]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/super.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | * | |
6 | * super.c contains code to handle: - mount structures | |
7 | * - super-block tables | |
8 | * - filesystem drivers list | |
9 | * - mount system call | |
10 | * - umount system call | |
11 | * - ustat system call | |
12 | * | |
13 | * GK 2/5/95 - Changed to support mounting the root fs via NFS | |
14 | * | |
15 | * Added kerneld support: Jacques Gelinas and Bjorn Ekwall | |
16 | * Added change_root: Werner Almesberger & Hans Lermen, Feb '96 | |
17 | * Added options to /proc/mounts: | |
96de0e25 | 18 | * Torbjörn Lindh (torbjorn.lindh@gopta.se), April 14, 1996. |
1da177e4 LT |
19 | * Added devfs support: Richard Gooch <rgooch@atnf.csiro.au>, 13-JAN-1998 |
20 | * Heavily rewritten for 'one fs - one tree' dcache architecture. AV, Mar 2000 | |
21 | */ | |
22 | ||
630d9c47 | 23 | #include <linux/export.h> |
1da177e4 | 24 | #include <linux/slab.h> |
1da177e4 | 25 | #include <linux/blkdev.h> |
1da177e4 LT |
26 | #include <linux/mount.h> |
27 | #include <linux/security.h> | |
1da177e4 LT |
28 | #include <linux/writeback.h> /* for the emergency remount stuff */ |
29 | #include <linux/idr.h> | |
353ab6e9 | 30 | #include <linux/mutex.h> |
5477d0fa | 31 | #include <linux/backing-dev.h> |
ceb5bdc2 | 32 | #include <linux/rculist_bl.h> |
c515e1fd | 33 | #include <linux/cleancache.h> |
40401530 | 34 | #include <linux/fsnotify.h> |
5accdf82 | 35 | #include <linux/lockdep.h> |
b6500fda | 36 | #include <linux/user_namespace.h> |
6d59e7f5 | 37 | #include "internal.h" |
1da177e4 LT |
38 | |
39 | ||
15d0f5ea AV |
40 | static LIST_HEAD(super_blocks); |
41 | static DEFINE_SPINLOCK(sb_lock); | |
1da177e4 | 42 | |
5accdf82 JK |
43 | static char *sb_writers_name[SB_FREEZE_LEVELS] = { |
44 | "sb_writers", | |
45 | "sb_pagefaults", | |
46 | "sb_internal", | |
47 | }; | |
48 | ||
b0d40c92 DC |
49 | /* |
50 | * One thing we have to be careful of with a per-sb shrinker is that we don't | |
51 | * drop the last active reference to the superblock from within the shrinker. | |
52 | * If that happens we could trigger unregistering the shrinker from within the | |
53 | * shrinker path and that leads to deadlock on the shrinker_rwsem. Hence we | |
54 | * take a passive reference to the superblock to avoid this from occurring. | |
55 | */ | |
0a234c6d DC |
56 | static unsigned long super_cache_scan(struct shrinker *shrink, |
57 | struct shrink_control *sc) | |
b0d40c92 DC |
58 | { |
59 | struct super_block *sb; | |
0a234c6d DC |
60 | long fs_objects = 0; |
61 | long total_objects; | |
62 | long freed = 0; | |
63 | long dentries; | |
64 | long inodes; | |
b0d40c92 DC |
65 | |
66 | sb = container_of(shrink, struct super_block, s_shrink); | |
67 | ||
68 | /* | |
69 | * Deadlock avoidance. We may hold various FS locks, and we don't want | |
70 | * to recurse into the FS that called us in clear_inode() and friends.. | |
71 | */ | |
0a234c6d DC |
72 | if (!(sc->gfp_mask & __GFP_FS)) |
73 | return SHRINK_STOP; | |
b0d40c92 | 74 | |
eb6ef3df | 75 | if (!trylock_super(sb)) |
0a234c6d | 76 | return SHRINK_STOP; |
b0d40c92 | 77 | |
d0407903 | 78 | if (sb->s_op->nr_cached_objects) |
4101b624 | 79 | fs_objects = sb->s_op->nr_cached_objects(sb, sc); |
0e1fdafd | 80 | |
503c358c VD |
81 | inodes = list_lru_shrink_count(&sb->s_inode_lru, sc); |
82 | dentries = list_lru_shrink_count(&sb->s_dentry_lru, sc); | |
f6041567 | 83 | total_objects = dentries + inodes + fs_objects + 1; |
475d0db7 TH |
84 | if (!total_objects) |
85 | total_objects = 1; | |
0e1fdafd | 86 | |
0a234c6d | 87 | /* proportion the scan between the caches */ |
f6041567 | 88 | dentries = mult_frac(sc->nr_to_scan, dentries, total_objects); |
bc3b14cb | 89 | inodes = mult_frac(sc->nr_to_scan, inodes, total_objects); |
503c358c | 90 | fs_objects = mult_frac(sc->nr_to_scan, fs_objects, total_objects); |
b0d40c92 | 91 | |
0a234c6d DC |
92 | /* |
93 | * prune the dcache first as the icache is pinned by it, then | |
94 | * prune the icache, followed by the filesystem specific caches | |
49e7e7ff VD |
95 | * |
96 | * Ensure that we always scan at least one object - memcg kmem | |
97 | * accounting uses this to fully empty the caches. | |
0a234c6d | 98 | */ |
49e7e7ff | 99 | sc->nr_to_scan = dentries + 1; |
503c358c | 100 | freed = prune_dcache_sb(sb, sc); |
49e7e7ff | 101 | sc->nr_to_scan = inodes + 1; |
503c358c | 102 | freed += prune_icache_sb(sb, sc); |
0a234c6d DC |
103 | |
104 | if (fs_objects) { | |
49e7e7ff | 105 | sc->nr_to_scan = fs_objects + 1; |
4101b624 | 106 | freed += sb->s_op->free_cached_objects(sb, sc); |
b0d40c92 DC |
107 | } |
108 | ||
eb6ef3df | 109 | up_read(&sb->s_umount); |
0a234c6d DC |
110 | return freed; |
111 | } | |
112 | ||
113 | static unsigned long super_cache_count(struct shrinker *shrink, | |
114 | struct shrink_control *sc) | |
115 | { | |
116 | struct super_block *sb; | |
117 | long total_objects = 0; | |
118 | ||
119 | sb = container_of(shrink, struct super_block, s_shrink); | |
120 | ||
d23da150 | 121 | /* |
eb6ef3df | 122 | * Don't call trylock_super as it is a potential |
d23da150 TC |
123 | * scalability bottleneck. The counts could get updated |
124 | * between super_cache_count and super_cache_scan anyway. | |
125 | * Call to super_cache_count with shrinker_rwsem held | |
503c358c | 126 | * ensures the safety of call to list_lru_shrink_count() and |
d23da150 TC |
127 | * s_op->nr_cached_objects(). |
128 | */ | |
0a234c6d | 129 | if (sb->s_op && sb->s_op->nr_cached_objects) |
4101b624 | 130 | total_objects = sb->s_op->nr_cached_objects(sb, sc); |
0a234c6d | 131 | |
503c358c VD |
132 | total_objects += list_lru_shrink_count(&sb->s_dentry_lru, sc); |
133 | total_objects += list_lru_shrink_count(&sb->s_inode_lru, sc); | |
0a234c6d | 134 | |
55f841ce | 135 | total_objects = vfs_pressure_ratio(total_objects); |
0e1fdafd | 136 | return total_objects; |
b0d40c92 DC |
137 | } |
138 | ||
853b39a7 ON |
139 | static void destroy_super_work(struct work_struct *work) |
140 | { | |
141 | struct super_block *s = container_of(work, struct super_block, | |
142 | destroy_work); | |
143 | int i; | |
144 | ||
145 | for (i = 0; i < SB_FREEZE_LEVELS; i++) | |
8129ed29 | 146 | percpu_free_rwsem(&s->s_writers.rw_sem[i]); |
853b39a7 ON |
147 | kfree(s); |
148 | } | |
149 | ||
150 | static void destroy_super_rcu(struct rcu_head *head) | |
151 | { | |
152 | struct super_block *s = container_of(head, struct super_block, rcu); | |
153 | INIT_WORK(&s->destroy_work, destroy_super_work); | |
154 | schedule_work(&s->destroy_work); | |
155 | } | |
156 | ||
7eb5e882 AV |
157 | /** |
158 | * destroy_super - frees a superblock | |
159 | * @s: superblock to free | |
160 | * | |
161 | * Frees a superblock. | |
162 | */ | |
163 | static void destroy_super(struct super_block *s) | |
5accdf82 | 164 | { |
7eb5e882 AV |
165 | list_lru_destroy(&s->s_dentry_lru); |
166 | list_lru_destroy(&s->s_inode_lru); | |
7eb5e882 AV |
167 | security_sb_free(s); |
168 | WARN_ON(!list_empty(&s->s_mounts)); | |
