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> |
6d59e7f5 | 35 | #include "internal.h" |
1da177e4 LT |
36 | |
37 | ||
1da177e4 LT |
38 | LIST_HEAD(super_blocks); |
39 | DEFINE_SPINLOCK(sb_lock); | |
40 | ||
b0d40c92 DC |
41 | /* |
42 | * One thing we have to be careful of with a per-sb shrinker is that we don't | |
43 | * drop the last active reference to the superblock from within the shrinker. | |
44 | * If that happens we could trigger unregistering the shrinker from within the | |
45 | * shrinker path and that leads to deadlock on the shrinker_rwsem. Hence we | |
46 | * take a passive reference to the superblock to avoid this from occurring. | |
47 | */ | |
48 | static int prune_super(struct shrinker *shrink, struct shrink_control *sc) | |
49 | { | |
50 | struct super_block *sb; | |
51 | int count; | |
52 | ||
53 | sb = container_of(shrink, struct super_block, s_shrink); | |
54 | ||
55 | /* | |
56 | * Deadlock avoidance. We may hold various FS locks, and we don't want | |
57 | * to recurse into the FS that called us in clear_inode() and friends.. | |
58 | */ | |
59 | if (sc->nr_to_scan && !(sc->gfp_mask & __GFP_FS)) | |
60 | return -1; | |
61 | ||
62 | if (!grab_super_passive(sb)) | |
63 | return -1; | |
64 | ||
65 | if (sc->nr_to_scan) { | |
66 | /* proportion the scan between the two caches */ | |
67 | int total; | |
68 | ||
69 | total = sb->s_nr_dentry_unused + sb->s_nr_inodes_unused + 1; | |
70 | count = (sc->nr_to_scan * sb->s_nr_dentry_unused) / total; | |
71 | ||
72 | /* prune dcache first as icache is pinned by it */ | |
73 | prune_dcache_sb(sb, count); | |
74 | prune_icache_sb(sb, sc->nr_to_scan - count); | |
75 | } | |
76 | ||
77 | count = ((sb->s_nr_dentry_unused + sb->s_nr_inodes_unused) / 100) | |
78 | * sysctl_vfs_cache_pressure; | |
79 | drop_super(sb); | |
80 | return count; | |
81 | } | |
82 | ||
1da177e4 LT |
83 | /** |
84 | * alloc_super - create new superblock | |
fe2bbc48 | 85 | * @type: filesystem type superblock should belong to |
1da177e4 LT |
86 | * |
87 | * Allocates and initializes a new &struct super_block. alloc_super() | |
88 | * returns a pointer new superblock or %NULL if allocation had failed. | |
89 | */ | |
cf516249 | 90 | static struct super_block *alloc_super(struct file_system_type *type) |
1da177e4 | 91 | { |
11b0b5ab | 92 | struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER); |
b87221de | 93 | static const struct super_operations default_op; |
1da177e4 LT |
94 | |
95 | if (s) { | |
1da177e4 LT |
96 | if (security_sb_alloc(s)) { |
97 | kfree(s); | |
98 | s = NULL; | |
99 | goto out; | |
100 | } | |
6416ccb7 NP |
101 | #ifdef CONFIG_SMP |
102 | s->s_files = alloc_percpu(struct list_head); | |
103 | if (!s->s_files) { | |
104 | security_sb_free(s); | |
105 | kfree(s); | |
106 | s = NULL; | |
107 | goto out; | |
108 | } else { | |
109 | int i; | |
110 | ||
111 | for_each_possible_cpu(i) | |
112 | INIT_LIST_HEAD(per_cpu_ptr(s->s_files, i)); | |
113 | } | |
114 | #else | |
1da177e4 | 115 | INIT_LIST_HEAD(&s->s_files); |
6416ccb7 | 116 | #endif |
95f28604 | 117 | s->s_bdi = &default_backing_dev_info; |
1da177e4 | 118 | INIT_LIST_HEAD(&s->s_instances); |
ceb5bdc2 | 119 | INIT_HLIST_BL_HEAD(&s->s_anon); |
1da177e4 | 120 | INIT_LIST_HEAD(&s->s_inodes); |
da3bbdd4 | 121 | INIT_LIST_HEAD(&s->s_dentry_lru); |
98b745c6 | 122 | INIT_LIST_HEAD(&s->s_inode_lru); |
09cc9fc7 | 123 | spin_lock_init(&s->s_inode_lru_lock); |
1da177e4 | 124 | init_rwsem(&s->s_umount); |
7892f2f4 | 125 | mutex_init(&s->s_lock); |
897c6ff9 | 126 | lockdep_set_class(&s->s_umount, &type->s_umount_key); |
cf516249 IM |
127 | /* |
128 | * The locking rules for s_lock are up to the | |
129 | * filesystem. For example ext3fs has different | |
130 | * lock ordering than usbfs: | |
131 | */ | |
132 | lockdep_set_class(&s->s_lock, &type->s_lock_key); | |
ada723dc PZ |
133 | /* |
134 | * sget() can have s_umount recursion. | |
135 | * | |
136 | * When it cannot find a suitable sb, it allocates a new | |
137 | * one (this one), and tries again to find a suitable old | |
138 | * one. | |
139 | * | |
140 | * In case that succeeds, it will acquire the s_umount | |
141 | * lock of the old one. Since these are clearly distrinct | |
142 | * locks, and this object isn't exposed yet, there's no | |
143 | * risk of deadlocks. | |
144 | * | |
145 | * Annotate this by putting this lock in a different | |
146 | * subclass. | |
147 | */ | |
148 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); | |
b20bd1a5 | 149 | s->s_count = 1; |
1da177e4 | 150 | atomic_set(&s->s_active, 1); |
a11f3a05 | 151 | mutex_init(&s->s_vfs_rename_mutex); |
51ee049e | 152 | lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key); |
d3be915f IM |
153 | mutex_init(&s->s_dquot.dqio_mutex); |
154 | mutex_init(&s->s_dquot.dqonoff_mutex); | |
1da177e4 LT |
155 | init_rwsem(&s->s_dquot.dqptr_sem); |
156 | init_waitqueue_head(&s->s_wait_unfrozen); | |
157 | s->s_maxbytes = MAX_NON_LFS; | |
1da177e4 LT |
158 | s->s_op = &default_op; |
159 | s->s_time_gran = 1000000000; | |
c515e1fd | 160 | s->cleancache_poolid = -1; |
b0d40c92 DC |
161 | |
162 | s->s_shrink.seeks = DEFAULT_SEEKS; | |
163 | s->s_shrink.shrink = prune_super; | |
