]>
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> |
22d94f49 | 35 | #include <linux/fscrypt.h> |
40401530 | 36 | #include <linux/fsnotify.h> |
5accdf82 | 37 | #include <linux/lockdep.h> |
6e4eab57 | 38 | #include <linux/user_namespace.h> |
9bc61ab1 | 39 | #include <linux/fs_context.h> |
e262e32d | 40 | #include <uapi/linux/mount.h> |
6d59e7f5 | 41 | #include "internal.h" |
1da177e4 | 42 | |
08fdc8a0 | 43 | static int thaw_super_locked(struct super_block *sb); |
1da177e4 | 44 | |
15d0f5ea AV |
45 | static LIST_HEAD(super_blocks); |
46 | static DEFINE_SPINLOCK(sb_lock); | |
1da177e4 | 47 | |
5accdf82 JK |
48 | static char *sb_writers_name[SB_FREEZE_LEVELS] = { |
49 | "sb_writers", | |
50 | "sb_pagefaults", | |
51 | "sb_internal", | |
52 | }; | |
53 | ||
b0d40c92 DC |
54 | /* |
55 | * One thing we have to be careful of with a per-sb shrinker is that we don't | |
56 | * drop the last active reference to the superblock from within the shrinker. | |
57 | * If that happens we could trigger unregistering the shrinker from within the | |
58 | * shrinker path and that leads to deadlock on the shrinker_rwsem. Hence we | |
59 | * take a passive reference to the superblock to avoid this from occurring. | |
60 | */ | |
0a234c6d DC |
61 | static unsigned long super_cache_scan(struct shrinker *shrink, |
62 | struct shrink_control *sc) | |
b0d40c92 DC |
63 | { |
64 | struct super_block *sb; | |
0a234c6d DC |
65 | long fs_objects = 0; |
66 | long total_objects; | |
67 | long freed = 0; | |
68 | long dentries; | |
69 | long inodes; | |
b0d40c92 DC |
70 | |
71 | sb = container_of(shrink, struct super_block, s_shrink); | |
72 | ||
73 | /* | |
74 | * Deadlock avoidance. We may hold various FS locks, and we don't want | |
75 | * to recurse into the FS that called us in clear_inode() and friends.. | |
76 | */ | |
0a234c6d DC |
77 | if (!(sc->gfp_mask & __GFP_FS)) |
78 | return SHRINK_STOP; | |
b0d40c92 | 79 | |
eb6ef3df | 80 | if (!trylock_super(sb)) |
0a234c6d | 81 | return SHRINK_STOP; |
b0d40c92 | 82 | |
d0407903 | 83 | if (sb->s_op->nr_cached_objects) |
4101b624 | 84 | fs_objects = sb->s_op->nr_cached_objects(sb, sc); |
0e1fdafd | 85 | |
503c358c VD |
86 | inodes = list_lru_shrink_count(&sb->s_inode_lru, sc); |
87 | dentries = list_lru_shrink_count(&sb->s_dentry_lru, sc); | |
f6041567 | 88 | total_objects = dentries + inodes + fs_objects + 1; |
475d0db7 TH |
89 | if (!total_objects) |
90 | total_objects = 1; | |
0e1fdafd | 91 | |
0a234c6d | 92 | /* proportion the scan between the caches */ |
f6041567 | 93 | dentries = mult_frac(sc->nr_to_scan, dentries, total_objects); |
bc3b14cb | 94 | inodes = mult_frac(sc->nr_to_scan, inodes, total_objects); |
503c358c | 95 | fs_objects = mult_frac(sc->nr_to_scan, fs_objects, total_objects); |
b0d40c92 | 96 | |
0a234c6d DC |
97 | /* |
98 | * prune the dcache first as the icache is pinned by it, then | |
99 | * prune the icache, followed by the filesystem specific caches | |
49e7e7ff VD |
100 | * |
101 | * Ensure that we always scan at least one object - memcg kmem | |
102 | * accounting uses this to fully empty the caches. | |
0a234c6d | 103 | */ |
49e7e7ff | 104 | sc->nr_to_scan = dentries + 1; |
503c358c | 105 | freed = prune_dcache_sb(sb, sc); |
49e7e7ff | 106 | sc->nr_to_scan = inodes + 1; |
503c358c | 107 | freed += prune_icache_sb(sb, sc); |
0a234c6d DC |
108 | |
109 | if (fs_objects) { | |
49e7e7ff | 110 | sc->nr_to_scan = fs_objects + 1; |
4101b624 | 111 | freed += sb->s_op->free_cached_objects(sb, sc); |
b0d40c92 DC |
112 | } |
113 | ||
eb6ef3df | 114 | up_read(&sb->s_umount); |
0a234c6d DC |
115 | return freed; |
116 | } | |
117 | ||
118 | static unsigned long super_cache_count(struct shrinker *shrink, | |
119 | struct shrink_control *sc) | |
120 | { | |
121 | struct super_block *sb; | |
122 | long total_objects = 0; | |
123 | ||
124 | sb = container_of(shrink, struct super_block, s_shrink); | |
125 | ||
d23da150 | 126 | /* |
79f546a6 DC |
127 | * We don't call trylock_super() here as it is a scalability bottleneck, |
128 | * so we're exposed to partial setup state. The shrinker rwsem does not | |
129 | * protect filesystem operations backing list_lru_shrink_count() or | |
130 | * s_op->nr_cached_objects(). Counts can change between | |
131 | * super_cache_count and super_cache_scan, so we really don't need locks | |
132 | * here. | |
133 | * | |
134 | * However, if we are currently mounting the superblock, the underlying | |
135 | * filesystem might be in a state of partial construction and hence it | |
136 | * is dangerous to access it. trylock_super() uses a SB_BORN check to | |
137 | * avoid this situation, so do the same here. The memory barrier is | |
138 | * matched with the one in mount_fs() as we don't hold locks here. | |
d23da150 | 139 | */ |
79f546a6 DC |
140 | if (!(sb->s_flags & SB_BORN)) |
141 | return 0; | |
142 | smp_rmb(); | |
143 | ||
0a234c6d | 144 | if (sb->s_op && sb->s_op->nr_cached_objects) |
4101b624 | 145 | total_objects = sb->s_op->nr_cached_objects(sb, sc); |
0a234c6d | 146 | |
503c358c VD |
147 | total_objects += list_lru_shrink_count(&sb->s_dentry_lru, sc); |
148 | total_objects += list_lru_shrink_count(&sb->s_inode_lru, sc); | |
0a234c6d | 149 | |
9b996468 KT |
150 | if (!total_objects) |
151 | return SHRINK_EMPTY; | |
152 | ||
55f841ce | 153 | total_objects = vfs_pressure_ratio(total_objects); |
0e1fdafd | 154 | return total_objects; |
b0d40c92 DC |
155 | } |
156 | ||
853b39a7 ON |
157 | static void destroy_super_work(struct work_struct *work) |
158 | { | |
159 | struct super_block *s = container_of(work, struct super_block, | |
160 | destroy_work); | |
161 | int i; | |
162 | ||
163 | for (i = 0; i < SB_FREEZE_LEVELS; i++) | |
8129ed29 | 164 | percpu_free_rwsem(&s->s_writers.rw_sem[i]); |
853b39a7 ON |
165 | kfree(s); |
166 | } | |
167 | ||
168 | static void destroy_super_rcu(struct rcu_head *head) | |
169 | { | |
170 | struct super_block *s = container_of(head, struct super_block, rcu); | |
171 | INIT_WORK(&s->destroy_work, destroy_super_work); | |
172 | schedule_work(&s->destroy_work); | |
173 | } | |
174 | ||
0200894d AV |
175 | /* Free a superblock that has never been seen by anyone */ |
176 | static void destroy_unused_super(struct super_block *s) | |
5accdf82 | 177 | { |
0200894d AV |
178 | if (!s) |
179 | return; | |
180 | up_write(&s->s_umount); | |
7eb5e882 AV |
181 | list_lru_destroy(&s->s_dentry_lru); |
182 | list_lru_destroy(&s->s_inode_lru); | |
7eb5e882 | 183 | security_sb_free(s); |
6e4eab57 | 184 | put_user_ns(s->s_user_ns); |
7eb5e882 | 185 | kfree(s->s_subtype); |
8e04944f | 186 | free_prealloced_shrinker(&s->s_shrink); |
0200894d AV |
187 | /* no delays needed */ |
188 | destroy_super_work(&s->destroy_work); | |
5accdf82 JK |
189 | } |
190 | ||
1da177e4 LT |
191 | /** |
192 | * alloc_super - create new superblock | |
fe2bbc48 | 193 | * @type: filesystem type superblock should belong to |
9249e17f | 194 | * @flags: the mount flags |
6e4eab57 | 195 | * @user_ns: User namespace for the super_block |
1da177e4 LT |
196 | * |
197 | * Allocates and initializes a new &struct super_block. alloc_super() | |
198 | * returns a pointer new superblock or %NULL if allocation had failed. | |
199 | */ | |
6e4eab57 EB |
200 | static struct super_block *alloc_super(struct file_system_type *type, int flags, |
201 | struct user_namespace *user_ns) | |
1da177e4 | 202 | { |
11b0b5ab | 203 | struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER); |
b87221de | 204 | static const struct super_operations default_op; |
7eb5e882 AV |
205 | int i; |
206 | ||
207 | if (!s) | |
208 | return NULL; | |
1da177e4 | 209 | |
b5bd856a | 210 | INIT_LIST_HEAD(&s->s_mounts); |
6e4eab57 | 211 | s->s_user_ns = get_user_ns(user_ns); |
ca0168e8 AV |
212 | init_rwsem(&s->s_umount); |
213 | lockdep_set_class(&s->s_umount, &type->s_umount_key); | |
214 | /* | |
215 | * sget() can have s_umount recursion. | |
216 | * | |
217 | * When it cannot find a suitable sb, it allocates a new | |
218 | * one (this one), and tries again to find a suitable old | |
219 | * one. | |
220 | * | |
221 | * In case that succeeds, it will acquire the s_umount | |
222 | * lock of the old one. Since these are clearly distrinct | |
223 | * locks, and this object isn't exposed yet, there's no | |
224 | * risk of deadlocks. | |
225 | * | |
226 | * Annotate this by putting this lock in a different | |
227 | * subclass. | |
228 | */ | |
229 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); | |
b5bd856a | 230 | |
7eb5e882 AV |
231 | if (security_sb_alloc(s)) |
232 | goto fail; | |
7b7a8665 | 233 | |
7eb5e882 | 234 | for (i = 0; i < SB_FREEZE_LEVELS; i++) { |
8129ed29 ON |
235 | if (__percpu_init_rwsem(&s->s_writers.rw_sem[i], |
236 | sb_writers_name[i], | |
237 | &type->s_writers_key[i])) | |
7eb5e882 | 238 | goto fail; |
1da177e4 | 239 | } |
7eb5e882 | 240 | init_waitqueue_head(&s->s_writers.wait_unfrozen); |