b6500fda | 169 | put_user_ns(s->s_user_ns); |
7eb5e882 AV |
170 | kfree(s->s_subtype); |
171 | kfree(s->s_options); | |
853b39a7 | 172 | call_rcu(&s->rcu, destroy_super_rcu); |
5accdf82 JK |
173 | } |
174 | ||
1da177e4 LT |
175 | /** |
176 | * alloc_super - create new superblock | |
fe2bbc48 | 177 | * @type: filesystem type superblock should belong to |
9249e17f | 178 | * @flags: the mount flags |
b6500fda | 179 | * @user_ns: User namespace for the super_block |
1da177e4 LT |
180 | * |
181 | * Allocates and initializes a new &struct super_block. alloc_super() | |
182 | * returns a pointer new superblock or %NULL if allocation had failed. | |
183 | */ | |
b6500fda SF |
184 | static struct super_block *alloc_super(struct file_system_type *type, int flags, |
185 | struct user_namespace *user_ns) | |
1da177e4 | 186 | { |
11b0b5ab | 187 | struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER); |
b87221de | 188 | static const struct super_operations default_op; |
7eb5e882 AV |
189 | int i; |
190 | ||
191 | if (!s) | |
192 | return NULL; | |
1da177e4 | 193 | |
b5bd856a | 194 | INIT_LIST_HEAD(&s->s_mounts); |
b6500fda | 195 | s->s_user_ns = get_user_ns(user_ns); |
b5bd856a | 196 | |
7eb5e882 AV |
197 | if (security_sb_alloc(s)) |
198 | goto fail; | |
7b7a8665 | 199 | |
7eb5e882 | 200 | for (i = 0; i < SB_FREEZE_LEVELS; i++) { |
8129ed29 ON |
201 | if (__percpu_init_rwsem(&s->s_writers.rw_sem[i], |
202 | sb_writers_name[i], | |
203 | &type->s_writers_key[i])) | |
7eb5e882 | 204 | goto fail; |
1da177e4 | 205 | } |
7eb5e882 | 206 | init_waitqueue_head(&s->s_writers.wait_unfrozen); |
df0ce26c | 207 | s->s_bdi = &noop_backing_dev_info; |
7eb5e882 | 208 | s->s_flags = flags; |
7eb5e882 AV |
209 | INIT_HLIST_NODE(&s->s_instances); |
210 | INIT_HLIST_BL_HEAD(&s->s_anon); | |
e97fedb9 | 211 | mutex_init(&s->s_sync_lock); |
7eb5e882 | 212 | INIT_LIST_HEAD(&s->s_inodes); |
74278da9 | 213 | spin_lock_init(&s->s_inode_list_lock); |
7eb5e882 | 214 | |
2acb60a0 | 215 | if (list_lru_init_memcg(&s->s_dentry_lru)) |
7eb5e882 | 216 | goto fail; |
2acb60a0 | 217 | if (list_lru_init_memcg(&s->s_inode_lru)) |
7eb5e882 AV |
218 | goto fail; |
219 | ||
7eb5e882 AV |
220 | init_rwsem(&s->s_umount); |
221 | lockdep_set_class(&s->s_umount, &type->s_umount_key); | |
222 | /* | |
223 | * sget() can have s_umount recursion. | |
224 | * | |
225 | * When it cannot find a suitable sb, it allocates a new | |
226 | * one (this one), and tries again to find a suitable old | |
227 | * one. | |
228 | * | |
229 | * In case that succeeds, it will acquire the s_umount | |
230 | * lock of the old one. Since these are clearly distrinct | |
231 | * locks, and this object isn't exposed yet, there's no | |
232 | * risk of deadlocks. | |
233 | * | |
234 | * Annotate this by putting this lock in a different | |
235 | * subclass. | |
236 | */ | |
237 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); | |
238 | s->s_count = 1; | |
239 | atomic_set(&s->s_active, 1); | |
240 | mutex_init(&s->s_vfs_rename_mutex); | |
241 | lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key); | |
242 | mutex_init(&s->s_dquot.dqio_mutex); | |
243 | mutex_init(&s->s_dquot.dqonoff_mutex); | |
7eb5e882 AV |
244 | s->s_maxbytes = MAX_NON_LFS; |
245 | s->s_op = &default_op; | |
246 | s->s_time_gran = 1000000000; | |
3cb29d11 | 247 | s->cleancache_poolid = CLEANCACHE_NO_POOL; |
7eb5e882 AV |
248 | |
249 | s->s_shrink.seeks = DEFAULT_SEEKS; | |
250 | s->s_shrink.scan_objects = super_cache_scan; | |
251 | s->s_shrink.count_objects = super_cache_count; | |
252 | s->s_shrink.batch = 1024; | |
2acb60a0 | 253 | s->s_shrink.flags = SHRINKER_NUMA_AWARE | SHRINKER_MEMCG_AWARE; |
1da177e4 | 254 | return s; |
5ca302c8 | 255 | |
7eb5e882 AV |
256 | fail: |
257 | destroy_super(s); | |
258 | return NULL; | |
1da177e4 LT |
259 | } |
260 | ||
261 | /* Superblock refcounting */ | |
262 | ||
263 | /* | |
35cf7ba0 | 264 | * Drop a superblock's refcount. The caller must hold sb_lock. |
1da177e4 | 265 | */ |
f47ec3f2 | 266 | static void __put_super(struct super_block *sb) |
1da177e4 | 267 | { |
1da177e4 | 268 | if (!--sb->s_count) { |
551de6f3 | 269 | list_del_init(&sb->s_list); |
1da177e4 | 270 | destroy_super(sb); |
1da177e4 | 271 | } |
1da177e4 LT |
272 | } |
273 | ||
274 | /** | |
275 | * put_super - drop a temporary reference to superblock | |
276 | * @sb: superblock in question | |
277 | * | |
278 | * Drops a temporary reference, frees superblock if there's no | |
279 | * references left. | |
280 | */ | |
f47ec3f2 | 281 | static void put_super(struct super_block *sb) |
1da177e4 LT |
282 | { |
283 | spin_lock(&sb_lock); | |
284 | __put_super(sb); | |
285 | spin_unlock(&sb_lock); | |
286 | } | |
287 | ||
288 | ||
289 | /** | |
1712ac8f | 290 | * deactivate_locked_super - drop an active reference to superblock |
1da177e4 LT |
291 | * @s: superblock to deactivate |
292 | * | |
1712ac8f AV |
293 | * Drops an active reference to superblock, converting it into a temprory |
294 | * one if there is no other active references left. In that case we | |
1da177e4 LT |
295 | * tell fs driver to shut it down and drop the temporary reference we |
296 | * had just acquired. | |
1712ac8f AV |
297 | * |
298 | * Caller holds exclusive lock on superblock; that lock is released. | |
1da177e4 | 299 | */ |
1712ac8f | 300 | void deactivate_locked_super(struct super_block *s) |
1da177e4 LT |
301 | { |
302 | struct file_system_type *fs = s->s_type; | |
b20bd1a5 | 303 | if (atomic_dec_and_test(&s->s_active)) { |
3167760f | 304 | cleancache_invalidate_fs(s); |
b0d40c92 | 305 | unregister_shrinker(&s->s_shrink); |
28f2cd4f | 306 | fs->kill_sb(s); |
f5e1dd34 | 307 | |
c0a5b560 VD |
308 | /* |
309 | * Since list_lru_destroy() may sleep, we cannot call it from | |
310 | * put_super(), where we hold the sb_lock. Therefore we destroy | |
311 | * the lru lists right now. | |
312 | */ | |
313 | list_lru_destroy(&s->s_dentry_lru); | |
314 | list_lru_destroy(&s->s_inode_lru); | |
315 | ||
1da177e4 LT |
316 | put_filesystem(fs); |
317 | put_super(s); | |
1712ac8f AV |
318 | } else { |
319 | up_write(&s->s_umount); | |
1da177e4 LT |
320 | } |
321 | } | |
322 | ||
1712ac8f | 323 | EXPORT_SYMBOL(deactivate_locked_super); |
1da177e4 | 324 | |
74dbbdd7 | 325 | /** |
1712ac8f | 326 | * deactivate_super - drop an active reference to superblock |
74dbbdd7 AV |
327 | * @s: superblock to deactivate |
328 | * | |
1712ac8f AV |
329 | * Variant of deactivate_locked_super(), except that superblock is *not* |
330 | * locked by caller. If we are going to drop the final active reference, | |
331 | * lock will be acquired prior to that. | |
74dbbdd7 | 332 | */ |
1712ac8f | 333 | void deactivate_super(struct super_block *s) |
74dbbdd7 | 334 | { |
1712ac8f AV |
335 | if (!atomic_add_unless(&s->s_active, -1, 1)) { |
336 | down_write(&s->s_umount); | |
337 | deactivate_locked_super(s); | |
74dbbdd7 AV |
338 | } |
339 | } | |
340 | ||
1712ac8f | 341 | EXPORT_SYMBOL(deactivate_super); |
74dbbdd7 | 342 | |
1da177e4 LT |
343 | /** |
