Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 LT |
2 | /* |
3 | * linux/fs/super.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992 Linus Torvalds | |
6 | * | |
7 | * super.c contains code to handle: - mount structures | |
8 | * - super-block tables | |
9 | * - filesystem drivers list | |
10 | * - mount system call | |
11 | * - umount system call | |
12 | * - ustat system call | |
13 | * | |
14 | * GK 2/5/95 - Changed to support mounting the root fs via NFS | |
15 | * | |
16 | * Added kerneld support: Jacques Gelinas and Bjorn Ekwall | |
17 | * Added change_root: Werner Almesberger & Hans Lermen, Feb '96 | |
18 | * Added options to /proc/mounts: | |
96de0e25 | 19 | * Torbjörn Lindh (torbjorn.lindh@gopta.se), April 14, 1996. |
1da177e4 LT |
20 | * Added devfs support: Richard Gooch <rgooch@atnf.csiro.au>, 13-JAN-1998 |
21 | * Heavily rewritten for 'one fs - one tree' dcache architecture. AV, Mar 2000 | |
22 | */ | |
23 | ||
630d9c47 | 24 | #include <linux/export.h> |
1da177e4 | 25 | #include <linux/slab.h> |
1da177e4 | 26 | #include <linux/blkdev.h> |
1da177e4 LT |
27 | #include <linux/mount.h> |
28 | #include <linux/security.h> | |
1da177e4 LT |
29 | #include <linux/writeback.h> /* for the emergency remount stuff */ |
30 | #include <linux/idr.h> | |
353ab6e9 | 31 | #include <linux/mutex.h> |
5477d0fa | 32 | #include <linux/backing-dev.h> |
ceb5bdc2 | 33 | #include <linux/rculist_bl.h> |
22d94f49 | 34 | #include <linux/fscrypt.h> |
40401530 | 35 | #include <linux/fsnotify.h> |
5accdf82 | 36 | #include <linux/lockdep.h> |
6e4eab57 | 37 | #include <linux/user_namespace.h> |
9bc61ab1 | 38 | #include <linux/fs_context.h> |
e262e32d | 39 | #include <uapi/linux/mount.h> |
6d59e7f5 | 40 | #include "internal.h" |
1da177e4 | 41 | |
08fdc8a0 | 42 | static int thaw_super_locked(struct super_block *sb); |
1da177e4 | 43 | |
15d0f5ea AV |
44 | static LIST_HEAD(super_blocks); |
45 | static DEFINE_SPINLOCK(sb_lock); | |
1da177e4 | 46 | |
5accdf82 JK |
47 | static char *sb_writers_name[SB_FREEZE_LEVELS] = { |
48 | "sb_writers", | |
49 | "sb_pagefaults", | |
50 | "sb_internal", | |
51 | }; | |
52 | ||
0ed33598 CB |
53 | static inline void super_lock(struct super_block *sb, bool excl) |
54 | { | |
55 | if (excl) | |
56 | down_write(&sb->s_umount); | |
57 | else | |
58 | down_read(&sb->s_umount); | |
59 | } | |
60 | ||
61 | static inline void super_unlock(struct super_block *sb, bool excl) | |
62 | { | |
63 | if (excl) | |
64 | up_write(&sb->s_umount); | |
65 | else | |
66 | up_read(&sb->s_umount); | |
67 | } | |
68 | ||
69 | static inline void super_lock_excl(struct super_block *sb) | |
70 | { | |
71 | super_lock(sb, true); | |
72 | } | |
73 | ||
74 | static inline void super_lock_shared(struct super_block *sb) | |
75 | { | |
76 | super_lock(sb, false); | |
77 | } | |
78 | ||
79 | static inline void super_unlock_excl(struct super_block *sb) | |
80 | { | |
81 | super_unlock(sb, true); | |
82 | } | |
83 | ||
84 | static inline void super_unlock_shared(struct super_block *sb) | |
85 | { | |
86 | super_unlock(sb, false); | |
87 | } | |
88 | ||
b0d40c92 DC |
89 | /* |
90 | * One thing we have to be careful of with a per-sb shrinker is that we don't | |
91 | * drop the last active reference to the superblock from within the shrinker. | |
92 | * If that happens we could trigger unregistering the shrinker from within the | |
47a7c01c | 93 | * shrinker path and that leads to deadlock on the shrinker_rwsem. Hence we |
b0d40c92 DC |
94 | * take a passive reference to the superblock to avoid this from occurring. |
95 | */ | |
0a234c6d DC |
96 | static unsigned long super_cache_scan(struct shrinker *shrink, |
97 | struct shrink_control *sc) | |
b0d40c92 DC |
98 | { |
99 | struct super_block *sb; | |
0a234c6d DC |
100 | long fs_objects = 0; |
101 | long total_objects; | |
102 | long freed = 0; | |
103 | long dentries; | |
104 | long inodes; | |
b0d40c92 DC |
105 | |
106 | sb = container_of(shrink, struct super_block, s_shrink); | |
107 | ||
108 | /* | |
109 | * Deadlock avoidance. We may hold various FS locks, and we don't want | |
110 | * to recurse into the FS that called us in clear_inode() and friends.. | |
111 | */ | |
0a234c6d DC |
112 | if (!(sc->gfp_mask & __GFP_FS)) |
113 | return SHRINK_STOP; | |
b0d40c92 | 114 | |
eb6ef3df | 115 | if (!trylock_super(sb)) |
0a234c6d | 116 | return SHRINK_STOP; |
b0d40c92 | 117 | |
d0407903 | 118 | if (sb->s_op->nr_cached_objects) |
4101b624 | 119 | fs_objects = sb->s_op->nr_cached_objects(sb, sc); |
0e1fdafd | 120 | |
503c358c VD |
121 | inodes = list_lru_shrink_count(&sb->s_inode_lru, sc); |
122 | dentries = list_lru_shrink_count(&sb->s_dentry_lru, sc); | |
f6041567 | 123 | total_objects = dentries + inodes + fs_objects + 1; |
475d0db7 TH |
124 | if (!total_objects) |
125 | total_objects = 1; | |
0e1fdafd | 126 | |
0a234c6d | 127 | /* proportion the scan between the caches */ |
f6041567 | 128 | dentries = mult_frac(sc->nr_to_scan, dentries, total_objects); |
bc3b14cb | 129 | inodes = mult_frac(sc->nr_to_scan, inodes, total_objects); |
503c358c | 130 | fs_objects = mult_frac(sc->nr_to_scan, fs_objects, total_objects); |
b0d40c92 | 131 | |
0a234c6d DC |
132 | /* |
133 | * prune the dcache first as the icache is pinned by it, then | |
134 | * prune the icache, followed by the filesystem specific caches | |
49e7e7ff VD |
135 | * |
136 | * Ensure that we always scan at least one object - memcg kmem | |
137 | * accounting uses this to fully empty the caches. | |
0a234c6d | 138 | */ |
49e7e7ff | 139 | sc->nr_to_scan = dentries + 1; |
503c358c | 140 | freed = prune_dcache_sb(sb, sc); |
49e7e7ff | 141 | sc->nr_to_scan = inodes + 1; |
503c358c | 142 | freed += prune_icache_sb(sb, sc); |
0a234c6d DC |
143 | |
144 | if (fs_objects) { | |
49e7e7ff | 145 | sc->nr_to_scan = fs_objects + 1; |
4101b624 | 146 | freed += sb->s_op->free_cached_objects(sb, sc); |
b0d40c92 DC |
147 | } |
148 | ||
0ed33598 | 149 | super_unlock_shared(sb); |
0a234c6d DC |
150 | return freed; |
151 | } | |
152 | ||
153 | static unsigned long super_cache_count(struct shrinker *shrink, | |
154 | struct shrink_control *sc) | |
155 | { | |
156 | struct super_block *sb; | |
157 | long total_objects = 0; | |
158 | ||
159 | sb = container_of(shrink, struct super_block, s_shrink); | |
160 | ||
d23da150 | 161 | /* |
79f546a6 DC |
162 | * We don't call trylock_super() here as it is a scalability bottleneck, |
163 | * so we're exposed to partial setup state. The shrinker rwsem does not | |
164 | * protect filesystem operations backing list_lru_shrink_count() or | |
165 | * s_op->nr_cached_objects(). Counts can change between | |
166 | * super_cache_count and super_cache_scan, so we really don't need locks | |
167 | * here. | |
168 | * | |
169 | * However, if we are currently mounting the superblock, the underlying | |
170 | * filesystem might be in a state of partial construction and hence it | |
171 | * is dangerous to access it. trylock_super() uses a SB_BORN check to | |
172 | * avoid this situation, so do the same here. The memory barrier is | |
173 | * matched with the one in mount_fs() as we don't hold locks here. | |
d23da150 | 174 | */ |
79f546a6 DC |
175 | if (!(sb->s_flags & SB_BORN)) |
176 | return 0; | |
177 | smp_rmb(); | |
178 | ||
0a234c6d | 179 | if (sb->s_op && sb->s_op->nr_cached_objects) |
4101b624 | 180 | total_objects = sb->s_op->nr_cached_objects(sb, sc); |
0a234c6d | 181 | |
503c358c VD |
182 | total_objects += list_lru_shrink_count(&sb->s_dentry_lru, sc); |
183 | total_objects += list_lru_shrink_count(&sb->s_inode_lru, sc); | |
0a234c6d | 184 | |
9b996468 KT |
185 | if (!total_objects) |
186 | return SHRINK_EMPTY; | |
187 | ||
55f841ce | 188 | total_objects = vfs_pressure_ratio(total_objects); |
0e1fdafd | 189 | return total_objects; |
b0d40c92 DC |
190 | } |
191 | ||
853b39a7 ON |
192 | static void destroy_super_work(struct work_struct *work) |
193 | { | |
194 | struct super_block *s = container_of(work, struct super_block, | |
195 | destroy_work); | |
196 | int i; | |
197 | ||
198 | for (i = 0; i < SB_FREEZE_LEVELS; i++) | |
8129ed29 | 199 | percpu_free_rwsem(&s->s_writers.rw_sem[i]); |
853b39a7 ON |
200 | kfree(s); |
201 | } | |
202 | ||
203 | static void destroy_super_rcu(struct rcu_head *head) | |
204 | { | |
205 | struct super_block *s = container_of(head, struct super_block, rcu); | |
206 | INIT_WORK(&s->destroy_work, destroy_super_work); | |
207 | schedule_work(&s->destroy_work); | |
208 | } | |
209 | ||
0200894d AV |
210 | /* Free a superblock that has never been seen by anyone */ |
211 | static void destroy_unused_super(struct super_block *s) | |
5accdf82 | 212 | { |
0200894d AV |
213 | if (!s) |
214 | return; | |
0ed33598 | 215 | super_unlock_excl(s); |
7eb5e882 AV |
216 | list_lru_destroy(&s->s_dentry_lru); |
217 | list_lru_destroy(&s->s_inode_lru); | |
7eb5e882 | 218 | security_sb_free(s); |
6e4eab57 | 219 | put_user_ns(s->s_user_ns); |
7eb5e882 | 220 | kfree(s->s_subtype); |
8e04944f | 221 | free_prealloced_shrinker(&s->s_shrink); |