df0ce26c | 241 | s->s_bdi = &noop_backing_dev_info; |
7eb5e882 | 242 | s->s_flags = flags; |
cc50a07a | 243 | if (s->s_user_ns != &init_user_ns) |
67690f93 | 244 | s->s_iflags |= SB_I_NODEV; |
7eb5e882 | 245 | INIT_HLIST_NODE(&s->s_instances); |
f1ee6162 | 246 | INIT_HLIST_BL_HEAD(&s->s_roots); |
e97fedb9 | 247 | mutex_init(&s->s_sync_lock); |
7eb5e882 | 248 | INIT_LIST_HEAD(&s->s_inodes); |
74278da9 | 249 | spin_lock_init(&s->s_inode_list_lock); |
6c60d2b5 DC |
250 | INIT_LIST_HEAD(&s->s_inodes_wb); |
251 | spin_lock_init(&s->s_inode_wblist_lock); | |
7eb5e882 | 252 | |
7eb5e882 AV |
253 | s->s_count = 1; |
254 | atomic_set(&s->s_active, 1); | |
255 | mutex_init(&s->s_vfs_rename_mutex); | |
256 | lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key); | |
bc8230ee | 257 | init_rwsem(&s->s_dquot.dqio_sem); |
7eb5e882 AV |
258 | s->s_maxbytes = MAX_NON_LFS; |
259 | s->s_op = &default_op; | |
260 | s->s_time_gran = 1000000000; | |
188d20bc DD |
261 | s->s_time_min = TIME64_MIN; |
262 | s->s_time_max = TIME64_MAX; | |
3cb29d11 | 263 | s->cleancache_poolid = CLEANCACHE_NO_POOL; |
7eb5e882 AV |
264 | |
265 | s->s_shrink.seeks = DEFAULT_SEEKS; | |
266 | s->s_shrink.scan_objects = super_cache_scan; | |
267 | s->s_shrink.count_objects = super_cache_count; | |
268 | s->s_shrink.batch = 1024; | |
2acb60a0 | 269 | s->s_shrink.flags = SHRINKER_NUMA_AWARE | SHRINKER_MEMCG_AWARE; |
8e04944f TH |
270 | if (prealloc_shrinker(&s->s_shrink)) |
271 | goto fail; | |
c92e8e10 | 272 | if (list_lru_init_memcg(&s->s_dentry_lru, &s->s_shrink)) |
2b3648a6 | 273 | goto fail; |
c92e8e10 | 274 | if (list_lru_init_memcg(&s->s_inode_lru, &s->s_shrink)) |
2b3648a6 | 275 | goto fail; |
1da177e4 | 276 | return s; |
5ca302c8 | 277 | |
7eb5e882 | 278 | fail: |
0200894d | 279 | destroy_unused_super(s); |
7eb5e882 | 280 | return NULL; |
1da177e4 LT |
281 | } |
282 | ||
283 | /* Superblock refcounting */ | |
284 | ||
285 | /* | |
35cf7ba0 | 286 | * Drop a superblock's refcount. The caller must hold sb_lock. |
1da177e4 | 287 | */ |
c645b930 | 288 | static void __put_super(struct super_block *s) |
1da177e4 | 289 | { |
c645b930 AV |
290 | if (!--s->s_count) { |
291 | list_del_init(&s->s_list); | |
292 | WARN_ON(s->s_dentry_lru.node); | |
293 | WARN_ON(s->s_inode_lru.node); | |
294 | WARN_ON(!list_empty(&s->s_mounts)); | |
295 | security_sb_free(s); | |
22d94f49 | 296 | fscrypt_sb_free(s); |
c645b930 AV |
297 | put_user_ns(s->s_user_ns); |
298 | kfree(s->s_subtype); | |
299 | call_rcu(&s->rcu, destroy_super_rcu); | |
1da177e4 | 300 | } |
1da177e4 LT |
301 | } |
302 | ||
303 | /** | |
304 | * put_super - drop a temporary reference to superblock | |
305 | * @sb: superblock in question | |
306 | * | |
307 | * Drops a temporary reference, frees superblock if there's no | |
308 | * references left. | |
309 | */ | |
f47ec3f2 | 310 | static void put_super(struct super_block *sb) |
1da177e4 LT |
311 | { |
312 | spin_lock(&sb_lock); | |
313 | __put_super(sb); | |
314 | spin_unlock(&sb_lock); | |
315 | } | |
316 | ||
317 | ||
318 | /** | |
1712ac8f | 319 | * deactivate_locked_super - drop an active reference to superblock |
1da177e4 LT |
320 | * @s: superblock to deactivate |
321 | * | |
bd7ced98 | 322 | * Drops an active reference to superblock, converting it into a temporary |
1712ac8f | 323 | * one if there is no other active references left. In that case we |
1da177e4 LT |
324 | * tell fs driver to shut it down and drop the temporary reference we |
325 | * had just acquired. | |
1712ac8f AV |
326 | * |
327 | * Caller holds exclusive lock on superblock; that lock is released. | |
1da177e4 | 328 | */ |
1712ac8f | 329 | void deactivate_locked_super(struct super_block *s) |
1da177e4 LT |
330 | { |
331 | struct file_system_type *fs = s->s_type; | |
b20bd1a5 | 332 | if (atomic_dec_and_test(&s->s_active)) { |
3167760f | 333 | cleancache_invalidate_fs(s); |
b0d40c92 | 334 | unregister_shrinker(&s->s_shrink); |
28f2cd4f | 335 | fs->kill_sb(s); |
f5e1dd34 | 336 | |
c0a5b560 VD |
337 | /* |
338 | * Since list_lru_destroy() may sleep, we cannot call it from | |
339 | * put_super(), where we hold the sb_lock. Therefore we destroy | |
340 | * the lru lists right now. | |
341 | */ | |
342 | list_lru_destroy(&s->s_dentry_lru); | |
343 | list_lru_destroy(&s->s_inode_lru); | |
344 | ||
1da177e4 LT |
345 | put_filesystem(fs); |
346 | put_super(s); | |
1712ac8f AV |
347 | } else { |
348 | up_write(&s->s_umount); | |
1da177e4 LT |
349 | } |
350 | } | |
351 | ||
1712ac8f | 352 | EXPORT_SYMBOL(deactivate_locked_super); |
1da177e4 | 353 | |
74dbbdd7 | 354 | /** |
1712ac8f | 355 | * deactivate_super - drop an active reference to superblock |
74dbbdd7 AV |
356 | * @s: superblock to deactivate |
357 | * | |
1712ac8f AV |
358 | * Variant of deactivate_locked_super(), except that superblock is *not* |
359 | * locked by caller. If we are going to drop the final active reference, | |
360 | * lock will be acquired prior to that. | |
74dbbdd7 | 361 | */ |
1712ac8f | 362 | void deactivate_super(struct super_block *s) |
74dbbdd7 | 363 | { |
1712ac8f AV |
364 | if (!atomic_add_unless(&s->s_active, -1, 1)) { |
365 | down_write(&s->s_umount); | |
366 | deactivate_locked_super(s); | |
74dbbdd7 AV |
367 | } |
368 | } | |
369 | ||
1712ac8f | 370 | EXPORT_SYMBOL(deactivate_super); |
74dbbdd7 | 371 | |
1da177e4 LT |
372 | /** |
373 | * grab_super - acquire an active reference | |
374 | * @s: reference we are trying to make active | |
375 | * | |
376 | * Tries to acquire an active reference. grab_super() is used when we | |
377 | * had just found a superblock in super_blocks or fs_type->fs_supers | |
378 | * and want to turn it into a full-blown active reference. grab_super() | |
379 | * is called with sb_lock held and drops it. Returns 1 in case of | |
380 | * success, 0 if we had failed (superblock contents was already dead or | |
acfec9a5 AV |
381 | * dying when grab_super() had been called). Note that this is only |
382 | * called for superblocks not in rundown mode (== ones still on ->fs_supers | |
383 | * of their type), so increment of ->s_count is OK here. | |
1da177e4 | 384 | */ |
9c4dbee7 | 385 | static int grab_super(struct super_block *s) __releases(sb_lock) |
1da177e4 LT |
386 | { |
387 | s->s_count++; | |
388 | spin_unlock(&sb_lock); | |
389 | down_write(&s->s_umount); | |
e462ec50 | 390 | if ((s->s_flags & SB_BORN) && atomic_inc_not_zero(&s->s_active)) { |
acfec9a5 AV |
391 | put_super(s); |
392 | return 1; | |
393 | } | |
1da177e4 LT |
394 | up_write(&s->s_umount); |
395 | put_super(s); | |
1da177e4 LT |
396 | return 0; |
397 | } | |
398 | ||
12ad3ab6 | 399 | /* |
eb6ef3df | 400 | * trylock_super - try to grab ->s_umount shared |
331cbdee | 401 | * @sb: reference we are trying to grab |
12ad3ab6 | 402 | * |
eb6ef3df | 403 | * Try to prevent fs shutdown. This is used in places where we |
12ad3ab6 | 404 | * cannot take an active reference but we need to ensure that the |
eb6ef3df KK |
405 | * filesystem is not shut down while we are working on it. It returns |
406 | * false if we cannot acquire s_umount or if we lose the race and | |
407 | * filesystem already got into shutdown, and returns true with the s_umount | |
408 | * lock held in read mode in case of success. On successful return, | |
409 | * the caller must drop the s_umount lock when done. | |
410 | * | |
411 | * Note that unlike get_super() et.al. this one does *not* bump ->s_count. | |
412 | * The reason why it's safe is that we are OK with doing trylock instead | |
413 | * of down_read(). There's a couple of places that are OK with that, but | |
414 | * it's very much not a general-purpose interface. | |
12ad3ab6 | 415 | */ |
eb6ef3df | 416 | bool trylock_super(struct super_block *sb) |
12ad3ab6 | 417 | { |
12ad3ab6 | 418 | if (down_read_trylock(&sb->s_umount)) { |
eb6ef3df | 419 | if (!hlist_unhashed(&sb->s_instances) && |
e462ec50 | 420 | sb->s_root && (sb->s_flags & SB_BORN)) |
12ad3ab6 DC |
421 | return true; |
422 | up_read(&sb->s_umount); | |
423 | } | |
424 | ||
12ad3ab6 DC |
425 | return false; |
426 | } | |
427 | ||
1da177e4 LT |
428 | /** |
429 | * generic_shutdown_super - common helper for ->kill_sb() | |
430 | * @sb: superblock to kill | |
431 | * | |
432 | * generic_shutdown_super() does all fs-independent work on superblock | |
433 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects | |
434 | * that need destruction out of superblock, call generic_shutdown_super() | |
435 | * and release aforementioned objects. Note: dentries and inodes _are_ | |
436 | * taken care of and do not need specific handling. | |
c636ebdb DH |
437 | * |
438 | * Upon calling this function, the filesystem may no longer alter or | |
439 | * rearrange the set of dentries belonging to this super_block, nor may it | |
440 | * change the attachments of dentries to inodes. | |
1da177e4 LT |
441 | */ |