1da177e4 LT |
164 | } |
165 | out: | |
166 | return s; | |
167 | } | |
168 | ||
169 | /** | |
170 | * destroy_super - frees a superblock | |
171 | * @s: superblock to free | |
172 | * | |
173 | * Frees a superblock. | |
174 | */ | |
175 | static inline void destroy_super(struct super_block *s) | |
176 | { | |
6416ccb7 NP |
177 | #ifdef CONFIG_SMP |
178 | free_percpu(s->s_files); | |
179 | #endif | |
1da177e4 | 180 | security_sb_free(s); |
79c0b2df | 181 | kfree(s->s_subtype); |
b3b304a2 | 182 | kfree(s->s_options); |
1da177e4 LT |
183 | kfree(s); |
184 | } | |
185 | ||
186 | /* Superblock refcounting */ | |
187 | ||
188 | /* | |
35cf7ba0 | 189 | * Drop a superblock's refcount. The caller must hold sb_lock. |
1da177e4 | 190 | */ |
35cf7ba0 | 191 | void __put_super(struct super_block *sb) |
1da177e4 | 192 | { |
1da177e4 | 193 | if (!--sb->s_count) { |
551de6f3 | 194 | list_del_init(&sb->s_list); |
1da177e4 | 195 | destroy_super(sb); |
1da177e4 | 196 | } |
1da177e4 LT |
197 | } |
198 | ||
199 | /** | |
200 | * put_super - drop a temporary reference to superblock | |
201 | * @sb: superblock in question | |
202 | * | |
203 | * Drops a temporary reference, frees superblock if there's no | |
204 | * references left. | |
205 | */ | |
03ba3782 | 206 | void put_super(struct super_block *sb) |
1da177e4 LT |
207 | { |
208 | spin_lock(&sb_lock); | |
209 | __put_super(sb); | |
210 | spin_unlock(&sb_lock); | |
211 | } | |
212 | ||
213 | ||
214 | /** | |
1712ac8f | 215 | * deactivate_locked_super - drop an active reference to superblock |
1da177e4 LT |
216 | * @s: superblock to deactivate |
217 | * | |
1712ac8f AV |
218 | * Drops an active reference to superblock, converting it into a temprory |
219 | * one if there is no other active references left. In that case we | |
1da177e4 LT |
220 | * tell fs driver to shut it down and drop the temporary reference we |
221 | * had just acquired. | |
1712ac8f AV |
222 | * |
223 | * Caller holds exclusive lock on superblock; that lock is released. | |
1da177e4 | 224 | */ |
1712ac8f | 225 | void deactivate_locked_super(struct super_block *s) |
1da177e4 LT |
226 | { |
227 | struct file_system_type *fs = s->s_type; | |
b20bd1a5 | 228 | if (atomic_dec_and_test(&s->s_active)) { |
c515e1fd | 229 | cleancache_flush_fs(s); |
1da177e4 | 230 | fs->kill_sb(s); |
b0d40c92 DC |
231 | |
232 | /* caches are now gone, we can safely kill the shrinker now */ | |
233 | unregister_shrinker(&s->s_shrink); | |
234 | ||
d863b50a BH |
235 | /* |
236 | * We need to call rcu_barrier so all the delayed rcu free | |
237 | * inodes are flushed before we release the fs module. | |
238 | */ | |
239 | rcu_barrier(); | |
1da177e4 LT |
240 | put_filesystem(fs); |
241 | put_super(s); | |
1712ac8f AV |
242 | } else { |
243 | up_write(&s->s_umount); | |
1da177e4 LT |
244 | } |
245 | } | |
246 | ||
1712ac8f | 247 | EXPORT_SYMBOL(deactivate_locked_super); |
1da177e4 | 248 | |
74dbbdd7 | 249 | /** |
1712ac8f | 250 | * deactivate_super - drop an active reference to superblock |
74dbbdd7 AV |
251 | * @s: superblock to deactivate |
252 | * | |
1712ac8f AV |
253 | * Variant of deactivate_locked_super(), except that superblock is *not* |
254 | * locked by caller. If we are going to drop the final active reference, | |
255 | * lock will be acquired prior to that. | |
74dbbdd7 | 256 | */ |
1712ac8f | 257 | void deactivate_super(struct super_block *s) |
74dbbdd7 | 258 | { |
1712ac8f AV |
259 | if (!atomic_add_unless(&s->s_active, -1, 1)) { |
260 | down_write(&s->s_umount); | |
261 | deactivate_locked_super(s); | |
74dbbdd7 AV |
262 | } |
263 | } | |
264 | ||
1712ac8f | 265 | EXPORT_SYMBOL(deactivate_super); |
74dbbdd7 | 266 | |
1da177e4 LT |
267 | /** |
268 | * grab_super - acquire an active reference | |
269 | * @s: reference we are trying to make active | |
270 | * | |
271 | * Tries to acquire an active reference. grab_super() is used when we | |
272 | * had just found a superblock in super_blocks or fs_type->fs_supers | |
273 | * and want to turn it into a full-blown active reference. grab_super() | |
274 | * is called with sb_lock held and drops it. Returns 1 in case of | |
275 | * success, 0 if we had failed (superblock contents was already dead or | |
276 | * dying when grab_super() had been called). | |
277 | */ | |
9c4dbee7 | 278 | static int grab_super(struct super_block *s) __releases(sb_lock) |
1da177e4 | 279 | { |
b20bd1a5 AV |
280 | if (atomic_inc_not_zero(&s->s_active)) { |
281 | spin_unlock(&sb_lock); | |
b20bd1a5 AV |
282 | return 1; |
283 | } | |
284 | /* it's going away */ | |
1da177e4 LT |
285 | s->s_count++; |
286 | spin_unlock(&sb_lock); | |
1712ac8f | 287 | /* wait for it to die */ |
1da177e4 | 288 | down_write(&s->s_umount); |
1da177e4 LT |
289 | up_write(&s->s_umount); |
290 | put_super(s); | |
1da177e4 LT |
291 | return 0; |
292 | } | |
293 | ||
12ad3ab6 DC |
294 | /* |
295 | * grab_super_passive - acquire a passive reference | |
296 | * @s: reference we are trying to grab | |
297 | * | |
298 | * Tries to acquire a passive reference. This is used in places where we | |
299 | * cannot take an active reference but we need to ensure that the | |
300 | * superblock does not go away while we are working on it. It returns | |
