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 | ||
1da177e4 LT |
23 | #include <linux/module.h> |
24 | #include <linux/slab.h> | |
1da177e4 LT |
25 | #include <linux/acct.h> |
26 | #include <linux/blkdev.h> | |
1da177e4 LT |
27 | #include <linux/mount.h> |
28 | #include <linux/security.h> | |
1da177e4 LT |
29 | #include <linux/writeback.h> /* for the emergency remount stuff */ |
30 | #include <linux/idr.h> | |
353ab6e9 | 31 | #include <linux/mutex.h> |
5477d0fa | 32 | #include <linux/backing-dev.h> |
ceb5bdc2 | 33 | #include <linux/rculist_bl.h> |
c515e1fd | 34 | #include <linux/cleancache.h> |
40401530 | 35 | #include <linux/fsnotify.h> |
6d59e7f5 | 36 | #include "internal.h" |
1da177e4 LT |
37 | |
38 | ||
1da177e4 LT |
39 | LIST_HEAD(super_blocks); |
40 | DEFINE_SPINLOCK(sb_lock); | |
41 | ||
b0d40c92 DC |
42 | /* |
43 | * One thing we have to be careful of with a per-sb shrinker is that we don't | |
44 | * drop the last active reference to the superblock from within the shrinker. | |
45 | * If that happens we could trigger unregistering the shrinker from within the | |
46 | * shrinker path and that leads to deadlock on the shrinker_rwsem. Hence we | |
47 | * take a passive reference to the superblock to avoid this from occurring. | |
48 | */ | |
49 | static int prune_super(struct shrinker *shrink, struct shrink_control *sc) | |
50 | { | |
51 | struct super_block *sb; | |
0e1fdafd DC |
52 | int fs_objects = 0; |
53 | int total_objects; | |
b0d40c92 DC |
54 | |
55 | sb = container_of(shrink, struct super_block, s_shrink); | |
56 | ||
57 | /* | |
58 | * Deadlock avoidance. We may hold various FS locks, and we don't want | |
59 | * to recurse into the FS that called us in clear_inode() and friends.. | |
60 | */ | |
61 | if (sc->nr_to_scan && !(sc->gfp_mask & __GFP_FS)) | |
62 | return -1; | |
63 | ||
64 | if (!grab_super_passive(sb)) | |
09f363c7 | 65 | return !sc->nr_to_scan ? 0 : -1; |
b0d40c92 | 66 | |
0e1fdafd DC |
67 | if (sb->s_op && sb->s_op->nr_cached_objects) |
68 | fs_objects = sb->s_op->nr_cached_objects(sb); | |
69 | ||
70 | total_objects = sb->s_nr_dentry_unused + | |
71 | sb->s_nr_inodes_unused + fs_objects + 1; | |
72 | ||
b0d40c92 | 73 | if (sc->nr_to_scan) { |
0e1fdafd DC |
74 | int dentries; |
75 | int inodes; | |
b0d40c92 | 76 | |
0e1fdafd DC |
77 | /* proportion the scan between the caches */ |
78 | dentries = (sc->nr_to_scan * sb->s_nr_dentry_unused) / | |
79 | total_objects; | |
80 | inodes = (sc->nr_to_scan * sb->s_nr_inodes_unused) / | |
81 | total_objects; | |
82 | if (fs_objects) | |
83 | fs_objects = (sc->nr_to_scan * fs_objects) / | |
84 | total_objects; | |
85 | /* | |
86 | * prune the dcache first as the icache is pinned by it, then | |
87 | * prune the icache, followed by the filesystem specific caches | |
88 | */ | |
89 | prune_dcache_sb(sb, dentries); | |
90 | prune_icache_sb(sb, inodes); | |
b0d40c92 | 91 | |
0e1fdafd DC |
92 | if (fs_objects && sb->s_op->free_cached_objects) { |
93 | sb->s_op->free_cached_objects(sb, fs_objects); | |
94 | fs_objects = sb->s_op->nr_cached_objects(sb); | |
95 | } | |
96 | total_objects = sb->s_nr_dentry_unused + | |
97 | sb->s_nr_inodes_unused + fs_objects; | |
b0d40c92 DC |
98 | } |
99 | ||
0e1fdafd | 100 | total_objects = (total_objects / 100) * sysctl_vfs_cache_pressure; |
b0d40c92 | 101 | drop_super(sb); |
0e1fdafd | 102 | return total_objects; |
b0d40c92 DC |
103 | } |
104 | ||
1da177e4 LT |
105 | /** |
106 | * alloc_super - create new superblock | |
fe2bbc48 | 107 | * @type: filesystem type superblock should belong to |
1da177e4 LT |
108 | * |
109 | * Allocates and initializes a new &struct super_block. alloc_super() | |
110 | * returns a pointer new superblock or %NULL if allocation had failed. | |
111 | */ | |
cf516249 | 112 | static struct super_block *alloc_super(struct file_system_type *type) |
1da177e4 | 113 | { |
11b0b5ab | 114 | struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER); |
b87221de | 115 | static const struct super_operations default_op; |
1da177e4 LT |
116 | |
117 | if (s) { | |
1da177e4 LT |
118 | if (security_sb_alloc(s)) { |
119 | kfree(s); | |
120 | s = NULL; | |
121 | goto out; | |
122 | } | |
6416ccb7 NP |
123 | #ifdef CONFIG_SMP |
124 | s->s_files = alloc_percpu(struct list_head); | |
125 | if (!s->s_files) { | |
126 | security_sb_free(s); | |
127 | kfree(s); | |
128 | s = NULL; | |
129 | goto out; | |
130 | } else { | |
131 | int i; | |
132 | ||
133 | for_each_possible_cpu(i) | |
134 | INIT_LIST_HEAD(per_cpu_ptr(s->s_files, i)); | |
135 | } | |
136 | #else | |
1da177e4 | 137 | INIT_LIST_HEAD(&s->s_files); |
6416ccb7 | 138 | #endif |
95f28604 | 139 | s->s_bdi = &default_backing_dev_info; |
a5166169 | 140 | INIT_HLIST_NODE(&s->s_instances); |
ceb5bdc2 | 141 | INIT_HLIST_BL_HEAD(&s->s_anon); |
1da177e4 | 142 | INIT_LIST_HEAD(&s->s_inodes); |
da3bbdd4 | 143 | INIT_LIST_HEAD(&s->s_dentry_lru); |
98b745c6 | 144 | INIT_LIST_HEAD(&s->s_inode_lru); |
09cc9fc7 | 145 | spin_lock_init(&s->s_inode_lru_lock); |
39f7c4db | 146 | INIT_LIST_HEAD(&s->s_mounts); |
1da177e4 | 147 | init_rwsem(&s->s_umount); |
7892f2f4 | 148 | mutex_init(&s->s_lock); |
897c6ff9 | 149 | lockdep_set_class(&s->s_umount, &type->s_umount_key); |
cf516249 IM |
150 | /* |
151 | * The locking rules for s_lock are up to the | |
152 | * filesystem. For example ext3fs has different | |
153 | * lock ordering than usbfs: | |
154 | */ | |
155 | lockdep_set_class(&s->s_lock, &type->s_lock_key); | |
ada723dc PZ |
156 | /* |
157 | * sget() can have s_umount recursion. | |
158 | * | |