0200894d AV |
222 | /* no delays needed */ |
223 | destroy_super_work(&s->destroy_work); | |
5accdf82 JK |
224 | } |
225 | ||
1da177e4 LT |
226 | /** |
227 | * alloc_super - create new superblock | |
fe2bbc48 | 228 | * @type: filesystem type superblock should belong to |
9249e17f | 229 | * @flags: the mount flags |
6e4eab57 | 230 | * @user_ns: User namespace for the super_block |
1da177e4 LT |
231 | * |
232 | * Allocates and initializes a new &struct super_block. alloc_super() | |
233 | * returns a pointer new superblock or %NULL if allocation had failed. | |
234 | */ | |
6e4eab57 EB |
235 | static struct super_block *alloc_super(struct file_system_type *type, int flags, |
236 | struct user_namespace *user_ns) | |
1da177e4 | 237 | { |
11b0b5ab | 238 | struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER); |
b87221de | 239 | static const struct super_operations default_op; |
7eb5e882 AV |
240 | int i; |
241 | ||
242 | if (!s) | |
243 | return NULL; | |
1da177e4 | 244 | |
b5bd856a | 245 | INIT_LIST_HEAD(&s->s_mounts); |
6e4eab57 | 246 | s->s_user_ns = get_user_ns(user_ns); |
ca0168e8 AV |
247 | init_rwsem(&s->s_umount); |
248 | lockdep_set_class(&s->s_umount, &type->s_umount_key); | |
249 | /* | |
250 | * sget() can have s_umount recursion. | |
251 | * | |
252 | * When it cannot find a suitable sb, it allocates a new | |
253 | * one (this one), and tries again to find a suitable old | |
254 | * one. | |
255 | * | |
256 | * In case that succeeds, it will acquire the s_umount | |
257 | * lock of the old one. Since these are clearly distrinct | |
258 | * locks, and this object isn't exposed yet, there's no | |
259 | * risk of deadlocks. | |
260 | * | |
261 | * Annotate this by putting this lock in a different | |
262 | * subclass. | |
263 | */ | |
264 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); | |
b5bd856a | 265 | |
7eb5e882 AV |
266 | if (security_sb_alloc(s)) |
267 | goto fail; | |
7b7a8665 | 268 | |
7eb5e882 | 269 | for (i = 0; i < SB_FREEZE_LEVELS; i++) { |
8129ed29 ON |
270 | if (__percpu_init_rwsem(&s->s_writers.rw_sem[i], |
271 | sb_writers_name[i], | |
272 | &type->s_writers_key[i])) | |
7eb5e882 | 273 | goto fail; |
1da177e4 | 274 | } |
df0ce26c | 275 | s->s_bdi = &noop_backing_dev_info; |
7eb5e882 | 276 | s->s_flags = flags; |
cc50a07a | 277 | if (s->s_user_ns != &init_user_ns) |
67690f93 | 278 | s->s_iflags |= SB_I_NODEV; |
7eb5e882 | 279 | INIT_HLIST_NODE(&s->s_instances); |
f1ee6162 | 280 | INIT_HLIST_BL_HEAD(&s->s_roots); |
e97fedb9 | 281 | mutex_init(&s->s_sync_lock); |
7eb5e882 | 282 | INIT_LIST_HEAD(&s->s_inodes); |
74278da9 | 283 | spin_lock_init(&s->s_inode_list_lock); |
6c60d2b5 DC |
284 | INIT_LIST_HEAD(&s->s_inodes_wb); |
285 | spin_lock_init(&s->s_inode_wblist_lock); | |
7eb5e882 | 286 | |
7eb5e882 AV |
287 | s->s_count = 1; |
288 | atomic_set(&s->s_active, 1); | |
289 | mutex_init(&s->s_vfs_rename_mutex); | |
290 | lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key); | |
bc8230ee | 291 | init_rwsem(&s->s_dquot.dqio_sem); |
7eb5e882 AV |
292 | s->s_maxbytes = MAX_NON_LFS; |
293 | s->s_op = &default_op; | |
294 | s->s_time_gran = 1000000000; | |
188d20bc DD |
295 | s->s_time_min = TIME64_MIN; |
296 | s->s_time_max = TIME64_MAX; | |
7eb5e882 AV |
297 | |
298 | s->s_shrink.seeks = DEFAULT_SEEKS; | |
299 | s->s_shrink.scan_objects = super_cache_scan; | |
300 | s->s_shrink.count_objects = super_cache_count; | |
301 | s->s_shrink.batch = 1024; | |
2acb60a0 | 302 | s->s_shrink.flags = SHRINKER_NUMA_AWARE | SHRINKER_MEMCG_AWARE; |
e33c267a | 303 | if (prealloc_shrinker(&s->s_shrink, "sb-%s", type->name)) |
8e04944f | 304 | goto fail; |
c92e8e10 | 305 | if (list_lru_init_memcg(&s->s_dentry_lru, &s->s_shrink)) |
2b3648a6 | 306 | goto fail; |
c92e8e10 | 307 | if (list_lru_init_memcg(&s->s_inode_lru, &s->s_shrink)) |
2b3648a6 | 308 | goto fail; |
1da177e4 | 309 | return s; |
5ca302c8 | 310 | |
7eb5e882 | 311 | fail: |
0200894d | 312 | destroy_unused_super(s); |
7eb5e882 | 313 | return NULL; |
1da177e4 LT |
314 | } |
315 | ||
316 | /* Superblock refcounting */ | |
317 | ||
318 | /* | |
35cf7ba0 | 319 | * Drop a superblock's refcount. The caller must hold sb_lock. |
1da177e4 | 320 | */ |
c645b930 | 321 | static void __put_super(struct super_block *s) |
1da177e4 | 322 | { |
c645b930 AV |
323 | if (!--s->s_count) { |
324 | list_del_init(&s->s_list); | |
325 | WARN_ON(s->s_dentry_lru.node); | |
326 | WARN_ON(s->s_inode_lru.node); | |
327 | WARN_ON(!list_empty(&s->s_mounts)); | |
328 | security_sb_free(s); | |
329 | put_user_ns(s->s_user_ns); | |
330 | kfree(s->s_subtype); | |
331 | call_rcu(&s->rcu, destroy_super_rcu); | |
1da177e4 | 332 | } |
1da177e4 LT |
333 | } |
334 | ||
335 | /** | |
336 | * put_super - drop a temporary reference to superblock | |
337 | * @sb: superblock in question | |
338 | * | |
339 | * Drops a temporary reference, frees superblock if there's no | |
340 | * references left. | |
341 | */ | |
60b49885 | 342 | void put_super(struct super_block *sb) |
1da177e4 LT |
343 | { |
344 | spin_lock(&sb_lock); | |
345 | __put_super(sb); | |
346 | spin_unlock(&sb_lock); | |
347 | } | |
348 | ||
349 | ||
350 | /** | |
1712ac8f | 351 | * deactivate_locked_super - drop an active reference to superblock |
1da177e4 LT |
352 | * @s: superblock to deactivate |
353 | * | |
bd7ced98 | 354 | * Drops an active reference to superblock, converting it into a temporary |
1712ac8f | 355 | * one if there is no other active references left. In that case we |
1da177e4 LT |
356 | * tell fs driver to shut it down and drop the temporary reference we |
357 | * had just acquired. | |
1712ac8f AV |
358 | * |
359 | * Caller holds exclusive lock on superblock; that lock is released. | |
1da177e4 | 360 | */ |
1712ac8f | 361 | void deactivate_locked_super(struct super_block *s) |
1da177e4 LT |
362 | { |
363 | struct file_system_type *fs = s->s_type; | |
b20bd1a5 | 364 | if (atomic_dec_and_test(&s->s_active)) { |
b0d40c92 | 365 | unregister_shrinker(&s->s_shrink); |
28f2cd4f | 366 | fs->kill_sb(s); |
f5e1dd34 | 367 | |
c0a5b560 VD |
368 | /* |
369 | * Since list_lru_destroy() may sleep, we cannot call it from | |
370 | * put_super(), where we hold the sb_lock. Therefore we destroy | |
371 | * the lru lists right now. | |
372 | */ | |
373 | list_lru_destroy(&s->s_dentry_lru); | |
374 | list_lru_destroy(&s->s_inode_lru); | |
375 | ||
1da177e4 LT |
376 | put_filesystem(fs); |
377 | put_super(s); | |
1712ac8f | 378 | } else { |
0ed33598 | 379 | super_unlock_excl(s); |
1da177e4 LT |
380 | } |
381 | } | |
382 | ||
1712ac8f | 383 | EXPORT_SYMBOL(deactivate_locked_super); |
1da177e4 | 384 | |
74dbbdd7 | 385 | /** |
1712ac8f | 386 | * deactivate_super - drop an active reference to superblock |
74dbbdd7 AV |
387 | * @s: superblock to deactivate |
388 | * | |
1712ac8f AV |
389 | * Variant of deactivate_locked_super(), except that superblock is *not* |
390 | * locked by caller. If we are going to drop the final active reference, | |
391 | * lock will be acquired prior to that. | |
74dbbdd7 | 392 | */ |
1712ac8f | 393 | void deactivate_super(struct super_block *s) |
74dbbdd7 | 394 | { |
cc23402c | 395 | if (!atomic_add_unless(&s->s_active, -1, 1)) { |
0ed33598 | 396 | super_lock_excl(s); |
1712ac8f | 397 | deactivate_locked_super(s); |
74dbbdd7 AV |
398 | } |
399 | } | |
400 | ||
1712ac8f | 401 | EXPORT_SYMBOL(deactivate_super); |
74dbbdd7 | 402 | |
1da177e4 LT |
403 | /** |
404 | * grab_super - acquire an active reference | |
405 | * @s: reference we are trying to make active | |
406 | * | |
407 | * Tries to acquire an active reference. grab_super() is used when we | |
408 | * had just found a superblock in super_blocks or fs_type->fs_supers | |
409 | * and want to turn it into a full-blown active reference. grab_super() | |
410 | * is called with sb_lock held and drops it. Returns 1 in case of | |
411 | * success, 0 if we had failed (superblock contents was already dead or | |
acfec9a5 AV |
412 | * dying when grab_super() had been called). Note that this is only |
413 | * called for superblocks not in rundown mode (== ones still on ->fs_supers | |
414 | * of their type), so increment of ->s_count is OK here. | |
1da177e4 | 415 | */ |
9c4dbee7 | 416 | static int grab_super(struct super_block *s) __releases(sb_lock) |
1da177e4 LT |
417 | { |
418 | s->s_count++; | |
419 | spin_unlock(&sb_lock); | |
0ed33598 | 420 | super_lock_excl(s); |
e462ec50 | 421 | if ((s->s_flags & SB_BORN) && atomic_inc_not_zero(&s->s_active)) { |
acfec9a5 AV |
422 | put_super(s); |
423 | return 1; | |
424 | } | |
0ed33598 | 425 | super_unlock_excl(s); |
1da177e4 | 426 | put_super(s); |
1da177e4 LT |
427 | return 0; |
428 | } | |
429 | ||
12ad3ab6 | 430 | /* |
eb6ef3df | 431 | * trylock_super - try to grab ->s_umount shared |
331cbdee | 432 | * @sb: reference we are trying to grab |
12ad3ab6 | 433 | * |
eb6ef3df | 434 | * Try to prevent fs shutdown. This is used in places where we |
12ad3ab6 | 435 | * cannot take an active reference but we need to ensure that the |