344 | * grab_super - acquire an active reference | |
345 | * @s: reference we are trying to make active | |
346 | * | |
347 | * Tries to acquire an active reference. grab_super() is used when we | |
348 | * had just found a superblock in super_blocks or fs_type->fs_supers | |
349 | * and want to turn it into a full-blown active reference. grab_super() | |
350 | * is called with sb_lock held and drops it. Returns 1 in case of | |
351 | * success, 0 if we had failed (superblock contents was already dead or | |
acfec9a5 AV |
352 | * dying when grab_super() had been called). Note that this is only |
353 | * called for superblocks not in rundown mode (== ones still on ->fs_supers | |
354 | * of their type), so increment of ->s_count is OK here. | |
1da177e4 | 355 | */ |
9c4dbee7 | 356 | static int grab_super(struct super_block *s) __releases(sb_lock) |
1da177e4 LT |
357 | { |
358 | s->s_count++; | |
359 | spin_unlock(&sb_lock); | |
360 | down_write(&s->s_umount); | |
acfec9a5 AV |
361 | if ((s->s_flags & MS_BORN) && atomic_inc_not_zero(&s->s_active)) { |
362 | put_super(s); | |
363 | return 1; | |
364 | } | |
1da177e4 LT |
365 | up_write(&s->s_umount); |
366 | put_super(s); | |
1da177e4 LT |
367 | return 0; |
368 | } | |
369 | ||
12ad3ab6 | 370 | /* |
eb6ef3df | 371 | * trylock_super - try to grab ->s_umount shared |
331cbdee | 372 | * @sb: reference we are trying to grab |
12ad3ab6 | 373 | * |
eb6ef3df | 374 | * Try to prevent fs shutdown. This is used in places where we |
12ad3ab6 | 375 | * cannot take an active reference but we need to ensure that the |
eb6ef3df KK |
376 | * filesystem is not shut down while we are working on it. It returns |
377 | * false if we cannot acquire s_umount or if we lose the race and | |
378 | * filesystem already got into shutdown, and returns true with the s_umount | |
379 | * lock held in read mode in case of success. On successful return, | |
380 | * the caller must drop the s_umount lock when done. | |
381 | * | |
382 | * Note that unlike get_super() et.al. this one does *not* bump ->s_count. | |
383 | * The reason why it's safe is that we are OK with doing trylock instead | |
384 | * of down_read(). There's a couple of places that are OK with that, but | |
385 | * it's very much not a general-purpose interface. | |
12ad3ab6 | 386 | */ |
eb6ef3df | 387 | bool trylock_super(struct super_block *sb) |
12ad3ab6 | 388 | { |
12ad3ab6 | 389 | if (down_read_trylock(&sb->s_umount)) { |
eb6ef3df KK |
390 | if (!hlist_unhashed(&sb->s_instances) && |
391 | sb->s_root && (sb->s_flags & MS_BORN)) | |
12ad3ab6 DC |
392 | return true; |
393 | up_read(&sb->s_umount); | |
394 | } | |
395 | ||
12ad3ab6 DC |
396 | return false; |
397 | } | |
398 | ||
1da177e4 LT |
399 | /** |
400 | * generic_shutdown_super - common helper for ->kill_sb() | |
401 | * @sb: superblock to kill | |
402 | * | |
403 | * generic_shutdown_super() does all fs-independent work on superblock | |
404 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects | |
405 | * that need destruction out of superblock, call generic_shutdown_super() | |
406 | * and release aforementioned objects. Note: dentries and inodes _are_ | |
407 | * taken care of and do not need specific handling. | |
c636ebdb DH |
408 | * |
409 | * Upon calling this function, the filesystem may no longer alter or | |
410 | * rearrange the set of dentries belonging to this super_block, nor may it | |
411 | * change the attachments of dentries to inodes. | |
1da177e4 LT |
412 | */ |
413 | void generic_shutdown_super(struct super_block *sb) | |
414 | { | |
ee9b6d61 | 415 | const struct super_operations *sop = sb->s_op; |
1da177e4 | 416 | |
c636ebdb DH |
417 | if (sb->s_root) { |
418 | shrink_dcache_for_umount(sb); | |
60b0680f | 419 | sync_filesystem(sb); |
1da177e4 | 420 | sb->s_flags &= ~MS_ACTIVE; |
efaee192 | 421 | |
74278da9 | 422 | fsnotify_unmount_inodes(sb); |
4875f773 | 423 | cgroup_writeback_umount(); |
63997e98 AV |
424 | |
425 | evict_inodes(sb); | |
1da177e4 | 426 | |
7b7a8665 CH |
427 | if (sb->s_dio_done_wq) { |
428 | destroy_workqueue(sb->s_dio_done_wq); | |
429 | sb->s_dio_done_wq = NULL; | |
430 | } | |
431 | ||
1da177e4 LT |
432 | if (sop->put_super) |
433 | sop->put_super(sb); | |
434 | ||
63997e98 | 435 | if (!list_empty(&sb->s_inodes)) { |
7b4fe29e DJ |
436 | printk("VFS: Busy inodes after unmount of %s. " |
437 | "Self-destruct in 5 seconds. Have a nice day...\n", | |
438 | sb->s_id); | |
1da177e4 | 439 | } |
1da177e4 LT |
440 | } |
441 | spin_lock(&sb_lock); | |
442 | /* should be initialized for __put_super_and_need_restart() */ | |
a5166169 | 443 | hlist_del_init(&sb->s_instances); |
1da177e4 LT |
444 | spin_unlock(&sb_lock); |
445 | up_write(&sb->s_umount); | |
446 | } | |
447 | ||
448 | EXPORT_SYMBOL(generic_shutdown_super); | |
449 | ||
450 | /** | |
b6500fda | 451 | * sget_userns - find or create a superblock |
1da177e4 LT |
452 | * @type: filesystem type superblock should belong to |
453 | * @test: comparison callback | |
454 | * @set: setup callback | |
9249e17f | 455 | * @flags: mount flags |
b6500fda | 456 | * @user_ns: User namespace for the super_block |
1da177e4 LT |
457 | * @data: argument to each of them |
458 | */ | |
b6500fda | 459 | struct super_block *sget_userns(struct file_system_type *type, |
1da177e4 LT |
460 | int (*test)(struct super_block *,void *), |
461 | int (*set)(struct super_block *,void *), | |
b6500fda | 462 | int flags, struct user_namespace *user_ns, |
1da177e4 LT |
463 | void *data) |
464 | { | |
465 | struct super_block *s = NULL; | |
d4730127 | 466 | struct super_block *old; |
1da177e4 LT |
467 | int err; |
468 | ||
469 | retry: | |
470 | spin_lock(&sb_lock); | |
d4730127 | 471 | if (test) { |
b67bfe0d | 472 | hlist_for_each_entry(old, &type->fs_supers, s_instances) { |
d4730127 MK |
473 | if (!test(old, data)) |
474 | continue; | |
b6500fda SF |
475 | if (user_ns != old->s_user_ns) { |
476 | spin_unlock(&sb_lock); | |
ed2833f8 PT |
477 | if (s) { |
478 | up_write(&s->s_umount); | |
479 | destroy_super(s); | |
480 | } | |
b6500fda SF |
481 | return ERR_PTR(-EBUSY); |
482 | } | |
d4730127 MK |
483 | if (!grab_super(old)) |
484 | goto retry; | |
a3cfbb53 LZ |
485 | if (s) { |
486 | up_write(&s->s_umount); | |
d4730127 | 487 | destroy_super(s); |
7a4dec53 | 488 | s = NULL; |
a3cfbb53 | 489 | } |
d4730127 MK |
490 | return old; |
491 | } | |
1da177e4 LT |
492 | } |
493 | if (!s) { | |
494 | spin_unlock(&sb_lock); | |
b6500fda | 495 | s = alloc_super(type, flags, user_ns); |
1da177e4 LT |
496 | if (!s) |
497 | return ERR_PTR(-ENOMEM); | |
498 | goto retry; | |
499 | } | |
500 | ||
501 | err = set(s, data); | |
502 | if (err) { | |
503 | spin_unlock(&sb_lock); | |
a3cfbb53 | 504 | up_write(&s->s_umount); |
1da177e4 LT |
505 | destroy_super(s); |
506 | return ERR_PTR(err); | |
507 | } | |
508 | s->s_type = type; | |
509 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); | |
510 | list_add_tail(&s->s_list, &super_blocks); | |
a5166169 | 511 | hlist_add_head(&s->s_instances, &type->fs_supers); |
1da177e4 LT |
512 | spin_unlock(&sb_lock); |