159 | * When it cannot find a suitable sb, it allocates a new | |
160 | * one (this one), and tries again to find a suitable old | |
161 | * one. | |
162 | * | |
163 | * In case that succeeds, it will acquire the s_umount | |
164 | * lock of the old one. Since these are clearly distrinct | |
165 | * locks, and this object isn't exposed yet, there's no | |
166 | * risk of deadlocks. | |
167 | * | |
168 | * Annotate this by putting this lock in a different | |
169 | * subclass. | |
170 | */ | |
171 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); | |
b20bd1a5 | 172 | s->s_count = 1; |
1da177e4 | 173 | atomic_set(&s->s_active, 1); |
a11f3a05 | 174 | mutex_init(&s->s_vfs_rename_mutex); |
51ee049e | 175 | lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key); |
d3be915f IM |
176 | mutex_init(&s->s_dquot.dqio_mutex); |
177 | mutex_init(&s->s_dquot.dqonoff_mutex); | |
1da177e4 LT |
178 | init_rwsem(&s->s_dquot.dqptr_sem); |
179 | init_waitqueue_head(&s->s_wait_unfrozen); | |
180 | s->s_maxbytes = MAX_NON_LFS; | |
1da177e4 LT |
181 | s->s_op = &default_op; |
182 | s->s_time_gran = 1000000000; | |
c515e1fd | 183 | s->cleancache_poolid = -1; |
b0d40c92 DC |
184 | |
185 | s->s_shrink.seeks = DEFAULT_SEEKS; | |
186 | s->s_shrink.shrink = prune_super; | |
8ab47664 | 187 | s->s_shrink.batch = 1024; |
1da177e4 LT |
188 | } |
189 | out: | |
190 | return s; | |
191 | } | |
192 | ||
193 | /** | |
194 | * destroy_super - frees a superblock | |
195 | * @s: superblock to free | |
196 | * | |
197 | * Frees a superblock. | |
198 | */ | |
199 | static inline void destroy_super(struct super_block *s) | |
200 | { | |
6416ccb7 NP |
201 | #ifdef CONFIG_SMP |
202 | free_percpu(s->s_files); | |
203 | #endif | |
1da177e4 | 204 | security_sb_free(s); |
39f7c4db | 205 | WARN_ON(!list_empty(&s->s_mounts)); |
79c0b2df | 206 | kfree(s->s_subtype); |
b3b304a2 | 207 | kfree(s->s_options); |
1da177e4 LT |
208 | kfree(s); |
209 | } | |
210 | ||
211 | /* Superblock refcounting */ | |
212 | ||
213 | /* | |
35cf7ba0 | 214 | * Drop a superblock's refcount. The caller must hold sb_lock. |
1da177e4 | 215 | */ |
f47ec3f2 | 216 | static void __put_super(struct super_block *sb) |
1da177e4 | 217 | { |
1da177e4 | 218 | if (!--sb->s_count) { |
551de6f3 | 219 | list_del_init(&sb->s_list); |
1da177e4 | 220 | destroy_super(sb); |
1da177e4 | 221 | } |
1da177e4 LT |
222 | } |
223 | ||
224 | /** | |
225 | * put_super - drop a temporary reference to superblock | |
226 | * @sb: superblock in question | |
227 | * | |
228 | * Drops a temporary reference, frees superblock if there's no | |
229 | * references left. | |
230 | */ | |
f47ec3f2 | 231 | static void put_super(struct super_block *sb) |
1da177e4 LT |
232 | { |
233 | spin_lock(&sb_lock); | |
234 | __put_super(sb); | |
235 | spin_unlock(&sb_lock); | |
236 | } | |
237 | ||
238 | ||
239 | /** | |
1712ac8f | 240 | * deactivate_locked_super - drop an active reference to superblock |
1da177e4 LT |
241 | * @s: superblock to deactivate |
242 | * | |
1712ac8f AV |
243 | * Drops an active reference to superblock, converting it into a temprory |
244 | * one if there is no other active references left. In that case we | |
1da177e4 LT |
245 | * tell fs driver to shut it down and drop the temporary reference we |
246 | * had just acquired. | |
1712ac8f AV |
247 | * |
248 | * Caller holds exclusive lock on superblock; that lock is released. | |
1da177e4 | 249 | */ |
1712ac8f | 250 | void deactivate_locked_super(struct super_block *s) |
1da177e4 LT |
251 | { |
252 | struct file_system_type *fs = s->s_type; | |
b20bd1a5 | 253 | if (atomic_dec_and_test(&s->s_active)) { |
c515e1fd | 254 | cleancache_flush_fs(s); |
1da177e4 | 255 | fs->kill_sb(s); |
b0d40c92 DC |
256 | |
257 | /* caches are now gone, we can safely kill the shrinker now */ | |
258 | unregister_shrinker(&s->s_shrink); | |
259 | ||
d863b50a BH |
260 | /* |
261 | * We need to call rcu_barrier so all the delayed rcu free | |
262 | * inodes are flushed before we release the fs module. | |
263 | */ | |
264 | rcu_barrier(); | |
1da177e4 LT |
265 | put_filesystem(fs); |
266 | put_super(s); | |
1712ac8f AV |
267 | } else { |
268 | up_write(&s->s_umount); | |
1da177e4 LT |
269 | } |
270 | } | |
271 | ||
1712ac8f | 272 | EXPORT_SYMBOL(deactivate_locked_super); |
1da177e4 | 273 | |
74dbbdd7 | 274 | /** |
1712ac8f | 275 | * deactivate_super - drop an active reference to superblock |
74dbbdd7 AV |
276 | * @s: superblock to deactivate |
277 | * | |
1712ac8f AV |
278 | * Variant of deactivate_locked_super(), except that superblock is *not* |
279 | * locked by caller. If we are going to drop the final active reference, | |
280 | * lock will be acquired prior to that. | |
74dbbdd7 | 281 | */ |
1712ac8f | 282 | void deactivate_super(struct super_block *s) |
74dbbdd7 | 283 | { |
1712ac8f AV |
284 | if (!atomic_add_unless(&s->s_active, -1, 1)) { |
285 | down_write(&s->s_umount); | |
286 | deactivate_locked_super(s); | |
74dbbdd7 AV |
287 | } |
288 | } | |
289 | ||
1712ac8f | 290 | EXPORT_SYMBOL(deactivate_super); |
74dbbdd7 | 291 | |
1da177e4 LT |
292 | /** |
293 | * grab_super - acquire an active reference | |
294 | * @s: reference we are trying to make active | |
295 | * | |
296 | * Tries to acquire an active reference. grab_super() is used when we | |
297 | * had just found a superblock in super_blocks or fs_type->fs_supers | |
298 | * and want to turn it into a full-blown active reference. grab_super() | |
299 | * is called with sb_lock held and drops it. Returns 1 in case of | |
300 | * success, 0 if we had failed (superblock contents was already dead or | |
