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 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> |
5accdf82 | 36 | #include <linux/lockdep.h> |
6d59e7f5 | 37 | #include "internal.h" |
1da177e4 LT |
38 | |
39 | ||
1da177e4 | 40 | LIST_HEAD(super_blocks); |
6fa3eb70 S |
41 | EXPORT_SYMBOL_GPL(super_blocks); |
42 | ||
1da177e4 LT |
43 | DEFINE_SPINLOCK(sb_lock); |
44 | ||
5accdf82 JK |
45 | static char *sb_writers_name[SB_FREEZE_LEVELS] = { |
46 | "sb_writers", | |
47 | "sb_pagefaults", | |
48 | "sb_internal", | |
49 | }; | |
50 | ||
b0d40c92 DC |
51 | /* |
52 | * One thing we have to be careful of with a per-sb shrinker is that we don't | |
53 | * drop the last active reference to the superblock from within the shrinker. | |
54 | * If that happens we could trigger unregistering the shrinker from within the | |
55 | * shrinker path and that leads to deadlock on the shrinker_rwsem. Hence we | |
56 | * take a passive reference to the superblock to avoid this from occurring. | |
57 | */ | |
58 | static int prune_super(struct shrinker *shrink, struct shrink_control *sc) | |
59 | { | |
60 | struct super_block *sb; | |
0e1fdafd DC |
61 | int fs_objects = 0; |
62 | int total_objects; | |
b0d40c92 DC |
63 | |
64 | sb = container_of(shrink, struct super_block, s_shrink); | |
65 | ||
66 | /* | |
67 | * Deadlock avoidance. We may hold various FS locks, and we don't want | |
68 | * to recurse into the FS that called us in clear_inode() and friends.. | |
69 | */ | |
70 | if (sc->nr_to_scan && !(sc->gfp_mask & __GFP_FS)) | |
71 | return -1; | |
72 | ||
73 | if (!grab_super_passive(sb)) | |
8e125cd8 | 74 | return -1; |
b0d40c92 | 75 | |
0e1fdafd DC |
76 | if (sb->s_op && sb->s_op->nr_cached_objects) |
77 | fs_objects = sb->s_op->nr_cached_objects(sb); | |
78 | ||
79 | total_objects = sb->s_nr_dentry_unused + | |
80 | sb->s_nr_inodes_unused + fs_objects + 1; | |
d016a08a TH |
81 | if (!total_objects) |
82 | total_objects = 1; | |
0e1fdafd | 83 | |
b0d40c92 | 84 | if (sc->nr_to_scan) { |
0e1fdafd DC |
85 | int dentries; |
86 | int inodes; | |
b0d40c92 | 87 | |
0e1fdafd DC |
88 | /* proportion the scan between the caches */ |
89 | dentries = (sc->nr_to_scan * sb->s_nr_dentry_unused) / | |
90 | total_objects; | |
91 | inodes = (sc->nr_to_scan * sb->s_nr_inodes_unused) / | |
92 | total_objects; | |
93 | if (fs_objects) | |
94 | fs_objects = (sc->nr_to_scan * fs_objects) / | |
95 | total_objects; | |
96 | /* | |
97 | * prune the dcache first as the icache is pinned by it, then | |
98 | * prune the icache, followed by the filesystem specific caches | |
99 | */ | |
100 | prune_dcache_sb(sb, dentries); | |
101 | prune_icache_sb(sb, inodes); | |
b0d40c92 | 102 | |
0e1fdafd DC |
103 | if (fs_objects && sb->s_op->free_cached_objects) { |
104 | sb->s_op->free_cached_objects(sb, fs_objects); | |
105 | fs_objects = sb->s_op->nr_cached_objects(sb); | |
106 | } | |
107 | total_objects = sb->s_nr_dentry_unused + | |
108 | sb->s_nr_inodes_unused + fs_objects; | |
b0d40c92 DC |
109 | } |
110 | ||
0e1fdafd | 111 | total_objects = (total_objects / 100) * sysctl_vfs_cache_pressure; |
b0d40c92 | 112 | drop_super(sb); |
0e1fdafd | 113 | return total_objects; |
b0d40c92 DC |
114 | } |
115 | ||
5accdf82 JK |
116 | static int init_sb_writers(struct super_block *s, struct file_system_type *type) |
117 | { | |
118 | int err; | |
119 | int i; | |
120 | ||
121 | for (i = 0; i < SB_FREEZE_LEVELS; i++) { | |