513 | get_filesystem(type); | |
b0d40c92 | 514 | register_shrinker(&s->s_shrink); |
1da177e4 LT |
515 | return s; |
516 | } | |
517 | ||
b6500fda SF |
518 | EXPORT_SYMBOL(sget_userns); |
519 | ||
520 | /** | |
521 | * sget - find or create a superblock | |
522 | * @type: filesystem type superblock should belong to | |
523 | * @test: comparison callback | |
524 | * @set: setup callback | |
525 | * @flags: mount flags | |
526 | * @data: argument to each of them | |
527 | */ | |
528 | struct super_block *sget(struct file_system_type *type, | |
529 | int (*test)(struct super_block *,void *), | |
530 | int (*set)(struct super_block *,void *), | |
531 | int flags, | |
532 | void *data) | |
533 | { | |
534 | return sget_userns(type, test, set, flags, current_user_ns(), data); | |
535 | } | |
536 | ||
1da177e4 LT |
537 | EXPORT_SYMBOL(sget); |
538 | ||
539 | void drop_super(struct super_block *sb) | |
540 | { | |
541 | up_read(&sb->s_umount); | |
542 | put_super(sb); | |
543 | } | |
544 | ||
545 | EXPORT_SYMBOL(drop_super); | |
546 | ||
01a05b33 AV |
547 | /** |
548 | * iterate_supers - call function for all active superblocks | |
549 | * @f: function to call | |
550 | * @arg: argument to pass to it | |
551 | * | |
552 | * Scans the superblock list and calls given function, passing it | |
553 | * locked superblock and given argument. | |
554 | */ | |
555 | void iterate_supers(void (*f)(struct super_block *, void *), void *arg) | |
556 | { | |
dca33252 | 557 | struct super_block *sb, *p = NULL; |
01a05b33 AV |
558 | |
559 | spin_lock(&sb_lock); | |
dca33252 | 560 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 561 | if (hlist_unhashed(&sb->s_instances)) |
01a05b33 AV |
562 | continue; |
563 | sb->s_count++; | |
564 | spin_unlock(&sb_lock); | |
565 | ||
566 | down_read(&sb->s_umount); | |
dabe0dc1 | 567 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
01a05b33 AV |
568 | f(sb, arg); |
569 | up_read(&sb->s_umount); | |
570 | ||
571 | spin_lock(&sb_lock); | |
dca33252 AV |
572 | if (p) |
573 | __put_super(p); | |
574 | p = sb; | |
01a05b33 | 575 | } |
dca33252 AV |
576 | if (p) |
577 | __put_super(p); | |
01a05b33 AV |
578 | spin_unlock(&sb_lock); |
579 | } | |
580 | ||
43e15cdb AV |
581 | /** |
582 | * iterate_supers_type - call function for superblocks of given type | |
583 | * @type: fs type | |
584 | * @f: function to call | |
585 | * @arg: argument to pass to it | |
586 | * | |
587 | * Scans the superblock list and calls given function, passing it | |
588 | * locked superblock and given argument. | |
589 | */ | |
590 | void iterate_supers_type(struct file_system_type *type, | |
591 | void (*f)(struct super_block *, void *), void *arg) | |
592 | { | |
593 | struct super_block *sb, *p = NULL; | |
594 | ||
595 | spin_lock(&sb_lock); | |
b67bfe0d | 596 | hlist_for_each_entry(sb, &type->fs_supers, s_instances) { |
43e15cdb AV |
597 | sb->s_count++; |
598 | spin_unlock(&sb_lock); | |
599 | ||
600 | down_read(&sb->s_umount); | |
dabe0dc1 | 601 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
43e15cdb AV |
602 | f(sb, arg); |
603 | up_read(&sb->s_umount); | |
604 | ||
605 | spin_lock(&sb_lock); | |
606 | if (p) | |
607 | __put_super(p); | |
608 | p = sb; | |
609 | } | |
610 | if (p) | |
611 | __put_super(p); | |
612 | spin_unlock(&sb_lock); | |
613 | } | |
614 | ||
615 | EXPORT_SYMBOL(iterate_supers_type); | |
616 | ||
1da177e4 LT |
617 | /** |
618 | * get_super - get the superblock of a device | |
619 | * @bdev: device to get the superblock for | |
620 | * | |
621 | * Scans the superblock list and finds the superblock of the file system | |
622 | * mounted on the device given. %NULL is returned if no match is found. | |
623 | */ | |
624 | ||
df40c01a | 625 | struct super_block *get_super(struct block_device *bdev) |
1da177e4 | 626 | { |
618f0636 KK |
627 | struct super_block *sb; |
628 | ||
1da177e4 LT |
629 | if (!bdev) |
630 | return NULL; | |
618f0636 | 631 | |
1da177e4 | 632 | spin_lock(&sb_lock); |
618f0636 KK |
633 | rescan: |
634 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 635 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 636 | continue; |
618f0636 KK |
637 | if (sb->s_bdev == bdev) { |
638 | sb->s_count++; | |
1da177e4 | 639 | spin_unlock(&sb_lock); |
618f0636 | 640 | down_read(&sb->s_umount); |
df40c01a | 641 | /* still alive? */ |
dabe0dc1 | 642 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
618f0636 KK |
643 | return sb; |
644 | up_read(&sb->s_umount); | |
df40c01a | 645 | /* nope, got unmounted */ |
618f0636 | 646 | spin_lock(&sb_lock); |
df40c01a AV |
647 | __put_super(sb); |
648 | goto rescan; | |
1da177e4 LT |
649 | } |
650 | } | |
651 | spin_unlock(&sb_lock); | |
652 | return NULL; | |
653 | } | |
654 | ||
655 | EXPORT_SYMBOL(get_super); | |
4504230a | 656 | |
6b6dc836 JK |
657 | /** |
658 | * get_super_thawed - get thawed superblock of a device | |
659 | * @bdev: device to get the superblock for | |
660 | * | |
661 | * Scans the superblock list and finds the superblock of the file system | |
662 | * mounted on the device. The superblock is returned once it is thawed | |
663 | * (or immediately if it was not frozen). %NULL is returned if no match | |
664 | * is found. | |
665 | */ | |
666 | struct super_block *get_super_thawed(struct block_device *bdev) | |
667 | { | |
668 | while (1) { | |
669 | struct super_block *s = get_super(bdev); | |
5accdf82 | 670 | if (!s || s->s_writers.frozen == SB_UNFROZEN) |
6b6dc836 JK |
671 | return s; |
672 | up_read(&s->s_umount); | |
5accdf82 JK |
673 | wait_event(s->s_writers.wait_unfrozen, |
674 | s->s_writers.frozen == SB_UNFROZEN); | |
6b6dc836 JK |
675 | put_super(s); |
676 | } | |
677 | } | |
678 | EXPORT_SYMBOL(get_super_thawed); | |
679 | ||
4504230a CH |
680 | /** |
681 | * get_active_super - get an active reference to the superblock of a device | |
682 | * @bdev: device to get the superblock for | |
683 | * | |
684 | * Scans the superblock list and finds the superblock of the file system | |
685 | * mounted on the device given. Returns the superblock with an active | |
d3f21473 | 686 | * reference or %NULL if none was found. |
4504230a CH |
687 | */ |
688 | struct super_block *get_active_super(struct block_device *bdev) | |
689 | { | |
690 | struct super_block *sb; | |
691 | ||
692 | if (!bdev) | |
693 | return NULL; | |
694 | ||
1494583d | 695 | restart: |
4504230a CH |
696 | spin_lock(&sb_lock); |
697 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 698 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 699 | continue; |
1494583d | 700 | if (sb->s_bdev == bdev) { |
acfec9a5 | 701 | if (!grab_super(sb)) |
1494583d | 702 | goto restart; |
acfec9a5 AV |
703 | up_write(&sb->s_umount); |
704 | return sb; | |
1494583d | 705 | } |
4504230a CH |
706 | } |
707 | spin_unlock(&sb_lock); | |
708 | return NULL; | |
709 | } | |
1da177e4 | 710 | |
df40c01a | 711 | struct super_block *user_get_super(dev_t dev) |
1da177e4 | 712 | { |
618f0636 | 713 | struct super_block *sb; |
1da177e4 | 714 | |
1da177e4 | 715 | spin_lock(&sb_lock); |