301 | * dying when grab_super() had been called). | |
302 | */ | |
9c4dbee7 | 303 | static int grab_super(struct super_block *s) __releases(sb_lock) |
1da177e4 | 304 | { |
b20bd1a5 AV |
305 | if (atomic_inc_not_zero(&s->s_active)) { |
306 | spin_unlock(&sb_lock); | |
b20bd1a5 AV |
307 | return 1; |
308 | } | |
309 | /* it's going away */ | |
1da177e4 LT |
310 | s->s_count++; |
311 | spin_unlock(&sb_lock); | |
1712ac8f | 312 | /* wait for it to die */ |
1da177e4 | 313 | down_write(&s->s_umount); |
1da177e4 LT |
314 | up_write(&s->s_umount); |
315 | put_super(s); | |
1da177e4 LT |
316 | return 0; |
317 | } | |
318 | ||
12ad3ab6 DC |
319 | /* |
320 | * grab_super_passive - acquire a passive reference | |
321 | * @s: reference we are trying to grab | |
322 | * | |
323 | * Tries to acquire a passive reference. This is used in places where we | |
324 | * cannot take an active reference but we need to ensure that the | |
325 | * superblock does not go away while we are working on it. It returns | |
326 | * false if a reference was not gained, and returns true with the s_umount | |
327 | * lock held in read mode if a reference is gained. On successful return, | |
328 | * the caller must drop the s_umount lock and the passive reference when | |
329 | * done. | |
330 | */ | |
331 | bool grab_super_passive(struct super_block *sb) | |
332 | { | |
333 | spin_lock(&sb_lock); | |
a5166169 | 334 | if (hlist_unhashed(&sb->s_instances)) { |
12ad3ab6 DC |
335 | spin_unlock(&sb_lock); |
336 | return false; | |
337 | } | |
338 | ||
339 | sb->s_count++; | |
340 | spin_unlock(&sb_lock); | |
341 | ||
342 | if (down_read_trylock(&sb->s_umount)) { | |
dabe0dc1 | 343 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
12ad3ab6 DC |
344 | return true; |
345 | up_read(&sb->s_umount); | |
346 | } | |
347 | ||
348 | put_super(sb); | |
349 | return false; | |
350 | } | |
351 | ||
914e2637 AV |
352 | /* |
353 | * Superblock locking. We really ought to get rid of these two. | |
354 | */ | |
355 | void lock_super(struct super_block * sb) | |
356 | { | |
914e2637 AV |
357 | mutex_lock(&sb->s_lock); |
358 | } | |
359 | ||
360 | void unlock_super(struct super_block * sb) | |
361 | { | |
914e2637 AV |
362 | mutex_unlock(&sb->s_lock); |
363 | } | |
364 | ||
365 | EXPORT_SYMBOL(lock_super); | |
366 | EXPORT_SYMBOL(unlock_super); | |
367 | ||
1da177e4 LT |
368 | /** |
369 | * generic_shutdown_super - common helper for ->kill_sb() | |
370 | * @sb: superblock to kill | |
371 | * | |
372 | * generic_shutdown_super() does all fs-independent work on superblock | |
373 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects | |
374 | * that need destruction out of superblock, call generic_shutdown_super() | |
375 | * and release aforementioned objects. Note: dentries and inodes _are_ | |
376 | * taken care of and do not need specific handling. | |
c636ebdb DH |
377 | * |
378 | * Upon calling this function, the filesystem may no longer alter or | |
379 | * rearrange the set of dentries belonging to this super_block, nor may it | |
380 | * change the attachments of dentries to inodes. | |
1da177e4 LT |
381 | */ |
382 | void generic_shutdown_super(struct super_block *sb) | |
383 | { | |
ee9b6d61 | 384 | const struct super_operations *sop = sb->s_op; |
1da177e4 | 385 | |
c636ebdb DH |
386 | if (sb->s_root) { |
387 | shrink_dcache_for_umount(sb); | |
60b0680f | 388 | sync_filesystem(sb); |
1da177e4 | 389 | sb->s_flags &= ~MS_ACTIVE; |
efaee192 | 390 | |
63997e98 AV |
391 | fsnotify_unmount_inodes(&sb->s_inodes); |
392 | ||
393 | evict_inodes(sb); | |
1da177e4 | 394 | |
1da177e4 LT |
395 | if (sop->put_super) |
396 | sop->put_super(sb); | |
397 | ||
63997e98 | 398 | if (!list_empty(&sb->s_inodes)) { |
7b4fe29e DJ |
399 | printk("VFS: Busy inodes after unmount of %s. " |
400 | "Self-destruct in 5 seconds. Have a nice day...\n", | |
401 | sb->s_id); | |
1da177e4 | 402 | } |
1da177e4 LT |
403 | } |
404 | spin_lock(&sb_lock); | |
405 | /* should be initialized for __put_super_and_need_restart() */ | |
a5166169 | 406 | hlist_del_init(&sb->s_instances); |
1da177e4 LT |
407 | spin_unlock(&sb_lock); |
408 | up_write(&sb->s_umount); | |
409 | } | |
410 | ||
411 | EXPORT_SYMBOL(generic_shutdown_super); | |
412 | ||
413 | /** | |
414 | * sget - find or create a superblock | |
415 | * @type: filesystem type superblock should belong to | |
416 | * @test: comparison callback | |
417 | * @set: setup callback | |
418 | * @data: argument to each of them | |
419 | */ | |
420 | struct super_block *sget(struct file_system_type *type, | |
421 | int (*test)(struct super_block *,void *), | |
422 | int (*set)(struct super_block *,void *), | |
423 | void *data) | |
424 | { | |
425 | struct super_block *s = NULL; | |
a5166169 | 426 | struct hlist_node *node; |
d4730127 | 427 | struct super_block *old; |
1da177e4 LT |
428 | int err; |
429 | ||
430 | retry: | |
431 | spin_lock(&sb_lock); | |
d4730127 | 432 | if (test) { |
a5166169 | 433 | hlist_for_each_entry(old, node, &type->fs_supers, s_instances) { |
d4730127 MK |
434 | if (!test(old, data)) |
435 | continue; | |
436 | if (!grab_super(old)) | |
437 | goto retry; | |
a3cfbb53 LZ |
438 | if (s) { |
439 | up_write(&s->s_umount); | |
d4730127 | 440 | destroy_super(s); |
7a4dec53 | 441 | s = NULL; |
a3cfbb53 | 442 | } |
d3f21473 | 443 | down_write(&old->s_umount); |
7a4dec53 AV |
444 | if (unlikely(!(old->s_flags & MS_BORN))) { |
445 | deactivate_locked_super(old); | |
446 | goto retry; | |
447 | } | |
d4730127 MK |
448 | return old; |
449 | } | |
1da177e4 LT |
450 | } |
451 | if (!s) { | |