eb6ef3df KK |
436 | * filesystem is not shut down while we are working on it. It returns |
437 | * false if we cannot acquire s_umount or if we lose the race and | |
438 | * filesystem already got into shutdown, and returns true with the s_umount | |
439 | * lock held in read mode in case of success. On successful return, | |
440 | * the caller must drop the s_umount lock when done. | |
441 | * | |
442 | * Note that unlike get_super() et.al. this one does *not* bump ->s_count. | |
443 | * The reason why it's safe is that we are OK with doing trylock instead | |
444 | * of down_read(). There's a couple of places that are OK with that, but | |
445 | * it's very much not a general-purpose interface. | |
12ad3ab6 | 446 | */ |
eb6ef3df | 447 | bool trylock_super(struct super_block *sb) |
12ad3ab6 | 448 | { |
12ad3ab6 | 449 | if (down_read_trylock(&sb->s_umount)) { |
eb6ef3df | 450 | if (!hlist_unhashed(&sb->s_instances) && |
e462ec50 | 451 | sb->s_root && (sb->s_flags & SB_BORN)) |
12ad3ab6 | 452 | return true; |
0ed33598 | 453 | super_unlock_shared(sb); |
12ad3ab6 DC |
454 | } |
455 | ||
12ad3ab6 DC |
456 | return false; |
457 | } | |
458 | ||
04b94071 DL |
459 | /** |
460 | * retire_super - prevents superblock from being reused | |
461 | * @sb: superblock to retire | |
462 | * | |
463 | * The function marks superblock to be ignored in superblock test, which | |
464 | * prevents it from being reused for any new mounts. If the superblock has | |
465 | * a private bdi, it also unregisters it, but doesn't reduce the refcount | |
466 | * of the superblock to prevent potential races. The refcount is reduced | |
467 | * by generic_shutdown_super(). The function can not be called | |
468 | * concurrently with generic_shutdown_super(). It is safe to call the | |
469 | * function multiple times, subsequent calls have no effect. | |
470 | * | |
471 | * The marker will affect the re-use only for block-device-based | |
472 | * superblocks. Other superblocks will still get marked if this function | |
473 | * is used, but that will not affect their reusability. | |
474 | */ | |
475 | void retire_super(struct super_block *sb) | |
476 | { | |
477 | WARN_ON(!sb->s_bdev); | |
0ed33598 | 478 | super_lock_excl(sb); |
04b94071 DL |
479 | if (sb->s_iflags & SB_I_PERSB_BDI) { |
480 | bdi_unregister(sb->s_bdi); | |
481 | sb->s_iflags &= ~SB_I_PERSB_BDI; | |
482 | } | |
483 | sb->s_iflags |= SB_I_RETIRED; | |
0ed33598 | 484 | super_unlock_excl(sb); |
04b94071 DL |
485 | } |
486 | EXPORT_SYMBOL(retire_super); | |
487 | ||
1da177e4 LT |
488 | /** |
489 | * generic_shutdown_super - common helper for ->kill_sb() | |
490 | * @sb: superblock to kill | |
491 | * | |
492 | * generic_shutdown_super() does all fs-independent work on superblock | |
493 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects | |
494 | * that need destruction out of superblock, call generic_shutdown_super() | |
495 | * and release aforementioned objects. Note: dentries and inodes _are_ | |
496 | * taken care of and do not need specific handling. | |
c636ebdb DH |
497 | * |
498 | * Upon calling this function, the filesystem may no longer alter or | |
499 | * rearrange the set of dentries belonging to this super_block, nor may it | |
500 | * change the attachments of dentries to inodes. | |
1da177e4 LT |
501 | */ |
502 | void generic_shutdown_super(struct super_block *sb) | |
503 | { | |
ee9b6d61 | 504 | const struct super_operations *sop = sb->s_op; |
1da177e4 | 505 | |
c636ebdb DH |
506 | if (sb->s_root) { |
507 | shrink_dcache_for_umount(sb); | |
60b0680f | 508 | sync_filesystem(sb); |
e462ec50 | 509 | sb->s_flags &= ~SB_ACTIVE; |
efaee192 | 510 | |
a1a0e23e | 511 | cgroup_writeback_umount(); |
63997e98 | 512 | |
ccb820dc | 513 | /* Evict all inodes with zero refcount. */ |
63997e98 | 514 | evict_inodes(sb); |
ccb820dc EB |
515 | |
516 | /* | |
517 | * Clean up and evict any inodes that still have references due | |
518 | * to fsnotify or the security policy. | |
519 | */ | |
1edc8eb2 | 520 | fsnotify_sb_delete(sb); |
83e804f0 | 521 | security_sb_delete(sb); |
1da177e4 | 522 | |
ccb820dc EB |
523 | /* |
524 | * Now that all potentially-encrypted inodes have been evicted, | |
525 | * the fscrypt keyring can be destroyed. | |
526 | */ | |
527 | fscrypt_destroy_keyring(sb); | |
528 | ||
7b7a8665 CH |
529 | if (sb->s_dio_done_wq) { |
530 | destroy_workqueue(sb->s_dio_done_wq); | |
531 | sb->s_dio_done_wq = NULL; | |
532 | } | |
533 | ||
1da177e4 LT |
534 | if (sop->put_super) |
535 | sop->put_super(sb); | |
536 | ||
47d58691 JH |
537 | if (CHECK_DATA_CORRUPTION(!list_empty(&sb->s_inodes), |
538 | "VFS: Busy inodes after unmount of %s (%s)", | |
539 | sb->s_id, sb->s_type->name)) { | |
540 | /* | |
541 | * Adding a proper bailout path here would be hard, but | |
542 | * we can at least make it more likely that a later | |
543 | * iput_final() or such crashes cleanly. | |
544 | */ | |
545 | struct inode *inode; | |
546 | ||
547 | spin_lock(&sb->s_inode_list_lock); | |
548 | list_for_each_entry(inode, &sb->s_inodes, i_sb_list) { | |
549 | inode->i_op = VFS_PTR_POISON; | |
550 | inode->i_sb = VFS_PTR_POISON; | |
551 | inode->i_mapping = VFS_PTR_POISON; | |
552 | } | |
553 | spin_unlock(&sb->s_inode_list_lock); | |
1da177e4 | 554 | } |
1da177e4 LT |
555 | } |
556 | spin_lock(&sb_lock); | |
557 | /* should be initialized for __put_super_and_need_restart() */ | |
a5166169 | 558 | hlist_del_init(&sb->s_instances); |
1da177e4 | 559 | spin_unlock(&sb_lock); |
0ed33598 | 560 | super_unlock_excl(sb); |
c1844d53 | 561 | if (sb->s_bdi != &noop_backing_dev_info) { |
0b3ea092 CH |
562 | if (sb->s_iflags & SB_I_PERSB_BDI) |
563 | bdi_unregister(sb->s_bdi); | |
fca39346 JK |
564 | bdi_put(sb->s_bdi); |
565 | sb->s_bdi = &noop_backing_dev_info; | |
fca39346 | 566 | } |
1da177e4 LT |
567 | } |
568 | ||
569 | EXPORT_SYMBOL(generic_shutdown_super); | |
570 | ||
20284ab7 | 571 | bool mount_capable(struct fs_context *fc) |
0ce0cf12 | 572 | { |
20284ab7 | 573 | if (!(fc->fs_type->fs_flags & FS_USERNS_MOUNT)) |
0ce0cf12 AV |
574 | return capable(CAP_SYS_ADMIN); |
575 | else | |
c2c44ec2 | 576 | return ns_capable(fc->user_ns, CAP_SYS_ADMIN); |
0ce0cf12 AV |
577 | } |
578 | ||
cb50b348 AV |
579 | /** |
580 | * sget_fc - Find or create a superblock | |
581 | * @fc: Filesystem context. | |
582 | * @test: Comparison callback | |
583 | * @set: Setup callback | |
584 | * | |
585 | * Find or create a superblock using the parameters stored in the filesystem | |
586 | * context and the two callback functions. | |
587 | * | |
588 | * If an extant superblock is matched, then that will be returned with an | |
589 | * elevated reference count that the caller must transfer or discard. | |
590 | * | |
591 | * If no match is made, a new superblock will be allocated and basic | |
592 | * initialisation will be performed (s_type, s_fs_info and s_id will be set and | |
593 | * the set() callback will be invoked), the superblock will be published and it | |
594 | * will be returned in a partially constructed state with SB_BORN and SB_ACTIVE | |
595 | * as yet unset. | |
596 | */ | |
597 | struct super_block *sget_fc(struct fs_context *fc, | |
598 | int (*test)(struct super_block *, struct fs_context *), | |
599 | int (*set)(struct super_block *, struct fs_context *)) | |
600 | { | |
601 | struct super_block *s = NULL; | |
602 | struct super_block *old; | |
603 | struct user_namespace *user_ns = fc->global ? &init_user_ns : fc->user_ns; | |
604 | int err; | |
605 | ||
cb50b348 AV |
606 | retry: |
607 | spin_lock(&sb_lock); | |
608 | if (test) { | |
609 | hlist_for_each_entry(old, &fc->fs_type->fs_supers, s_instances) { | |
610 | if (test(old, fc)) | |
611 | goto share_extant_sb; | |
612 | } | |
613 | } | |
614 | if (!s) { | |
615 | spin_unlock(&sb_lock); | |
616 | s = alloc_super(fc->fs_type, fc->sb_flags, user_ns); | |
617 | if (!s) | |
618 | return ERR_PTR(-ENOMEM); | |
619 | goto retry; | |
620 | } | |
621 | ||
622 | s->s_fs_info = fc->s_fs_info; | |
623 | err = set(s, fc); | |
624 | if (err) { | |
625 | s->s_fs_info = NULL; | |
626 | spin_unlock(&sb_lock); | |
627 | destroy_unused_super(s); | |
628 | return ERR_PTR(err); | |
629 | } | |
630 | fc->s_fs_info = NULL; | |
631 | s->s_type = fc->fs_type; | |
c80fa7c8 | 632 | s->s_iflags |= fc->s_iflags; |
c642256b | 633 | strscpy(s->s_id, s->s_type->name, sizeof(s->s_id)); |
cb50b348 AV |
634 | list_add_tail(&s->s_list, &super_blocks); |
635 | hlist_add_head(&s->s_instances, &s->s_type->fs_supers); | |
636 | spin_unlock(&sb_lock); | |
637 | get_filesystem(s->s_type); | |
638 | register_shrinker_prepared(&s->s_shrink); | |
639 | return s; | |
640 | ||
641 | share_extant_sb: | |
642 | if (user_ns != old->s_user_ns) { | |
643 | spin_unlock(&sb_lock); | |
644 | destroy_unused_super(s); | |
645 | return ERR_PTR(-EBUSY); | |
646 | } | |
647 | if (!grab_super(old)) | |
648 | goto retry; | |
649 | destroy_unused_super(s); | |
650 | return old; | |
651 | } | |