301 | * false if a reference was not gained, and returns true with the s_umount | |
302 | * lock held in read mode if a reference is gained. On successful return, | |
303 | * the caller must drop the s_umount lock and the passive reference when | |
304 | * done. | |
305 | */ | |
306 | bool grab_super_passive(struct super_block *sb) | |
307 | { | |
308 | spin_lock(&sb_lock); | |
309 | if (list_empty(&sb->s_instances)) { | |
310 | spin_unlock(&sb_lock); | |
311 | return false; | |
312 | } | |
313 | ||
314 | sb->s_count++; | |
315 | spin_unlock(&sb_lock); | |
316 | ||
317 | if (down_read_trylock(&sb->s_umount)) { | |
318 | if (sb->s_root) | |
319 | return true; | |
320 | up_read(&sb->s_umount); | |
321 | } | |
322 | ||
323 | put_super(sb); | |
324 | return false; | |
325 | } | |
326 | ||
914e2637 AV |
327 | /* |
328 | * Superblock locking. We really ought to get rid of these two. | |
329 | */ | |
330 | void lock_super(struct super_block * sb) | |
331 | { | |
332 | get_fs_excl(); | |
333 | mutex_lock(&sb->s_lock); | |
334 | } | |
335 | ||
336 | void unlock_super(struct super_block * sb) | |
337 | { | |
338 | put_fs_excl(); | |
339 | mutex_unlock(&sb->s_lock); | |
340 | } | |
341 | ||
342 | EXPORT_SYMBOL(lock_super); | |
343 | EXPORT_SYMBOL(unlock_super); | |
344 | ||
1da177e4 LT |
345 | /** |
346 | * generic_shutdown_super - common helper for ->kill_sb() | |
347 | * @sb: superblock to kill | |
348 | * | |
349 | * generic_shutdown_super() does all fs-independent work on superblock | |
350 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects | |
351 | * that need destruction out of superblock, call generic_shutdown_super() | |
352 | * and release aforementioned objects. Note: dentries and inodes _are_ | |
353 | * taken care of and do not need specific handling. | |
c636ebdb DH |
354 | * |
355 | * Upon calling this function, the filesystem may no longer alter or | |
356 | * rearrange the set of dentries belonging to this super_block, nor may it | |
357 | * change the attachments of dentries to inodes. | |
1da177e4 LT |
358 | */ |
359 | void generic_shutdown_super(struct super_block *sb) | |
360 | { | |
ee9b6d61 | 361 | const struct super_operations *sop = sb->s_op; |
1da177e4 | 362 | |
c636ebdb DH |
363 | if (sb->s_root) { |
364 | shrink_dcache_for_umount(sb); | |
60b0680f | 365 | sync_filesystem(sb); |
a9e220f8 | 366 | get_fs_excl(); |
1da177e4 | 367 | sb->s_flags &= ~MS_ACTIVE; |
efaee192 | 368 | |
63997e98 AV |
369 | fsnotify_unmount_inodes(&sb->s_inodes); |
370 | ||
371 | evict_inodes(sb); | |
1da177e4 | 372 | |
1da177e4 LT |
373 | if (sop->put_super) |
374 | sop->put_super(sb); | |
375 | ||
63997e98 | 376 | if (!list_empty(&sb->s_inodes)) { |
7b4fe29e DJ |
377 | printk("VFS: Busy inodes after unmount of %s. " |
378 | "Self-destruct in 5 seconds. Have a nice day...\n", | |
379 | sb->s_id); | |
1da177e4 | 380 | } |
a9e220f8 | 381 | put_fs_excl(); |
1da177e4 LT |
382 | } |
383 | spin_lock(&sb_lock); | |
384 | /* should be initialized for __put_super_and_need_restart() */ | |
551de6f3 | 385 | list_del_init(&sb->s_instances); |
1da177e4 LT |
386 | spin_unlock(&sb_lock); |
387 | up_write(&sb->s_umount); | |
388 | } | |
389 | ||
390 | EXPORT_SYMBOL(generic_shutdown_super); | |
391 | ||
392 | /** | |
393 | * sget - find or create a superblock | |
394 | * @type: filesystem type superblock should belong to | |
395 | * @test: comparison callback | |
396 | * @set: setup callback | |
397 | * @data: argument to each of them | |
398 | */ | |
399 | struct super_block *sget(struct file_system_type *type, | |
400 | int (*test)(struct super_block *,void *), | |
401 | int (*set)(struct super_block *,void *), | |
402 | void *data) | |
403 | { | |
404 | struct super_block *s = NULL; | |
d4730127 | 405 | struct super_block *old; |
1da177e4 LT |
406 | int err; |
407 | ||
408 | retry: | |
409 | spin_lock(&sb_lock); | |
d4730127 MK |
410 | if (test) { |
411 | list_for_each_entry(old, &type->fs_supers, s_instances) { | |
412 | if (!test(old, data)) | |
413 | continue; | |
414 | if (!grab_super(old)) | |
415 | goto retry; | |
a3cfbb53 LZ |
416 | if (s) { |
417 | up_write(&s->s_umount); | |
d4730127 | 418 | destroy_super(s); |
7a4dec53 | 419 | s = NULL; |
a3cfbb53 | 420 | } |
d3f21473 | 421 | down_write(&old->s_umount); |
7a4dec53 AV |
422 | if (unlikely(!(old->s_flags & MS_BORN))) { |
423 | deactivate_locked_super(old); | |
424 | goto retry; | |
425 | } | |
d4730127 MK |
426 | return old; |
427 | } | |
1da177e4 LT |
428 | } |
429 | if (!s) { | |
430 | spin_unlock(&sb_lock); | |
cf516249 | 431 | s = alloc_super(type); |
1da177e4 LT |
432 | if (!s) |
433 | return ERR_PTR(-ENOMEM); | |
434 | goto retry; | |
435 | } | |
436 | ||
437 | err = set(s, data); | |
438 | if (err) { | |
439 | spin_unlock(&sb_lock); | |
a3cfbb53 | 440 | up_write(&s->s_umount); |
1da177e4 LT |
441 | destroy_super(s); |
442 | return ERR_PTR(err); | |
443 | } | |
444 | s->s_type = type; | |
445 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); | |
446 | list_add_tail(&s->s_list, &super_blocks); | |
447 | list_add(&s->s_instances, &type->fs_supers); | |
448 | spin_unlock(&sb_lock); | |
449 | get_filesystem(type); | |