452 | spin_unlock(&sb_lock); | |
cf516249 | 453 | s = alloc_super(type); |
1da177e4 LT |
454 | if (!s) |
455 | return ERR_PTR(-ENOMEM); | |
456 | goto retry; | |
457 | } | |
458 | ||
459 | err = set(s, data); | |
460 | if (err) { | |
461 | spin_unlock(&sb_lock); | |
a3cfbb53 | 462 | up_write(&s->s_umount); |
1da177e4 LT |
463 | destroy_super(s); |
464 | return ERR_PTR(err); | |
465 | } | |
466 | s->s_type = type; | |
467 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); | |
468 | list_add_tail(&s->s_list, &super_blocks); | |
a5166169 | 469 | hlist_add_head(&s->s_instances, &type->fs_supers); |
1da177e4 LT |
470 | spin_unlock(&sb_lock); |
471 | get_filesystem(type); | |
b0d40c92 | 472 | register_shrinker(&s->s_shrink); |
1da177e4 LT |
473 | return s; |
474 | } | |
475 | ||
476 | EXPORT_SYMBOL(sget); | |
477 | ||
478 | void drop_super(struct super_block *sb) | |
479 | { | |
480 | up_read(&sb->s_umount); | |
481 | put_super(sb); | |
482 | } | |
483 | ||
484 | EXPORT_SYMBOL(drop_super); | |
485 | ||
e5004753 CH |
486 | /** |
487 | * sync_supers - helper for periodic superblock writeback | |
488 | * | |
489 | * Call the write_super method if present on all dirty superblocks in | |
490 | * the system. This is for the periodic writeback used by most older | |
491 | * filesystems. For data integrity superblock writeback use | |
492 | * sync_filesystems() instead. | |
493 | * | |
1da177e4 LT |
494 | * Note: check the dirty flag before waiting, so we don't |
495 | * hold up the sync while mounting a device. (The newly | |
496 | * mounted device won't need syncing.) | |
497 | */ | |
498 | void sync_supers(void) | |
499 | { | |
dca33252 | 500 | struct super_block *sb, *p = NULL; |
618f0636 | 501 | |
1da177e4 | 502 | spin_lock(&sb_lock); |
dca33252 | 503 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 504 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 505 | continue; |
e5004753 | 506 | if (sb->s_op->write_super && sb->s_dirt) { |
1da177e4 LT |
507 | sb->s_count++; |
508 | spin_unlock(&sb_lock); | |
e5004753 | 509 | |
1da177e4 | 510 | down_read(&sb->s_umount); |
dabe0dc1 | 511 | if (sb->s_root && sb->s_dirt && (sb->s_flags & MS_BORN)) |
e5004753 | 512 | sb->s_op->write_super(sb); |
618f0636 | 513 | up_read(&sb->s_umount); |
e5004753 | 514 | |
618f0636 | 515 | spin_lock(&sb_lock); |
dca33252 AV |
516 | if (p) |
517 | __put_super(p); | |
518 | p = sb; | |
618f0636 KK |
519 | } |
520 | } | |
dca33252 AV |
521 | if (p) |
522 | __put_super(p); | |
1da177e4 LT |
523 | spin_unlock(&sb_lock); |
524 | } | |
525 | ||
01a05b33 AV |
526 | /** |
527 | * iterate_supers - call function for all active superblocks | |
528 | * @f: function to call | |
529 | * @arg: argument to pass to it | |
530 | * | |
531 | * Scans the superblock list and calls given function, passing it | |
532 | * locked superblock and given argument. | |
533 | */ | |
534 | void iterate_supers(void (*f)(struct super_block *, void *), void *arg) | |
535 | { | |
dca33252 | 536 | struct super_block *sb, *p = NULL; |
01a05b33 AV |
537 | |
538 | spin_lock(&sb_lock); | |
dca33252 | 539 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 540 | if (hlist_unhashed(&sb->s_instances)) |
01a05b33 AV |
541 | continue; |
542 | sb->s_count++; | |
543 | spin_unlock(&sb_lock); | |
544 | ||
545 | down_read(&sb->s_umount); | |
dabe0dc1 | 546 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
01a05b33 AV |
547 | f(sb, arg); |
548 | up_read(&sb->s_umount); | |
549 | ||
550 | spin_lock(&sb_lock); | |
dca33252 AV |
551 | if (p) |
552 | __put_super(p); | |
553 | p = sb; | |
01a05b33 | 554 | } |
dca33252 AV |
555 | if (p) |
556 | __put_super(p); | |
01a05b33 AV |
557 | spin_unlock(&sb_lock); |
558 | } | |
559 | ||
43e15cdb AV |
560 | /** |
561 | * iterate_supers_type - call function for superblocks of given type | |
562 | * @type: fs type | |
563 | * @f: function to call | |
564 | * @arg: argument to pass to it | |
565 | * | |
566 | * Scans the superblock list and calls given function, passing it | |
567 | * locked superblock and given argument. | |
568 | */ | |
569 | void iterate_supers_type(struct file_system_type *type, | |
570 | void (*f)(struct super_block *, void *), void *arg) | |
571 | { | |
572 | struct super_block *sb, *p = NULL; | |
a5166169 | 573 | struct hlist_node *node; |
43e15cdb AV |
574 | |
575 | spin_lock(&sb_lock); | |
a5166169 | 576 | hlist_for_each_entry(sb, node, &type->fs_supers, s_instances) { |
43e15cdb AV |
577 | sb->s_count++; |
578 | spin_unlock(&sb_lock); | |
579 | ||
580 | down_read(&sb->s_umount); | |
dabe0dc1 | 581 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
43e15cdb AV |
582 | f(sb, arg); |
583 | up_read(&sb->s_umount); | |
584 | ||
585 | spin_lock(&sb_lock); | |
586 | if (p) | |
587 | __put_super(p); | |
588 | p = sb; | |
589 | } | |
590 | if (p) | |
591 | __put_super(p); | |
592 | spin_unlock(&sb_lock); | |
593 | } | |
594 | ||
595 | EXPORT_SYMBOL(iterate_supers_type); | |
596 | ||
1da177e4 LT |
597 | /** |
598 | * get_super - get the superblock of a device | |
599 | * @bdev: device to get the superblock for | |
600 | * | |
601 | * Scans the superblock list and finds the superblock of the file system | |
602 | * mounted on the device given. %NULL is returned if no match is found. | |
603 | */ | |
604 | ||
df40c01a | 605 | struct super_block *get_super(struct block_device *bdev) |
1da177e4 | 606 | { |
618f0636 KK |
607 | struct super_block *sb; |
608 | ||
1da177e4 LT |
609 | if (!bdev) |
610 | return NULL; | |
618f0636 | 611 | |
1da177e4 | 612 | spin_lock(&sb_lock); |
618f0636 KK |