652 | EXPORT_SYMBOL(sget_fc); | |
653 | ||
1da177e4 | 654 | /** |
023d066a DH |
655 | * sget - find or create a superblock |
656 | * @type: filesystem type superblock should belong to | |
657 | * @test: comparison callback | |
658 | * @set: setup callback | |
659 | * @flags: mount flags | |
660 | * @data: argument to each of them | |
1da177e4 | 661 | */ |
023d066a | 662 | struct super_block *sget(struct file_system_type *type, |
1da177e4 LT |
663 | int (*test)(struct super_block *,void *), |
664 | int (*set)(struct super_block *,void *), | |
023d066a | 665 | int flags, |
1da177e4 LT |
666 | void *data) |
667 | { | |
023d066a | 668 | struct user_namespace *user_ns = current_user_ns(); |
1da177e4 | 669 | struct super_block *s = NULL; |
d4730127 | 670 | struct super_block *old; |
1da177e4 LT |
671 | int err; |
672 | ||
023d066a DH |
673 | /* We don't yet pass the user namespace of the parent |
674 | * mount through to here so always use &init_user_ns | |
675 | * until that changes. | |
676 | */ | |
677 | if (flags & SB_SUBMOUNT) | |
678 | user_ns = &init_user_ns; | |
679 | ||
1da177e4 LT |
680 | retry: |
681 | spin_lock(&sb_lock); | |
d4730127 | 682 | if (test) { |
b67bfe0d | 683 | hlist_for_each_entry(old, &type->fs_supers, s_instances) { |
d4730127 MK |
684 | if (!test(old, data)) |
685 | continue; | |
6e4eab57 EB |
686 | if (user_ns != old->s_user_ns) { |
687 | spin_unlock(&sb_lock); | |
0200894d | 688 | destroy_unused_super(s); |
6e4eab57 EB |
689 | return ERR_PTR(-EBUSY); |
690 | } | |
d4730127 MK |
691 | if (!grab_super(old)) |
692 | goto retry; | |
0200894d | 693 | destroy_unused_super(s); |
d4730127 MK |
694 | return old; |
695 | } | |
1da177e4 LT |
696 | } |
697 | if (!s) { | |
698 | spin_unlock(&sb_lock); | |
e462ec50 | 699 | s = alloc_super(type, (flags & ~SB_SUBMOUNT), user_ns); |
1da177e4 LT |
700 | if (!s) |
701 | return ERR_PTR(-ENOMEM); | |
702 | goto retry; | |
703 | } | |
dd111b31 | 704 | |
1da177e4 LT |
705 | err = set(s, data); |
706 | if (err) { | |
707 | spin_unlock(&sb_lock); | |
0200894d | 708 | destroy_unused_super(s); |
1da177e4 LT |
709 | return ERR_PTR(err); |
710 | } | |
711 | s->s_type = type; | |
c642256b | 712 | strscpy(s->s_id, type->name, sizeof(s->s_id)); |
1da177e4 | 713 | list_add_tail(&s->s_list, &super_blocks); |
a5166169 | 714 | hlist_add_head(&s->s_instances, &type->fs_supers); |
1da177e4 LT |
715 | spin_unlock(&sb_lock); |
716 | get_filesystem(type); | |
8e04944f | 717 | register_shrinker_prepared(&s->s_shrink); |
1da177e4 LT |
718 | return s; |
719 | } | |
1da177e4 LT |
720 | EXPORT_SYMBOL(sget); |
721 | ||
722 | void drop_super(struct super_block *sb) | |
723 | { | |
0ed33598 | 724 | super_unlock_shared(sb); |
1da177e4 LT |
725 | put_super(sb); |
726 | } | |
727 | ||
728 | EXPORT_SYMBOL(drop_super); | |
729 | ||
ba6379f7 JK |
730 | void drop_super_exclusive(struct super_block *sb) |
731 | { | |
0ed33598 | 732 | super_unlock_excl(sb); |
ba6379f7 JK |
733 | put_super(sb); |
734 | } | |
735 | EXPORT_SYMBOL(drop_super_exclusive); | |
736 | ||
fa7c1d50 MG |
737 | static void __iterate_supers(void (*f)(struct super_block *)) |
738 | { | |
739 | struct super_block *sb, *p = NULL; | |
740 | ||
741 | spin_lock(&sb_lock); | |
742 | list_for_each_entry(sb, &super_blocks, s_list) { | |
743 | if (hlist_unhashed(&sb->s_instances)) | |
744 | continue; | |
745 | sb->s_count++; | |
746 | spin_unlock(&sb_lock); | |
747 | ||
748 | f(sb); | |
749 | ||
750 | spin_lock(&sb_lock); | |
751 | if (p) | |
752 | __put_super(p); | |
753 | p = sb; | |
754 | } | |
755 | if (p) | |
756 | __put_super(p); | |
757 | spin_unlock(&sb_lock); | |
758 | } | |
01a05b33 AV |
759 | /** |
760 | * iterate_supers - call function for all active superblocks | |
761 | * @f: function to call | |
762 | * @arg: argument to pass to it | |
763 | * | |
764 | * Scans the superblock list and calls given function, passing it | |
765 | * locked superblock and given argument. | |
766 | */ | |
767 | void iterate_supers(void (*f)(struct super_block *, void *), void *arg) | |
768 | { | |
dca33252 | 769 | struct super_block *sb, *p = NULL; |
01a05b33 AV |
770 | |
771 | spin_lock(&sb_lock); | |
dca33252 | 772 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 773 | if (hlist_unhashed(&sb->s_instances)) |
01a05b33 AV |
774 | continue; |
775 | sb->s_count++; | |
776 | spin_unlock(&sb_lock); | |
777 | ||
0ed33598 | 778 | super_lock_shared(sb); |
e462ec50 | 779 | if (sb->s_root && (sb->s_flags & SB_BORN)) |
01a05b33 | 780 | f(sb, arg); |
0ed33598 | 781 | super_unlock_shared(sb); |
01a05b33 AV |
782 | |
783 | spin_lock(&sb_lock); | |
dca33252 AV |
784 | if (p) |
785 | __put_super(p); | |
786 | p = sb; | |
01a05b33 | 787 | } |
dca33252 AV |
788 | if (p) |
789 | __put_super(p); | |
01a05b33 AV |
790 | spin_unlock(&sb_lock); |
791 | } | |
792 | ||
43e15cdb AV |
793 | /** |
794 | * iterate_supers_type - call function for superblocks of given type | |
795 | * @type: fs type | |
796 | * @f: function to call | |
797 | * @arg: argument to pass to it | |
798 | * | |
799 | * Scans the superblock list and calls given function, passing it | |
800 | * locked superblock and given argument. | |
801 | */ | |
802 | void iterate_supers_type(struct file_system_type *type, | |
803 | void (*f)(struct super_block *, void *), void *arg) | |
804 | { | |
805 | struct super_block *sb, *p = NULL; | |
806 | ||
807 | spin_lock(&sb_lock); | |
b67bfe0d | 808 | hlist_for_each_entry(sb, &type->fs_supers, s_instances) { |
43e15cdb AV |
809 | sb->s_count++; |
810 | spin_unlock(&sb_lock); | |
811 | ||
0ed33598 | 812 | super_lock_shared(sb); |
e462ec50 | 813 | if (sb->s_root && (sb->s_flags & SB_BORN)) |
43e15cdb | 814 | f(sb, arg); |
0ed33598 | 815 | super_unlock_shared(sb); |
43e15cdb AV |
816 | |
817 | spin_lock(&sb_lock); | |
818 | if (p) | |
819 | __put_super(p); | |
820 | p = sb; | |
821 | } | |
822 | if (p) | |
823 | __put_super(p); | |
824 | spin_unlock(&sb_lock); | |
825 | } | |
826 | ||
827 | EXPORT_SYMBOL(iterate_supers_type); | |
828 | ||
4504230a CH |
829 | /** |
830 | * get_active_super - get an active reference to the superblock of a device | |
831 | * @bdev: device to get the superblock for | |
832 | * | |
833 | * Scans the superblock list and finds the superblock of the file system | |
834 | * mounted on the device given. Returns the superblock with an active | |
d3f21473 | 835 | * reference or %NULL if none was found. |
4504230a CH |
836 | */ |
837 | struct super_block *get_active_super(struct block_device *bdev) | |
838 | { | |
839 | struct super_block *sb; | |
840 | ||
841 | if (!bdev) | |
842 | return NULL; | |
843 | ||
1494583d | 844 | restart: |
4504230a CH |
845 | spin_lock(&sb_lock); |
846 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 847 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 848 | continue; |
1494583d | 849 | if (sb->s_bdev == bdev) { |
acfec9a5 | 850 | if (!grab_super(sb)) |
1494583d | 851 | goto restart; |
0ed33598 | 852 | super_unlock_excl(sb); |
acfec9a5 | 853 | return sb; |
1494583d | 854 | } |
4504230a CH |
855 | } |
856 | spin_unlock(&sb_lock); | |
857 | return NULL; | |
858 | } | |
dd111b31 | 859 | |
4e7b5671 | 860 | struct super_block *user_get_super(dev_t dev, bool excl) |
1da177e4 | 861 | { |
618f0636 | 862 | struct super_block *sb; |
1da177e4 | 863 | |
1da177e4 | 864 | spin_lock(&sb_lock); |
618f0636 KK |
865 | rescan: |
866 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 867 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 868 | continue; |
618f0636 KK |
869 | if (sb->s_dev == dev) { |
870 | sb->s_count++; | |
1da177e4 | 871 | spin_unlock(&sb_lock); |
0ed33598 | 872 | super_lock(sb, excl); |
df40c01a | 873 | /* still alive? */ |
e462ec50 | 874 | if (sb->s_root && (sb->s_flags & SB_BORN)) |
618f0636 | 875 | return sb; |
0ed33598 | 876 | super_unlock(sb, excl); |
df40c01a | 877 | /* nope, got unmounted */ |
618f0636 | 878 | spin_lock(&sb_lock); |
df40c01a AV |
879 | __put_super(sb); |
880 | goto rescan; | |
1da177e4 LT |
881 | } |
882 | } | |
883 | spin_unlock(&sb_lock); | |
884 | return NULL; | |
885 | } | |
886 | ||
1da177e4 | 887 | /** |
8d0347f6 DH |
888 | * reconfigure_super - asks filesystem to change superblock parameters |
889 | * @fc: The superblock and configuration | |
1da177e4 | 890 | * |
8d0347f6 | 891 | * Alters the configuration parameters of a live superblock. |
1da177e4 | 892 | */ |
8d0347f6 | 893 | int reconfigure_super(struct fs_context *fc) |
1da177e4 | 894 | { |
8d0347f6 | 895 | struct super_block *sb = fc->root->d_sb; |
1da177e4 | 896 | int retval; |
8d0347f6 | 897 | bool remount_ro = false; |
c541dce8 | 898 | bool remount_rw = false; |
8d0347f6 | 899 | bool force = fc->sb_flags & SB_FORCE; |
4504230a | 900 | |
8d0347f6 DH |
901 | if (fc->sb_flags_mask & ~MS_RMT_MASK) |
902 | return -EINVAL; | |
5accdf82 | 903 | if (sb->s_writers.frozen != SB_UNFROZEN) |
4504230a CH |
904 | return -EBUSY; |
905 | ||
8d0347f6 DH |
906 | retval = security_sb_remount(sb, fc->security); |
907 | if (retval) | |