b0d40c92 | 450 | register_shrinker(&s->s_shrink); |
1da177e4 LT |
451 | return s; |
452 | } | |
453 | ||
454 | EXPORT_SYMBOL(sget); | |
455 | ||
456 | void drop_super(struct super_block *sb) | |
457 | { | |
458 | up_read(&sb->s_umount); | |
459 | put_super(sb); | |
460 | } | |
461 | ||
462 | EXPORT_SYMBOL(drop_super); | |
463 | ||
e5004753 CH |
464 | /** |
465 | * sync_supers - helper for periodic superblock writeback | |
466 | * | |
467 | * Call the write_super method if present on all dirty superblocks in | |
468 | * the system. This is for the periodic writeback used by most older | |
469 | * filesystems. For data integrity superblock writeback use | |
470 | * sync_filesystems() instead. | |
471 | * | |
1da177e4 LT |
472 | * Note: check the dirty flag before waiting, so we don't |
473 | * hold up the sync while mounting a device. (The newly | |
474 | * mounted device won't need syncing.) | |
475 | */ | |
476 | void sync_supers(void) | |
477 | { | |
dca33252 | 478 | struct super_block *sb, *p = NULL; |
618f0636 | 479 | |
1da177e4 | 480 | spin_lock(&sb_lock); |
dca33252 | 481 | list_for_each_entry(sb, &super_blocks, s_list) { |
551de6f3 AV |
482 | if (list_empty(&sb->s_instances)) |
483 | continue; | |
e5004753 | 484 | if (sb->s_op->write_super && sb->s_dirt) { |
1da177e4 LT |
485 | sb->s_count++; |
486 | spin_unlock(&sb_lock); | |
e5004753 | 487 | |
1da177e4 | 488 | down_read(&sb->s_umount); |
e5004753 CH |
489 | if (sb->s_root && sb->s_dirt) |
490 | sb->s_op->write_super(sb); | |
618f0636 | 491 | up_read(&sb->s_umount); |
e5004753 | 492 | |
618f0636 | 493 | spin_lock(&sb_lock); |
dca33252 AV |
494 | if (p) |
495 | __put_super(p); | |
496 | p = sb; | |
618f0636 KK |
497 | } |
498 | } | |
dca33252 AV |
499 | if (p) |
500 | __put_super(p); | |
1da177e4 LT |
501 | spin_unlock(&sb_lock); |
502 | } | |
503 | ||
01a05b33 AV |
504 | /** |
505 | * iterate_supers - call function for all active superblocks | |
506 | * @f: function to call | |
507 | * @arg: argument to pass to it | |
508 | * | |
509 | * Scans the superblock list and calls given function, passing it | |
510 | * locked superblock and given argument. | |
511 | */ | |
512 | void iterate_supers(void (*f)(struct super_block *, void *), void *arg) | |
513 | { | |
dca33252 | 514 | struct super_block *sb, *p = NULL; |
01a05b33 AV |
515 | |
516 | spin_lock(&sb_lock); | |
dca33252 | 517 | list_for_each_entry(sb, &super_blocks, s_list) { |
01a05b33 AV |
518 | if (list_empty(&sb->s_instances)) |
519 | continue; | |
520 | sb->s_count++; | |
521 | spin_unlock(&sb_lock); | |
522 | ||
523 | down_read(&sb->s_umount); | |
524 | if (sb->s_root) | |
525 | f(sb, arg); | |
526 | up_read(&sb->s_umount); | |
527 | ||
528 | spin_lock(&sb_lock); | |
dca33252 AV |
529 | if (p) |
530 | __put_super(p); | |
531 | p = sb; | |
01a05b33 | 532 | } |
dca33252 AV |
533 | if (p) |
534 | __put_super(p); | |
01a05b33 AV |
535 | spin_unlock(&sb_lock); |
536 | } | |
537 | ||
43e15cdb AV |
538 | /** |
539 | * iterate_supers_type - call function for superblocks of given type | |
540 | * @type: fs type | |
541 | * @f: function to call | |
542 | * @arg: argument to pass to it | |
543 | * | |
544 | * Scans the superblock list and calls given function, passing it | |
545 | * locked superblock and given argument. | |
546 | */ | |
547 | void iterate_supers_type(struct file_system_type *type, | |
548 | void (*f)(struct super_block *, void *), void *arg) | |
549 | { | |
550 | struct super_block *sb, *p = NULL; | |
551 | ||
552 | spin_lock(&sb_lock); | |
553 | list_for_each_entry(sb, &type->fs_supers, s_instances) { | |
554 | sb->s_count++; | |
555 | spin_unlock(&sb_lock); | |
556 | ||
557 | down_read(&sb->s_umount); | |
558 | if (sb->s_root) | |
559 | f(sb, arg); | |
560 | up_read(&sb->s_umount); | |
561 | ||
562 | spin_lock(&sb_lock); | |
563 | if (p) | |
564 | __put_super(p); | |
565 | p = sb; | |
566 | } | |
567 | if (p) | |
568 | __put_super(p); | |
569 | spin_unlock(&sb_lock); | |
570 | } | |
571 | ||
572 | EXPORT_SYMBOL(iterate_supers_type); | |
573 | ||
1da177e4 LT |
574 | /** |
575 | * get_super - get the superblock of a device | |
576 | * @bdev: device to get the superblock for | |
577 | * | |
578 | * Scans the superblock list and finds the superblock of the file system | |
579 | * mounted on the device given. %NULL is returned if no match is found. | |
580 | */ | |
581 | ||
df40c01a | 582 | struct super_block *get_super(struct block_device *bdev) |
1da177e4 | 583 | { |
618f0636 KK |
584 | struct super_block *sb; |
585 | ||
1da177e4 LT |
586 | if (!bdev) |
587 | return NULL; | |
618f0636 | 588 | |
1da177e4 | 589 | spin_lock(&sb_lock); |
618f0636 KK |
590 | rescan: |
591 | list_for_each_entry(sb, &super_blocks, s_list) { | |
551de6f3 AV |
592 | if (list_empty(&sb->s_instances)) |
593 | continue; | |
618f0636 KK |
594 | if (sb->s_bdev == bdev) { |
595 | sb->s_count++; | |
1da177e4 | 596 | spin_unlock(&sb_lock); |
618f0636 | 597 | down_read(&sb->s_umount); |
df40c01a | 598 | /* still alive? */ |
618f0636 KK |
599 | if (sb->s_root) |
600 | return sb; | |
601 | up_read(&sb->s_umount); | |
df40c01a | 602 | /* nope, got unmounted */ |
618f0636 | 603 | spin_lock(&sb_lock); |