442 | void generic_shutdown_super(struct super_block *sb) | |
443 | { | |
ee9b6d61 | 444 | const struct super_operations *sop = sb->s_op; |
1da177e4 | 445 | |
c636ebdb DH |
446 | if (sb->s_root) { |
447 | shrink_dcache_for_umount(sb); | |
60b0680f | 448 | sync_filesystem(sb); |
e462ec50 | 449 | sb->s_flags &= ~SB_ACTIVE; |
efaee192 | 450 | |
1e6cb723 | 451 | fsnotify_sb_delete(sb); |
a1a0e23e | 452 | cgroup_writeback_umount(); |
63997e98 AV |
453 | |
454 | evict_inodes(sb); | |
1da177e4 | 455 | |
7b7a8665 CH |
456 | if (sb->s_dio_done_wq) { |
457 | destroy_workqueue(sb->s_dio_done_wq); | |
458 | sb->s_dio_done_wq = NULL; | |
459 | } | |
460 | ||
1da177e4 LT |
461 | if (sop->put_super) |
462 | sop->put_super(sb); | |
463 | ||
63997e98 | 464 | if (!list_empty(&sb->s_inodes)) { |
7b4fe29e DJ |
465 | printk("VFS: Busy inodes after unmount of %s. " |
466 | "Self-destruct in 5 seconds. Have a nice day...\n", | |
467 | sb->s_id); | |
1da177e4 | 468 | } |
1da177e4 LT |
469 | } |
470 | spin_lock(&sb_lock); | |
471 | /* should be initialized for __put_super_and_need_restart() */ | |
a5166169 | 472 | hlist_del_init(&sb->s_instances); |
1da177e4 LT |
473 | spin_unlock(&sb_lock); |
474 | up_write(&sb->s_umount); | |
c1844d53 | 475 | if (sb->s_bdi != &noop_backing_dev_info) { |
fca39346 JK |
476 | bdi_put(sb->s_bdi); |
477 | sb->s_bdi = &noop_backing_dev_info; | |
fca39346 | 478 | } |
1da177e4 LT |
479 | } |
480 | ||
481 | EXPORT_SYMBOL(generic_shutdown_super); | |
482 | ||
20284ab7 | 483 | bool mount_capable(struct fs_context *fc) |
0ce0cf12 | 484 | { |
20284ab7 | 485 | if (!(fc->fs_type->fs_flags & FS_USERNS_MOUNT)) |
0ce0cf12 AV |
486 | return capable(CAP_SYS_ADMIN); |
487 | else | |
c2c44ec2 | 488 | return ns_capable(fc->user_ns, CAP_SYS_ADMIN); |
0ce0cf12 AV |
489 | } |
490 | ||
cb50b348 AV |
491 | /** |
492 | * sget_fc - Find or create a superblock | |
493 | * @fc: Filesystem context. | |
494 | * @test: Comparison callback | |
495 | * @set: Setup callback | |
496 | * | |
497 | * Find or create a superblock using the parameters stored in the filesystem | |
498 | * context and the two callback functions. | |
499 | * | |
500 | * If an extant superblock is matched, then that will be returned with an | |
501 | * elevated reference count that the caller must transfer or discard. | |
502 | * | |
503 | * If no match is made, a new superblock will be allocated and basic | |
504 | * initialisation will be performed (s_type, s_fs_info and s_id will be set and | |
505 | * the set() callback will be invoked), the superblock will be published and it | |
506 | * will be returned in a partially constructed state with SB_BORN and SB_ACTIVE | |
507 | * as yet unset. | |
508 | */ | |
509 | struct super_block *sget_fc(struct fs_context *fc, | |
510 | int (*test)(struct super_block *, struct fs_context *), | |
511 | int (*set)(struct super_block *, struct fs_context *)) | |
512 | { | |
513 | struct super_block *s = NULL; | |
514 | struct super_block *old; | |
515 | struct user_namespace *user_ns = fc->global ? &init_user_ns : fc->user_ns; | |
516 | int err; | |
517 | ||
cb50b348 AV |
518 | retry: |
519 | spin_lock(&sb_lock); | |
520 | if (test) { | |
521 | hlist_for_each_entry(old, &fc->fs_type->fs_supers, s_instances) { | |
522 | if (test(old, fc)) | |
523 | goto share_extant_sb; | |
524 | } | |
525 | } | |
526 | if (!s) { | |
527 | spin_unlock(&sb_lock); | |
528 | s = alloc_super(fc->fs_type, fc->sb_flags, user_ns); | |
529 | if (!s) | |
530 | return ERR_PTR(-ENOMEM); | |
531 | goto retry; | |
532 | } | |
533 | ||
534 | s->s_fs_info = fc->s_fs_info; | |
535 | err = set(s, fc); | |
536 | if (err) { | |
537 | s->s_fs_info = NULL; | |
538 | spin_unlock(&sb_lock); | |
539 | destroy_unused_super(s); | |
540 | return ERR_PTR(err); | |
541 | } | |
542 | fc->s_fs_info = NULL; | |
543 | s->s_type = fc->fs_type; | |
c80fa7c8 | 544 | s->s_iflags |= fc->s_iflags; |
cb50b348 AV |
545 | strlcpy(s->s_id, s->s_type->name, sizeof(s->s_id)); |
546 | list_add_tail(&s->s_list, &super_blocks); | |
547 | hlist_add_head(&s->s_instances, &s->s_type->fs_supers); | |
548 | spin_unlock(&sb_lock); | |
549 | get_filesystem(s->s_type); | |
550 | register_shrinker_prepared(&s->s_shrink); | |
551 | return s; | |
552 | ||
553 | share_extant_sb: | |
554 | if (user_ns != old->s_user_ns) { | |
555 | spin_unlock(&sb_lock); | |
556 | destroy_unused_super(s); | |
557 | return ERR_PTR(-EBUSY); | |
558 | } | |
559 | if (!grab_super(old)) | |
560 | goto retry; | |
561 | destroy_unused_super(s); | |
562 | return old; | |
563 | } | |
564 | EXPORT_SYMBOL(sget_fc); | |
565 | ||
1da177e4 | 566 | /** |
023d066a DH |
567 | * sget - find or create a superblock |
568 | * @type: filesystem type superblock should belong to | |
569 | * @test: comparison callback | |
570 | * @set: setup callback | |
571 | * @flags: mount flags | |
572 | * @data: argument to each of them | |
1da177e4 | 573 | */ |
023d066a | 574 | struct super_block *sget(struct file_system_type *type, |
1da177e4 LT |
575 | int (*test)(struct super_block *,void *), |
576 | int (*set)(struct super_block *,void *), | |
023d066a | 577 | int flags, |
1da177e4 LT |
578 | void *data) |
579 | { | |
023d066a | 580 | struct user_namespace *user_ns = current_user_ns(); |
1da177e4 | 581 | struct super_block *s = NULL; |
d4730127 | 582 | struct super_block *old; |
1da177e4 LT |
583 | int err; |
584 | ||
023d066a DH |
585 | /* We don't yet pass the user namespace of the parent |
586 | * mount through to here so always use &init_user_ns | |
587 | * until that changes. | |
588 | */ | |
589 | if (flags & SB_SUBMOUNT) | |
590 | user_ns = &init_user_ns; | |
591 | ||
1da177e4 LT |
592 | retry: |
593 | spin_lock(&sb_lock); | |
d4730127 | 594 | if (test) { |
b67bfe0d | 595 | hlist_for_each_entry(old, &type->fs_supers, s_instances) { |
d4730127 MK |
596 | if (!test(old, data)) |
597 | continue; | |
6e4eab57 EB |
598 | if (user_ns != old->s_user_ns) { |
599 | spin_unlock(&sb_lock); | |
0200894d | 600 | destroy_unused_super(s); |
6e4eab57 EB |
601 | return ERR_PTR(-EBUSY); |
602 | } | |
d4730127 MK |
603 | if (!grab_super(old)) |
604 | goto retry; | |
0200894d | 605 | destroy_unused_super(s); |
d4730127 MK |
606 | return old; |
607 | } | |
1da177e4 LT |
608 | } |
609 | if (!s) { | |
610 | spin_unlock(&sb_lock); | |
e462ec50 | 611 | s = alloc_super(type, (flags & ~SB_SUBMOUNT), user_ns); |
1da177e4 LT |
612 | if (!s) |
613 | return ERR_PTR(-ENOMEM); | |
614 | goto retry; | |
615 | } | |
dd111b31 | 616 | |
1da177e4 LT |
617 | err = set(s, data); |
618 | if (err) { | |
619 | spin_unlock(&sb_lock); | |
0200894d | 620 | destroy_unused_super(s); |
1da177e4 LT |
621 | return ERR_PTR(err); |
622 | } | |
623 | s->s_type = type; | |
624 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); | |
625 | list_add_tail(&s->s_list, &super_blocks); | |
a5166169 | 626 | hlist_add_head(&s->s_instances, &type->fs_supers); |
1da177e4 LT |
627 | spin_unlock(&sb_lock); |
628 | get_filesystem(type); | |
8e04944f | 629 | register_shrinker_prepared(&s->s_shrink); |
1da177e4 LT |
630 | return s; |
631 | } | |
1da177e4 LT |
632 | EXPORT_SYMBOL(sget); |
633 | ||
634 | void drop_super(struct super_block *sb) | |
635 | { | |
636 | up_read(&sb->s_umount); | |
637 | put_super(sb); | |
638 | } | |
639 | ||
640 | EXPORT_SYMBOL(drop_super); | |
641 | ||
ba6379f7 JK |
642 | void drop_super_exclusive(struct super_block *sb) |
643 | { | |
644 | up_write(&sb->s_umount); | |
645 | put_super(sb); | |
646 | } | |
647 | EXPORT_SYMBOL(drop_super_exclusive); | |
648 | ||
fa7c1d50 MG |
649 | static void __iterate_supers(void (*f)(struct super_block *)) |
650 | { | |
651 | struct super_block *sb, *p = NULL; | |
652 | ||
653 | spin_lock(&sb_lock); | |
654 | list_for_each_entry(sb, &super_blocks, s_list) { | |
655 | if (hlist_unhashed(&sb->s_instances)) | |
656 | continue; | |
657 | sb->s_count++; | |
658 | spin_unlock(&sb_lock); | |
659 | ||
660 | f(sb); | |
661 | ||
662 | spin_lock(&sb_lock); | |
663 | if (p) | |
664 | __put_super(p); | |
665 | p = sb; | |
666 | } | |
667 | if (p) | |
668 | __put_super(p); | |
669 | spin_unlock(&sb_lock); | |
670 | } | |
01a05b33 AV |
671 | /** |
672 | * iterate_supers - call function for all active superblocks | |
673 | * @f: function to call | |
674 | * @arg: argument to pass to it | |
675 | * | |
676 | * Scans the superblock list and calls given function, passing it | |
677 | * locked superblock and given argument. | |
678 | */ | |
679 | void iterate_supers(void (*f)(struct super_block *, void *), void *arg) | |
680 | { | |
dca33252 | 681 | struct super_block *sb, *p = NULL; |
01a05b33 AV |
682 | |
683 | spin_lock(&sb_lock); | |
dca33252 | 684 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 685 | if (hlist_unhashed(&sb->s_instances)) |
01a05b33 AV |
686 | continue; |
687 | sb->s_count++; | |
688 | spin_unlock(&sb_lock); | |
689 | ||
690 | down_read(&sb->s_umount); | |