908 | return retval; | |
909 | ||
910 | if (fc->sb_flags_mask & SB_RDONLY) { | |
9361401e | 911 | #ifdef CONFIG_BLOCK |
6f0d9689 CH |
912 | if (!(fc->sb_flags & SB_RDONLY) && sb->s_bdev && |
913 | bdev_read_only(sb->s_bdev)) | |
8d0347f6 | 914 | return -EACCES; |
9361401e | 915 | #endif |
c541dce8 | 916 | remount_rw = !(fc->sb_flags & SB_RDONLY) && sb_rdonly(sb); |
8d0347f6 DH |
917 | remount_ro = (fc->sb_flags & SB_RDONLY) && !sb_rdonly(sb); |
918 | } | |
d208bbdd | 919 | |
0aec09d0 | 920 | if (remount_ro) { |
fdab684d | 921 | if (!hlist_empty(&sb->s_pins)) { |
0ed33598 | 922 | super_unlock_excl(sb); |
fdab684d | 923 | group_pin_kill(&sb->s_pins); |
0ed33598 | 924 | super_lock_excl(sb); |
0aec09d0 AV |
925 | if (!sb->s_root) |
926 | return 0; | |
927 | if (sb->s_writers.frozen != SB_UNFROZEN) | |
928 | return -EBUSY; | |
8d0347f6 | 929 | remount_ro = !sb_rdonly(sb); |
0aec09d0 AV |
930 | } |
931 | } | |
932 | shrink_dcache_sb(sb); | |
933 | ||
8d0347f6 DH |
934 | /* If we are reconfiguring to RDONLY and current sb is read/write, |
935 | * make sure there are no files open for writing. | |
936 | */ | |
d208bbdd | 937 | if (remount_ro) { |
4ed5e82f | 938 | if (force) { |
d7439fb1 | 939 | sb_start_ro_state_change(sb); |
4ed5e82f MS |
940 | } else { |
941 | retval = sb_prepare_remount_readonly(sb); | |
942 | if (retval) | |
943 | return retval; | |
4ed5e82f | 944 | } |
c541dce8 JK |
945 | } else if (remount_rw) { |
946 | /* | |
d7439fb1 JK |
947 | * Protect filesystem's reconfigure code from writes from |
948 | * userspace until reconfigure finishes. | |
c541dce8 | 949 | */ |
d7439fb1 | 950 | sb_start_ro_state_change(sb); |
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))); | |
d7439fb1 | 966 | sb_end_ro_state_change(sb); |
c79d967d | 967 | |
d208bbdd NP |
968 | /* |
969 | * Some filesystems modify their metadata via some other path than the | |
970 | * bdev buffer cache (eg. use a private mapping, or directories in | |
971 | * pagecache, etc). Also file data modifications go via their own | |
972 | * mappings. So If we try to mount readonly then copy the filesystem | |
973 | * from bdev, we could get stale data, so invalidate it to give a best | |
974 | * effort at coherency. | |
975 | */ | |
976 | if (remount_ro && sb->s_bdev) | |
977 | invalidate_bdev(sb->s_bdev); | |
1da177e4 | 978 | return 0; |
4ed5e82f MS |
979 | |
980 | cancel_readonly: | |
d7439fb1 | 981 | sb_end_ro_state_change(sb); |
4ed5e82f | 982 | return retval; |
1da177e4 LT |
983 | } |
984 | ||
fa7c1d50 | 985 | static void do_emergency_remount_callback(struct super_block *sb) |
1da177e4 | 986 | { |
0ed33598 | 987 | super_lock_excl(sb); |
fa7c1d50 MG |
988 | if (sb->s_root && sb->s_bdev && (sb->s_flags & SB_BORN) && |
989 | !sb_rdonly(sb)) { | |
8d0347f6 DH |
990 | struct fs_context *fc; |
991 | ||
992 | fc = fs_context_for_reconfigure(sb->s_root, | |
993 | SB_RDONLY | SB_FORCE, SB_RDONLY); | |
994 | if (!IS_ERR(fc)) { | |
995 | if (parse_monolithic_mount_data(fc, NULL) == 0) | |
996 | (void)reconfigure_super(fc); | |
997 | put_fs_context(fc); | |
998 | } | |
1da177e4 | 999 | } |
0ed33598 | 1000 | super_unlock_excl(sb); |
fa7c1d50 MG |
1001 | } |
1002 | ||
1003 | static void do_emergency_remount(struct work_struct *work) | |
1004 | { | |
1005 | __iterate_supers(do_emergency_remount_callback); | |
a2a9537a | 1006 | kfree(work); |
1da177e4 LT |
1007 | printk("Emergency Remount complete\n"); |
1008 | } | |
1009 | ||
1010 | void emergency_remount(void) | |
1011 | { | |
a2a9537a JA |
1012 | struct work_struct *work; |
1013 | ||
1014 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | |
1015 | if (work) { | |
1016 | INIT_WORK(work, do_emergency_remount); | |
1017 | schedule_work(work); | |
1018 | } | |
1da177e4 LT |
1019 | } |
1020 | ||
08fdc8a0 MG |
1021 | static void do_thaw_all_callback(struct super_block *sb) |
1022 | { | |
0ed33598 | 1023 | super_lock_excl(sb); |
1c18d2a1 | 1024 | if (sb->s_root && sb->s_flags & SB_BORN) { |
08fdc8a0 MG |
1025 | emergency_thaw_bdev(sb); |
1026 | thaw_super_locked(sb); | |
1027 | } else { | |
0ed33598 | 1028 | super_unlock_excl(sb); |
08fdc8a0 MG |
1029 | } |
1030 | } | |
1031 | ||
1032 | static void do_thaw_all(struct work_struct *work) | |
1033 | { | |
1034 | __iterate_supers(do_thaw_all_callback); | |
1035 | kfree(work); | |
1036 | printk(KERN_WARNING "Emergency Thaw complete\n"); | |
1037 | } | |
1038 | ||
1039 | /** | |
1040 | * emergency_thaw_all -- forcibly thaw every frozen filesystem | |
1041 | * | |
1042 | * Used for emergency unfreeze of all filesystems via SysRq | |
1043 | */ | |
1044 | void emergency_thaw_all(void) | |
1045 | { | |
1046 | struct work_struct *work; | |
1047 | ||
1048 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | |
1049 | if (work) { | |
1050 | INIT_WORK(work, do_thaw_all); | |
1051 | schedule_work(work); | |
1052 | } | |
1053 | } | |
1054 | ||
ad76cbc6 | 1055 | static DEFINE_IDA(unnamed_dev_ida); |
1da177e4 | 1056 | |
5a66847e MW |
1057 | /** |
1058 | * get_anon_bdev - Allocate a block device for filesystems which don't have one. | |
1059 | * @p: Pointer to a dev_t. | |
1060 | * | |
1061 | * Filesystems which don't use real block devices can call this function | |
1062 | * to allocate a virtual block device. | |
1063 | * | |
1064 | * Context: Any context. Frequently called while holding sb_lock. | |
1065 | * Return: 0 on success, -EMFILE if there are no anonymous bdevs left | |
1066 | * or -ENOMEM if memory allocation failed. | |
1067 | */ | |
0ee5dc67 | 1068 | int get_anon_bdev(dev_t *p) |
1da177e4 LT |
1069 | { |
1070 | int dev; | |
5a66847e MW |
1071 | |
1072 | /* | |
1073 | * Many userspace utilities consider an FSID of 0 invalid. | |
1074 | * Always return at least 1 from get_anon_bdev. | |
1075 | */ | |
1076 | dev = ida_alloc_range(&unnamed_dev_ida, 1, (1 << MINORBITS) - 1, | |
1077 | GFP_ATOMIC); | |
1078 | if (dev == -ENOSPC) | |
1079 | dev = -EMFILE; | |
1080 | if (dev < 0) | |
1081 | return dev; | |
1082 | ||
1083 | *p = MKDEV(0, dev); | |
1da177e4 LT |
1084 | return 0; |
1085 | } | |
0ee5dc67 | 1086 | EXPORT_SYMBOL(get_anon_bdev); |
1da177e4 | 1087 | |
0ee5dc67 | 1088 | void free_anon_bdev(dev_t dev) |
1da177e4 | 1089 | { |
5a66847e | 1090 | ida_free(&unnamed_dev_ida, MINOR(dev)); |
1da177e4 | 1091 | } |
0ee5dc67 AV |
1092 | EXPORT_SYMBOL(free_anon_bdev); |
1093 | ||
1094 | int set_anon_super(struct super_block *s, void *data) | |
1095 | { | |
df0ce26c | 1096 | return get_anon_bdev(&s->s_dev); |
0ee5dc67 | 1097 | } |
0ee5dc67 AV |
1098 | EXPORT_SYMBOL(set_anon_super); |
1099 | ||
1100 | void kill_anon_super(struct super_block *sb) | |
1101 | { | |
1102 | dev_t dev = sb->s_dev; | |
1103 | generic_shutdown_super(sb); | |
1104 | free_anon_bdev(dev); | |
1105 | } | |
1da177e4 LT |
1106 | EXPORT_SYMBOL(kill_anon_super); |
1107 | ||
1da177e4 LT |
1108 | void kill_litter_super(struct super_block *sb) |
1109 | { | |
1110 | if (sb->s_root) | |
1111 | d_genocide(sb->s_root); | |
1112 | kill_anon_super(sb); | |
1113 | } | |
1da177e4 LT |
1114 | EXPORT_SYMBOL(kill_litter_super); |
1115 | ||
cb50b348 AV |
1116 | int set_anon_super_fc(struct super_block *sb, struct fs_context *fc) |
1117 | { | |
1118 | return set_anon_super(sb, NULL); | |
1119 | } | |
1120 | EXPORT_SYMBOL(set_anon_super_fc); | |
1121 | ||
1122 | static int test_keyed_super(struct super_block *sb, struct fs_context *fc) | |
1123 | { | |
1124 | return sb->s_fs_info == fc->s_fs_info; | |
1125 | } | |
1126 | ||
1127 | static int test_single_super(struct super_block *s, struct fs_context *fc) | |
1128 | { | |
1129 | return 1; | |
1130 | } | |
1131 | ||
cda2ed05 CH |
1132 | static int vfs_get_super(struct fs_context *fc, bool reconf, |
1133 | int (*test)(struct super_block *, struct fs_context *), | |
1134 | int (*fill_super)(struct super_block *sb, | |
1135 | struct fs_context *fc)) | |
cb50b348 | 1136 | { |
cb50b348 | 1137 | struct super_block *sb; |
43ce4c1f | 1138 | int err; |
cb50b348 | 1139 | |
cb50b348 AV |
1140 | sb = sget_fc(fc, test, set_anon_super_fc); |
1141 | if (IS_ERR(sb)) | |
1142 | return PTR_ERR(sb); | |
1143 | ||
1144 | if (!sb->s_root) { | |
43ce4c1f DH |
1145 | err = fill_super(sb, fc); |
1146 | if (err) | |
1147 | goto error; | |
cb50b348 AV |
1148 | |
1149 | sb->s_flags |= SB_ACTIVE; | |
43ce4c1f DH |
1150 | fc->root = dget(sb->s_root); |
1151 | } else { | |
1152 | fc->root = dget(sb->s_root); | |
cda2ed05 | 1153 | if (reconf) { |
43ce4c1f DH |
1154 | err = reconfigure_super(fc); |
1155 | if (err < 0) { | |
1156 | dput(fc->root); | |
1157 | fc->root = NULL; | |
1158 | goto error; | |
1159 | } | |
1160 | } | |
cb50b348 AV |
1161 | } |
1162 | ||
cb50b348 | 1163 | return 0; |
43ce4c1f DH |
1164 | |
1165 | error: | |
1166 | deactivate_locked_super(sb); | |
1167 | return err; | |
cb50b348 | 1168 | } |
cb50b348 | 1169 | |
2ac295d4 AV |
1170 | int get_tree_nodev(struct fs_context *fc, |
1171 | int (*fill_super)(struct super_block *sb, | |
1172 | struct fs_context *fc)) | |
1173 | { | |