df40c01a AV |
604 | __put_super(sb); |
605 | goto rescan; | |
1da177e4 LT |
606 | } |
607 | } | |
608 | spin_unlock(&sb_lock); | |
609 | return NULL; | |
610 | } | |
611 | ||
612 | EXPORT_SYMBOL(get_super); | |
4504230a CH |
613 | |
614 | /** | |
615 | * get_active_super - get an active reference to the superblock of a device | |
616 | * @bdev: device to get the superblock for | |
617 | * | |
618 | * Scans the superblock list and finds the superblock of the file system | |
619 | * mounted on the device given. Returns the superblock with an active | |
d3f21473 | 620 | * reference or %NULL if none was found. |
4504230a CH |
621 | */ |
622 | struct super_block *get_active_super(struct block_device *bdev) | |
623 | { | |
624 | struct super_block *sb; | |
625 | ||
626 | if (!bdev) | |
627 | return NULL; | |
628 | ||
1494583d | 629 | restart: |
4504230a CH |
630 | spin_lock(&sb_lock); |
631 | list_for_each_entry(sb, &super_blocks, s_list) { | |
551de6f3 AV |
632 | if (list_empty(&sb->s_instances)) |
633 | continue; | |
1494583d AV |
634 | if (sb->s_bdev == bdev) { |
635 | if (grab_super(sb)) /* drops sb_lock */ | |
636 | return sb; | |
637 | else | |
638 | goto restart; | |
639 | } | |
4504230a CH |
640 | } |
641 | spin_unlock(&sb_lock); | |
642 | return NULL; | |
643 | } | |
1da177e4 | 644 | |
df40c01a | 645 | struct super_block *user_get_super(dev_t dev) |
1da177e4 | 646 | { |
618f0636 | 647 | struct super_block *sb; |
1da177e4 | 648 | |
1da177e4 | 649 | spin_lock(&sb_lock); |
618f0636 KK |
650 | rescan: |
651 | list_for_each_entry(sb, &super_blocks, s_list) { | |
551de6f3 AV |
652 | if (list_empty(&sb->s_instances)) |
653 | continue; | |
618f0636 KK |
654 | if (sb->s_dev == dev) { |
655 | sb->s_count++; | |
1da177e4 | 656 | spin_unlock(&sb_lock); |
618f0636 | 657 | down_read(&sb->s_umount); |
df40c01a | 658 | /* still alive? */ |
618f0636 KK |
659 | if (sb->s_root) |
660 | return sb; | |
661 | up_read(&sb->s_umount); | |
df40c01a | 662 | /* nope, got unmounted */ |
618f0636 | 663 | spin_lock(&sb_lock); |
df40c01a AV |
664 | __put_super(sb); |
665 | goto rescan; | |
1da177e4 LT |
666 | } |
667 | } | |
668 | spin_unlock(&sb_lock); | |
669 | return NULL; | |
670 | } | |
671 | ||
1da177e4 LT |
672 | /** |
673 | * do_remount_sb - asks filesystem to change mount options. | |
674 | * @sb: superblock in question | |
675 | * @flags: numeric part of options | |
676 | * @data: the rest of options | |
677 | * @force: whether or not to force the change | |
678 | * | |
679 | * Alters the mount options of a mounted file system. | |
680 | */ | |
681 | int do_remount_sb(struct super_block *sb, int flags, void *data, int force) | |
682 | { | |
683 | int retval; | |
c79d967d | 684 | int remount_ro; |
4504230a CH |
685 | |
686 | if (sb->s_frozen != SB_UNFROZEN) | |
687 | return -EBUSY; | |
688 | ||
9361401e | 689 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
690 | if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev)) |
691 | return -EACCES; | |
9361401e | 692 | #endif |
4504230a | 693 | |
1da177e4 LT |
694 | if (flags & MS_RDONLY) |
695 | acct_auto_close(sb); | |
696 | shrink_dcache_sb(sb); | |
60b0680f | 697 | sync_filesystem(sb); |
1da177e4 | 698 | |
d208bbdd | 699 | remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY); |
d208bbdd | 700 | |
1da177e4 LT |
701 | /* If we are remounting RDONLY and current sb is read/write, |
702 | make sure there are no rw files opened */ | |
d208bbdd | 703 | if (remount_ro) { |
1da177e4 LT |
704 | if (force) |
705 | mark_files_ro(sb); | |
b0895513 | 706 | else if (!fs_may_remount_ro(sb)) |
1da177e4 LT |
707 | return -EBUSY; |
708 | } | |
709 | ||
710 | if (sb->s_op->remount_fs) { | |
1da177e4 | 711 | retval = sb->s_op->remount_fs(sb, &flags, data); |
b0895513 | 712 | if (retval) |
1da177e4 LT |
713 | return retval; |
714 | } | |
715 | sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK); | |
c79d967d | 716 | |
d208bbdd NP |
717 | /* |
718 | * Some filesystems modify their metadata via some other path than the | |
719 | * bdev buffer cache (eg. use a private mapping, or directories in | |
720 | * pagecache, etc). Also file data modifications go via their own | |
721 | * mappings. So If we try to mount readonly then copy the filesystem | |
722 | * from bdev, we could get stale data, so invalidate it to give a best | |
723 | * effort at coherency. | |
724 | */ | |
725 | if (remount_ro && sb->s_bdev) | |
726 | invalidate_bdev(sb->s_bdev); | |
1da177e4 LT |
727 | return 0; |
728 | } | |
729 | ||
a2a9537a | 730 | static void do_emergency_remount(struct work_struct *work) |
1da177e4 | 731 | { |
dca33252 | 732 | struct super_block *sb, *p = NULL; |
1da177e4 LT |
733 | |
734 | spin_lock(&sb_lock); | |
dca33252 | 735 | list_for_each_entry(sb, &super_blocks, s_list) { |
551de6f3 AV |
736 | if (list_empty(&sb->s_instances)) |
737 | continue; | |
1da177e4 LT |
738 | sb->s_count++; |
739 | spin_unlock(&sb_lock); | |
443b94ba | 740 | down_write(&sb->s_umount); |
1da177e4 LT |
741 | if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) { |
742 | /* | |
1da177e4 LT |
743 | * What lock protects sb->s_flags?? |
744 | */ | |
1da177e4 | 745 | do_remount_sb(sb, MS_RDONLY, NULL, 1); |