e462ec50 | 691 | if (sb->s_root && (sb->s_flags & SB_BORN)) |
01a05b33 AV |
692 | f(sb, arg); |
693 | up_read(&sb->s_umount); | |
694 | ||
695 | spin_lock(&sb_lock); | |
dca33252 AV |
696 | if (p) |
697 | __put_super(p); | |
698 | p = sb; | |
01a05b33 | 699 | } |
dca33252 AV |
700 | if (p) |
701 | __put_super(p); | |
01a05b33 AV |
702 | spin_unlock(&sb_lock); |
703 | } | |
704 | ||
43e15cdb AV |
705 | /** |
706 | * iterate_supers_type - call function for superblocks of given type | |
707 | * @type: fs type | |
708 | * @f: function to call | |
709 | * @arg: argument to pass to it | |
710 | * | |
711 | * Scans the superblock list and calls given function, passing it | |
712 | * locked superblock and given argument. | |
713 | */ | |
714 | void iterate_supers_type(struct file_system_type *type, | |
715 | void (*f)(struct super_block *, void *), void *arg) | |
716 | { | |
717 | struct super_block *sb, *p = NULL; | |
718 | ||
719 | spin_lock(&sb_lock); | |
b67bfe0d | 720 | hlist_for_each_entry(sb, &type->fs_supers, s_instances) { |
43e15cdb AV |
721 | sb->s_count++; |
722 | spin_unlock(&sb_lock); | |
723 | ||
724 | down_read(&sb->s_umount); | |
e462ec50 | 725 | if (sb->s_root && (sb->s_flags & SB_BORN)) |
43e15cdb AV |
726 | f(sb, arg); |
727 | up_read(&sb->s_umount); | |
728 | ||
729 | spin_lock(&sb_lock); | |
730 | if (p) | |
731 | __put_super(p); | |
732 | p = sb; | |
733 | } | |
734 | if (p) | |
735 | __put_super(p); | |
736 | spin_unlock(&sb_lock); | |
737 | } | |
738 | ||
739 | EXPORT_SYMBOL(iterate_supers_type); | |
740 | ||
ba6379f7 | 741 | static struct super_block *__get_super(struct block_device *bdev, bool excl) |
1da177e4 | 742 | { |
618f0636 KK |
743 | struct super_block *sb; |
744 | ||
1da177e4 LT |
745 | if (!bdev) |
746 | return NULL; | |
618f0636 | 747 | |
1da177e4 | 748 | spin_lock(&sb_lock); |
618f0636 KK |
749 | rescan: |
750 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 751 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 752 | continue; |
618f0636 KK |
753 | if (sb->s_bdev == bdev) { |
754 | sb->s_count++; | |
1da177e4 | 755 | spin_unlock(&sb_lock); |
ba6379f7 JK |
756 | if (!excl) |
757 | down_read(&sb->s_umount); | |
758 | else | |
759 | down_write(&sb->s_umount); | |
df40c01a | 760 | /* still alive? */ |
e462ec50 | 761 | if (sb->s_root && (sb->s_flags & SB_BORN)) |
618f0636 | 762 | return sb; |
ba6379f7 JK |
763 | if (!excl) |
764 | up_read(&sb->s_umount); | |
765 | else | |
766 | up_write(&sb->s_umount); | |
df40c01a | 767 | /* nope, got unmounted */ |
618f0636 | 768 | spin_lock(&sb_lock); |
df40c01a AV |
769 | __put_super(sb); |
770 | goto rescan; | |
1da177e4 LT |
771 | } |
772 | } | |
773 | spin_unlock(&sb_lock); | |
774 | return NULL; | |
775 | } | |
776 | ||
6b6dc836 | 777 | /** |
ba6379f7 | 778 | * get_super - get the superblock of a device |
6b6dc836 JK |
779 | * @bdev: device to get the superblock for |
780 | * | |
781 | * Scans the superblock list and finds the superblock of the file system | |
ba6379f7 | 782 | * mounted on the device given. %NULL is returned if no match is found. |
6b6dc836 | 783 | */ |
ba6379f7 JK |
784 | struct super_block *get_super(struct block_device *bdev) |
785 | { | |
786 | return __get_super(bdev, false); | |
787 | } | |
788 | EXPORT_SYMBOL(get_super); | |
789 | ||
790 | static struct super_block *__get_super_thawed(struct block_device *bdev, | |
791 | bool excl) | |
6b6dc836 JK |
792 | { |
793 | while (1) { | |
ba6379f7 | 794 | struct super_block *s = __get_super(bdev, excl); |
5accdf82 | 795 | if (!s || s->s_writers.frozen == SB_UNFROZEN) |
6b6dc836 | 796 | return s; |
ba6379f7 JK |
797 | if (!excl) |
798 | up_read(&s->s_umount); | |
799 | else | |
800 | up_write(&s->s_umount); | |
5accdf82 JK |
801 | wait_event(s->s_writers.wait_unfrozen, |
802 | s->s_writers.frozen == SB_UNFROZEN); | |
6b6dc836 JK |
803 | put_super(s); |
804 | } | |
805 | } | |
ba6379f7 JK |
806 | |
807 | /** | |
808 | * get_super_thawed - get thawed superblock of a device | |
809 | * @bdev: device to get the superblock for | |
810 | * | |
811 | * Scans the superblock list and finds the superblock of the file system | |
812 | * mounted on the device. The superblock is returned once it is thawed | |
813 | * (or immediately if it was not frozen). %NULL is returned if no match | |
814 | * is found. | |
815 | */ | |
816 | struct super_block *get_super_thawed(struct block_device *bdev) | |
817 | { | |
818 | return __get_super_thawed(bdev, false); | |
819 | } | |
6b6dc836 JK |
820 | EXPORT_SYMBOL(get_super_thawed); |
821 | ||
ba6379f7 JK |
822 | /** |
823 | * get_super_exclusive_thawed - get thawed superblock of a device | |
824 | * @bdev: device to get the superblock for | |
825 | * | |
826 | * Scans the superblock list and finds the superblock of the file system | |
827 | * mounted on the device. The superblock is returned once it is thawed | |
828 | * (or immediately if it was not frozen) and s_umount semaphore is held | |
829 | * in exclusive mode. %NULL is returned if no match is found. | |
830 | */ | |
831 | struct super_block *get_super_exclusive_thawed(struct block_device *bdev) | |
832 | { | |
833 | return __get_super_thawed(bdev, true); | |
834 | } | |
835 | EXPORT_SYMBOL(get_super_exclusive_thawed); | |
836 | ||
4504230a CH |
837 | /** |
838 | * get_active_super - get an active reference to the superblock of a device | |
839 | * @bdev: device to get the superblock for | |
840 | * | |
841 | * Scans the superblock list and finds the superblock of the file system | |
842 | * mounted on the device given. Returns the superblock with an active | |
d3f21473 | 843 | * reference or %NULL if none was found. |
4504230a CH |
844 | */ |
845 | struct super_block *get_active_super(struct block_device *bdev) | |
846 | { | |
847 | struct super_block *sb; | |
848 | ||
849 | if (!bdev) | |
850 | return NULL; | |
851 | ||
1494583d | 852 | restart: |
4504230a CH |
853 | spin_lock(&sb_lock); |
854 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 855 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 856 | continue; |
1494583d | 857 | if (sb->s_bdev == bdev) { |
acfec9a5 | 858 | if (!grab_super(sb)) |
1494583d | 859 | goto restart; |
acfec9a5 AV |
860 | up_write(&sb->s_umount); |
861 | return sb; | |
1494583d | 862 | } |
4504230a CH |
863 | } |
864 | spin_unlock(&sb_lock); | |
865 | return NULL; | |
866 | } | |
dd111b31 | 867 | |
df40c01a | 868 | struct super_block *user_get_super(dev_t dev) |
1da177e4 | 869 | { |
618f0636 | 870 | struct super_block *sb; |
1da177e4 | 871 | |
1da177e4 | 872 | spin_lock(&sb_lock); |
618f0636 KK |
873 | rescan: |
874 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 875 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 876 | continue; |
618f0636 KK |
877 | if (sb->s_dev == dev) { |
878 | sb->s_count++; | |
1da177e4 | 879 | spin_unlock(&sb_lock); |
618f0636 | 880 | down_read(&sb->s_umount); |
df40c01a | 881 | /* still alive? */ |
e462ec50 | 882 | if (sb->s_root && (sb->s_flags & SB_BORN)) |
618f0636 KK |
883 | return sb; |
884 | up_read(&sb->s_umount); | |
df40c01a | 885 | /* nope, got unmounted */ |
618f0636 | 886 | spin_lock(&sb_lock); |
df40c01a AV |
887 | __put_super(sb); |
888 | goto rescan; | |
1da177e4 LT |
889 | } |
890 | } | |
891 | spin_unlock(&sb_lock); | |
892 | return NULL; | |
893 | } | |
894 | ||
1da177e4 | 895 | /** |
8d0347f6 DH |
896 | * reconfigure_super - asks filesystem to change superblock parameters |
897 | * @fc: The superblock and configuration | |
1da177e4 | 898 | * |
8d0347f6 | 899 | * Alters the configuration parameters of a live superblock. |
1da177e4 | 900 | */ |
8d0347f6 | 901 | int reconfigure_super(struct fs_context *fc) |
1da177e4 | 902 | { |
8d0347f6 | 903 | struct super_block *sb = fc->root->d_sb; |
1da177e4 | 904 | int retval; |
8d0347f6 DH |
905 | bool remount_ro = false; |
906 | bool force = fc->sb_flags & SB_FORCE; | |
4504230a | 907 | |
8d0347f6 DH |
908 | if (fc->sb_flags_mask & ~MS_RMT_MASK) |
909 | return -EINVAL; | |
5accdf82 | 910 | if (sb->s_writers.frozen != SB_UNFROZEN) |
4504230a CH |
911 | return -EBUSY; |
912 | ||
8d0347f6 DH |
913 | retval = security_sb_remount(sb, fc->security); |
914 | if (retval) | |
915 | return retval; | |
916 | ||
917 | if (fc->sb_flags_mask & SB_RDONLY) { | |
9361401e | 918 | #ifdef CONFIG_BLOCK |
8d0347f6 DH |
919 | if (!(fc->sb_flags & SB_RDONLY) && bdev_read_only(sb->s_bdev)) |
920 | return -EACCES; | |
9361401e | 921 | #endif |
4504230a | 922 | |
8d0347f6 DH |
923 | remount_ro = (fc->sb_flags & SB_RDONLY) && !sb_rdonly(sb); |
924 | } | |
d208bbdd | 925 | |
0aec09d0 | 926 | if (remount_ro) { |
fdab684d | 927 | if (!hlist_empty(&sb->s_pins)) { |
0aec09d0 | 928 | up_write(&sb->s_umount); |
fdab684d | 929 | group_pin_kill(&sb->s_pins); |
0aec09d0 AV |
930 | down_write(&sb->s_umount); |
931 | if (!sb->s_root) | |
932 | return 0; | |
933 | if (sb->s_writers.frozen != SB_UNFROZEN) | |