cda2ed05 | 1174 | return vfs_get_super(fc, false, NULL, fill_super); |
2ac295d4 AV |
1175 | } |
1176 | EXPORT_SYMBOL(get_tree_nodev); | |
1177 | ||
c23a0bba AV |
1178 | int get_tree_single(struct fs_context *fc, |
1179 | int (*fill_super)(struct super_block *sb, | |
1180 | struct fs_context *fc)) | |
1181 | { | |
cda2ed05 | 1182 | return vfs_get_super(fc, false, test_single_super, fill_super); |
c23a0bba AV |
1183 | } |
1184 | EXPORT_SYMBOL(get_tree_single); | |
1185 | ||
43ce4c1f DH |
1186 | int get_tree_single_reconf(struct fs_context *fc, |
1187 | int (*fill_super)(struct super_block *sb, | |
1188 | struct fs_context *fc)) | |
1189 | { | |
cda2ed05 | 1190 | return vfs_get_super(fc, true, test_single_super, fill_super); |
43ce4c1f DH |
1191 | } |
1192 | EXPORT_SYMBOL(get_tree_single_reconf); | |
1193 | ||
533770cc AV |
1194 | int get_tree_keyed(struct fs_context *fc, |
1195 | int (*fill_super)(struct super_block *sb, | |
1196 | struct fs_context *fc), | |
1197 | void *key) | |
1198 | { | |
1199 | fc->s_fs_info = key; | |
cda2ed05 | 1200 | return vfs_get_super(fc, false, test_keyed_super, fill_super); |
533770cc AV |
1201 | } |
1202 | EXPORT_SYMBOL(get_tree_keyed); | |
1203 | ||
9361401e | 1204 | #ifdef CONFIG_BLOCK |
9c09a7cf CH |
1205 | /* |
1206 | * Lock a super block that the callers holds a reference to. | |
1207 | * | |
1208 | * The caller needs to ensure that the super_block isn't being freed while | |
1209 | * calling this function, e.g. by holding a lock over the call to this function | |
1210 | * and the place that clears the pointer to the superblock used by this function | |
1211 | * before freeing the superblock. | |
1212 | */ | |
1213 | static bool lock_active_super(struct super_block *sb) | |
1214 | { | |
0ed33598 | 1215 | super_lock_shared(sb); |
9c09a7cf CH |
1216 | if (!sb->s_root || |
1217 | (sb->s_flags & (SB_ACTIVE | SB_BORN)) != (SB_ACTIVE | SB_BORN)) { | |
0ed33598 | 1218 | super_unlock_shared(sb); |
9c09a7cf CH |
1219 | return false; |
1220 | } | |
1221 | return true; | |
1222 | } | |
1223 | ||
d8530de5 | 1224 | static void fs_bdev_mark_dead(struct block_device *bdev, bool surprise) |
87efb390 | 1225 | { |
9c09a7cf | 1226 | struct super_block *sb = bdev->bd_holder; |
87efb390 | 1227 | |
9c09a7cf CH |
1228 | /* bd_holder_lock ensures that the sb isn't freed */ |
1229 | lockdep_assert_held(&bdev->bd_holder_lock); | |
1230 | ||
1231 | if (!lock_active_super(sb)) | |
87efb390 CH |
1232 | return; |
1233 | ||
d8530de5 CH |
1234 | if (!surprise) |
1235 | sync_filesystem(sb); | |
1236 | shrink_dcache_sb(sb); | |
e127b9bc | 1237 | invalidate_inodes(sb); |
87efb390 CH |
1238 | if (sb->s_op->shutdown) |
1239 | sb->s_op->shutdown(sb); | |
9c09a7cf | 1240 | |
0ed33598 | 1241 | super_unlock_shared(sb); |
87efb390 CH |
1242 | } |
1243 | ||
2142b88c CH |
1244 | static void fs_bdev_sync(struct block_device *bdev) |
1245 | { | |
1246 | struct super_block *sb = bdev->bd_holder; | |
1247 | ||
1248 | lockdep_assert_held(&bdev->bd_holder_lock); | |
1249 | ||
1250 | if (!lock_active_super(sb)) | |
1251 | return; | |
1252 | sync_filesystem(sb); | |
0ed33598 | 1253 | super_unlock_shared(sb); |
2142b88c CH |
1254 | } |
1255 | ||
7ecd0b6f | 1256 | const struct blk_holder_ops fs_holder_ops = { |
d8530de5 | 1257 | .mark_dead = fs_bdev_mark_dead, |
2142b88c | 1258 | .sync = fs_bdev_sync, |
87efb390 | 1259 | }; |
7ecd0b6f | 1260 | EXPORT_SYMBOL_GPL(fs_holder_ops); |
fe62c3a4 | 1261 | |
1da177e4 LT |
1262 | static int set_bdev_super(struct super_block *s, void *data) |
1263 | { | |
aca740ce | 1264 | s->s_dev = *(dev_t *)data; |
1da177e4 LT |
1265 | return 0; |
1266 | } | |
1267 | ||
fe62c3a4 DH |
1268 | static int set_bdev_super_fc(struct super_block *s, struct fs_context *fc) |
1269 | { | |
1270 | return set_bdev_super(s, fc->sget_key); | |
1271 | } | |
1272 | ||
1273 | static int test_bdev_super_fc(struct super_block *s, struct fs_context *fc) | |
1274 | { | |
aca740ce JK |
1275 | return !(s->s_iflags & SB_I_RETIRED) && |
1276 | s->s_dev == *(dev_t *)fc->sget_key; | |
1277 | } | |
1278 | ||
cf6da236 | 1279 | int setup_bdev_super(struct super_block *sb, int sb_flags, |
aca740ce JK |
1280 | struct fs_context *fc) |
1281 | { | |
1282 | blk_mode_t mode = sb_open_mode(sb_flags); | |
1283 | struct block_device *bdev; | |
1284 | ||
2ea6f689 | 1285 | bdev = blkdev_get_by_dev(sb->s_dev, mode, sb, &fs_holder_ops); |
aca740ce JK |
1286 | if (IS_ERR(bdev)) { |
1287 | if (fc) | |
1288 | errorf(fc, "%s: Can't open blockdev", fc->source); | |
1289 | return PTR_ERR(bdev); | |
1290 | } | |
1291 | ||
1292 | /* | |
1293 | * This really should be in blkdev_get_by_dev, but right now can't due | |
1294 | * to legacy issues that require us to allow opening a block device node | |
1295 | * writable from userspace even for a read-only block device. | |
1296 | */ | |
1297 | if ((mode & BLK_OPEN_WRITE) && bdev_read_only(bdev)) { | |
2ea6f689 | 1298 | blkdev_put(bdev, sb); |
aca740ce JK |
1299 | return -EACCES; |
1300 | } | |
1301 | ||
1302 | /* | |
1303 | * Until SB_BORN flag is set, there can be no active superblock | |
1304 | * references and thus no filesystem freezing. get_active_super() will | |
1305 | * just loop waiting for SB_BORN so even freeze_bdev() cannot proceed. | |
1306 | * | |
1307 | * It is enough to check bdev was not frozen before we set s_bdev. | |
1308 | */ | |
1309 | mutex_lock(&bdev->bd_fsfreeze_mutex); | |
1310 | if (bdev->bd_fsfreeze_count > 0) { | |
1311 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
1312 | if (fc) | |
1313 | warnf(fc, "%pg: Can't mount, blockdev is frozen", bdev); | |
2ea6f689 | 1314 | blkdev_put(bdev, sb); |
aca740ce JK |
1315 | return -EBUSY; |
1316 | } | |
1317 | spin_lock(&sb_lock); | |
1318 | sb->s_bdev = bdev; | |
1319 | sb->s_bdi = bdi_get(bdev->bd_disk->bdi); | |
1320 | if (bdev_stable_writes(bdev)) | |
1321 | sb->s_iflags |= SB_I_STABLE_WRITES; | |
1322 | spin_unlock(&sb_lock); | |
1323 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
1324 | ||
1325 | snprintf(sb->s_id, sizeof(sb->s_id), "%pg", bdev); | |
1326 | shrinker_debugfs_rename(&sb->s_shrink, "sb-%s:%s", sb->s_type->name, | |
1327 | sb->s_id); | |
1328 | sb_set_blocksize(sb, block_size(bdev)); | |
1329 | return 0; | |
fe62c3a4 | 1330 | } |
cf6da236 | 1331 | EXPORT_SYMBOL_GPL(setup_bdev_super); |
fe62c3a4 DH |
1332 | |
1333 | /** | |
1334 | * get_tree_bdev - Get a superblock based on a single block device | |
1335 | * @fc: The filesystem context holding the parameters | |
1336 | * @fill_super: Helper to initialise a new superblock | |
1337 | */ | |
1338 | int get_tree_bdev(struct fs_context *fc, | |
1339 | int (*fill_super)(struct super_block *, | |
1340 | struct fs_context *)) | |
1341 | { | |
fe62c3a4 | 1342 | struct super_block *s; |
fe62c3a4 | 1343 | int error = 0; |
aca740ce | 1344 | dev_t dev; |
fe62c3a4 | 1345 | |
fe62c3a4 DH |
1346 | if (!fc->source) |
1347 | return invalf(fc, "No source specified"); | |
1348 | ||
aca740ce JK |
1349 | error = lookup_bdev(fc->source, &dev); |
1350 | if (error) { | |
1351 | errorf(fc, "%s: Can't lookup blockdev", fc->source); | |
1352 | return error; | |
fe62c3a4 DH |
1353 | } |
1354 | ||
1355 | fc->sb_flags |= SB_NOSEC; | |
aca740ce | 1356 | fc->sget_key = &dev; |
fe62c3a4 | 1357 | s = sget_fc(fc, test_bdev_super_fc, set_bdev_super_fc); |
aca740ce | 1358 | if (IS_ERR(s)) |
fe62c3a4 DH |
1359 | return PTR_ERR(s); |
1360 | ||
1361 | if (s->s_root) { | |
1362 | /* Don't summarily change the RO/RW state. */ | |
1363 | if ((fc->sb_flags ^ s->s_flags) & SB_RDONLY) { | |
aca740ce | 1364 | warnf(fc, "%pg: Can't mount, would change RO state", s->s_bdev); |
fe62c3a4 | 1365 | deactivate_locked_super(s); |
fe62c3a4 DH |
1366 | return -EBUSY; |
1367 | } | |
aca740ce | 1368 | } else { |
fe62c3a4 | 1369 | /* |
aca740ce JK |
1370 | * We drop s_umount here because we need to open the bdev and |
1371 | * bdev->open_mutex ranks above s_umount (blkdev_put() -> | |
560e20e4 | 1372 | * bdev_mark_dead()). It is safe because we have active sb |
aca740ce | 1373 | * reference and SB_BORN is not set yet. |
fe62c3a4 | 1374 | */ |
0ed33598 | 1375 | super_unlock_excl(s); |
aca740ce | 1376 | error = setup_bdev_super(s, fc->sb_flags, fc); |
0ed33598 | 1377 | super_lock_excl(s); |
aca740ce JK |
1378 | if (!error) |
1379 | error = fill_super(s, fc); | |
fe62c3a4 DH |
1380 | if (error) { |
1381 | deactivate_locked_super(s); | |
1382 | return error; | |
1383 | } | |
fe62c3a4 | 1384 | s->s_flags |= SB_ACTIVE; |
fe62c3a4 DH |
1385 | } |
1386 | ||
1387 | BUG_ON(fc->root); | |
1388 | fc->root = dget(s->s_root); | |
1389 | return 0; | |
1390 | } | |
1391 | EXPORT_SYMBOL(get_tree_bdev); | |
1392 | ||
1da177e4 LT |
1393 | static int test_bdev_super(struct super_block *s, void *data) |
1394 | { | |
aca740ce | 1395 | return !(s->s_iflags & SB_I_RETIRED) && s->s_dev == *(dev_t *)data; |
1da177e4 LT |
1396 | } |
1397 | ||
152a0836 | 1398 | struct dentry *mount_bdev(struct file_system_type *fs_type, |
1da177e4 | 1399 | int flags, const char *dev_name, void *data, |