1da177e4 | 746 | } |
443b94ba | 747 | up_write(&sb->s_umount); |
1da177e4 | 748 | spin_lock(&sb_lock); |
dca33252 AV |
749 | if (p) |
750 | __put_super(p); | |
751 | p = sb; | |
1da177e4 | 752 | } |
dca33252 AV |
753 | if (p) |
754 | __put_super(p); | |
1da177e4 | 755 | spin_unlock(&sb_lock); |
a2a9537a | 756 | kfree(work); |
1da177e4 LT |
757 | printk("Emergency Remount complete\n"); |
758 | } | |
759 | ||
760 | void emergency_remount(void) | |
761 | { | |
a2a9537a JA |
762 | struct work_struct *work; |
763 | ||
764 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | |
765 | if (work) { | |
766 | INIT_WORK(work, do_emergency_remount); | |
767 | schedule_work(work); | |
768 | } | |
1da177e4 LT |
769 | } |
770 | ||
771 | /* | |
772 | * Unnamed block devices are dummy devices used by virtual | |
773 | * filesystems which don't use real block-devices. -- jrs | |
774 | */ | |
775 | ||
ad76cbc6 | 776 | static DEFINE_IDA(unnamed_dev_ida); |
1da177e4 | 777 | static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */ |
c63e09ec | 778 | static int unnamed_dev_start = 0; /* don't bother trying below it */ |
1da177e4 | 779 | |
0ee5dc67 | 780 | int get_anon_bdev(dev_t *p) |
1da177e4 LT |
781 | { |
782 | int dev; | |
783 | int error; | |
784 | ||
785 | retry: | |
ad76cbc6 | 786 | if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0) |
1da177e4 LT |
787 | return -ENOMEM; |
788 | spin_lock(&unnamed_dev_lock); | |
c63e09ec | 789 | error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev); |
f21f6220 AV |
790 | if (!error) |
791 | unnamed_dev_start = dev + 1; | |
1da177e4 LT |
792 | spin_unlock(&unnamed_dev_lock); |
793 | if (error == -EAGAIN) | |
794 | /* We raced and lost with another CPU. */ | |
795 | goto retry; | |
796 | else if (error) | |
797 | return -EAGAIN; | |
798 | ||
799 | if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) { | |
800 | spin_lock(&unnamed_dev_lock); | |
ad76cbc6 | 801 | ida_remove(&unnamed_dev_ida, dev); |
f21f6220 AV |
802 | if (unnamed_dev_start > dev) |
803 | unnamed_dev_start = dev; | |
1da177e4 LT |
804 | spin_unlock(&unnamed_dev_lock); |
805 | return -EMFILE; | |
806 | } | |
0ee5dc67 | 807 | *p = MKDEV(0, dev & MINORMASK); |
1da177e4 LT |
808 | return 0; |
809 | } | |
0ee5dc67 | 810 | EXPORT_SYMBOL(get_anon_bdev); |
1da177e4 | 811 | |
0ee5dc67 | 812 | void free_anon_bdev(dev_t dev) |
1da177e4 | 813 | { |
0ee5dc67 | 814 | int slot = MINOR(dev); |
1da177e4 | 815 | spin_lock(&unnamed_dev_lock); |
ad76cbc6 | 816 | ida_remove(&unnamed_dev_ida, slot); |
c63e09ec AV |
817 | if (slot < unnamed_dev_start) |
818 | unnamed_dev_start = slot; | |
1da177e4 LT |
819 | spin_unlock(&unnamed_dev_lock); |
820 | } | |
0ee5dc67 AV |
821 | EXPORT_SYMBOL(free_anon_bdev); |
822 | ||
823 | int set_anon_super(struct super_block *s, void *data) | |
824 | { | |
825 | int error = get_anon_bdev(&s->s_dev); | |
826 | if (!error) | |
827 | s->s_bdi = &noop_backing_dev_info; | |
828 | return error; | |
829 | } | |
830 | ||
831 | EXPORT_SYMBOL(set_anon_super); | |
832 | ||
833 | void kill_anon_super(struct super_block *sb) | |
834 | { | |
835 | dev_t dev = sb->s_dev; | |
836 | generic_shutdown_super(sb); | |
837 | free_anon_bdev(dev); | |
838 | } | |
1da177e4 LT |
839 | |
840 | EXPORT_SYMBOL(kill_anon_super); | |
841 | ||
1da177e4 LT |
842 | void kill_litter_super(struct super_block *sb) |
843 | { | |
844 | if (sb->s_root) | |
845 | d_genocide(sb->s_root); | |
846 | kill_anon_super(sb); | |
847 | } | |
848 | ||
849 | EXPORT_SYMBOL(kill_litter_super); | |
850 | ||
909e6d94 SH |
851 | static int ns_test_super(struct super_block *sb, void *data) |
852 | { | |
853 | return sb->s_fs_info == data; | |
854 | } | |
855 | ||
856 | static int ns_set_super(struct super_block *sb, void *data) | |
857 | { | |
858 | sb->s_fs_info = data; | |
859 | return set_anon_super(sb, NULL); | |
860 | } | |
861 | ||
ceefda69 AV |
862 | struct dentry *mount_ns(struct file_system_type *fs_type, int flags, |
863 | void *data, int (*fill_super)(struct super_block *, void *, int)) | |
909e6d94 SH |
864 | { |
865 | struct super_block *sb; | |
866 | ||
867 | sb = sget(fs_type, ns_test_super, ns_set_super, data); | |
868 | if (IS_ERR(sb)) | |
ceefda69 | 869 | return ERR_CAST(sb); |
909e6d94 SH |
870 | |
871 | if (!sb->s_root) { | |
872 | int err; | |
873 | sb->s_flags = flags; | |
874 | err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0); | |
875 | if (err) { | |
74dbbdd7 | 876 | deactivate_locked_super(sb); |
ceefda69 | 877 | return ERR_PTR(err); |
909e6d94 SH |
878 | } |
879 | ||
880 | sb->s_flags |= MS_ACTIVE; | |
881 | } | |
882 | ||
ceefda69 | 883 | return dget(sb->s_root); |
909e6d94 SH |
884 | } |
885 | ||
ceefda69 | 886 | EXPORT_SYMBOL(mount_ns); |
909e6d94 | 887 | |
9361401e | 888 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
889 | static int set_bdev_super(struct super_block *s, void *data) |
890 | { | |
891 | s->s_bdev = data; | |
892 | s->s_dev = s->s_bdev->bd_dev; | |
32a88aa1 JA |
893 | |
894 | /* | |
895 | * We set the bdi here to the queue backing, file systems can | |
896 | * overwrite this in ->fill_super() | |
897 | */ | |
898 | s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info; | |