934 | return -EBUSY; | |
8d0347f6 | 935 | remount_ro = !sb_rdonly(sb); |
0aec09d0 AV |
936 | } |
937 | } | |
938 | shrink_dcache_sb(sb); | |
939 | ||
8d0347f6 DH |
940 | /* If we are reconfiguring to RDONLY and current sb is read/write, |
941 | * make sure there are no files open for writing. | |
942 | */ | |
d208bbdd | 943 | if (remount_ro) { |
4ed5e82f | 944 | if (force) { |
eee5cc27 AV |
945 | sb->s_readonly_remount = 1; |
946 | smp_wmb(); | |
4ed5e82f MS |
947 | } else { |
948 | retval = sb_prepare_remount_readonly(sb); | |
949 | if (retval) | |
950 | return retval; | |
4ed5e82f | 951 | } |
1da177e4 LT |
952 | } |
953 | ||
f3a09c92 AV |
954 | if (fc->ops->reconfigure) { |
955 | retval = fc->ops->reconfigure(fc); | |
956 | if (retval) { | |
957 | if (!force) | |
958 | goto cancel_readonly; | |
959 | /* If forced remount, go ahead despite any errors */ | |
960 | WARN(1, "forced remount of a %s fs returned %i\n", | |
961 | sb->s_type->name, retval); | |
962 | } | |
1da177e4 | 963 | } |
8d0347f6 DH |
964 | |
965 | WRITE_ONCE(sb->s_flags, ((sb->s_flags & ~fc->sb_flags_mask) | | |
966 | (fc->sb_flags & fc->sb_flags_mask))); | |
4ed5e82f MS |
967 | /* Needs to be ordered wrt mnt_is_readonly() */ |
968 | smp_wmb(); | |
969 | sb->s_readonly_remount = 0; | |
c79d967d | 970 | |
d208bbdd NP |
971 | /* |
972 | * Some filesystems modify their metadata via some other path than the | |
973 | * bdev buffer cache (eg. use a private mapping, or directories in | |
974 | * pagecache, etc). Also file data modifications go via their own | |
975 | * mappings. So If we try to mount readonly then copy the filesystem | |
976 | * from bdev, we could get stale data, so invalidate it to give a best | |
977 | * effort at coherency. | |
978 | */ | |
979 | if (remount_ro && sb->s_bdev) | |
980 | invalidate_bdev(sb->s_bdev); | |
1da177e4 | 981 | return 0; |
4ed5e82f MS |
982 | |
983 | cancel_readonly: | |
984 | sb->s_readonly_remount = 0; | |
985 | return retval; | |
1da177e4 LT |
986 | } |
987 | ||
fa7c1d50 | 988 | static void do_emergency_remount_callback(struct super_block *sb) |
1da177e4 | 989 | { |
fa7c1d50 MG |
990 | down_write(&sb->s_umount); |
991 | if (sb->s_root && sb->s_bdev && (sb->s_flags & SB_BORN) && | |
992 | !sb_rdonly(sb)) { | |
8d0347f6 DH |
993 | struct fs_context *fc; |
994 | ||
995 | fc = fs_context_for_reconfigure(sb->s_root, | |
996 | SB_RDONLY | SB_FORCE, SB_RDONLY); | |
997 | if (!IS_ERR(fc)) { | |
998 | if (parse_monolithic_mount_data(fc, NULL) == 0) | |
999 | (void)reconfigure_super(fc); | |
1000 | put_fs_context(fc); | |
1001 | } | |
1da177e4 | 1002 | } |
fa7c1d50 MG |
1003 | up_write(&sb->s_umount); |
1004 | } | |
1005 | ||
1006 | static void do_emergency_remount(struct work_struct *work) | |
1007 | { | |
1008 | __iterate_supers(do_emergency_remount_callback); | |
a2a9537a | 1009 | kfree(work); |
1da177e4 LT |
1010 | printk("Emergency Remount complete\n"); |
1011 | } | |
1012 | ||
1013 | void emergency_remount(void) | |
1014 | { | |
a2a9537a JA |
1015 | struct work_struct *work; |
1016 | ||
1017 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | |
1018 | if (work) { | |
1019 | INIT_WORK(work, do_emergency_remount); | |
1020 | schedule_work(work); | |
1021 | } | |
1da177e4 LT |
1022 | } |
1023 | ||
08fdc8a0 MG |
1024 | static void do_thaw_all_callback(struct super_block *sb) |
1025 | { | |
1026 | down_write(&sb->s_umount); | |
1c18d2a1 | 1027 | if (sb->s_root && sb->s_flags & SB_BORN) { |
08fdc8a0 MG |
1028 | emergency_thaw_bdev(sb); |
1029 | thaw_super_locked(sb); | |
1030 | } else { | |
1031 | up_write(&sb->s_umount); | |
1032 | } | |
1033 | } | |
1034 | ||
1035 | static void do_thaw_all(struct work_struct *work) | |
1036 | { | |
1037 | __iterate_supers(do_thaw_all_callback); | |
1038 | kfree(work); | |
1039 | printk(KERN_WARNING "Emergency Thaw complete\n"); | |
1040 | } | |
1041 | ||
1042 | /** | |
1043 | * emergency_thaw_all -- forcibly thaw every frozen filesystem | |
1044 | * | |
1045 | * Used for emergency unfreeze of all filesystems via SysRq | |
1046 | */ | |
1047 | void emergency_thaw_all(void) | |
1048 | { | |
1049 | struct work_struct *work; | |
1050 | ||
1051 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | |
1052 | if (work) { | |
1053 | INIT_WORK(work, do_thaw_all); | |
1054 | schedule_work(work); | |
1055 | } | |
1056 | } | |
1057 | ||
ad76cbc6 | 1058 | static DEFINE_IDA(unnamed_dev_ida); |
1da177e4 | 1059 | |
5a66847e MW |
1060 | /** |
1061 | * get_anon_bdev - Allocate a block device for filesystems which don't have one. | |
1062 | * @p: Pointer to a dev_t. | |
1063 | * | |
1064 | * Filesystems which don't use real block devices can call this function | |
1065 | * to allocate a virtual block device. | |
1066 | * | |
1067 | * Context: Any context. Frequently called while holding sb_lock. | |
1068 | * Return: 0 on success, -EMFILE if there are no anonymous bdevs left | |
1069 | * or -ENOMEM if memory allocation failed. | |
1070 | */ | |
0ee5dc67 | 1071 | int get_anon_bdev(dev_t *p) |
1da177e4 LT |
1072 | { |
1073 | int dev; | |
5a66847e MW |
1074 | |
1075 | /* | |
1076 | * Many userspace utilities consider an FSID of 0 invalid. | |
1077 | * Always return at least 1 from get_anon_bdev. | |
1078 | */ | |
1079 | dev = ida_alloc_range(&unnamed_dev_ida, 1, (1 << MINORBITS) - 1, | |
1080 | GFP_ATOMIC); | |
1081 | if (dev == -ENOSPC) | |
1082 | dev = -EMFILE; | |
1083 | if (dev < 0) | |
1084 | return dev; | |
1085 | ||
1086 | *p = MKDEV(0, dev); | |
1da177e4 LT |
1087 | return 0; |
1088 | } | |
0ee5dc67 | 1089 | EXPORT_SYMBOL(get_anon_bdev); |
1da177e4 | 1090 | |
0ee5dc67 | 1091 | void free_anon_bdev(dev_t dev) |
1da177e4 | 1092 | { |
5a66847e | 1093 | ida_free(&unnamed_dev_ida, MINOR(dev)); |
1da177e4 | 1094 | } |
0ee5dc67 AV |
1095 | EXPORT_SYMBOL(free_anon_bdev); |
1096 | ||
1097 | int set_anon_super(struct super_block *s, void *data) | |
1098 | { | |
df0ce26c | 1099 | return get_anon_bdev(&s->s_dev); |
0ee5dc67 | 1100 | } |
0ee5dc67 AV |
1101 | EXPORT_SYMBOL(set_anon_super); |
1102 | ||
1103 | void kill_anon_super(struct super_block *sb) | |
1104 | { | |
1105 | dev_t dev = sb->s_dev; | |
1106 | generic_shutdown_super(sb); | |
1107 | free_anon_bdev(dev); | |
1108 | } | |
1da177e4 LT |
1109 | EXPORT_SYMBOL(kill_anon_super); |
1110 | ||
1da177e4 LT |
1111 | void kill_litter_super(struct super_block *sb) |
1112 | { | |
1113 | if (sb->s_root) | |
1114 | d_genocide(sb->s_root); | |
1115 | kill_anon_super(sb); | |
1116 | } | |
1da177e4 LT |
1117 | EXPORT_SYMBOL(kill_litter_super); |
1118 | ||
cb50b348 AV |
1119 | int set_anon_super_fc(struct super_block *sb, struct fs_context *fc) |
1120 | { | |
1121 | return set_anon_super(sb, NULL); | |
1122 | } | |
1123 | EXPORT_SYMBOL(set_anon_super_fc); | |
1124 | ||
1125 | static int test_keyed_super(struct super_block *sb, struct fs_context *fc) | |
1126 | { | |
1127 | return sb->s_fs_info == fc->s_fs_info; | |
1128 | } | |
1129 | ||
1130 | static int test_single_super(struct super_block *s, struct fs_context *fc) | |
1131 | { | |
1132 | return 1; | |
1133 | } | |
1134 | ||
1135 | /** | |
1136 | * vfs_get_super - Get a superblock with a search key set in s_fs_info. | |
1137 | * @fc: The filesystem context holding the parameters | |
1138 | * @keying: How to distinguish superblocks | |
1139 | * @fill_super: Helper to initialise a new superblock | |
1140 | * | |
1141 | * Search for a superblock and create a new one if not found. The search | |
1142 | * criterion is controlled by @keying. If the search fails, a new superblock | |
1143 | * is created and @fill_super() is called to initialise it. | |
1144 | * | |
1145 | * @keying can take one of a number of values: | |
1146 | * | |
1147 | * (1) vfs_get_single_super - Only one superblock of this type may exist on the | |
1148 | * system. This is typically used for special system filesystems. | |
1149 | * | |
1150 | * (2) vfs_get_keyed_super - Multiple superblocks may exist, but they must have | |
1151 | * distinct keys (where the key is in s_fs_info). Searching for the same | |
1152 | * key again will turn up the superblock for that key. | |
1153 | * | |
1154 | * (3) vfs_get_independent_super - Multiple superblocks may exist and are | |
1155 | * unkeyed. Each call will get a new superblock. | |
1156 | * | |
1157 | * A permissions check is made by sget_fc() unless we're getting a superblock | |
1158 | * for a kernel-internal mount or a submount. | |
1159 | */ | |
1160 | int vfs_get_super(struct fs_context *fc, | |
1161 | enum vfs_get_super_keying keying, | |
1162 | int (*fill_super)(struct super_block *sb, | |
1163 | struct fs_context *fc)) | |
1164 | { | |
1165 | int (*test)(struct super_block *, struct fs_context *); | |
1166 | struct super_block *sb; | |
43ce4c1f | 1167 | int err; |
cb50b348 AV |
1168 | |
1169 | switch (keying) { | |
1170 | case vfs_get_single_super: | |