152a0836 | 1400 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 | 1401 | { |
1da177e4 | 1402 | struct super_block *s; |
aca740ce JK |
1403 | int error; |
1404 | dev_t dev; | |
1da177e4 | 1405 | |
aca740ce JK |
1406 | error = lookup_bdev(dev_name, &dev); |
1407 | if (error) | |
1408 | return ERR_PTR(error); | |
1da177e4 | 1409 | |
aca740ce JK |
1410 | flags |= SB_NOSEC; |
1411 | s = sget(fs_type, test_bdev_super, set_bdev_super, flags, &dev); | |
1da177e4 | 1412 | if (IS_ERR(s)) |
aca740ce | 1413 | return ERR_CAST(s); |
1da177e4 LT |
1414 | |
1415 | if (s->s_root) { | |
e462ec50 | 1416 | if ((flags ^ s->s_flags) & SB_RDONLY) { |
74dbbdd7 | 1417 | deactivate_locked_super(s); |
aca740ce | 1418 | return ERR_PTR(-EBUSY); |
1da177e4 | 1419 | } |
aca740ce | 1420 | } else { |
4f331f01 | 1421 | /* |
aca740ce JK |
1422 | * We drop s_umount here because we need to open the bdev and |
1423 | * bdev->open_mutex ranks above s_umount (blkdev_put() -> | |
560e20e4 | 1424 | * bdev_mark_dead()). It is safe because we have active sb |
aca740ce | 1425 | * reference and SB_BORN is not set yet. |
4f331f01 | 1426 | */ |
0ed33598 | 1427 | super_unlock_excl(s); |
aca740ce | 1428 | error = setup_bdev_super(s, flags, NULL); |
0ed33598 | 1429 | super_lock_excl(s); |
aca740ce JK |
1430 | if (!error) |
1431 | error = fill_super(s, data, flags & SB_SILENT ? 1 : 0); | |
1da177e4 | 1432 | if (error) { |
74dbbdd7 | 1433 | deactivate_locked_super(s); |
aca740ce | 1434 | return ERR_PTR(error); |
fa675765 | 1435 | } |
454e2398 | 1436 | |
e462ec50 | 1437 | s->s_flags |= SB_ACTIVE; |
1da177e4 LT |
1438 | } |
1439 | ||
152a0836 | 1440 | return dget(s->s_root); |
152a0836 AV |
1441 | } |
1442 | EXPORT_SYMBOL(mount_bdev); | |
1443 | ||
1da177e4 LT |
1444 | void kill_block_super(struct super_block *sb) |
1445 | { | |
1446 | struct block_device *bdev = sb->s_bdev; | |
1447 | ||
1da177e4 | 1448 | generic_shutdown_super(sb); |
aca740ce JK |
1449 | if (bdev) { |
1450 | sync_blockdev(bdev); | |
2ea6f689 | 1451 | blkdev_put(bdev, sb); |
aca740ce | 1452 | } |
1da177e4 LT |
1453 | } |
1454 | ||
1455 | EXPORT_SYMBOL(kill_block_super); | |
9361401e | 1456 | #endif |
1da177e4 | 1457 | |
3c26ff6e | 1458 | struct dentry *mount_nodev(struct file_system_type *fs_type, |
1da177e4 | 1459 | int flags, void *data, |
3c26ff6e | 1460 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1461 | { |
1462 | int error; | |
9249e17f | 1463 | struct super_block *s = sget(fs_type, NULL, set_anon_super, flags, NULL); |
1da177e4 LT |
1464 | |
1465 | if (IS_ERR(s)) | |
3c26ff6e | 1466 | return ERR_CAST(s); |
1da177e4 | 1467 | |
e462ec50 | 1468 | error = fill_super(s, data, flags & SB_SILENT ? 1 : 0); |
1da177e4 | 1469 | if (error) { |
74dbbdd7 | 1470 | deactivate_locked_super(s); |
3c26ff6e | 1471 | return ERR_PTR(error); |
1da177e4 | 1472 | } |
e462ec50 | 1473 | s->s_flags |= SB_ACTIVE; |
3c26ff6e | 1474 | return dget(s->s_root); |
1da177e4 | 1475 | } |
3c26ff6e AV |
1476 | EXPORT_SYMBOL(mount_nodev); |
1477 | ||
a6097180 N |
1478 | int reconfigure_single(struct super_block *s, |
1479 | int flags, void *data) | |
8d0347f6 DH |
1480 | { |
1481 | struct fs_context *fc; | |
1482 | int ret; | |
1483 | ||
1484 | /* The caller really need to be passing fc down into mount_single(), | |
1485 | * then a chunk of this can be removed. [Bollocks -- AV] | |
1486 | * Better yet, reconfiguration shouldn't happen, but rather the second | |
1487 | * mount should be rejected if the parameters are not compatible. | |
1488 | */ | |
1489 | fc = fs_context_for_reconfigure(s->s_root, flags, MS_RMT_MASK); | |
1490 | if (IS_ERR(fc)) | |
1491 | return PTR_ERR(fc); | |
1492 | ||
1493 | ret = parse_monolithic_mount_data(fc, data); | |
1494 | if (ret < 0) | |
1495 | goto out; | |
1496 | ||
1497 | ret = reconfigure_super(fc); | |
1498 | out: | |
1499 | put_fs_context(fc); | |
1500 | return ret; | |
1501 | } | |
1502 | ||
1da177e4 LT |
1503 | static int compare_single(struct super_block *s, void *p) |
1504 | { | |
1505 | return 1; | |
1506 | } | |
1507 | ||
fc14f2fe | 1508 | struct dentry *mount_single(struct file_system_type *fs_type, |
1da177e4 | 1509 | int flags, void *data, |
fc14f2fe | 1510 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1511 | { |
1512 | struct super_block *s; | |
1513 | int error; | |
1514 | ||
9249e17f | 1515 | s = sget(fs_type, compare_single, set_anon_super, flags, NULL); |
1da177e4 | 1516 | if (IS_ERR(s)) |
fc14f2fe | 1517 | return ERR_CAST(s); |
1da177e4 | 1518 | if (!s->s_root) { |
e462ec50 | 1519 | error = fill_super(s, data, flags & SB_SILENT ? 1 : 0); |
8d0347f6 DH |
1520 | if (!error) |
1521 | s->s_flags |= SB_ACTIVE; | |
9329d1be | 1522 | } else { |
8d0347f6 DH |
1523 | error = reconfigure_single(s, flags, data); |
1524 | } | |
1525 | if (unlikely(error)) { | |
1526 | deactivate_locked_super(s); | |
1527 | return ERR_PTR(error); | |
1da177e4 | 1528 | } |
fc14f2fe AV |
1529 | return dget(s->s_root); |
1530 | } | |
1531 | EXPORT_SYMBOL(mount_single); | |
1532 | ||
9bc61ab1 DH |
1533 | /** |
1534 | * vfs_get_tree - Get the mountable root | |
1535 | * @fc: The superblock configuration context. | |
1536 | * | |
1537 | * The filesystem is invoked to get or create a superblock which can then later | |
1538 | * be used for mounting. The filesystem places a pointer to the root to be | |
1539 | * used for mounting in @fc->root. | |
1540 | */ | |
1541 | int vfs_get_tree(struct fs_context *fc) | |
1da177e4 | 1542 | { |
9d412a43 | 1543 | struct super_block *sb; |
9bc61ab1 | 1544 | int error; |
8089352a | 1545 | |
f3a09c92 AV |
1546 | if (fc->root) |
1547 | return -EBUSY; | |
1548 | ||
1549 | /* Get the mountable root in fc->root, with a ref on the root and a ref | |
1550 | * on the superblock. | |
1551 | */ | |
1552 | error = fc->ops->get_tree(fc); | |
9bc61ab1 DH |
1553 | if (error < 0) |
1554 | return error; | |
1da177e4 | 1555 | |
f3a09c92 AV |
1556 | if (!fc->root) { |
1557 | pr_err("Filesystem %s get_tree() didn't set fc->root\n", | |
1558 | fc->fs_type->name); | |
1559 | /* We don't know what the locking state of the superblock is - | |
1560 | * if there is a superblock. | |
1561 | */ | |
1562 | BUG(); | |
1563 | } | |
1564 | ||
9bc61ab1 | 1565 | sb = fc->root->d_sb; |
9d412a43 | 1566 | WARN_ON(!sb->s_bdi); |
79f546a6 DC |
1567 | |
1568 | /* | |
1569 | * Write barrier is for super_cache_count(). We place it before setting | |
1570 | * SB_BORN as the data dependency between the two functions is the | |
1571 | * superblock structure contents that we just set up, not the SB_BORN | |
1572 | * flag. | |
1573 | */ | |
1574 | smp_wmb(); | |
e462ec50 | 1575 | sb->s_flags |= SB_BORN; |
454e2398 | 1576 | |
9bc61ab1 | 1577 | error = security_sb_set_mnt_opts(sb, fc->security, 0, NULL); |
c9ce29ed AV |
1578 | if (unlikely(error)) { |
1579 | fc_drop_locked(fc); | |
1580 | return error; | |
a10d7c22 AV |
1581 | } |
1582 | ||
42cb56ae JL |
1583 | /* |
1584 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE | |
1585 | * but s_maxbytes was an unsigned long long for many releases. Throw | |
1586 | * this warning for a little while to try and catch filesystems that | |
4358b567 | 1587 | * violate this rule. |
42cb56ae | 1588 | */ |
9d412a43 | 1589 | WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " |
9bc61ab1 | 1590 | "negative value (%lld)\n", fc->fs_type->name, sb->s_maxbytes); |
42cb56ae | 1591 | |
9bc61ab1 | 1592 | return 0; |
1da177e4 | 1593 | } |
9bc61ab1 | 1594 | EXPORT_SYMBOL(vfs_get_tree); |
1da177e4 | 1595 | |
fca39346 JK |
1596 | /* |
1597 | * Setup private BDI for given superblock. It gets automatically cleaned up | |
1598 | * in generic_shutdown_super(). | |
1599 | */ | |
1600 | int super_setup_bdi_name(struct super_block *sb, char *fmt, ...) | |
1601 | { | |
1602 | struct backing_dev_info *bdi; | |
1603 | int err; | |
1604 | va_list args; | |
1605 | ||
aef33c2f | 1606 | bdi = bdi_alloc(NUMA_NO_NODE); |
fca39346 JK |
1607 | if (!bdi) |
1608 | return -ENOMEM; | |
1609 | ||
fca39346 | 1610 | va_start(args, fmt); |
7c4cc300 | 1611 | err = bdi_register_va(bdi, fmt, args); |
fca39346 JK |
1612 | va_end(args); |
1613 | if (err) { | |
1614 | bdi_put(bdi); | |
1615 | return err; | |
1616 | } | |
1617 | WARN_ON(sb->s_bdi != &noop_backing_dev_info); | |
1618 | sb->s_bdi = bdi; | |
0b3ea092 | 1619 | sb->s_iflags |= SB_I_PERSB_BDI; |
fca39346 JK |
1620 | |
1621 | return 0; | |
1622 | } | |
1623 | EXPORT_SYMBOL(super_setup_bdi_name); | |
1624 | ||
1625 | /* | |
1626 | * Setup private BDI for given superblock. I gets automatically cleaned up | |
1627 | * in generic_shutdown_super(). | |
1628 | */ | |
1629 | int super_setup_bdi(struct super_block *sb) | |
1630 | { | |
1631 | static atomic_long_t bdi_seq = ATOMIC_LONG_INIT(0); | |
1632 | ||
1633 | return super_setup_bdi_name(sb, "%.28s-%ld", sb->s_type->name, | |
1634 | atomic_long_inc_return(&bdi_seq)); | |
1635 | } | |
1636 | EXPORT_SYMBOL(super_setup_bdi); | |
1637 | ||
5accdf82 JK |
1638 | /** |
1639 | * sb_wait_write - wait until all writers to given file system finish | |