1da177e4 LT |
899 | return 0; |
900 | } | |
901 | ||
902 | static int test_bdev_super(struct super_block *s, void *data) | |
903 | { | |
904 | return (void *)s->s_bdev == data; | |
905 | } | |
906 | ||
152a0836 | 907 | struct dentry *mount_bdev(struct file_system_type *fs_type, |
1da177e4 | 908 | int flags, const char *dev_name, void *data, |
152a0836 | 909 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
910 | { |
911 | struct block_device *bdev; | |
912 | struct super_block *s; | |
d4d77629 | 913 | fmode_t mode = FMODE_READ | FMODE_EXCL; |
1da177e4 LT |
914 | int error = 0; |
915 | ||
30c40d2c AV |
916 | if (!(flags & MS_RDONLY)) |
917 | mode |= FMODE_WRITE; | |
918 | ||
d4d77629 | 919 | bdev = blkdev_get_by_path(dev_name, mode, fs_type); |
1da177e4 | 920 | if (IS_ERR(bdev)) |
152a0836 | 921 | return ERR_CAST(bdev); |
1da177e4 LT |
922 | |
923 | /* | |
924 | * once the super is inserted into the list by sget, s_umount | |
925 | * will protect the lockfs code from trying to start a snapshot | |
926 | * while we are mounting | |
927 | */ | |
4fadd7bb CH |
928 | mutex_lock(&bdev->bd_fsfreeze_mutex); |
929 | if (bdev->bd_fsfreeze_count > 0) { | |
930 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
931 | error = -EBUSY; | |
932 | goto error_bdev; | |
933 | } | |
1da177e4 | 934 | s = sget(fs_type, test_bdev_super, set_bdev_super, bdev); |
4fadd7bb | 935 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
1da177e4 | 936 | if (IS_ERR(s)) |
454e2398 | 937 | goto error_s; |
1da177e4 LT |
938 | |
939 | if (s->s_root) { | |
940 | if ((flags ^ s->s_flags) & MS_RDONLY) { | |
74dbbdd7 | 941 | deactivate_locked_super(s); |
454e2398 DH |
942 | error = -EBUSY; |
943 | goto error_bdev; | |
1da177e4 | 944 | } |
454e2398 | 945 | |
4f331f01 TH |
946 | /* |
947 | * s_umount nests inside bd_mutex during | |
e525fd89 TH |
948 | * __invalidate_device(). blkdev_put() acquires |
949 | * bd_mutex and can't be called under s_umount. Drop | |
950 | * s_umount temporarily. This is safe as we're | |
951 | * holding an active reference. | |
4f331f01 TH |
952 | */ |
953 | up_write(&s->s_umount); | |
d4d77629 | 954 | blkdev_put(bdev, mode); |
4f331f01 | 955 | down_write(&s->s_umount); |
1da177e4 LT |
956 | } else { |
957 | char b[BDEVNAME_SIZE]; | |
958 | ||
9e1f1de0 | 959 | s->s_flags = flags | MS_NOSEC; |
30c40d2c | 960 | s->s_mode = mode; |
1da177e4 | 961 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); |
e78c9a00 | 962 | sb_set_blocksize(s, block_size(bdev)); |
9b04c997 | 963 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 964 | if (error) { |
74dbbdd7 | 965 | deactivate_locked_super(s); |
454e2398 | 966 | goto error; |
fa675765 | 967 | } |
454e2398 DH |
968 | |
969 | s->s_flags |= MS_ACTIVE; | |
87d8fe1e | 970 | bdev->bd_super = s; |
1da177e4 LT |
971 | } |
972 | ||
152a0836 | 973 | return dget(s->s_root); |
1da177e4 | 974 | |
454e2398 DH |
975 | error_s: |
976 | error = PTR_ERR(s); | |
977 | error_bdev: | |
d4d77629 | 978 | blkdev_put(bdev, mode); |
454e2398 | 979 | error: |
152a0836 AV |
980 | return ERR_PTR(error); |
981 | } | |
982 | EXPORT_SYMBOL(mount_bdev); | |
983 | ||
1da177e4 LT |
984 | void kill_block_super(struct super_block *sb) |
985 | { | |
986 | struct block_device *bdev = sb->s_bdev; | |
30c40d2c | 987 | fmode_t mode = sb->s_mode; |
1da177e4 | 988 | |
ddbaaf30 | 989 | bdev->bd_super = NULL; |
1da177e4 LT |
990 | generic_shutdown_super(sb); |
991 | sync_blockdev(bdev); | |
d4d77629 | 992 | WARN_ON_ONCE(!(mode & FMODE_EXCL)); |
e525fd89 | 993 | blkdev_put(bdev, mode | FMODE_EXCL); |
1da177e4 LT |
994 | } |
995 | ||
996 | EXPORT_SYMBOL(kill_block_super); | |
9361401e | 997 | #endif |
1da177e4 | 998 | |
3c26ff6e | 999 | struct dentry *mount_nodev(struct file_system_type *fs_type, |
1da177e4 | 1000 | int flags, void *data, |
3c26ff6e | 1001 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1002 | { |
1003 | int error; | |
1004 | struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL); | |
1005 | ||
1006 | if (IS_ERR(s)) | |
3c26ff6e | 1007 | return ERR_CAST(s); |
1da177e4 LT |
1008 | |
1009 | s->s_flags = flags; | |
1010 | ||
9b04c997 | 1011 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1012 | if (error) { |
74dbbdd7 | 1013 | deactivate_locked_super(s); |
3c26ff6e | 1014 | return ERR_PTR(error); |
1da177e4 LT |
1015 | } |
1016 | s->s_flags |= MS_ACTIVE; | |
3c26ff6e | 1017 | return dget(s->s_root); |
1da177e4 | 1018 | } |
3c26ff6e AV |
1019 | EXPORT_SYMBOL(mount_nodev); |
1020 | ||
1da177e4 LT |
1021 | static int compare_single(struct super_block *s, void *p) |
1022 | { | |
1023 | return 1; | |
1024 | } | |
1025 | ||
fc14f2fe | 1026 | struct dentry *mount_single(struct file_system_type *fs_type, |
1da177e4 | 1027 | int flags, void *data, |
fc14f2fe | 1028 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1029 | { |
1030 | struct super_block *s; | |
1031 | int error; | |
1032 | ||
1033 | s = sget(fs_type, compare_single, set_anon_super, NULL); | |
1034 | if (IS_ERR(s)) | |
fc14f2fe | 1035 | return ERR_CAST(s); |
1da177e4 LT |
1036 | if (!s->s_root) { |
1037 | s->s_flags = flags; | |
9b04c997 | 1038 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1039 | if (error) { |