613 | rescan: |
614 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 615 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 616 | continue; |
618f0636 KK |
617 | if (sb->s_bdev == bdev) { |
618 | sb->s_count++; | |
1da177e4 | 619 | spin_unlock(&sb_lock); |
618f0636 | 620 | down_read(&sb->s_umount); |
df40c01a | 621 | /* still alive? */ |
dabe0dc1 | 622 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
618f0636 KK |
623 | return sb; |
624 | up_read(&sb->s_umount); | |
df40c01a | 625 | /* nope, got unmounted */ |
618f0636 | 626 | spin_lock(&sb_lock); |
df40c01a AV |
627 | __put_super(sb); |
628 | goto rescan; | |
1da177e4 LT |
629 | } |
630 | } | |
631 | spin_unlock(&sb_lock); | |
632 | return NULL; | |
633 | } | |
634 | ||
635 | EXPORT_SYMBOL(get_super); | |
4504230a CH |
636 | |
637 | /** | |
638 | * get_active_super - get an active reference to the superblock of a device | |
639 | * @bdev: device to get the superblock for | |
640 | * | |
641 | * Scans the superblock list and finds the superblock of the file system | |
642 | * mounted on the device given. Returns the superblock with an active | |
d3f21473 | 643 | * reference or %NULL if none was found. |
4504230a CH |
644 | */ |
645 | struct super_block *get_active_super(struct block_device *bdev) | |
646 | { | |
647 | struct super_block *sb; | |
648 | ||
649 | if (!bdev) | |
650 | return NULL; | |
651 | ||
1494583d | 652 | restart: |
4504230a CH |
653 | spin_lock(&sb_lock); |
654 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 655 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 656 | continue; |
1494583d AV |
657 | if (sb->s_bdev == bdev) { |
658 | if (grab_super(sb)) /* drops sb_lock */ | |
659 | return sb; | |
660 | else | |
661 | goto restart; | |
662 | } | |
4504230a CH |
663 | } |
664 | spin_unlock(&sb_lock); | |
665 | return NULL; | |
666 | } | |
1da177e4 | 667 | |
df40c01a | 668 | struct super_block *user_get_super(dev_t dev) |
1da177e4 | 669 | { |
618f0636 | 670 | struct super_block *sb; |
1da177e4 | 671 | |
1da177e4 | 672 | spin_lock(&sb_lock); |
618f0636 KK |
673 | rescan: |
674 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 675 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 676 | continue; |
618f0636 KK |
677 | if (sb->s_dev == dev) { |
678 | sb->s_count++; | |
1da177e4 | 679 | spin_unlock(&sb_lock); |
618f0636 | 680 | down_read(&sb->s_umount); |
df40c01a | 681 | /* still alive? */ |
dabe0dc1 | 682 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
618f0636 KK |
683 | return sb; |
684 | up_read(&sb->s_umount); | |
df40c01a | 685 | /* nope, got unmounted */ |
618f0636 | 686 | spin_lock(&sb_lock); |
df40c01a AV |
687 | __put_super(sb); |
688 | goto rescan; | |
1da177e4 LT |
689 | } |
690 | } | |
691 | spin_unlock(&sb_lock); | |
692 | return NULL; | |
693 | } | |
694 | ||
1da177e4 LT |
695 | /** |
696 | * do_remount_sb - asks filesystem to change mount options. | |
697 | * @sb: superblock in question | |
698 | * @flags: numeric part of options | |
699 | * @data: the rest of options | |
700 | * @force: whether or not to force the change | |
701 | * | |
702 | * Alters the mount options of a mounted file system. | |
703 | */ | |
704 | int do_remount_sb(struct super_block *sb, int flags, void *data, int force) | |
705 | { | |
706 | int retval; | |
c79d967d | 707 | int remount_ro; |
4504230a CH |
708 | |
709 | if (sb->s_frozen != SB_UNFROZEN) | |
710 | return -EBUSY; | |
711 | ||
9361401e | 712 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
713 | if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev)) |
714 | return -EACCES; | |
9361401e | 715 | #endif |
4504230a | 716 | |
1da177e4 LT |
717 | if (flags & MS_RDONLY) |
718 | acct_auto_close(sb); | |
719 | shrink_dcache_sb(sb); | |
60b0680f | 720 | sync_filesystem(sb); |
1da177e4 | 721 | |
d208bbdd | 722 | remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY); |
d208bbdd | 723 | |
1da177e4 LT |
724 | /* If we are remounting RDONLY and current sb is read/write, |
725 | make sure there are no rw files opened */ | |
d208bbdd | 726 | if (remount_ro) { |
4ed5e82f | 727 | if (force) { |
1da177e4 | 728 | mark_files_ro(sb); |
4ed5e82f MS |
729 | } else { |
730 | retval = sb_prepare_remount_readonly(sb); | |
731 | if (retval) | |
732 | return retval; | |
4ed5e82f | 733 | } |
1da177e4 LT |
734 | } |
735 | ||
736 | if (sb->s_op->remount_fs) { | |
1da177e4 | 737 | retval = sb->s_op->remount_fs(sb, &flags, data); |
2833eb2b MS |
738 | if (retval) { |
739 | if (!force) | |
4ed5e82f | 740 | goto cancel_readonly; |
2833eb2b MS |
741 | /* If forced remount, go ahead despite any errors */ |
742 | WARN(1, "forced remount of a %s fs returned %i\n", | |
743 | sb->s_type->name, retval); | |
744 | } | |
1da177e4 LT |
745 | } |
746 | sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK); | |
4ed5e82f MS |
747 | /* Needs to be ordered wrt mnt_is_readonly() */ |
748 | smp_wmb(); | |
749 | sb->s_readonly_remount = 0; | |
c79d967d | 750 | |
d208bbdd NP |
751 | /* |
752 | * Some filesystems modify their metadata via some other path than the | |
753 | * bdev buffer cache (eg. use a private mapping, or directories in | |
754 | * pagecache, etc). Also file data modifications go via their own | |
755 | * mappings. So If we try to mount readonly then copy the filesystem | |
756 | * from bdev, we could get stale data, so invalidate it to give a best | |
757 | * effort at coherency. | |
758 | */ | |
759 | if (remount_ro && sb->s_bdev) | |
760 | invalidate_bdev(sb->s_bdev); | |
1da177e4 | 761 | return 0; |
4ed5e82f MS |
762 | |