618f0636 KK |
716 | rescan: |
717 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 718 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 719 | continue; |
618f0636 KK |
720 | if (sb->s_dev == dev) { |
721 | sb->s_count++; | |
1da177e4 | 722 | spin_unlock(&sb_lock); |
618f0636 | 723 | down_read(&sb->s_umount); |
df40c01a | 724 | /* still alive? */ |
dabe0dc1 | 725 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
618f0636 KK |
726 | return sb; |
727 | up_read(&sb->s_umount); | |
df40c01a | 728 | /* nope, got unmounted */ |
618f0636 | 729 | spin_lock(&sb_lock); |
df40c01a AV |
730 | __put_super(sb); |
731 | goto rescan; | |
1da177e4 LT |
732 | } |
733 | } | |
734 | spin_unlock(&sb_lock); | |
735 | return NULL; | |
736 | } | |
737 | ||
1da177e4 LT |
738 | /** |
739 | * do_remount_sb - asks filesystem to change mount options. | |
740 | * @sb: superblock in question | |
741 | * @flags: numeric part of options | |
742 | * @data: the rest of options | |
743 | * @force: whether or not to force the change | |
744 | * | |
745 | * Alters the mount options of a mounted file system. | |
746 | */ | |
747 | int do_remount_sb(struct super_block *sb, int flags, void *data, int force) | |
748 | { | |
749 | int retval; | |
c79d967d | 750 | int remount_ro; |
4504230a | 751 | |
5accdf82 | 752 | if (sb->s_writers.frozen != SB_UNFROZEN) |
4504230a CH |
753 | return -EBUSY; |
754 | ||
9361401e | 755 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
756 | if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev)) |
757 | return -EACCES; | |
9361401e | 758 | #endif |
4504230a | 759 | |
d208bbdd | 760 | remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY); |
d208bbdd | 761 | |
0aec09d0 | 762 | if (remount_ro) { |
fdab684d | 763 | if (!hlist_empty(&sb->s_pins)) { |
0aec09d0 | 764 | up_write(&sb->s_umount); |
fdab684d | 765 | group_pin_kill(&sb->s_pins); |
0aec09d0 AV |
766 | down_write(&sb->s_umount); |
767 | if (!sb->s_root) | |
768 | return 0; | |
769 | if (sb->s_writers.frozen != SB_UNFROZEN) | |
770 | return -EBUSY; | |
771 | remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY); | |
772 | } | |
773 | } | |
774 | shrink_dcache_sb(sb); | |
775 | ||
1da177e4 LT |
776 | /* If we are remounting RDONLY and current sb is read/write, |
777 | make sure there are no rw files opened */ | |
d208bbdd | 778 | if (remount_ro) { |
4ed5e82f | 779 | if (force) { |
eee5cc27 AV |
780 | sb->s_readonly_remount = 1; |
781 | smp_wmb(); | |
4ed5e82f MS |
782 | } else { |
783 | retval = sb_prepare_remount_readonly(sb); | |
784 | if (retval) | |
785 | return retval; | |
4ed5e82f | 786 | } |
1da177e4 LT |
787 | } |
788 | ||
789 | if (sb->s_op->remount_fs) { | |
1da177e4 | 790 | retval = sb->s_op->remount_fs(sb, &flags, data); |
2833eb2b MS |
791 | if (retval) { |
792 | if (!force) | |
4ed5e82f | 793 | goto cancel_readonly; |
2833eb2b MS |
794 | /* If forced remount, go ahead despite any errors */ |
795 | WARN(1, "forced remount of a %s fs returned %i\n", | |
796 | sb->s_type->name, retval); | |
797 | } | |
1da177e4 LT |
798 | } |
799 | sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK); | |
4ed5e82f MS |
800 | /* Needs to be ordered wrt mnt_is_readonly() */ |
801 | smp_wmb(); | |
802 | sb->s_readonly_remount = 0; | |
c79d967d | 803 | |
d208bbdd NP |
804 | /* |
805 | * Some filesystems modify their metadata via some other path than the | |
806 | * bdev buffer cache (eg. use a private mapping, or directories in | |
807 | * pagecache, etc). Also file data modifications go via their own | |
808 | * mappings. So If we try to mount readonly then copy the filesystem | |
809 | * from bdev, we could get stale data, so invalidate it to give a best | |
810 | * effort at coherency. | |
811 | */ | |
812 | if (remount_ro && sb->s_bdev) | |
813 | invalidate_bdev(sb->s_bdev); | |
1da177e4 | 814 | return 0; |
4ed5e82f MS |
815 | |
816 | cancel_readonly: | |
817 | sb->s_readonly_remount = 0; | |
818 | return retval; | |
1da177e4 LT |
819 | } |
820 | ||
a2a9537a | 821 | static void do_emergency_remount(struct work_struct *work) |
1da177e4 | 822 | { |
dca33252 | 823 | struct super_block *sb, *p = NULL; |
1da177e4 LT |
824 | |
825 | spin_lock(&sb_lock); | |
dca33252 | 826 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 827 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 828 | continue; |
1da177e4 LT |
829 | sb->s_count++; |
830 | spin_unlock(&sb_lock); | |
443b94ba | 831 | down_write(&sb->s_umount); |
dabe0dc1 AV |
832 | if (sb->s_root && sb->s_bdev && (sb->s_flags & MS_BORN) && |
833 | !(sb->s_flags & MS_RDONLY)) { | |
1da177e4 | 834 | /* |
1da177e4 LT |
835 | * What lock protects sb->s_flags?? |
836 | */ | |
1da177e4 | 837 | do_remount_sb(sb, MS_RDONLY, NULL, 1); |
1da177e4 | 838 | } |
443b94ba | 839 | up_write(&sb->s_umount); |
1da177e4 | 840 | spin_lock(&sb_lock); |
dca33252 AV |
841 | if (p) |
842 | __put_super(p); | |
843 | p = sb; | |
1da177e4 | 844 | } |
dca33252 AV |
845 | if (p) |
846 | __put_super(p); | |
1da177e4 | 847 | spin_unlock(&sb_lock); |
a2a9537a | 848 | kfree(work); |
1da177e4 LT |
849 | printk("Emergency Remount complete\n"); |
850 | } | |
851 | ||
852 | void emergency_remount(void) | |
853 | { | |
a2a9537a JA |
854 | struct work_struct *work; |
855 | ||
856 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | |
857 | if (work) { | |
858 | INIT_WORK(work, do_emergency_remount); | |
859 | schedule_work(work); | |
860 | } | |
1da177e4 LT |
861 | } |
862 | ||
863 | /* | |
864 | * Unnamed block devices are dummy devices used by virtual | |
865 | * filesystems which don't use real block-devices. -- jrs | |
866 | */ | |
867 | ||
ad76cbc6 | 868 | static DEFINE_IDA(unnamed_dev_ida); |
1da177e4 | 869 | static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */ |
a2a4dc49 TB |
870 | /* Many userspace utilities consider an FSID of 0 invalid. |
871 | * Always return at least 1 from get_anon_bdev. | |
872 | */ | |
873 | static int unnamed_dev_start = 1; | |
1da177e4 | 874 | |
0ee5dc67 | 875 | int get_anon_bdev(dev_t *p) |
1da177e4 LT |
876 | { |
877 | int dev; | |
878 | int error; | |
879 | ||
880 | retry: | |
ad76cbc6 | 881 | if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0) |
1da177e4 LT |
882 | return -ENOMEM; |
883 | spin_lock(&unnamed_dev_lock); | |
c63e09ec | 884 | error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev); |
f21f6220 AV |
885 | if (!error) |
886 | unnamed_dev_start = dev + 1; | |
1da177e4 LT |
887 | spin_unlock(&unnamed_dev_lock); |
888 | if (error == -EAGAIN) | |
889 | /* We raced and lost with another CPU. */ | |
890 | goto retry; | |
891 | else if (error) | |
892 | return -EAGAIN; | |
893 | ||
1af95de6 | 894 | if (dev >= (1 << MINORBITS)) { |
1da177e4 | 895 | spin_lock(&unnamed_dev_lock); |
ad76cbc6 | 896 | ida_remove(&unnamed_dev_ida, dev); |
f21f6220 AV |
897 | if (unnamed_dev_start > dev) |
898 | unnamed_dev_start = dev; | |
1da177e4 LT |
899 | spin_unlock(&unnamed_dev_lock); |
900 | return -EMFILE; | |
901 | } | |
0ee5dc67 | 902 | *p = MKDEV(0, dev & MINORMASK); |
1da177e4 LT |