74dbbdd7 | 1040 | deactivate_locked_super(s); |
fc14f2fe | 1041 | return ERR_PTR(error); |
1da177e4 LT |
1042 | } |
1043 | s->s_flags |= MS_ACTIVE; | |
9329d1be KS |
1044 | } else { |
1045 | do_remount_sb(s, flags, data, 0); | |
1da177e4 | 1046 | } |
fc14f2fe AV |
1047 | return dget(s->s_root); |
1048 | } | |
1049 | EXPORT_SYMBOL(mount_single); | |
1050 | ||
9d412a43 AV |
1051 | struct dentry * |
1052 | mount_fs(struct file_system_type *type, int flags, const char *name, void *data) | |
1da177e4 | 1053 | { |
c96e41e9 | 1054 | struct dentry *root; |
9d412a43 | 1055 | struct super_block *sb; |
1da177e4 | 1056 | char *secdata = NULL; |
9d412a43 | 1057 | int error = -ENOMEM; |
8089352a | 1058 | |
e0007529 | 1059 | if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) { |
1da177e4 | 1060 | secdata = alloc_secdata(); |
454e2398 | 1061 | if (!secdata) |
9d412a43 | 1062 | goto out; |
1da177e4 | 1063 | |
e0007529 | 1064 | error = security_sb_copy_data(data, secdata); |
454e2398 | 1065 | if (error) |
1da177e4 | 1066 | goto out_free_secdata; |
1da177e4 LT |
1067 | } |
1068 | ||
1a102ff9 AV |
1069 | root = type->mount(type, flags, name, data); |
1070 | if (IS_ERR(root)) { | |
1071 | error = PTR_ERR(root); | |
1072 | goto out_free_secdata; | |
c96e41e9 | 1073 | } |
9d412a43 AV |
1074 | sb = root->d_sb; |
1075 | BUG_ON(!sb); | |
1076 | WARN_ON(!sb->s_bdi); | |
6c510389 | 1077 | WARN_ON(sb->s_bdi == &default_backing_dev_info); |
9d412a43 | 1078 | sb->s_flags |= MS_BORN; |
454e2398 | 1079 | |
9d412a43 | 1080 | error = security_sb_kern_mount(sb, flags, secdata); |
5129a469 JE |
1081 | if (error) |
1082 | goto out_sb; | |
454e2398 | 1083 | |
42cb56ae JL |
1084 | /* |
1085 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE | |
1086 | * but s_maxbytes was an unsigned long long for many releases. Throw | |
1087 | * this warning for a little while to try and catch filesystems that | |
4358b567 | 1088 | * violate this rule. |
42cb56ae | 1089 | */ |
9d412a43 AV |
1090 | WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " |
1091 | "negative value (%lld)\n", type->name, sb->s_maxbytes); | |
42cb56ae | 1092 | |
9d412a43 | 1093 | up_write(&sb->s_umount); |
8680e22f | 1094 | free_secdata(secdata); |
9d412a43 | 1095 | return root; |
1da177e4 | 1096 | out_sb: |
9d412a43 AV |
1097 | dput(root); |
1098 | deactivate_locked_super(sb); | |
1da177e4 LT |
1099 | out_free_secdata: |
1100 | free_secdata(secdata); | |
1da177e4 | 1101 | out: |
454e2398 | 1102 | return ERR_PTR(error); |
1da177e4 LT |
1103 | } |
1104 | ||
18e9e510 | 1105 | /** |
7000d3c4 RD |
1106 | * freeze_super - lock the filesystem and force it into a consistent state |
1107 | * @sb: the super to lock | |
18e9e510 JB |
1108 | * |
1109 | * Syncs the super to make sure the filesystem is consistent and calls the fs's | |
1110 | * freeze_fs. Subsequent calls to this without first thawing the fs will return | |
1111 | * -EBUSY. | |
1112 | */ | |
1113 | int freeze_super(struct super_block *sb) | |
1114 | { | |
1115 | int ret; | |
1116 | ||
1117 | atomic_inc(&sb->s_active); | |
1118 | down_write(&sb->s_umount); | |
1119 | if (sb->s_frozen) { | |
1120 | deactivate_locked_super(sb); | |
1121 | return -EBUSY; | |
1122 | } | |
1123 | ||
1124 | if (sb->s_flags & MS_RDONLY) { | |
1125 | sb->s_frozen = SB_FREEZE_TRANS; | |
1126 | smp_wmb(); | |
1127 | up_write(&sb->s_umount); | |
1128 | return 0; | |
1129 | } | |
1130 | ||
1131 | sb->s_frozen = SB_FREEZE_WRITE; | |
1132 | smp_wmb(); | |
1133 | ||
1134 | sync_filesystem(sb); | |
1135 | ||
1136 | sb->s_frozen = SB_FREEZE_TRANS; | |
1137 | smp_wmb(); | |
1138 | ||
1139 | sync_blockdev(sb->s_bdev); | |
1140 | if (sb->s_op->freeze_fs) { | |
1141 | ret = sb->s_op->freeze_fs(sb); | |
1142 | if (ret) { | |
1143 | printk(KERN_ERR | |
1144 | "VFS:Filesystem freeze failed\n"); | |
1145 | sb->s_frozen = SB_UNFROZEN; | |
1146 | deactivate_locked_super(sb); | |
1147 | return ret; | |
1148 | } | |
1149 | } | |
1150 | up_write(&sb->s_umount); | |
1151 | return 0; | |
1152 | } | |
1153 | EXPORT_SYMBOL(freeze_super); | |
1154 | ||
1155 | /** | |
1156 | * thaw_super -- unlock filesystem | |
1157 | * @sb: the super to thaw | |
1158 | * | |
1159 | * Unlocks the filesystem and marks it writeable again after freeze_super(). | |
1160 | */ | |
1161 | int thaw_super(struct super_block *sb) | |
1162 | { | |
1163 | int error; | |
1164 | ||
1165 | down_write(&sb->s_umount); | |
1166 | if (sb->s_frozen == SB_UNFROZEN) { | |
1167 | up_write(&sb->s_umount); | |
1168 | return -EINVAL; | |
1169 | } | |
1170 | ||
1171 | if (sb->s_flags & MS_RDONLY) | |
1172 | goto out; | |
1173 | ||
1174 | if (sb->s_op->unfreeze_fs) { | |
1175 | error = sb->s_op->unfreeze_fs(sb); | |
1176 | if (error) { | |
1177 | printk(KERN_ERR | |
1178 | "VFS:Filesystem thaw failed\n"); | |
1179 | sb->s_frozen = SB_FREEZE_TRANS; | |
1180 | up_write(&sb->s_umount); | |
1181 | return error; | |
1182 | } | |
1183 | } | |
1184 | ||
1185 | out: | |
1186 | sb->s_frozen = SB_UNFROZEN; | |
1187 | smp_wmb(); | |
1188 | wake_up(&sb->s_wait_unfrozen); | |
1189 | deactivate_locked_super(sb); | |
1190 | ||
1191 | return 0; | |
1192 | } | |
1193 | EXPORT_SYMBOL(thaw_super); |