763 | cancel_readonly: | |
764 | sb->s_readonly_remount = 0; | |
765 | return retval; | |
1da177e4 LT |
766 | } |
767 | ||
a2a9537a | 768 | static void do_emergency_remount(struct work_struct *work) |
1da177e4 | 769 | { |
dca33252 | 770 | struct super_block *sb, *p = NULL; |
1da177e4 LT |
771 | |
772 | spin_lock(&sb_lock); | |
dca33252 | 773 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 774 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 775 | continue; |
1da177e4 LT |
776 | sb->s_count++; |
777 | spin_unlock(&sb_lock); | |
443b94ba | 778 | down_write(&sb->s_umount); |
dabe0dc1 AV |
779 | if (sb->s_root && sb->s_bdev && (sb->s_flags & MS_BORN) && |
780 | !(sb->s_flags & MS_RDONLY)) { | |
1da177e4 | 781 | /* |
1da177e4 LT |
782 | * What lock protects sb->s_flags?? |
783 | */ | |
1da177e4 | 784 | do_remount_sb(sb, MS_RDONLY, NULL, 1); |
1da177e4 | 785 | } |
443b94ba | 786 | up_write(&sb->s_umount); |
1da177e4 | 787 | spin_lock(&sb_lock); |
dca33252 AV |
788 | if (p) |
789 | __put_super(p); | |
790 | p = sb; | |
1da177e4 | 791 | } |
dca33252 AV |
792 | if (p) |
793 | __put_super(p); | |
1da177e4 | 794 | spin_unlock(&sb_lock); |
a2a9537a | 795 | kfree(work); |
1da177e4 LT |
796 | printk("Emergency Remount complete\n"); |
797 | } | |
798 | ||
799 | void emergency_remount(void) | |
800 | { | |
a2a9537a JA |
801 | struct work_struct *work; |
802 | ||
803 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | |
804 | if (work) { | |
805 | INIT_WORK(work, do_emergency_remount); | |
806 | schedule_work(work); | |
807 | } | |
1da177e4 LT |
808 | } |
809 | ||
810 | /* | |
811 | * Unnamed block devices are dummy devices used by virtual | |
812 | * filesystems which don't use real block-devices. -- jrs | |
813 | */ | |
814 | ||
ad76cbc6 | 815 | static DEFINE_IDA(unnamed_dev_ida); |
1da177e4 | 816 | static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */ |
c63e09ec | 817 | static int unnamed_dev_start = 0; /* don't bother trying below it */ |
1da177e4 | 818 | |
0ee5dc67 | 819 | int get_anon_bdev(dev_t *p) |
1da177e4 LT |
820 | { |
821 | int dev; | |
822 | int error; | |
823 | ||
824 | retry: | |
ad76cbc6 | 825 | if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0) |
1da177e4 LT |
826 | return -ENOMEM; |
827 | spin_lock(&unnamed_dev_lock); | |
c63e09ec | 828 | error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev); |
f21f6220 AV |
829 | if (!error) |
830 | unnamed_dev_start = dev + 1; | |
1da177e4 LT |
831 | spin_unlock(&unnamed_dev_lock); |
832 | if (error == -EAGAIN) | |
833 | /* We raced and lost with another CPU. */ | |
834 | goto retry; | |
835 | else if (error) | |
836 | return -EAGAIN; | |
837 | ||
838 | if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) { | |
839 | spin_lock(&unnamed_dev_lock); | |
ad76cbc6 | 840 | ida_remove(&unnamed_dev_ida, dev); |
f21f6220 AV |
841 | if (unnamed_dev_start > dev) |
842 | unnamed_dev_start = dev; | |
1da177e4 LT |
843 | spin_unlock(&unnamed_dev_lock); |
844 | return -EMFILE; | |
845 | } | |
0ee5dc67 | 846 | *p = MKDEV(0, dev & MINORMASK); |
1da177e4 LT |
847 | return 0; |
848 | } | |
0ee5dc67 | 849 | EXPORT_SYMBOL(get_anon_bdev); |
1da177e4 | 850 | |
0ee5dc67 | 851 | void free_anon_bdev(dev_t dev) |
1da177e4 | 852 | { |
0ee5dc67 | 853 | int slot = MINOR(dev); |
1da177e4 | 854 | spin_lock(&unnamed_dev_lock); |
ad76cbc6 | 855 | ida_remove(&unnamed_dev_ida, slot); |
c63e09ec AV |
856 | if (slot < unnamed_dev_start) |
857 | unnamed_dev_start = slot; | |
1da177e4 LT |
858 | spin_unlock(&unnamed_dev_lock); |
859 | } | |
0ee5dc67 AV |
860 | EXPORT_SYMBOL(free_anon_bdev); |
861 | ||
862 | int set_anon_super(struct super_block *s, void *data) | |
863 | { | |
864 | int error = get_anon_bdev(&s->s_dev); | |
865 | if (!error) | |
866 | s->s_bdi = &noop_backing_dev_info; | |
867 | return error; | |
868 | } | |
869 | ||
870 | EXPORT_SYMBOL(set_anon_super); | |
871 | ||
872 | void kill_anon_super(struct super_block *sb) | |
873 | { | |
874 | dev_t dev = sb->s_dev; | |
875 | generic_shutdown_super(sb); | |
876 | free_anon_bdev(dev); | |
877 | } | |
1da177e4 LT |
878 | |
879 | EXPORT_SYMBOL(kill_anon_super); | |
880 | ||
1da177e4 LT |
881 | void kill_litter_super(struct super_block *sb) |
882 | { | |
883 | if (sb->s_root) | |
884 | d_genocide(sb->s_root); | |
885 | kill_anon_super(sb); | |
886 | } | |
887 | ||
888 | EXPORT_SYMBOL(kill_litter_super); | |
889 | ||
909e6d94 SH |
890 | static int ns_test_super(struct super_block *sb, void *data) |
891 | { | |
892 | return sb->s_fs_info == data; | |
893 | } | |
894 | ||
895 | static int ns_set_super(struct super_block *sb, void *data) | |
896 | { | |
897 | sb->s_fs_info = data; | |
898 | return set_anon_super(sb, NULL); | |
899 | } | |
900 | ||
ceefda69 AV |
901 | struct dentry *mount_ns(struct file_system_type *fs_type, int flags, |
902 | void *data, int (*fill_super)(struct super_block *, void *, int)) | |
909e6d94 SH |
903 | { |
904 | struct super_block *sb; | |
905 | ||
906 | sb = sget(fs_type, ns_test_super, ns_set_super, data); | |
907 | if (IS_ERR(sb)) | |
ceefda69 | 908 | return ERR_CAST(sb); |
909e6d94 SH |
909 | |
910 | if (!sb->s_root) { | |
911 | int err; | |
912 | sb->s_flags = flags; | |
913 | err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0); | |
914 | if (err) { | |
74dbbdd7 | 915 | deactivate_locked_super(sb); |
ceefda69 | 916 | return ERR_PTR(err); |
909e6d94 SH |
917 | } |
918 | ||
919 | sb->s_flags |= MS_ACTIVE; | |
920 | } | |
921 | ||
ceefda69 | 922 | return dget(sb->s_root); |
909e6d94 SH |
923 | } |
924 | ||
ceefda69 | 925 | EXPORT_SYMBOL(mount_ns); |