903 | return 0; |
904 | } | |
0ee5dc67 | 905 | EXPORT_SYMBOL(get_anon_bdev); |
1da177e4 | 906 | |
0ee5dc67 | 907 | void free_anon_bdev(dev_t dev) |
1da177e4 | 908 | { |
0ee5dc67 | 909 | int slot = MINOR(dev); |
1da177e4 | 910 | spin_lock(&unnamed_dev_lock); |
ad76cbc6 | 911 | ida_remove(&unnamed_dev_ida, slot); |
c63e09ec AV |
912 | if (slot < unnamed_dev_start) |
913 | unnamed_dev_start = slot; | |
1da177e4 LT |
914 | spin_unlock(&unnamed_dev_lock); |
915 | } | |
0ee5dc67 AV |
916 | EXPORT_SYMBOL(free_anon_bdev); |
917 | ||
918 | int set_anon_super(struct super_block *s, void *data) | |
919 | { | |
df0ce26c | 920 | return get_anon_bdev(&s->s_dev); |
0ee5dc67 AV |
921 | } |
922 | ||
923 | EXPORT_SYMBOL(set_anon_super); | |
924 | ||
925 | void kill_anon_super(struct super_block *sb) | |
926 | { | |
927 | dev_t dev = sb->s_dev; | |
928 | generic_shutdown_super(sb); | |
929 | free_anon_bdev(dev); | |
930 | } | |
1da177e4 LT |
931 | |
932 | EXPORT_SYMBOL(kill_anon_super); | |
933 | ||
1da177e4 LT |
934 | void kill_litter_super(struct super_block *sb) |
935 | { | |
936 | if (sb->s_root) | |
937 | d_genocide(sb->s_root); | |
938 | kill_anon_super(sb); | |
939 | } | |
940 | ||
941 | EXPORT_SYMBOL(kill_litter_super); | |
942 | ||
909e6d94 SH |
943 | static int ns_test_super(struct super_block *sb, void *data) |
944 | { | |
945 | return sb->s_fs_info == data; | |
946 | } | |
947 | ||
948 | static int ns_set_super(struct super_block *sb, void *data) | |
949 | { | |
950 | sb->s_fs_info = data; | |
951 | return set_anon_super(sb, NULL); | |
952 | } | |
953 | ||
ceefda69 AV |
954 | struct dentry *mount_ns(struct file_system_type *fs_type, int flags, |
955 | void *data, int (*fill_super)(struct super_block *, void *, int)) | |
909e6d94 SH |
956 | { |
957 | struct super_block *sb; | |
958 | ||
9249e17f | 959 | sb = sget(fs_type, ns_test_super, ns_set_super, flags, data); |
909e6d94 | 960 | if (IS_ERR(sb)) |
ceefda69 | 961 | return ERR_CAST(sb); |
909e6d94 SH |
962 | |
963 | if (!sb->s_root) { | |
964 | int err; | |
909e6d94 SH |
965 | err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0); |
966 | if (err) { | |
74dbbdd7 | 967 | deactivate_locked_super(sb); |
ceefda69 | 968 | return ERR_PTR(err); |
909e6d94 SH |
969 | } |
970 | ||
971 | sb->s_flags |= MS_ACTIVE; | |
972 | } | |
973 | ||
ceefda69 | 974 | return dget(sb->s_root); |
909e6d94 SH |
975 | } |
976 | ||
ceefda69 | 977 | EXPORT_SYMBOL(mount_ns); |
909e6d94 | 978 | |
9361401e | 979 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
980 | static int set_bdev_super(struct super_block *s, void *data) |
981 | { | |
982 | s->s_bdev = data; | |
983 | s->s_dev = s->s_bdev->bd_dev; | |
32a88aa1 JA |
984 | |
985 | /* | |
986 | * We set the bdi here to the queue backing, file systems can | |
987 | * overwrite this in ->fill_super() | |
988 | */ | |
989 | s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info; | |
1da177e4 LT |
990 | return 0; |
991 | } | |
992 | ||
993 | static int test_bdev_super(struct super_block *s, void *data) | |
994 | { | |
995 | return (void *)s->s_bdev == data; | |
996 | } | |
997 | ||
152a0836 | 998 | struct dentry *mount_bdev(struct file_system_type *fs_type, |
1da177e4 | 999 | int flags, const char *dev_name, void *data, |
152a0836 | 1000 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1001 | { |
1002 | struct block_device *bdev; | |
1003 | struct super_block *s; | |
d4d77629 | 1004 | fmode_t mode = FMODE_READ | FMODE_EXCL; |
1da177e4 LT |
1005 | int error = 0; |
1006 | ||
30c40d2c AV |
1007 | if (!(flags & MS_RDONLY)) |
1008 | mode |= FMODE_WRITE; | |
1009 | ||
d4d77629 | 1010 | bdev = blkdev_get_by_path(dev_name, mode, fs_type); |
1da177e4 | 1011 | if (IS_ERR(bdev)) |
152a0836 | 1012 | return ERR_CAST(bdev); |
1da177e4 | 1013 | |
4de0a612 SF |
1014 | if (current_user_ns() != &init_user_ns) { |
1015 | /* | |
1016 | * For userns mounts, disallow mounting if bdev is open for | |
1017 | * writing | |
1018 | */ | |
1019 | if (!atomic_dec_unless_positive(&bdev->bd_inode->i_writecount)) { | |
1020 | error = -EBUSY; | |
1021 | goto error_bdev; | |
1022 | } | |
1023 | if (bdev->bd_contains != bdev && | |
1024 | !atomic_dec_unless_positive(&bdev->bd_contains->bd_inode->i_writecount)) { | |
1025 | atomic_inc(&bdev->bd_inode->i_writecount); | |
1026 | error = -EBUSY; | |
1027 | goto error_bdev; | |
1028 | } | |
1029 | } | |
1030 | ||
1da177e4 LT |
1031 | /* |
1032 | * once the super is inserted into the list by sget, s_umount | |
1033 | * will protect the lockfs code from trying to start a snapshot | |
1034 | * while we are mounting | |
1035 | */ | |
4fadd7bb CH |
1036 | mutex_lock(&bdev->bd_fsfreeze_mutex); |
1037 | if (bdev->bd_fsfreeze_count > 0) { | |
1038 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
1039 | error = -EBUSY; | |
4de0a612 | 1040 | goto error_inc; |
4fadd7bb | 1041 | } |
9249e17f DH |
1042 | s = sget(fs_type, test_bdev_super, set_bdev_super, flags | MS_NOSEC, |
1043 | bdev); | |
4fadd7bb | 1044 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
1da177e4 | 1045 | if (IS_ERR(s)) |
454e2398 | 1046 | goto error_s; |
1da177e4 LT |
1047 | |
1048 | if (s->s_root) { | |
1049 | if ((flags ^ s->s_flags) & MS_RDONLY) { | |
74dbbdd7 | 1050 | deactivate_locked_super(s); |
454e2398 | 1051 | error = -EBUSY; |
4de0a612 | 1052 | goto error_inc; |
1da177e4 | 1053 | } |
454e2398 | 1054 | |
4f331f01 TH |
1055 | /* |
1056 | * s_umount nests inside bd_mutex during | |
e525fd89 TH |
1057 | * __invalidate_device(). blkdev_put() acquires |
1058 | * bd_mutex and can't be called under s_umount. Drop | |
1059 | * s_umount temporarily. This is safe as we're | |
1060 | * holding an active reference. | |
4f331f01 TH |
1061 | */ |
1062 | up_write(&s->s_umount); | |
d4d77629 | 1063 | blkdev_put(bdev, mode); |
4f331f01 | 1064 | down_write(&s->s_umount); |
1da177e4 LT |
1065 | } else { |
1066 | char b[BDEVNAME_SIZE]; | |
1067 | ||
30c40d2c | 1068 | s->s_mode = mode; |
1da177e4 | 1069 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); |
e78c9a00 | 1070 | sb_set_blocksize(s, block_size(bdev)); |
9b04c997 | 1071 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1072 | if (error) { |
74dbbdd7 | 1073 | deactivate_locked_super(s); |
454e2398 | 1074 | goto error; |
fa675765 | 1075 | } |
454e2398 DH |
1076 | |
1077 | s->s_flags |= MS_ACTIVE; | |
87d8fe1e | 1078 | bdev->bd_super = s; |
1da177e4 LT |
1079 | } |
1080 | ||
152a0836 | 1081 | return dget(s->s_root); |
1da177e4 | 1082 | |
454e2398 DH |
1083 | error_s: |
1084 | error = PTR_ERR(s); | |
4de0a612 SF |
1085 | error_inc: |
1086 | if (current_user_ns() != &init_user_ns) { | |
1087 | atomic_inc(&bdev->bd_inode->i_writecount); | |
1088 | if (bdev->bd_contains != bdev) | |
1089 | atomic_inc(&bdev->bd_contains->bd_inode->i_writecount); | |
1090 | } | |
454e2398 | 1091 | error_bdev: |
d4d77629 | 1092 | blkdev_put(bdev, mode); |
454e2398 | 1093 | error: |
152a0836 AV |
1094 | return ERR_PTR(error); |
1095 | } | |
1096 | EXPORT_SYMBOL(mount_bdev); | |
1097 | ||
1da177e4 LT |
1098 | void kill_block_super(struct super_block *sb) |