122 | err = percpu_counter_init(&s->s_writers.counter[i], 0); | |
123 | if (err < 0) | |
124 | goto err_out; | |
125 | lockdep_init_map(&s->s_writers.lock_map[i], sb_writers_name[i], | |
126 | &type->s_writers_key[i], 0); | |
127 | } | |
128 | init_waitqueue_head(&s->s_writers.wait); | |
129 | init_waitqueue_head(&s->s_writers.wait_unfrozen); | |
130 | return 0; | |
131 | err_out: | |
132 | while (--i >= 0) | |
133 | percpu_counter_destroy(&s->s_writers.counter[i]); | |
134 | return err; | |
135 | } | |
136 | ||
137 | static void destroy_sb_writers(struct super_block *s) | |
138 | { | |
139 | int i; | |
140 | ||
141 | for (i = 0; i < SB_FREEZE_LEVELS; i++) | |
142 | percpu_counter_destroy(&s->s_writers.counter[i]); | |
143 | } | |
144 | ||
1da177e4 LT |
145 | /** |
146 | * alloc_super - create new superblock | |
fe2bbc48 | 147 | * @type: filesystem type superblock should belong to |
9249e17f | 148 | * @flags: the mount flags |
1da177e4 LT |
149 | * |
150 | * Allocates and initializes a new &struct super_block. alloc_super() | |
151 | * returns a pointer new superblock or %NULL if allocation had failed. | |
152 | */ | |
9249e17f | 153 | static struct super_block *alloc_super(struct file_system_type *type, int flags) |
1da177e4 | 154 | { |
11b0b5ab | 155 | struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER); |
b87221de | 156 | static const struct super_operations default_op; |
1da177e4 LT |
157 | |
158 | if (s) { | |
1da177e4 | 159 | if (security_sb_alloc(s)) { |
5accdf82 JK |
160 | /* |
161 | * We cannot call security_sb_free() without | |
162 | * security_sb_alloc() succeeding. So bail out manually | |
163 | */ | |
1da177e4 LT |
164 | kfree(s); |
165 | s = NULL; | |
166 | goto out; | |
167 | } | |
5accdf82 JK |
168 | if (init_sb_writers(s, type)) |
169 | goto err_out; | |
9249e17f | 170 | s->s_flags = flags; |
95f28604 | 171 | s->s_bdi = &default_backing_dev_info; |
a5166169 | 172 | INIT_HLIST_NODE(&s->s_instances); |
ceb5bdc2 | 173 | INIT_HLIST_BL_HEAD(&s->s_anon); |
1da177e4 | 174 | INIT_LIST_HEAD(&s->s_inodes); |
da3bbdd4 | 175 | INIT_LIST_HEAD(&s->s_dentry_lru); |
98b745c6 | 176 | INIT_LIST_HEAD(&s->s_inode_lru); |
09cc9fc7 | 177 | spin_lock_init(&s->s_inode_lru_lock); |
39f7c4db | 178 | INIT_LIST_HEAD(&s->s_mounts); |
1da177e4 | 179 | init_rwsem(&s->s_umount); |
897c6ff9 | 180 | lockdep_set_class(&s->s_umount, &type->s_umount_key); |
ada723dc PZ |
181 | /* |
182 | * sget() can have s_umount recursion. | |
183 | * | |
184 | * When it cannot find a suitable sb, it allocates a new | |
185 | * one (this one), and tries again to find a suitable old | |
186 | * one. | |
187 | * | |
188 | * In case that succeeds, it will acquire the s_umount | |
189 | * lock of the old one. Since these are clearly distrinct | |
190 | * locks, and this object isn't exposed yet, there's no | |
191 | * risk of deadlocks. | |
192 | * | |
193 | * Annotate this by putting this lock in a different | |
194 | * subclass. | |
195 | */ | |
196 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); | |
b20bd1a5 | 197 | s->s_count = 1; |
1da177e4 | 198 | atomic_set(&s->s_active, 1); |
a11f3a05 | 199 | mutex_init(&s->s_vfs_rename_mutex); |
51ee049e | 200 | lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key); |
d3be915f IM |
201 | mutex_init(&s->s_dquot.dqio_mutex); |
202 | mutex_init(&s->s_dquot.dqonoff_mutex); | |
1da177e4 | 203 | init_rwsem(&s->s_dquot.dqptr_sem); |