43ce4c1f | 1171 | case vfs_get_single_reconf_super: |
cb50b348 AV |
1172 | test = test_single_super; |
1173 | break; | |
1174 | case vfs_get_keyed_super: | |
1175 | test = test_keyed_super; | |
1176 | break; | |
1177 | case vfs_get_independent_super: | |
1178 | test = NULL; | |
1179 | break; | |
1180 | default: | |
1181 | BUG(); | |
1182 | } | |
1183 | ||
1184 | sb = sget_fc(fc, test, set_anon_super_fc); | |
1185 | if (IS_ERR(sb)) | |
1186 | return PTR_ERR(sb); | |
1187 | ||
1188 | if (!sb->s_root) { | |
43ce4c1f DH |
1189 | err = fill_super(sb, fc); |
1190 | if (err) | |
1191 | goto error; | |
cb50b348 AV |
1192 | |
1193 | sb->s_flags |= SB_ACTIVE; | |
43ce4c1f DH |
1194 | fc->root = dget(sb->s_root); |
1195 | } else { | |
1196 | fc->root = dget(sb->s_root); | |
1197 | if (keying == vfs_get_single_reconf_super) { | |
1198 | err = reconfigure_super(fc); | |
1199 | if (err < 0) { | |
1200 | dput(fc->root); | |
1201 | fc->root = NULL; | |
1202 | goto error; | |
1203 | } | |
1204 | } | |
cb50b348 AV |
1205 | } |
1206 | ||
cb50b348 | 1207 | return 0; |
43ce4c1f DH |
1208 | |
1209 | error: | |
1210 | deactivate_locked_super(sb); | |
1211 | return err; | |
cb50b348 AV |
1212 | } |
1213 | EXPORT_SYMBOL(vfs_get_super); | |
1214 | ||
2ac295d4 AV |
1215 | int get_tree_nodev(struct fs_context *fc, |
1216 | int (*fill_super)(struct super_block *sb, | |
1217 | struct fs_context *fc)) | |
1218 | { | |
1219 | return vfs_get_super(fc, vfs_get_independent_super, fill_super); | |
1220 | } | |
1221 | EXPORT_SYMBOL(get_tree_nodev); | |
1222 | ||
c23a0bba AV |
1223 | int get_tree_single(struct fs_context *fc, |
1224 | int (*fill_super)(struct super_block *sb, | |
1225 | struct fs_context *fc)) | |
1226 | { | |
1227 | return vfs_get_super(fc, vfs_get_single_super, fill_super); | |
1228 | } | |
1229 | EXPORT_SYMBOL(get_tree_single); | |
1230 | ||
43ce4c1f DH |
1231 | int get_tree_single_reconf(struct fs_context *fc, |
1232 | int (*fill_super)(struct super_block *sb, | |
1233 | struct fs_context *fc)) | |
1234 | { | |
1235 | return vfs_get_super(fc, vfs_get_single_reconf_super, fill_super); | |
1236 | } | |
1237 | EXPORT_SYMBOL(get_tree_single_reconf); | |
1238 | ||
533770cc AV |
1239 | int get_tree_keyed(struct fs_context *fc, |
1240 | int (*fill_super)(struct super_block *sb, | |
1241 | struct fs_context *fc), | |
1242 | void *key) | |
1243 | { | |
1244 | fc->s_fs_info = key; | |
1245 | return vfs_get_super(fc, vfs_get_keyed_super, fill_super); | |
1246 | } | |
1247 | EXPORT_SYMBOL(get_tree_keyed); | |
1248 | ||
9361401e | 1249 | #ifdef CONFIG_BLOCK |
fe62c3a4 | 1250 | |
1da177e4 LT |
1251 | static int set_bdev_super(struct super_block *s, void *data) |
1252 | { | |
1253 | s->s_bdev = data; | |
1254 | s->s_dev = s->s_bdev->bd_dev; | |
13eec236 | 1255 | s->s_bdi = bdi_get(s->s_bdev->bd_bdi); |
32a88aa1 | 1256 | |
1da177e4 LT |
1257 | return 0; |
1258 | } | |
1259 | ||
fe62c3a4 DH |
1260 | static int set_bdev_super_fc(struct super_block *s, struct fs_context *fc) |
1261 | { | |
1262 | return set_bdev_super(s, fc->sget_key); | |
1263 | } | |
1264 | ||
1265 | static int test_bdev_super_fc(struct super_block *s, struct fs_context *fc) | |
1266 | { | |
1267 | return s->s_bdev == fc->sget_key; | |
1268 | } | |
1269 | ||
1270 | /** | |
1271 | * get_tree_bdev - Get a superblock based on a single block device | |
1272 | * @fc: The filesystem context holding the parameters | |
1273 | * @fill_super: Helper to initialise a new superblock | |
1274 | */ | |
1275 | int get_tree_bdev(struct fs_context *fc, | |
1276 | int (*fill_super)(struct super_block *, | |
1277 | struct fs_context *)) | |
1278 | { | |
1279 | struct block_device *bdev; | |
1280 | struct super_block *s; | |
1281 | fmode_t mode = FMODE_READ | FMODE_EXCL; | |
1282 | int error = 0; | |
1283 | ||
1284 | if (!(fc->sb_flags & SB_RDONLY)) | |
1285 | mode |= FMODE_WRITE; | |
1286 | ||
1287 | if (!fc->source) | |
1288 | return invalf(fc, "No source specified"); | |
1289 | ||
1290 | bdev = blkdev_get_by_path(fc->source, mode, fc->fs_type); | |
1291 | if (IS_ERR(bdev)) { | |
1292 | errorf(fc, "%s: Can't open blockdev", fc->source); | |
1293 | return PTR_ERR(bdev); | |
1294 | } | |
1295 | ||
1296 | /* Once the superblock is inserted into the list by sget_fc(), s_umount | |
1297 | * will protect the lockfs code from trying to start a snapshot while | |
1298 | * we are mounting | |
1299 | */ | |
1300 | mutex_lock(&bdev->bd_fsfreeze_mutex); | |
1301 | if (bdev->bd_fsfreeze_count > 0) { | |
1302 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
6fcf0c72 | 1303 | blkdev_put(bdev, mode); |
fe62c3a4 DH |
1304 | warnf(fc, "%pg: Can't mount, blockdev is frozen", bdev); |
1305 | return -EBUSY; | |
1306 | } | |
1307 | ||
1308 | fc->sb_flags |= SB_NOSEC; | |
1309 | fc->sget_key = bdev; | |
1310 | s = sget_fc(fc, test_bdev_super_fc, set_bdev_super_fc); | |
1311 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
6fcf0c72 IK |
1312 | if (IS_ERR(s)) { |
1313 | blkdev_put(bdev, mode); | |
fe62c3a4 | 1314 | return PTR_ERR(s); |
6fcf0c72 | 1315 | } |
fe62c3a4 DH |
1316 | |
1317 | if (s->s_root) { | |
1318 | /* Don't summarily change the RO/RW state. */ | |
1319 | if ((fc->sb_flags ^ s->s_flags) & SB_RDONLY) { | |
1320 | warnf(fc, "%pg: Can't mount, would change RO state", bdev); | |
1321 | deactivate_locked_super(s); | |
1322 | blkdev_put(bdev, mode); | |
1323 | return -EBUSY; | |
1324 | } | |
1325 | ||
1326 | /* | |
1327 | * s_umount nests inside bd_mutex during | |
1328 | * __invalidate_device(). blkdev_put() acquires | |
1329 | * bd_mutex and can't be called under s_umount. Drop | |
1330 | * s_umount temporarily. This is safe as we're | |
1331 | * holding an active reference. | |
1332 | */ | |
1333 | up_write(&s->s_umount); | |
1334 | blkdev_put(bdev, mode); | |
1335 | down_write(&s->s_umount); | |
1336 | } else { | |
1337 | s->s_mode = mode; | |
1338 | snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev); | |
1339 | sb_set_blocksize(s, block_size(bdev)); | |
1340 | error = fill_super(s, fc); | |
1341 | if (error) { | |
1342 | deactivate_locked_super(s); | |
1343 | return error; | |
1344 | } | |
1345 | ||
1346 | s->s_flags |= SB_ACTIVE; | |
1347 | bdev->bd_super = s; | |
1348 | } | |
1349 | ||
1350 | BUG_ON(fc->root); | |
1351 | fc->root = dget(s->s_root); | |
1352 | return 0; | |
1353 | } | |
1354 | EXPORT_SYMBOL(get_tree_bdev); | |
1355 | ||
1da177e4 LT |
1356 | static int test_bdev_super(struct super_block *s, void *data) |
1357 | { | |
1358 | return (void *)s->s_bdev == data; | |
1359 | } | |
1360 | ||
152a0836 | 1361 | struct dentry *mount_bdev(struct file_system_type *fs_type, |
1da177e4 | 1362 | int flags, const char *dev_name, void *data, |
152a0836 | 1363 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1364 | { |
1365 | struct block_device *bdev; | |
1366 | struct super_block *s; | |
d4d77629 | 1367 | fmode_t mode = FMODE_READ | FMODE_EXCL; |
1da177e4 LT |
1368 | int error = 0; |
1369 | ||
e462ec50 | 1370 | if (!(flags & SB_RDONLY)) |
30c40d2c AV |
1371 | mode |= FMODE_WRITE; |
1372 | ||
d4d77629 | 1373 | bdev = blkdev_get_by_path(dev_name, mode, fs_type); |
1da177e4 | 1374 | if (IS_ERR(bdev)) |
152a0836 | 1375 | return ERR_CAST(bdev); |
1da177e4 LT |
1376 | |
1377 | /* | |
1378 | * once the super is inserted into the list by sget, s_umount | |
1379 | * will protect the lockfs code from trying to start a snapshot | |
1380 | * while we are mounting | |
1381 | */ | |
4fadd7bb CH |
1382 | mutex_lock(&bdev->bd_fsfreeze_mutex); |
1383 | if (bdev->bd_fsfreeze_count > 0) { | |
1384 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
1385 | error = -EBUSY; | |
1386 | goto error_bdev; | |
1387 | } | |
e462ec50 | 1388 | s = sget(fs_type, test_bdev_super, set_bdev_super, flags | SB_NOSEC, |
9249e17f | 1389 | bdev); |
4fadd7bb | 1390 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
1da177e4 | 1391 | if (IS_ERR(s)) |
454e2398 | 1392 | goto error_s; |
1da177e4 LT |
1393 | |
1394 | if (s->s_root) { | |
e462ec50 | 1395 | if ((flags ^ s->s_flags) & SB_RDONLY) { |
74dbbdd7 | 1396 | deactivate_locked_super(s); |
454e2398 DH |
1397 | error = -EBUSY; |
1398 | goto error_bdev; | |
1da177e4 | 1399 | } |
454e2398 | 1400 | |
4f331f01 TH |
1401 | /* |
1402 | * s_umount nests inside bd_mutex during | |
e525fd89 TH |
1403 | * __invalidate_device(). blkdev_put() acquires |
1404 | * bd_mutex and can't be called under s_umount. Drop | |
1405 | * s_umount temporarily. This is safe as we're | |
1406 | * holding an active reference. | |
4f331f01 TH |
1407 | */ |
1408 | up_write(&s->s_umount); | |
d4d77629 | 1409 | blkdev_put(bdev, mode); |
4f331f01 | 1410 | down_write(&s->s_umount); |
1da177e4 | 1411 | } else { |
30c40d2c | 1412 | s->s_mode = mode; |
a1c6f057 | 1413 | snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev); |
e78c9a00 | 1414 | sb_set_blocksize(s, block_size(bdev)); |
e462ec50 | 1415 | error = fill_super(s, data, flags & SB_SILENT ? 1 : 0); |