1099 | { | |
1100 | struct block_device *bdev = sb->s_bdev; | |
30c40d2c | 1101 | fmode_t mode = sb->s_mode; |
1da177e4 | 1102 | |
ddbaaf30 | 1103 | bdev->bd_super = NULL; |
1da177e4 LT |
1104 | generic_shutdown_super(sb); |
1105 | sync_blockdev(bdev); | |
d4d77629 | 1106 | WARN_ON_ONCE(!(mode & FMODE_EXCL)); |
4de0a612 SF |
1107 | if (sb->s_user_ns != &init_user_ns) { |
1108 | atomic_inc(&bdev->bd_inode->i_writecount); | |
1109 | if (bdev->bd_contains != bdev) | |
1110 | atomic_inc(&bdev->bd_contains->bd_inode->i_writecount); | |
1111 | } | |
e525fd89 | 1112 | blkdev_put(bdev, mode | FMODE_EXCL); |
1da177e4 LT |
1113 | } |
1114 | ||
1115 | EXPORT_SYMBOL(kill_block_super); | |
9361401e | 1116 | #endif |
1da177e4 | 1117 | |
3c26ff6e | 1118 | struct dentry *mount_nodev(struct file_system_type *fs_type, |
1da177e4 | 1119 | int flags, void *data, |
3c26ff6e | 1120 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1121 | { |
1122 | int error; | |
9249e17f | 1123 | struct super_block *s = sget(fs_type, NULL, set_anon_super, flags, NULL); |
1da177e4 LT |
1124 | |
1125 | if (IS_ERR(s)) | |
3c26ff6e | 1126 | return ERR_CAST(s); |
1da177e4 | 1127 | |
9b04c997 | 1128 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1129 | if (error) { |
74dbbdd7 | 1130 | deactivate_locked_super(s); |
3c26ff6e | 1131 | return ERR_PTR(error); |
1da177e4 LT |
1132 | } |
1133 | s->s_flags |= MS_ACTIVE; | |
3c26ff6e | 1134 | return dget(s->s_root); |
1da177e4 | 1135 | } |
3c26ff6e AV |
1136 | EXPORT_SYMBOL(mount_nodev); |
1137 | ||
1da177e4 LT |
1138 | static int compare_single(struct super_block *s, void *p) |
1139 | { | |
1140 | return 1; | |
1141 | } | |
1142 | ||
fc14f2fe | 1143 | struct dentry *mount_single(struct file_system_type *fs_type, |
1da177e4 | 1144 | int flags, void *data, |
fc14f2fe | 1145 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1146 | { |
1147 | struct super_block *s; | |
1148 | int error; | |
1149 | ||
9249e17f | 1150 | s = sget(fs_type, compare_single, set_anon_super, flags, NULL); |
1da177e4 | 1151 | if (IS_ERR(s)) |
fc14f2fe | 1152 | return ERR_CAST(s); |
1da177e4 | 1153 | if (!s->s_root) { |
9b04c997 | 1154 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1155 | if (error) { |
74dbbdd7 | 1156 | deactivate_locked_super(s); |
fc14f2fe | 1157 | return ERR_PTR(error); |
1da177e4 LT |
1158 | } |
1159 | s->s_flags |= MS_ACTIVE; | |
9329d1be KS |
1160 | } else { |
1161 | do_remount_sb(s, flags, data, 0); | |
1da177e4 | 1162 | } |
fc14f2fe AV |
1163 | return dget(s->s_root); |
1164 | } | |
1165 | EXPORT_SYMBOL(mount_single); | |
1166 | ||
9d412a43 AV |
1167 | struct dentry * |
1168 | mount_fs(struct file_system_type *type, int flags, const char *name, void *data) | |
1da177e4 | 1169 | { |
c96e41e9 | 1170 | struct dentry *root; |
9d412a43 | 1171 | struct super_block *sb; |
1da177e4 | 1172 | char *secdata = NULL; |
9d412a43 | 1173 | int error = -ENOMEM; |
8089352a | 1174 | |
e0007529 | 1175 | if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) { |
1da177e4 | 1176 | secdata = alloc_secdata(); |
454e2398 | 1177 | if (!secdata) |
9d412a43 | 1178 | goto out; |
1da177e4 | 1179 | |
e0007529 | 1180 | error = security_sb_copy_data(data, secdata); |
454e2398 | 1181 | if (error) |
1da177e4 | 1182 | goto out_free_secdata; |
1da177e4 LT |
1183 | } |
1184 | ||
1a102ff9 AV |
1185 | root = type->mount(type, flags, name, data); |
1186 | if (IS_ERR(root)) { | |
1187 | error = PTR_ERR(root); | |
1188 | goto out_free_secdata; | |
c96e41e9 | 1189 | } |
9d412a43 AV |
1190 | sb = root->d_sb; |
1191 | BUG_ON(!sb); | |
1192 | WARN_ON(!sb->s_bdi); | |
1193 | sb->s_flags |= MS_BORN; | |
454e2398 | 1194 | |
9d412a43 | 1195 | error = security_sb_kern_mount(sb, flags, secdata); |
5129a469 JE |
1196 | if (error) |
1197 | goto out_sb; | |
454e2398 | 1198 | |
42cb56ae JL |
1199 | /* |
1200 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE | |
1201 | * but s_maxbytes was an unsigned long long for many releases. Throw | |
1202 | * this warning for a little while to try and catch filesystems that | |
4358b567 | 1203 | * violate this rule. |
42cb56ae | 1204 | */ |
9d412a43 AV |
1205 | WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " |
1206 | "negative value (%lld)\n", type->name, sb->s_maxbytes); | |
42cb56ae | 1207 | |
9d412a43 | 1208 | up_write(&sb->s_umount); |
8680e22f | 1209 | free_secdata(secdata); |
9d412a43 | 1210 | return root; |
1da177e4 | 1211 | out_sb: |
9d412a43 AV |
1212 | dput(root); |
1213 | deactivate_locked_super(sb); | |
1da177e4 LT |
1214 | out_free_secdata: |
1215 | free_secdata(secdata); | |
1da177e4 | 1216 | out: |
454e2398 | 1217 | return ERR_PTR(error); |
1da177e4 LT |
1218 | } |
1219 | ||
5accdf82 JK |
1220 | /* |
1221 | * This is an internal function, please use sb_end_{write,pagefault,intwrite} | |
1222 | * instead. | |
1223 | */ | |
1224 | void __sb_end_write(struct super_block *sb, int level) | |
1225 | { | |
8129ed29 | 1226 | percpu_up_read(sb->s_writers.rw_sem + level-1); |
5accdf82 JK |
1227 | } |
1228 | EXPORT_SYMBOL(__sb_end_write); | |
1229 | ||
f4b554af ON |
1230 | /* |
1231 | * This is an internal function, please use sb_start_{write,pagefault,intwrite} | |
1232 | * instead. | |
1233 | */ | |
1234 | int __sb_start_write(struct super_block *sb, int level, bool wait) | |
1235 | { | |
1236 | bool force_trylock = false; | |
8129ed29 | 1237 | int ret = 1; |
f4b554af ON |
1238 | |
1239 | #ifdef CONFIG_LOCKDEP | |
1240 | /* | |
1241 | * We want lockdep to tell us about possible deadlocks with freezing | |
1242 | * but it's it bit tricky to properly instrument it. Getting a freeze | |
1243 | * protection works as getting a read lock but there are subtle | |
1244 | * problems. XFS for example gets freeze protection on internal level | |
1245 | * twice in some cases, which is OK only because we already hold a | |
1246 | * freeze protection also on higher level. Due to these cases we have | |
1247 | * to use wait == F (trylock mode) which must not fail. | |
1248 | */ | |
1249 | if (wait) { | |
1250 | int i; | |
1251 | ||
1252 | for (i = 0; i < level - 1; i++) | |
8129ed29 | 1253 | if (percpu_rwsem_is_held(sb->s_writers.rw_sem + i)) { |
f4b554af ON |
1254 | force_trylock = true; |
1255 | break; | |
1256 | } | |
1257 | } | |
1258 | #endif | |
8129ed29 ON |
1259 | if (wait && !force_trylock) |
1260 | percpu_down_read(sb->s_writers.rw_sem + level-1); | |
1261 | else | |
1262 | ret = percpu_down_read_trylock(sb->s_writers.rw_sem + level-1); | |
1263 | ||
f4b554af ON |
1264 | WARN_ON(force_trylock & !ret); |
1265 | return ret; | |
1266 | } | |
5accdf82 JK |
1267 | EXPORT_SYMBOL(__sb_start_write); |
1268 | ||
1269 | /** | |
1270 | * sb_wait_write - wait until all writers to given file system finish | |
1271 | * @sb: the super for which we wait | |
1272 | * @level: type of writers we wait for (normal vs page fault) | |
1273 | * | |
1274 | * This function waits until there are no writers of given type to given file | |
8129ed29 | 1275 | * system. |
5accdf82 JK |
1276 | */ |