1da177e4 | 204 | s->s_maxbytes = MAX_NON_LFS; |
1da177e4 LT |
205 | s->s_op = &default_op; |
206 | s->s_time_gran = 1000000000; | |
c515e1fd | 207 | s->cleancache_poolid = -1; |
b0d40c92 DC |
208 | |
209 | s->s_shrink.seeks = DEFAULT_SEEKS; | |
210 | s->s_shrink.shrink = prune_super; | |
8ab47664 | 211 | s->s_shrink.batch = 1024; |
1da177e4 LT |
212 | } |
213 | out: | |
214 | return s; | |
5accdf82 JK |
215 | err_out: |
216 | security_sb_free(s); | |
5accdf82 JK |
217 | destroy_sb_writers(s); |
218 | kfree(s); | |
219 | s = NULL; | |
220 | goto out; | |
1da177e4 LT |
221 | } |
222 | ||
223 | /** | |
224 | * destroy_super - frees a superblock | |
225 | * @s: superblock to free | |
226 | * | |
227 | * Frees a superblock. | |
228 | */ | |
229 | static inline void destroy_super(struct super_block *s) | |
230 | { | |
5accdf82 | 231 | destroy_sb_writers(s); |
1da177e4 | 232 | security_sb_free(s); |
39f7c4db | 233 | WARN_ON(!list_empty(&s->s_mounts)); |
79c0b2df | 234 | kfree(s->s_subtype); |
b3b304a2 | 235 | kfree(s->s_options); |
1da177e4 LT |
236 | kfree(s); |
237 | } | |
238 | ||
239 | /* Superblock refcounting */ | |
240 | ||
241 | /* | |
35cf7ba0 | 242 | * Drop a superblock's refcount. The caller must hold sb_lock. |
1da177e4 | 243 | */ |
f47ec3f2 | 244 | static void __put_super(struct super_block *sb) |
1da177e4 | 245 | { |
1da177e4 | 246 | if (!--sb->s_count) { |
551de6f3 | 247 | list_del_init(&sb->s_list); |
1da177e4 | 248 | destroy_super(sb); |
1da177e4 | 249 | } |
1da177e4 LT |
250 | } |
251 | ||
252 | /** | |
253 | * put_super - drop a temporary reference to superblock | |
254 | * @sb: superblock in question | |
255 | * | |
256 | * Drops a temporary reference, frees superblock if there's no | |
257 | * references left. | |
258 | */ | |
f47ec3f2 | 259 | static void put_super(struct super_block *sb) |
1da177e4 LT |
260 | { |
261 | spin_lock(&sb_lock); | |
262 | __put_super(sb); | |
263 | spin_unlock(&sb_lock); | |
264 | } | |
265 | ||
266 | ||
267 | /** | |
1712ac8f | 268 | * deactivate_locked_super - drop an active reference to superblock |
1da177e4 LT |
269 | * @s: superblock to deactivate |
270 | * | |
1712ac8f AV |
271 | * Drops an active reference to superblock, converting it into a temprory |
272 | * one if there is no other active references left. In that case we | |
1da177e4 LT |
273 | * tell fs driver to shut it down and drop the temporary reference we |
274 | * had just acquired. | |
1712ac8f AV |
275 | * |
276 | * Caller holds exclusive lock on superblock; that lock is released. | |
1da177e4 | 277 | */ |
1712ac8f | 278 | void deactivate_locked_super(struct super_block *s) |
1da177e4 LT |
279 | { |
280 | struct file_system_type *fs = s->s_type; | |
b20bd1a5 | 281 | if (atomic_dec_and_test(&s->s_active)) { |
3167760f | 282 | cleancache_invalidate_fs(s); |
1da177e4 | 283 | fs->kill_sb(s); |
b0d40c92 DC |
284 | |
285 | /* caches are now gone, we can safely kill the shrinker now */ | |
286 | unregister_shrinker(&s->s_shrink); | |
1da177e4 LT |
287 | put_filesystem(fs); |
288 | put_super(s); | |
1712ac8f AV |
289 | } else { |
290 | up_write(&s->s_umount); | |
1da177e4 LT |
291 | } |
292 | } | |
293 | ||
1712ac8f | 294 | EXPORT_SYMBOL(deactivate_locked_super); |
1da177e4 | 295 | |
74dbbdd7 | 296 | /** |
1712ac8f | 297 | * deactivate_super - drop an active reference to superblock |
74dbbdd7 AV |
298 | * @s: superblock to deactivate |
299 | * | |
1712ac8f AV |