1da177e4 | 1416 | if (error) { |
74dbbdd7 | 1417 | deactivate_locked_super(s); |
454e2398 | 1418 | goto error; |
fa675765 | 1419 | } |
454e2398 | 1420 | |
e462ec50 | 1421 | s->s_flags |= SB_ACTIVE; |
87d8fe1e | 1422 | bdev->bd_super = s; |
1da177e4 LT |
1423 | } |
1424 | ||
152a0836 | 1425 | return dget(s->s_root); |
1da177e4 | 1426 | |
454e2398 DH |
1427 | error_s: |
1428 | error = PTR_ERR(s); | |
1429 | error_bdev: | |
d4d77629 | 1430 | blkdev_put(bdev, mode); |
454e2398 | 1431 | error: |
152a0836 AV |
1432 | return ERR_PTR(error); |
1433 | } | |
1434 | EXPORT_SYMBOL(mount_bdev); | |
1435 | ||
1da177e4 LT |
1436 | void kill_block_super(struct super_block *sb) |
1437 | { | |
1438 | struct block_device *bdev = sb->s_bdev; | |
30c40d2c | 1439 | fmode_t mode = sb->s_mode; |
1da177e4 | 1440 | |
ddbaaf30 | 1441 | bdev->bd_super = NULL; |
1da177e4 LT |
1442 | generic_shutdown_super(sb); |
1443 | sync_blockdev(bdev); | |
d4d77629 | 1444 | WARN_ON_ONCE(!(mode & FMODE_EXCL)); |
e525fd89 | 1445 | blkdev_put(bdev, mode | FMODE_EXCL); |
1da177e4 LT |
1446 | } |
1447 | ||
1448 | EXPORT_SYMBOL(kill_block_super); | |
9361401e | 1449 | #endif |
1da177e4 | 1450 | |
3c26ff6e | 1451 | struct dentry *mount_nodev(struct file_system_type *fs_type, |
1da177e4 | 1452 | int flags, void *data, |
3c26ff6e | 1453 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1454 | { |
1455 | int error; | |
9249e17f | 1456 | struct super_block *s = sget(fs_type, NULL, set_anon_super, flags, NULL); |
1da177e4 LT |
1457 | |
1458 | if (IS_ERR(s)) | |
3c26ff6e | 1459 | return ERR_CAST(s); |
1da177e4 | 1460 | |
e462ec50 | 1461 | error = fill_super(s, data, flags & SB_SILENT ? 1 : 0); |
1da177e4 | 1462 | if (error) { |
74dbbdd7 | 1463 | deactivate_locked_super(s); |
3c26ff6e | 1464 | return ERR_PTR(error); |
1da177e4 | 1465 | } |
e462ec50 | 1466 | s->s_flags |= SB_ACTIVE; |
3c26ff6e | 1467 | return dget(s->s_root); |
1da177e4 | 1468 | } |
3c26ff6e AV |
1469 | EXPORT_SYMBOL(mount_nodev); |
1470 | ||
8d0347f6 DH |
1471 | static int reconfigure_single(struct super_block *s, |
1472 | int flags, void *data) | |
1473 | { | |
1474 | struct fs_context *fc; | |
1475 | int ret; | |
1476 | ||
1477 | /* The caller really need to be passing fc down into mount_single(), | |
1478 | * then a chunk of this can be removed. [Bollocks -- AV] | |
1479 | * Better yet, reconfiguration shouldn't happen, but rather the second | |
1480 | * mount should be rejected if the parameters are not compatible. | |
1481 | */ | |
1482 | fc = fs_context_for_reconfigure(s->s_root, flags, MS_RMT_MASK); | |
1483 | if (IS_ERR(fc)) | |
1484 | return PTR_ERR(fc); | |
1485 | ||
1486 | ret = parse_monolithic_mount_data(fc, data); | |
1487 | if (ret < 0) | |
1488 | goto out; | |
1489 | ||
1490 | ret = reconfigure_super(fc); | |
1491 | out: | |
1492 | put_fs_context(fc); | |
1493 | return ret; | |
1494 | } | |
1495 | ||
1da177e4 LT |
1496 | static int compare_single(struct super_block *s, void *p) |
1497 | { | |
1498 | return 1; | |
1499 | } | |
1500 | ||
fc14f2fe | 1501 | struct dentry *mount_single(struct file_system_type *fs_type, |
1da177e4 | 1502 | int flags, void *data, |
fc14f2fe | 1503 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1504 | { |
1505 | struct super_block *s; | |
1506 | int error; | |
1507 | ||
9249e17f | 1508 | s = sget(fs_type, compare_single, set_anon_super, flags, NULL); |
1da177e4 | 1509 | if (IS_ERR(s)) |
fc14f2fe | 1510 | return ERR_CAST(s); |
1da177e4 | 1511 | if (!s->s_root) { |
e462ec50 | 1512 | error = fill_super(s, data, flags & SB_SILENT ? 1 : 0); |
8d0347f6 DH |
1513 | if (!error) |
1514 | s->s_flags |= SB_ACTIVE; | |
9329d1be | 1515 | } else { |
8d0347f6 DH |
1516 | error = reconfigure_single(s, flags, data); |
1517 | } | |
1518 | if (unlikely(error)) { | |
1519 | deactivate_locked_super(s); | |
1520 | return ERR_PTR(error); | |
1da177e4 | 1521 | } |
fc14f2fe AV |
1522 | return dget(s->s_root); |
1523 | } | |
1524 | EXPORT_SYMBOL(mount_single); | |
1525 | ||
9bc61ab1 DH |
1526 | /** |
1527 | * vfs_get_tree - Get the mountable root | |
1528 | * @fc: The superblock configuration context. | |
1529 | * | |
1530 | * The filesystem is invoked to get or create a superblock which can then later | |
1531 | * be used for mounting. The filesystem places a pointer to the root to be | |
1532 | * used for mounting in @fc->root. | |
1533 | */ | |
1534 | int vfs_get_tree(struct fs_context *fc) | |
1da177e4 | 1535 | { |
9d412a43 | 1536 | struct super_block *sb; |
9bc61ab1 | 1537 | int error; |
8089352a | 1538 | |
f3a09c92 AV |
1539 | if (fc->root) |
1540 | return -EBUSY; | |
1541 | ||
1542 | /* Get the mountable root in fc->root, with a ref on the root and a ref | |
1543 | * on the superblock. | |
1544 | */ | |
1545 | error = fc->ops->get_tree(fc); | |
9bc61ab1 DH |
1546 | if (error < 0) |
1547 | return error; | |
1da177e4 | 1548 | |
f3a09c92 AV |
1549 | if (!fc->root) { |
1550 | pr_err("Filesystem %s get_tree() didn't set fc->root\n", | |
1551 | fc->fs_type->name); | |
1552 | /* We don't know what the locking state of the superblock is - | |
1553 | * if there is a superblock. | |
1554 | */ | |
1555 | BUG(); | |
1556 | } | |
1557 | ||
9bc61ab1 | 1558 | sb = fc->root->d_sb; |
9d412a43 | 1559 | WARN_ON(!sb->s_bdi); |
79f546a6 DC |
1560 | |
1561 | /* | |
1562 | * Write barrier is for super_cache_count(). We place it before setting | |
1563 | * SB_BORN as the data dependency between the two functions is the | |
1564 | * superblock structure contents that we just set up, not the SB_BORN | |
1565 | * flag. | |
1566 | */ | |
1567 | smp_wmb(); | |
e462ec50 | 1568 | sb->s_flags |= SB_BORN; |
454e2398 | 1569 | |
9bc61ab1 | 1570 | error = security_sb_set_mnt_opts(sb, fc->security, 0, NULL); |
c9ce29ed AV |
1571 | if (unlikely(error)) { |
1572 | fc_drop_locked(fc); | |
1573 | return error; | |
a10d7c22 AV |
1574 | } |
1575 | ||
42cb56ae JL |
1576 | /* |
1577 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE | |
1578 | * but s_maxbytes was an unsigned long long for many releases. Throw | |
1579 | * this warning for a little while to try and catch filesystems that | |
4358b567 | 1580 | * violate this rule. |
42cb56ae | 1581 | */ |
9d412a43 | 1582 | WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " |
9bc61ab1 | 1583 | "negative value (%lld)\n", fc->fs_type->name, sb->s_maxbytes); |
42cb56ae | 1584 | |
9bc61ab1 | 1585 | return 0; |
1da177e4 | 1586 | } |
9bc61ab1 | 1587 | EXPORT_SYMBOL(vfs_get_tree); |
1da177e4 | 1588 | |
fca39346 JK |
1589 | /* |
1590 | * Setup private BDI for given superblock. It gets automatically cleaned up | |
1591 | * in generic_shutdown_super(). | |
1592 | */ | |
1593 | int super_setup_bdi_name(struct super_block *sb, char *fmt, ...) | |
1594 | { | |
1595 | struct backing_dev_info *bdi; | |
1596 | int err; | |
1597 | va_list args; | |
1598 | ||
1599 | bdi = bdi_alloc(GFP_KERNEL); | |
1600 | if (!bdi) | |
1601 | return -ENOMEM; | |
1602 | ||
1603 | bdi->name = sb->s_type->name; | |
1604 | ||
1605 | va_start(args, fmt); | |
7c4cc300 | 1606 | err = bdi_register_va(bdi, fmt, args); |
fca39346 JK |
1607 | va_end(args); |
1608 | if (err) { | |
1609 | bdi_put(bdi); | |
1610 | return err; | |
1611 | } | |
1612 | WARN_ON(sb->s_bdi != &noop_backing_dev_info); | |
1613 | sb->s_bdi = bdi; | |
fca39346 JK |
1614 | |
1615 | return 0; | |
1616 | } | |
1617 | EXPORT_SYMBOL(super_setup_bdi_name); | |
1618 | ||
1619 | /* | |
1620 | * Setup private BDI for given superblock. I gets automatically cleaned up | |
1621 | * in generic_shutdown_super(). | |
1622 | */ | |
1623 | int super_setup_bdi(struct super_block *sb) | |
1624 | { | |
1625 | static atomic_long_t bdi_seq = ATOMIC_LONG_INIT(0); | |
1626 | ||
1627 | return super_setup_bdi_name(sb, "%.28s-%ld", sb->s_type->name, | |
1628 | atomic_long_inc_return(&bdi_seq)); | |
1629 | } | |
1630 | EXPORT_SYMBOL(super_setup_bdi); | |
1631 | ||
5accdf82 JK |
1632 | /* |
1633 | * This is an internal function, please use sb_end_{write,pagefault,intwrite} | |
1634 | * instead. | |
1635 | */ | |
1636 | void __sb_end_write(struct super_block *sb, int level) | |
1637 | { | |
8129ed29 | 1638 | percpu_up_read(sb->s_writers.rw_sem + level-1); |
5accdf82 JK |
1639 | } |
1640 | EXPORT_SYMBOL(__sb_end_write); | |
1641 | ||
f4b554af ON |
1642 | /* |
1643 | * This is an internal function, please use sb_start_{write,pagefault,intwrite} | |
1644 | * instead. | |
1645 | */ | |
1646 | int __sb_start_write(struct super_block *sb, int level, bool wait) | |
1647 | { | |
1648 | bool force_trylock = false; | |
8129ed29 | 1649 | int ret = 1; |
f4b554af ON |
1650 | |
1651 | #ifdef CONFIG_LOCKDEP | |
1652 | /* | |
1653 | * We want lockdep to tell us about possible deadlocks with freezing | |
1654 | * but it's it bit tricky to properly instrument it. Getting a freeze | |
1655 | * protection works as getting a read lock but there are subtle | |