1277 | static void sb_wait_write(struct super_block *sb, int level) | |
1278 | { | |
8129ed29 | 1279 | percpu_down_write(sb->s_writers.rw_sem + level-1); |
5accdf82 | 1280 | /* |
0e28e01f ON |
1281 | * We are going to return to userspace and forget about this lock, the |
1282 | * ownership goes to the caller of thaw_super() which does unlock. | |
1283 | * | |
1284 | * FIXME: we should do this before return from freeze_super() after we | |
1285 | * called sync_filesystem(sb) and s_op->freeze_fs(sb), and thaw_super() | |
1286 | * should re-acquire these locks before s_op->unfreeze_fs(sb). However | |
1287 | * this leads to lockdep false-positives, so currently we do the early | |
1288 | * release right after acquire. | |
5accdf82 | 1289 | */ |
8129ed29 ON |
1290 | percpu_rwsem_release(sb->s_writers.rw_sem + level-1, 0, _THIS_IP_); |
1291 | } | |
5accdf82 | 1292 | |
8129ed29 ON |
1293 | static void sb_freeze_unlock(struct super_block *sb) |
1294 | { | |
1295 | int level; | |
5accdf82 | 1296 | |
8129ed29 ON |
1297 | for (level = 0; level < SB_FREEZE_LEVELS; ++level) |
1298 | percpu_rwsem_acquire(sb->s_writers.rw_sem + level, 0, _THIS_IP_); | |
5accdf82 | 1299 | |
8129ed29 ON |
1300 | for (level = SB_FREEZE_LEVELS - 1; level >= 0; level--) |
1301 | percpu_up_write(sb->s_writers.rw_sem + level); | |
5accdf82 JK |
1302 | } |
1303 | ||
18e9e510 | 1304 | /** |
7000d3c4 RD |
1305 | * freeze_super - lock the filesystem and force it into a consistent state |
1306 | * @sb: the super to lock | |
18e9e510 JB |
1307 | * |
1308 | * Syncs the super to make sure the filesystem is consistent and calls the fs's | |
1309 | * freeze_fs. Subsequent calls to this without first thawing the fs will return | |
1310 | * -EBUSY. | |
5accdf82 JK |
1311 | * |
1312 | * During this function, sb->s_writers.frozen goes through these values: | |
1313 | * | |
1314 | * SB_UNFROZEN: File system is normal, all writes progress as usual. | |
1315 | * | |
1316 | * SB_FREEZE_WRITE: The file system is in the process of being frozen. New | |
1317 | * writes should be blocked, though page faults are still allowed. We wait for | |
1318 | * all writes to complete and then proceed to the next stage. | |
1319 | * | |
1320 | * SB_FREEZE_PAGEFAULT: Freezing continues. Now also page faults are blocked | |
1321 | * but internal fs threads can still modify the filesystem (although they | |
1322 | * should not dirty new pages or inodes), writeback can run etc. After waiting | |
1323 | * for all running page faults we sync the filesystem which will clean all | |
1324 | * dirty pages and inodes (no new dirty pages or inodes can be created when | |
1325 | * sync is running). | |
1326 | * | |
1327 | * SB_FREEZE_FS: The file system is frozen. Now all internal sources of fs | |
1328 | * modification are blocked (e.g. XFS preallocation truncation on inode | |
1329 | * reclaim). This is usually implemented by blocking new transactions for | |
1330 | * filesystems that have them and need this additional guard. After all | |
1331 | * internal writers are finished we call ->freeze_fs() to finish filesystem | |
1332 | * freezing. Then we transition to SB_FREEZE_COMPLETE state. This state is | |
1333 | * mostly auxiliary for filesystems to verify they do not modify frozen fs. | |
1334 | * | |
1335 | * sb->s_writers.frozen is protected by sb->s_umount. | |
18e9e510 JB |
1336 | */ |
1337 | int freeze_super(struct super_block *sb) | |
1338 | { | |
1339 | int ret; | |
1340 | ||
1341 | atomic_inc(&sb->s_active); | |
1342 | down_write(&sb->s_umount); | |
5accdf82 | 1343 | if (sb->s_writers.frozen != SB_UNFROZEN) { |
18e9e510 JB |
1344 | deactivate_locked_super(sb); |
1345 | return -EBUSY; | |
1346 | } | |
1347 | ||
dabe0dc1 AV |
1348 | if (!(sb->s_flags & MS_BORN)) { |
1349 | up_write(&sb->s_umount); | |
1350 | return 0; /* sic - it's "nothing to do" */ | |
1351 | } | |
1352 | ||
18e9e510 | 1353 | if (sb->s_flags & MS_RDONLY) { |
5accdf82 JK |
1354 | /* Nothing to do really... */ |
1355 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
18e9e510 JB |
1356 | up_write(&sb->s_umount); |
1357 | return 0; | |
1358 | } | |
1359 | ||
5accdf82 | 1360 | sb->s_writers.frozen = SB_FREEZE_WRITE; |
5accdf82 JK |
1361 | /* Release s_umount to preserve sb_start_write -> s_umount ordering */ |
1362 | up_write(&sb->s_umount); | |
5accdf82 | 1363 | sb_wait_write(sb, SB_FREEZE_WRITE); |
8129ed29 | 1364 | down_write(&sb->s_umount); |
5accdf82 JK |
1365 | |
1366 | /* Now we go and block page faults... */ | |
5accdf82 | 1367 | sb->s_writers.frozen = SB_FREEZE_PAGEFAULT; |
5accdf82 JK |
1368 | sb_wait_write(sb, SB_FREEZE_PAGEFAULT); |
1369 | ||
1370 | /* All writers are done so after syncing there won't be dirty data */ | |
18e9e510 JB |
1371 | sync_filesystem(sb); |
1372 | ||
5accdf82 JK |
1373 | /* Now wait for internal filesystem counter */ |
1374 | sb->s_writers.frozen = SB_FREEZE_FS; | |
5accdf82 | 1375 | sb_wait_write(sb, SB_FREEZE_FS); |
18e9e510 | 1376 | |
18e9e510 JB |
1377 | if (sb->s_op->freeze_fs) { |
1378 | ret = sb->s_op->freeze_fs(sb); | |
1379 | if (ret) { | |
1380 | printk(KERN_ERR | |
1381 | "VFS:Filesystem freeze failed\n"); | |
5accdf82 | 1382 | sb->s_writers.frozen = SB_UNFROZEN; |
8129ed29 | 1383 | sb_freeze_unlock(sb); |
5accdf82 | 1384 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 JB |
1385 | deactivate_locked_super(sb); |
1386 | return ret; | |
1387 | } | |
1388 | } | |
5accdf82 JK |
1389 | /* |
1390 | * This is just for debugging purposes so that fs can warn if it | |
1391 | * sees write activity when frozen is set to SB_FREEZE_COMPLETE. | |
1392 | */ | |
1393 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
18e9e510 JB |
1394 | up_write(&sb->s_umount); |
1395 | return 0; | |
1396 | } | |
1397 | EXPORT_SYMBOL(freeze_super); | |
1398 | ||
1399 | /** | |
1400 | * thaw_super -- unlock filesystem | |
1401 | * @sb: the super to thaw | |
1402 | * | |
1403 | * Unlocks the filesystem and marks it writeable again after freeze_super(). | |
1404 | */ | |
1405 | int thaw_super(struct super_block *sb) | |
1406 | { | |
1407 | int error; | |
1408 | ||
1409 | down_write(&sb->s_umount); | |
5accdf82 | 1410 | if (sb->s_writers.frozen == SB_UNFROZEN) { |
18e9e510 JB |
1411 | up_write(&sb->s_umount); |
1412 | return -EINVAL; | |
1413 | } | |
1414 | ||
8129ed29 ON |
1415 | if (sb->s_flags & MS_RDONLY) { |
1416 | sb->s_writers.frozen = SB_UNFROZEN; | |
18e9e510 | 1417 | goto out; |
8129ed29 | 1418 | } |
18e9e510 JB |
1419 | |
1420 | if (sb->s_op->unfreeze_fs) { | |
1421 | error = sb->s_op->unfreeze_fs(sb); | |
1422 | if (error) { | |
1423 | printk(KERN_ERR | |
1424 | "VFS:Filesystem thaw failed\n"); | |
18e9e510 JB |
1425 | up_write(&sb->s_umount); |
1426 | return error; | |
1427 | } | |
1428 | } | |
1429 | ||
5accdf82 | 1430 | sb->s_writers.frozen = SB_UNFROZEN; |
8129ed29 ON |
1431 | sb_freeze_unlock(sb); |
1432 | out: | |
5accdf82 | 1433 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 | 1434 | deactivate_locked_super(sb); |
18e9e510 JB |
1435 | return 0; |
1436 | } | |
1437 | EXPORT_SYMBOL(thaw_super); |