909e6d94 | 926 | |
9361401e | 927 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
928 | static int set_bdev_super(struct super_block *s, void *data) |
929 | { | |
930 | s->s_bdev = data; | |
931 | s->s_dev = s->s_bdev->bd_dev; | |
32a88aa1 JA |
932 | |
933 | /* | |
934 | * We set the bdi here to the queue backing, file systems can | |
935 | * overwrite this in ->fill_super() | |
936 | */ | |
937 | s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info; | |
1da177e4 LT |
938 | return 0; |
939 | } | |
940 | ||
941 | static int test_bdev_super(struct super_block *s, void *data) | |
942 | { | |
943 | return (void *)s->s_bdev == data; | |
944 | } | |
945 | ||
152a0836 | 946 | struct dentry *mount_bdev(struct file_system_type *fs_type, |
1da177e4 | 947 | int flags, const char *dev_name, void *data, |
152a0836 | 948 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
949 | { |
950 | struct block_device *bdev; | |
951 | struct super_block *s; | |
d4d77629 | 952 | fmode_t mode = FMODE_READ | FMODE_EXCL; |
1da177e4 LT |
953 | int error = 0; |
954 | ||
30c40d2c AV |
955 | if (!(flags & MS_RDONLY)) |
956 | mode |= FMODE_WRITE; | |
957 | ||
d4d77629 | 958 | bdev = blkdev_get_by_path(dev_name, mode, fs_type); |
1da177e4 | 959 | if (IS_ERR(bdev)) |
152a0836 | 960 | return ERR_CAST(bdev); |
1da177e4 LT |
961 | |
962 | /* | |
963 | * once the super is inserted into the list by sget, s_umount | |
964 | * will protect the lockfs code from trying to start a snapshot | |
965 | * while we are mounting | |
966 | */ | |
4fadd7bb CH |
967 | mutex_lock(&bdev->bd_fsfreeze_mutex); |
968 | if (bdev->bd_fsfreeze_count > 0) { | |
969 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
970 | error = -EBUSY; | |
971 | goto error_bdev; | |
972 | } | |
1da177e4 | 973 | s = sget(fs_type, test_bdev_super, set_bdev_super, bdev); |
4fadd7bb | 974 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
1da177e4 | 975 | if (IS_ERR(s)) |
454e2398 | 976 | goto error_s; |
1da177e4 LT |
977 | |
978 | if (s->s_root) { | |
979 | if ((flags ^ s->s_flags) & MS_RDONLY) { | |
74dbbdd7 | 980 | deactivate_locked_super(s); |
454e2398 DH |
981 | error = -EBUSY; |
982 | goto error_bdev; | |
1da177e4 | 983 | } |
454e2398 | 984 | |
4f331f01 TH |
985 | /* |
986 | * s_umount nests inside bd_mutex during | |
e525fd89 TH |
987 | * __invalidate_device(). blkdev_put() acquires |
988 | * bd_mutex and can't be called under s_umount. Drop | |
989 | * s_umount temporarily. This is safe as we're | |
990 | * holding an active reference. | |
4f331f01 TH |
991 | */ |
992 | up_write(&s->s_umount); | |
d4d77629 | 993 | blkdev_put(bdev, mode); |
4f331f01 | 994 | down_write(&s->s_umount); |
1da177e4 LT |
995 | } else { |
996 | char b[BDEVNAME_SIZE]; | |
997 | ||
9e1f1de0 | 998 | s->s_flags = flags | MS_NOSEC; |
30c40d2c | 999 | s->s_mode = mode; |
1da177e4 | 1000 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); |
e78c9a00 | 1001 | sb_set_blocksize(s, block_size(bdev)); |
9b04c997 | 1002 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1003 | if (error) { |
74dbbdd7 | 1004 | deactivate_locked_super(s); |
454e2398 | 1005 | goto error; |
fa675765 | 1006 | } |
454e2398 DH |
1007 | |
1008 | s->s_flags |= MS_ACTIVE; | |
87d8fe1e | 1009 | bdev->bd_super = s; |
1da177e4 LT |
1010 | } |
1011 | ||
152a0836 | 1012 | return dget(s->s_root); |
1da177e4 | 1013 | |
454e2398 DH |
1014 | error_s: |
1015 | error = PTR_ERR(s); | |
1016 | error_bdev: | |
d4d77629 | 1017 | blkdev_put(bdev, mode); |
454e2398 | 1018 | error: |
152a0836 AV |
1019 | return ERR_PTR(error); |
1020 | } | |
1021 | EXPORT_SYMBOL(mount_bdev); | |
1022 | ||
1da177e4 LT |
1023 | void kill_block_super(struct super_block *sb) |
1024 | { | |
1025 | struct block_device *bdev = sb->s_bdev; | |
30c40d2c | 1026 | fmode_t mode = sb->s_mode; |
1da177e4 | 1027 | |
ddbaaf30 | 1028 | bdev->bd_super = NULL; |
1da177e4 LT |
1029 | generic_shutdown_super(sb); |
1030 | sync_blockdev(bdev); | |
d4d77629 | 1031 | WARN_ON_ONCE(!(mode & FMODE_EXCL)); |
e525fd89 | 1032 | blkdev_put(bdev, mode | FMODE_EXCL); |
1da177e4 LT |
1033 | } |
1034 | ||
1035 | EXPORT_SYMBOL(kill_block_super); | |
9361401e | 1036 | #endif |
1da177e4 | 1037 | |
3c26ff6e | 1038 | struct dentry *mount_nodev(struct file_system_type *fs_type, |
1da177e4 | 1039 | int flags, void *data, |
3c26ff6e | 1040 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1041 | { |
1042 | int error; | |
1043 | struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL); | |
1044 | ||
1045 | if (IS_ERR(s)) | |
3c26ff6e | 1046 | return ERR_CAST(s); |
1da177e4 LT |
1047 | |
1048 | s->s_flags = flags; | |
1049 | ||
9b04c997 | 1050 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1051 | if (error) { |
74dbbdd7 | 1052 | deactivate_locked_super(s); |
3c26ff6e | 1053 | return ERR_PTR(error); |
1da177e4 LT |
1054 | } |
1055 | s->s_flags |= MS_ACTIVE; | |
3c26ff6e | 1056 | return dget(s->s_root); |
1da177e4 | 1057 | } |
3c26ff6e AV |
1058 | EXPORT_SYMBOL(mount_nodev); |
1059 | ||
1da177e4 LT |
1060 | static int compare_single(struct super_block *s, void *p) |
1061 | { | |
1062 | return 1; | |
1063 | } | |
1064 | ||
fc14f2fe | 1065 | struct dentry *mount_single(struct file_system_type *fs_type, |
1da177e4 | 1066 | int flags, void *data, |
fc14f2fe | 1067 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1068 | { |
1069 | struct super_block *s; | |
1070 | int error; | |
1071 | ||
1072 | s = sget(fs_type, compare_single, set_anon_super, NULL); | |
1073 | if (IS_ERR(s)) | |
fc14f2fe | 1074 | return ERR_CAST(s); |
1da177e4 LT |
1075 | if (!s->s_root) { |