1656 | * problems. XFS for example gets freeze protection on internal level | |
1657 | * twice in some cases, which is OK only because we already hold a | |
1658 | * freeze protection also on higher level. Due to these cases we have | |
1659 | * to use wait == F (trylock mode) which must not fail. | |
1660 | */ | |
1661 | if (wait) { | |
1662 | int i; | |
1663 | ||
1664 | for (i = 0; i < level - 1; i++) | |
8129ed29 | 1665 | if (percpu_rwsem_is_held(sb->s_writers.rw_sem + i)) { |
f4b554af ON |
1666 | force_trylock = true; |
1667 | break; | |
1668 | } | |
1669 | } | |
1670 | #endif | |
8129ed29 ON |
1671 | if (wait && !force_trylock) |
1672 | percpu_down_read(sb->s_writers.rw_sem + level-1); | |
1673 | else | |
1674 | ret = percpu_down_read_trylock(sb->s_writers.rw_sem + level-1); | |
1675 | ||
22224a17 | 1676 | WARN_ON(force_trylock && !ret); |
f4b554af ON |
1677 | return ret; |
1678 | } | |
5accdf82 JK |
1679 | EXPORT_SYMBOL(__sb_start_write); |
1680 | ||
1681 | /** | |
1682 | * sb_wait_write - wait until all writers to given file system finish | |
1683 | * @sb: the super for which we wait | |
1684 | * @level: type of writers we wait for (normal vs page fault) | |
1685 | * | |
1686 | * This function waits until there are no writers of given type to given file | |
8129ed29 | 1687 | * system. |
5accdf82 JK |
1688 | */ |
1689 | static void sb_wait_write(struct super_block *sb, int level) | |
1690 | { | |
8129ed29 | 1691 | percpu_down_write(sb->s_writers.rw_sem + level-1); |
8129ed29 | 1692 | } |
5accdf82 | 1693 | |
f1a96220 ON |
1694 | /* |
1695 | * We are going to return to userspace and forget about these locks, the | |
1696 | * ownership goes to the caller of thaw_super() which does unlock(). | |
1697 | */ | |
1698 | static void lockdep_sb_freeze_release(struct super_block *sb) | |
1699 | { | |
1700 | int level; | |
1701 | ||
1702 | for (level = SB_FREEZE_LEVELS - 1; level >= 0; level--) | |
1703 | percpu_rwsem_release(sb->s_writers.rw_sem + level, 0, _THIS_IP_); | |
1704 | } | |
1705 | ||
1706 | /* | |
1707 | * Tell lockdep we are holding these locks before we call ->unfreeze_fs(sb). | |
1708 | */ | |
1709 | static void lockdep_sb_freeze_acquire(struct super_block *sb) | |
8129ed29 ON |
1710 | { |
1711 | int level; | |
5accdf82 | 1712 | |
8129ed29 ON |
1713 | for (level = 0; level < SB_FREEZE_LEVELS; ++level) |
1714 | percpu_rwsem_acquire(sb->s_writers.rw_sem + level, 0, _THIS_IP_); | |
f1a96220 ON |
1715 | } |
1716 | ||
1717 | static void sb_freeze_unlock(struct super_block *sb) | |
1718 | { | |
1719 | int level; | |
5accdf82 | 1720 | |
8129ed29 ON |
1721 | for (level = SB_FREEZE_LEVELS - 1; level >= 0; level--) |
1722 | percpu_up_write(sb->s_writers.rw_sem + level); | |
5accdf82 JK |
1723 | } |
1724 | ||
18e9e510 | 1725 | /** |
7000d3c4 RD |
1726 | * freeze_super - lock the filesystem and force it into a consistent state |
1727 | * @sb: the super to lock | |
18e9e510 JB |
1728 | * |
1729 | * Syncs the super to make sure the filesystem is consistent and calls the fs's | |
1730 | * freeze_fs. Subsequent calls to this without first thawing the fs will return | |
1731 | * -EBUSY. | |
5accdf82 JK |
1732 | * |
1733 | * During this function, sb->s_writers.frozen goes through these values: | |
1734 | * | |
1735 | * SB_UNFROZEN: File system is normal, all writes progress as usual. | |
1736 | * | |
1737 | * SB_FREEZE_WRITE: The file system is in the process of being frozen. New | |
1738 | * writes should be blocked, though page faults are still allowed. We wait for | |
1739 | * all writes to complete and then proceed to the next stage. | |
1740 | * | |
1741 | * SB_FREEZE_PAGEFAULT: Freezing continues. Now also page faults are blocked | |
1742 | * but internal fs threads can still modify the filesystem (although they | |
1743 | * should not dirty new pages or inodes), writeback can run etc. After waiting | |
1744 | * for all running page faults we sync the filesystem which will clean all | |
1745 | * dirty pages and inodes (no new dirty pages or inodes can be created when | |
1746 | * sync is running). | |
1747 | * | |
1748 | * SB_FREEZE_FS: The file system is frozen. Now all internal sources of fs | |
1749 | * modification are blocked (e.g. XFS preallocation truncation on inode | |
1750 | * reclaim). This is usually implemented by blocking new transactions for | |
1751 | * filesystems that have them and need this additional guard. After all | |
1752 | * internal writers are finished we call ->freeze_fs() to finish filesystem | |
1753 | * freezing. Then we transition to SB_FREEZE_COMPLETE state. This state is | |
1754 | * mostly auxiliary for filesystems to verify they do not modify frozen fs. | |
1755 | * | |
1756 | * sb->s_writers.frozen is protected by sb->s_umount. | |
18e9e510 JB |
1757 | */ |
1758 | int freeze_super(struct super_block *sb) | |
1759 | { | |
1760 | int ret; | |
1761 | ||
1762 | atomic_inc(&sb->s_active); | |
1763 | down_write(&sb->s_umount); | |
5accdf82 | 1764 | if (sb->s_writers.frozen != SB_UNFROZEN) { |
18e9e510 JB |
1765 | deactivate_locked_super(sb); |
1766 | return -EBUSY; | |
1767 | } | |
1768 | ||
e462ec50 | 1769 | if (!(sb->s_flags & SB_BORN)) { |
dabe0dc1 AV |
1770 | up_write(&sb->s_umount); |
1771 | return 0; /* sic - it's "nothing to do" */ | |
1772 | } | |
1773 | ||
bc98a42c | 1774 | if (sb_rdonly(sb)) { |
5accdf82 JK |
1775 | /* Nothing to do really... */ |
1776 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
18e9e510 JB |
1777 | up_write(&sb->s_umount); |
1778 | return 0; | |
1779 | } | |
1780 | ||
5accdf82 | 1781 | sb->s_writers.frozen = SB_FREEZE_WRITE; |
5accdf82 JK |
1782 | /* Release s_umount to preserve sb_start_write -> s_umount ordering */ |
1783 | up_write(&sb->s_umount); | |
5accdf82 | 1784 | sb_wait_write(sb, SB_FREEZE_WRITE); |
8129ed29 | 1785 | down_write(&sb->s_umount); |
5accdf82 JK |
1786 | |
1787 | /* Now we go and block page faults... */ | |
5accdf82 | 1788 | sb->s_writers.frozen = SB_FREEZE_PAGEFAULT; |
5accdf82 JK |
1789 | sb_wait_write(sb, SB_FREEZE_PAGEFAULT); |
1790 | ||
1791 | /* All writers are done so after syncing there won't be dirty data */ | |
18e9e510 JB |
1792 | sync_filesystem(sb); |
1793 | ||
5accdf82 JK |
1794 | /* Now wait for internal filesystem counter */ |
1795 | sb->s_writers.frozen = SB_FREEZE_FS; | |
5accdf82 | 1796 | sb_wait_write(sb, SB_FREEZE_FS); |
18e9e510 | 1797 | |
18e9e510 JB |
1798 | if (sb->s_op->freeze_fs) { |
1799 | ret = sb->s_op->freeze_fs(sb); | |
1800 | if (ret) { | |
1801 | printk(KERN_ERR | |
1802 | "VFS:Filesystem freeze failed\n"); | |
5accdf82 | 1803 | sb->s_writers.frozen = SB_UNFROZEN; |
8129ed29 | 1804 | sb_freeze_unlock(sb); |
5accdf82 | 1805 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 JB |
1806 | deactivate_locked_super(sb); |
1807 | return ret; | |
1808 | } | |
1809 | } | |
5accdf82 | 1810 | /* |
89f39af1 ON |
1811 | * For debugging purposes so that fs can warn if it sees write activity |
1812 | * when frozen is set to SB_FREEZE_COMPLETE, and for thaw_super(). | |
5accdf82 JK |
1813 | */ |
1814 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
f1a96220 | 1815 | lockdep_sb_freeze_release(sb); |
18e9e510 JB |
1816 | up_write(&sb->s_umount); |
1817 | return 0; | |
1818 | } | |
1819 | EXPORT_SYMBOL(freeze_super); | |
1820 | ||
1821 | /** | |
1822 | * thaw_super -- unlock filesystem | |
1823 | * @sb: the super to thaw | |
1824 | * | |
1825 | * Unlocks the filesystem and marks it writeable again after freeze_super(). | |
1826 | */ | |
08fdc8a0 | 1827 | static int thaw_super_locked(struct super_block *sb) |
18e9e510 JB |
1828 | { |
1829 | int error; | |
1830 | ||
89f39af1 | 1831 | if (sb->s_writers.frozen != SB_FREEZE_COMPLETE) { |
18e9e510 JB |
1832 | up_write(&sb->s_umount); |
1833 | return -EINVAL; | |
1834 | } | |
1835 | ||
bc98a42c | 1836 | if (sb_rdonly(sb)) { |
8129ed29 | 1837 | sb->s_writers.frozen = SB_UNFROZEN; |
18e9e510 | 1838 | goto out; |
8129ed29 | 1839 | } |
18e9e510 | 1840 | |
f1a96220 ON |
1841 | lockdep_sb_freeze_acquire(sb); |
1842 | ||
18e9e510 JB |
1843 | if (sb->s_op->unfreeze_fs) { |
1844 | error = sb->s_op->unfreeze_fs(sb); | |
1845 | if (error) { | |
1846 | printk(KERN_ERR | |
1847 | "VFS:Filesystem thaw failed\n"); | |
f1a96220 | 1848 | lockdep_sb_freeze_release(sb); |
18e9e510 JB |
1849 | up_write(&sb->s_umount); |
1850 | return error; | |
1851 | } | |
1852 | } | |
1853 | ||
5accdf82 | 1854 | sb->s_writers.frozen = SB_UNFROZEN; |
8129ed29 ON |
1855 | sb_freeze_unlock(sb); |
1856 | out: | |
5accdf82 | 1857 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 | 1858 | deactivate_locked_super(sb); |
18e9e510 JB |
1859 | return 0; |
1860 | } | |
08fdc8a0 MG |
1861 | |
1862 | int thaw_super(struct super_block *sb) | |
1863 | { | |
1864 | down_write(&sb->s_umount); | |
1865 | return thaw_super_locked(sb); | |
1866 | } | |
18e9e510 | 1867 | EXPORT_SYMBOL(thaw_super); |