300 | * Variant of deactivate_locked_super(), except that superblock is *not* |
301 | * locked by caller. If we are going to drop the final active reference, | |
302 | * lock will be acquired prior to that. | |
74dbbdd7 | 303 | */ |
1712ac8f | 304 | void deactivate_super(struct super_block *s) |
74dbbdd7 | 305 | { |
1712ac8f AV |
306 | if (!atomic_add_unless(&s->s_active, -1, 1)) { |
307 | down_write(&s->s_umount); | |
308 | deactivate_locked_super(s); | |
74dbbdd7 AV |
309 | } |
310 | } | |
311 | ||
1712ac8f | 312 | EXPORT_SYMBOL(deactivate_super); |
74dbbdd7 | 313 | |
1da177e4 LT |
314 | /** |
315 | * grab_super - acquire an active reference | |
316 | * @s: reference we are trying to make active | |
317 | * | |
318 | * Tries to acquire an active reference. grab_super() is used when we | |
319 | * had just found a superblock in super_blocks or fs_type->fs_supers | |
320 | * and want to turn it into a full-blown active reference. grab_super() | |
321 | * is called with sb_lock held and drops it. Returns 1 in case of | |
322 | * success, 0 if we had failed (superblock contents was already dead or | |
d0f1a6e5 AV |
323 | * dying when grab_super() had been called). Note that this is only |
324 | * called for superblocks not in rundown mode (== ones still on ->fs_supers | |
325 | * of their type), so increment of ->s_count is OK here. | |
1da177e4 | 326 | */ |
9c4dbee7 | 327 | static int grab_super(struct super_block *s) __releases(sb_lock) |
1da177e4 LT |
328 | { |
329 | s->s_count++; | |
330 | spin_unlock(&sb_lock); | |
331 | down_write(&s->s_umount); | |
d0f1a6e5 AV |
332 | if ((s->s_flags & MS_BORN) && atomic_inc_not_zero(&s->s_active)) { |
333 | put_super(s); | |
334 | return 1; | |
335 | } | |
1da177e4 LT |
336 | up_write(&s->s_umount); |
337 | put_super(s); | |
1da177e4 LT |
338 | return 0; |
339 | } | |
340 | ||
12ad3ab6 DC |
341 | /* |
342 | * grab_super_passive - acquire a passive reference | |
331cbdee | 343 | * @sb: reference we are trying to grab |
12ad3ab6 DC |
344 | * |
345 | * Tries to acquire a passive reference. This is used in places where we | |
346 | * cannot take an active reference but we need to ensure that the | |
347 | * superblock does not go away while we are working on it. It returns | |
348 | * false if a reference was not gained, and returns true with the s_umount | |
349 | * lock held in read mode if a reference is gained. On successful return, | |
350 | * the caller must drop the s_umount lock and the passive reference when | |
351 | * done. | |
352 | */ | |
353 | bool grab_super_passive(struct super_block *sb) | |
354 | { | |
355 | spin_lock(&sb_lock); | |
a5166169 | 356 | if (hlist_unhashed(&sb->s_instances)) { |
12ad3ab6 DC |
357 | spin_unlock(&sb_lock); |
358 | return false; | |
359 | } | |
360 | ||
361 | sb->s_count++; | |
362 | spin_unlock(&sb_lock); | |
363 | ||
364 | if (down_read_trylock(&sb->s_umount)) { | |
dabe0dc1 | 365 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
12ad3ab6 DC |
366 | return true; |
367 | up_read(&sb->s_umount); | |
368 | } | |
369 | ||
370 | put_super(sb); | |
371 | return false; | |
372 | } | |
373 | ||
1da177e4 LT |
374 | /** |
375 | * generic_shutdown_super - common helper for ->kill_sb() | |
376 | * @sb: superblock to kill | |
377 | * | |
378 | * generic_shutdown_super() does all fs-independent work on superblock | |
379 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects | |
