faults... */ | |
5accdf82 | 1664 | sb->s_writers.frozen = SB_FREEZE_PAGEFAULT; |
5accdf82 JK |
1665 | sb_wait_write(sb, SB_FREEZE_PAGEFAULT); |
1666 | ||
1667 | /* All writers are done so after syncing there won't be dirty data */ | |
18e9e510 JB |
1668 | sync_filesystem(sb); |
1669 | ||
5accdf82 JK |
1670 | /* Now wait for internal filesystem counter */ |
1671 | sb->s_writers.frozen = SB_FREEZE_FS; | |
5accdf82 | 1672 | sb_wait_write(sb, SB_FREEZE_FS); |
18e9e510 | 1673 | |
18e9e510 JB |
1674 | if (sb->s_op->freeze_fs) { |
1675 | ret = sb->s_op->freeze_fs(sb); | |
1676 | if (ret) { | |
1677 | printk(KERN_ERR | |
1678 | "VFS:Filesystem freeze failed\n"); | |
5accdf82 | 1679 | sb->s_writers.frozen = SB_UNFROZEN; |
8129ed29 | 1680 | sb_freeze_unlock(sb); |
5accdf82 | 1681 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 JB |
1682 | deactivate_locked_super(sb); |
1683 | return ret; | |
1684 | } | |
1685 | } | |
5accdf82 | 1686 | /* |
89f39af1 ON |
1687 | * For debugging purposes so that fs can warn if it sees write activity |
1688 | * when frozen is set to SB_FREEZE_COMPLETE, and for thaw_super(). | |
5accdf82 JK |
1689 | */ |
1690 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
f1a96220 | 1691 | lockdep_sb_freeze_release(sb); |
18e9e510 JB |
1692 | up_write(&sb->s_umount); |
1693 | return 0; | |
1694 | } | |
1695 | EXPORT_SYMBOL(freeze_super); | |
1696 | ||
1697 | /** | |
1698 | * thaw_super -- unlock filesystem | |
1699 | * @sb: the super to thaw | |
1700 | * | |
1701 | * Unlocks the filesystem and marks it writeable again after freeze_super(). | |
1702 | */ | |
08fdc8a0 | 1703 | static int thaw_super_locked(struct super_block *sb) |
18e9e510 JB |
1704 | { |
1705 | int error; | |
1706 | ||
89f39af1 | 1707 | if (sb->s_writers.frozen != SB_FREEZE_COMPLETE) { |
18e9e510 JB |
1708 | up_write(&sb->s_umount); |
1709 | return -EINVAL; | |
1710 | } | |
1711 | ||
bc98a42c | 1712 | if (sb_rdonly(sb)) { |
8129ed29 | 1713 | sb->s_writers.frozen = SB_UNFROZEN; |
18e9e510 | 1714 | goto out; |
8129ed29 | 1715 | } |
18e9e510 | 1716 | |
f1a96220 ON |
1717 | lockdep_sb_freeze_acquire(sb); |
1718 | ||
18e9e510 JB |
1719 | if (sb->s_op->unfreeze_fs) { |
1720 | error = sb->s_op->unfreeze_fs(sb); | |
1721 | if (error) { | |
1722 | printk(KERN_ERR | |
1723 | "VFS:Filesystem thaw failed\n"); | |
f1a96220 | 1724 | lockdep_sb_freeze_release(sb); |
18e9e510 JB |
1725 | up_write(&sb->s_umount); |
1726 | return error; | |
1727 | } | |
1728 | } | |
1729 | ||
5accdf82 | 1730 | sb->s_writers.frozen = SB_UNFROZEN; |
8129ed29 ON |
1731 | sb_freeze_unlock(sb); |
1732 | out: | |
5accdf82 | 1733 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 | 1734 | deactivate_locked_super(sb); |
18e9e510 JB |
1735 | return 0; |
1736 | } | |
08fdc8a0 MG |
1737 | |
1738 | int thaw_super(struct super_block *sb) | |
1739 | { | |
1740 | down_write(&sb->s_umount); | |
1741 | return thaw_super_locked(sb); | |
1742 | } | |
18e9e510 | 1743 | EXPORT_SYMBOL(thaw_super); |