1076 | s->s_flags = flags; | |
9b04c997 | 1077 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1078 | if (error) { |
74dbbdd7 | 1079 | deactivate_locked_super(s); |
fc14f2fe | 1080 | return ERR_PTR(error); |
1da177e4 LT |
1081 | } |
1082 | s->s_flags |= MS_ACTIVE; | |
9329d1be KS |
1083 | } else { |
1084 | do_remount_sb(s, flags, data, 0); | |
1da177e4 | 1085 | } |
fc14f2fe AV |
1086 | return dget(s->s_root); |
1087 | } | |
1088 | EXPORT_SYMBOL(mount_single); | |
1089 | ||
9d412a43 AV |
1090 | struct dentry * |
1091 | mount_fs(struct file_system_type *type, int flags, const char *name, void *data) | |
1da177e4 | 1092 | { |
c96e41e9 | 1093 | struct dentry *root; |
9d412a43 | 1094 | struct super_block *sb; |
1da177e4 | 1095 | char *secdata = NULL; |
9d412a43 | 1096 | int error = -ENOMEM; |
8089352a | 1097 | |
e0007529 | 1098 | if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) { |
1da177e4 | 1099 | secdata = alloc_secdata(); |
454e2398 | 1100 | if (!secdata) |
9d412a43 | 1101 | goto out; |
1da177e4 | 1102 | |
e0007529 | 1103 | error = security_sb_copy_data(data, secdata); |
454e2398 | 1104 | if (error) |
1da177e4 | 1105 | goto out_free_secdata; |
1da177e4 LT |
1106 | } |
1107 | ||
1a102ff9 AV |
1108 | root = type->mount(type, flags, name, data); |
1109 | if (IS_ERR(root)) { | |
1110 | error = PTR_ERR(root); | |
1111 | goto out_free_secdata; | |
c96e41e9 | 1112 | } |
9d412a43 AV |
1113 | sb = root->d_sb; |
1114 | BUG_ON(!sb); | |
1115 | WARN_ON(!sb->s_bdi); | |
6c510389 | 1116 | WARN_ON(sb->s_bdi == &default_backing_dev_info); |
9d412a43 | 1117 | sb->s_flags |= MS_BORN; |
454e2398 | 1118 | |
9d412a43 | 1119 | error = security_sb_kern_mount(sb, flags, secdata); |
5129a469 JE |
1120 | if (error) |
1121 | goto out_sb; | |
454e2398 | 1122 | |
42cb56ae JL |
1123 | /* |
1124 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE | |
1125 | * but s_maxbytes was an unsigned long long for many releases. Throw | |
1126 | * this warning for a little while to try and catch filesystems that | |
4358b567 | 1127 | * violate this rule. |
42cb56ae | 1128 | */ |
9d412a43 AV |
1129 | WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " |
1130 | "negative value (%lld)\n", type->name, sb->s_maxbytes); | |
42cb56ae | 1131 | |
9d412a43 | 1132 | up_write(&sb->s_umount); |
8680e22f | 1133 | free_secdata(secdata); |
9d412a43 | 1134 | return root; |
1da177e4 | 1135 | out_sb: |
9d412a43 AV |
1136 | dput(root); |
1137 | deactivate_locked_super(sb); | |
1da177e4 LT |
1138 | out_free_secdata: |
1139 | free_secdata(secdata); | |
1da177e4 | 1140 | out: |
454e2398 | 1141 | return ERR_PTR(error); |
1da177e4 LT |
1142 | } |
1143 | ||
18e9e510 | 1144 | /** |
7000d3c4 RD |
1145 | * freeze_super - lock the filesystem and force it into a consistent state |
1146 | * @sb: the super to lock | |
18e9e510 JB |
1147 | * |
1148 | * Syncs the super to make sure the filesystem is consistent and calls the fs's | |
1149 | * freeze_fs. Subsequent calls to this without first thawing the fs will return | |
1150 | * -EBUSY. | |
1151 | */ | |
1152 | int freeze_super(struct super_block *sb) | |
1153 | { | |
1154 | int ret; | |
1155 | ||
1156 | atomic_inc(&sb->s_active); | |
1157 | down_write(&sb->s_umount); | |
1158 | if (sb->s_frozen) { | |
1159 | deactivate_locked_super(sb); | |
1160 | return -EBUSY; | |
1161 | } | |
1162 | ||
dabe0dc1 AV |
1163 | if (!(sb->s_flags & MS_BORN)) { |
1164 | up_write(&sb->s_umount); | |
1165 | return 0; /* sic - it's "nothing to do" */ | |
1166 | } | |
1167 | ||
18e9e510 JB |
1168 | if (sb->s_flags & MS_RDONLY) { |
1169 | sb->s_frozen = SB_FREEZE_TRANS; | |
1170 | smp_wmb(); | |
1171 | up_write(&sb->s_umount); | |
1172 | return 0; | |
1173 | } | |
1174 | ||
1175 | sb->s_frozen = SB_FREEZE_WRITE; | |
1176 | smp_wmb(); | |
1177 | ||
1178 | sync_filesystem(sb); | |
1179 | ||
1180 | sb->s_frozen = SB_FREEZE_TRANS; | |
1181 | smp_wmb(); | |
1182 | ||
1183 | sync_blockdev(sb->s_bdev); | |
1184 | if (sb->s_op->freeze_fs) { | |
1185 | ret = sb->s_op->freeze_fs(sb); | |
1186 | if (ret) { | |
1187 | printk(KERN_ERR | |
1188 | "VFS:Filesystem freeze failed\n"); | |
1189 | sb->s_frozen = SB_UNFROZEN; | |
e1616300 KM |
1190 | smp_wmb(); |
1191 | wake_up(&sb->s_wait_unfrozen); | |
18e9e510 JB |
1192 | deactivate_locked_super(sb); |
1193 | return ret; | |
1194 | } | |
1195 | } | |
1196 | up_write(&sb->s_umount); | |
1197 | return 0; | |
1198 | } | |
1199 | EXPORT_SYMBOL(freeze_super); | |
1200 | ||
1201 | /** | |
1202 | * thaw_super -- unlock filesystem | |
1203 | * @sb: the super to thaw | |
1204 | * | |
1205 | * Unlocks the filesystem and marks it writeable again after freeze_super(). | |
1206 | */ | |
1207 | int thaw_super(struct super_block *sb) | |
1208 | { | |
1209 | int error; | |
1210 | ||
1211 | down_write(&sb->s_umount); | |
1212 | if (sb->s_frozen == SB_UNFROZEN) { | |
1213 | up_write(&sb->s_umount); | |
1214 | return -EINVAL; | |
1215 | } | |
1216 | ||
1217 | if (sb->s_flags & MS_RDONLY) | |
1218 | goto out; | |
1219 | ||
1220 | if (sb->s_op->unfreeze_fs) { | |
1221 | error = sb->s_op->unfreeze_fs(sb); | |
1222 | if (error) { | |
1223 | printk(KERN_ERR | |
1224 | "VFS:Filesystem thaw failed\n"); | |
1225 | sb->s_frozen = SB_FREEZE_TRANS; | |
1226 | up_write(&sb->s_umount); | |
1227 | return error; | |
1228 | } | |
1229 | } | |
1230 | ||
1231 | out: | |
1232 | sb->s_frozen = SB_UNFROZEN; | |
1233 | smp_wmb(); | |
1234 | wake_up(&sb->s_wait_unfrozen); | |
1235 | deactivate_locked_super(sb); | |
1236 | ||
1237 | return 0; | |
1238 | } | |
1239 | EXPORT_SYMBOL(thaw_super); |