1640 | * @sb: the super for which we wait | |
1641 | * @level: type of writers we wait for (normal vs page fault) | |
1642 | * | |
1643 | * This function waits until there are no writers of given type to given file | |
8129ed29 | 1644 | * system. |
5accdf82 JK |
1645 | */ |
1646 | static void sb_wait_write(struct super_block *sb, int level) | |
1647 | { | |
8129ed29 | 1648 | percpu_down_write(sb->s_writers.rw_sem + level-1); |
8129ed29 | 1649 | } |
5accdf82 | 1650 | |
f1a96220 ON |
1651 | /* |
1652 | * We are going to return to userspace and forget about these locks, the | |
1653 | * ownership goes to the caller of thaw_super() which does unlock(). | |
1654 | */ | |
1655 | static void lockdep_sb_freeze_release(struct super_block *sb) | |
1656 | { | |
1657 | int level; | |
1658 | ||
1659 | for (level = SB_FREEZE_LEVELS - 1; level >= 0; level--) | |
1660 | percpu_rwsem_release(sb->s_writers.rw_sem + level, 0, _THIS_IP_); | |
1661 | } | |
1662 | ||
1663 | /* | |
1664 | * Tell lockdep we are holding these locks before we call ->unfreeze_fs(sb). | |
1665 | */ | |
1666 | static void lockdep_sb_freeze_acquire(struct super_block *sb) | |
8129ed29 ON |
1667 | { |
1668 | int level; | |
5accdf82 | 1669 | |
8129ed29 ON |
1670 | for (level = 0; level < SB_FREEZE_LEVELS; ++level) |
1671 | percpu_rwsem_acquire(sb->s_writers.rw_sem + level, 0, _THIS_IP_); | |
f1a96220 ON |
1672 | } |
1673 | ||
2719c716 | 1674 | static void sb_freeze_unlock(struct super_block *sb, int level) |
f1a96220 | 1675 | { |
2719c716 | 1676 | for (level--; level >= 0; level--) |
8129ed29 | 1677 | percpu_up_write(sb->s_writers.rw_sem + level); |
5accdf82 JK |
1678 | } |
1679 | ||
18e9e510 | 1680 | /** |
7000d3c4 RD |
1681 | * freeze_super - lock the filesystem and force it into a consistent state |
1682 | * @sb: the super to lock | |
18e9e510 JB |
1683 | * |
1684 | * Syncs the super to make sure the filesystem is consistent and calls the fs's | |
1685 | * freeze_fs. Subsequent calls to this without first thawing the fs will return | |
1686 | * -EBUSY. | |
5accdf82 JK |
1687 | * |
1688 | * During this function, sb->s_writers.frozen goes through these values: | |
1689 | * | |
1690 | * SB_UNFROZEN: File system is normal, all writes progress as usual. | |
1691 | * | |
1692 | * SB_FREEZE_WRITE: The file system is in the process of being frozen. New | |
1693 | * writes should be blocked, though page faults are still allowed. We wait for | |
1694 | * all writes to complete and then proceed to the next stage. | |
1695 | * | |
1696 | * SB_FREEZE_PAGEFAULT: Freezing continues. Now also page faults are blocked | |
1697 | * but internal fs threads can still modify the filesystem (although they | |
1698 | * should not dirty new pages or inodes), writeback can run etc. After waiting | |
1699 | * for all running page faults we sync the filesystem which will clean all | |
1700 | * dirty pages and inodes (no new dirty pages or inodes can be created when | |
1701 | * sync is running). | |
1702 | * | |
1703 | * SB_FREEZE_FS: The file system is frozen. Now all internal sources of fs | |
1704 | * modification are blocked (e.g. XFS preallocation truncation on inode | |
1705 | * reclaim). This is usually implemented by blocking new transactions for | |
1706 | * filesystems that have them and need this additional guard. After all | |
1707 | * internal writers are finished we call ->freeze_fs() to finish filesystem | |
1708 | * freezing. Then we transition to SB_FREEZE_COMPLETE state. This state is | |
1709 | * mostly auxiliary for filesystems to verify they do not modify frozen fs. | |
1710 | * | |
1711 | * sb->s_writers.frozen is protected by sb->s_umount. | |
18e9e510 JB |
1712 | */ |
1713 | int freeze_super(struct super_block *sb) | |
1714 | { | |
1715 | int ret; | |
1716 | ||
1717 | atomic_inc(&sb->s_active); | |
0ed33598 | 1718 | super_lock_excl(sb); |
5accdf82 | 1719 | if (sb->s_writers.frozen != SB_UNFROZEN) { |
18e9e510 JB |
1720 | deactivate_locked_super(sb); |
1721 | return -EBUSY; | |
1722 | } | |
1723 | ||
e462ec50 | 1724 | if (!(sb->s_flags & SB_BORN)) { |
0ed33598 | 1725 | super_unlock_excl(sb); |
dabe0dc1 AV |
1726 | return 0; /* sic - it's "nothing to do" */ |
1727 | } | |
1728 | ||
bc98a42c | 1729 | if (sb_rdonly(sb)) { |
5accdf82 JK |
1730 | /* Nothing to do really... */ |
1731 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
0ed33598 | 1732 | super_unlock_excl(sb); |
18e9e510 JB |
1733 | return 0; |
1734 | } | |
1735 | ||
5accdf82 | 1736 | sb->s_writers.frozen = SB_FREEZE_WRITE; |
5accdf82 | 1737 | /* Release s_umount to preserve sb_start_write -> s_umount ordering */ |
0ed33598 | 1738 | super_unlock_excl(sb); |
5accdf82 | 1739 | sb_wait_write(sb, SB_FREEZE_WRITE); |
0ed33598 | 1740 | super_lock_excl(sb); |
5accdf82 JK |
1741 | |
1742 | /* Now we go and block page faults... */ | |
5accdf82 | 1743 | sb->s_writers.frozen = SB_FREEZE_PAGEFAULT; |
5accdf82 JK |
1744 | sb_wait_write(sb, SB_FREEZE_PAGEFAULT); |
1745 | ||
1746 | /* All writers are done so after syncing there won't be dirty data */ | |
2719c716 DW |
1747 | ret = sync_filesystem(sb); |
1748 | if (ret) { | |
1749 | sb->s_writers.frozen = SB_UNFROZEN; | |
1750 | sb_freeze_unlock(sb, SB_FREEZE_PAGEFAULT); | |
2719c716 DW |
1751 | deactivate_locked_super(sb); |
1752 | return ret; | |
1753 | } | |
18e9e510 | 1754 | |
5accdf82 JK |
1755 | /* Now wait for internal filesystem counter */ |
1756 | sb->s_writers.frozen = SB_FREEZE_FS; | |
5accdf82 | 1757 | sb_wait_write(sb, SB_FREEZE_FS); |
18e9e510 | 1758 | |
18e9e510 JB |
1759 | if (sb->s_op->freeze_fs) { |
1760 | ret = sb->s_op->freeze_fs(sb); | |
1761 | if (ret) { | |
1762 | printk(KERN_ERR | |
1763 | "VFS:Filesystem freeze failed\n"); | |
5accdf82 | 1764 | sb->s_writers.frozen = SB_UNFROZEN; |
2719c716 | 1765 | sb_freeze_unlock(sb, SB_FREEZE_FS); |
18e9e510 JB |
1766 | deactivate_locked_super(sb); |
1767 | return ret; | |
1768 | } | |
1769 | } | |
5accdf82 | 1770 | /* |
89f39af1 ON |
1771 | * For debugging purposes so that fs can warn if it sees write activity |
1772 | * when frozen is set to SB_FREEZE_COMPLETE, and for thaw_super(). | |
5accdf82 JK |
1773 | */ |
1774 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
f1a96220 | 1775 | lockdep_sb_freeze_release(sb); |
0ed33598 | 1776 | super_unlock_excl(sb); |
18e9e510 JB |
1777 | return 0; |
1778 | } | |
1779 | EXPORT_SYMBOL(freeze_super); | |
1780 | ||
08fdc8a0 | 1781 | static int thaw_super_locked(struct super_block *sb) |
18e9e510 JB |
1782 | { |
1783 | int error; | |
1784 | ||
89f39af1 | 1785 | if (sb->s_writers.frozen != SB_FREEZE_COMPLETE) { |
0ed33598 | 1786 | super_unlock_excl(sb); |
18e9e510 JB |
1787 | return -EINVAL; |
1788 | } | |
1789 | ||
bc98a42c | 1790 | if (sb_rdonly(sb)) { |
8129ed29 | 1791 | sb->s_writers.frozen = SB_UNFROZEN; |
18e9e510 | 1792 | goto out; |
8129ed29 | 1793 | } |
18e9e510 | 1794 | |
f1a96220 ON |
1795 | lockdep_sb_freeze_acquire(sb); |
1796 | ||
18e9e510 JB |
1797 | if (sb->s_op->unfreeze_fs) { |
1798 | error = sb->s_op->unfreeze_fs(sb); | |
1799 | if (error) { | |
1800 | printk(KERN_ERR | |
1801 | "VFS:Filesystem thaw failed\n"); | |
f1a96220 | 1802 | lockdep_sb_freeze_release(sb); |
0ed33598 | 1803 | super_unlock_excl(sb); |
18e9e510 JB |
1804 | return error; |
1805 | } | |
1806 | } | |
1807 | ||
5accdf82 | 1808 | sb->s_writers.frozen = SB_UNFROZEN; |
2719c716 | 1809 | sb_freeze_unlock(sb, SB_FREEZE_FS); |
8129ed29 | 1810 | out: |
18e9e510 | 1811 | deactivate_locked_super(sb); |
18e9e510 JB |
1812 | return 0; |
1813 | } | |
08fdc8a0 | 1814 | |
961f3c89 MCC |
1815 | /** |
1816 | * thaw_super -- unlock filesystem | |
1817 | * @sb: the super to thaw | |
1818 | * | |
1819 | * Unlocks the filesystem and marks it writeable again after freeze_super(). | |
1820 | */ | |
08fdc8a0 MG |
1821 | int thaw_super(struct super_block *sb) |
1822 | { | |
0ed33598 | 1823 | super_lock_excl(sb); |
08fdc8a0 MG |
1824 | return thaw_super_locked(sb); |
1825 | } | |
18e9e510 | 1826 | EXPORT_SYMBOL(thaw_super); |
439bc39b CH |
1827 | |
1828 | /* | |
1829 | * Create workqueue for deferred direct IO completions. We allocate the | |
1830 | * workqueue when it's first needed. This avoids creating workqueue for | |
1831 | * filesystems that don't need it and also allows us to create the workqueue | |
1832 | * late enough so the we can include s_id in the name of the workqueue. | |
1833 | */ | |
1834 | int sb_init_dio_done_wq(struct super_block *sb) | |
1835 | { | |
1836 | struct workqueue_struct *old; | |
1837 | struct workqueue_struct *wq = alloc_workqueue("dio/%s", | |
1838 | WQ_MEM_RECLAIM, 0, | |
1839 | sb->s_id); | |
1840 | if (!wq) | |
1841 | return -ENOMEM; | |
1842 | /* | |
1843 | * This has to be atomic as more DIOs can race to create the workqueue | |
1844 | */ | |
1845 | old = cmpxchg(&sb->s_dio_done_wq, NULL, wq); | |
1846 | /* Someone created workqueue before us? Free ours... */ | |
1847 | if (old) | |
1848 | destroy_workqueue(wq); | |
1849 | return 0; | |
1850 | } |