380 | * that need destruction out of superblock, call generic_shutdown_super() | |
381 | * and release aforementioned objects. Note: dentries and inodes _are_ | |
382 | * taken care of and do not need specific handling. | |
c636ebdb DH |
383 | * |
384 | * Upon calling this function, the filesystem may no longer alter or | |
385 | * rearrange the set of dentries belonging to this super_block, nor may it | |
386 | * change the attachments of dentries to inodes. | |
1da177e4 LT |
387 | */ |
388 | void generic_shutdown_super(struct super_block *sb) | |
389 | { | |
ee9b6d61 | 390 | const struct super_operations *sop = sb->s_op; |
1da177e4 | 391 | |
c636ebdb DH |
392 | if (sb->s_root) { |
393 | shrink_dcache_for_umount(sb); | |
60b0680f | 394 | sync_filesystem(sb); |
1da177e4 | 395 | sb->s_flags &= ~MS_ACTIVE; |
efaee192 | 396 | |
63997e98 AV |
397 | fsnotify_unmount_inodes(&sb->s_inodes); |
398 | ||
399 | evict_inodes(sb); | |
1da177e4 | 400 | |
1da177e4 LT |
401 | if (sop->put_super) |
402 | sop->put_super(sb); | |
403 | ||
63997e98 | 404 | if (!list_empty(&sb->s_inodes)) { |
7b4fe29e DJ |
405 | printk("VFS: Busy inodes after unmount of %s. " |
406 | "Self-destruct in 5 seconds. Have a nice day...\n", | |
407 | sb->s_id); | |
1da177e4 | 408 | } |
1da177e4 LT |
409 | } |
410 | spin_lock(&sb_lock); | |
411 | /* should be initialized for __put_super_and_need_restart() */ | |
a5166169 | 412 | hlist_del_init(&sb->s_instances); |
1da177e4 LT |
413 | spin_unlock(&sb_lock); |
414 | up_write(&sb->s_umount); | |
415 | } | |
416 | ||
417 | EXPORT_SYMBOL(generic_shutdown_super); | |
418 | ||
419 | /** | |
420 | * sget - find or create a superblock | |
421 | * @type: filesystem type superblock should belong to | |
422 | * @test: comparison callback | |
423 | * @set: setup callback | |
9249e17f | 424 | * @flags: mount flags |
1da177e4 LT |
425 | * @data: argument to each of them |
426 | */ | |
427 | struct super_block *sget(struct file_system_type *type, | |
428 | int (*test)(struct super_block *,void *), | |
429 | int (*set)(struct super_block *,void *), | |
9249e17f | 430 | int flags, |
1da177e4 LT |
431 | void *data) |
432 | { | |
433 | struct super_block *s = NULL; | |
d4730127 | 434 | struct super_block *old; |
1da177e4 LT |
435 | int err; |
436 | ||
437 | retry: | |
438 | spin_lock(&sb_lock); | |
d4730127 | 439 | if (test) { |
b67bfe0d | 440 | hlist_for_each_entry(old, &type->fs_supers, s_instances) { |
d4730127 MK |
441 | if (!test(old, data)) |
442 | continue; | |
443 | if (!grab_super(old)) | |
444 | goto retry; | |
a3cfbb53 LZ |
445 | if (s) { |
446 | up_write(&s->s_umount); | |
d4730127 | 447 | destroy_super(s); |
7a4dec53 | 448 | s = NULL; |
a3cfbb53 | 449 | } |
d4730127 MK |
450 | return old; |
451 | } | |
1da177e4 LT |
452 | } |
453 | if (!s) { | |
454 | spin_unlock(&sb_lock); | |
9249e17f | 455 | s = alloc_super(type, flags); |
1da177e4 LT |
456 | if (!s) |
457 | return ERR_PTR(-ENOMEM); | |
458 | goto retry; | |
459 | } | |
460 | ||
461 | err = set(s, data); | |
462 | if (err) { | |
463 | spin_unlock(&sb_lock); | |
a3cfbb53 | 464 | up_write(&s->s_umount); |
1da177e4 LT |
465 | destroy_super(s); |
466 | return ERR_PTR(err); | |
467 | } | |
468 | s->s_type = type; | |
469 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); | |
470 | list_add_tail(&s->s_list, &super_blocks); | |
a5166169 | 471 | hlist_add_head(&s->s_instances, &type->fs_supers); |
1da177e4 LT |
472 | spin_unlock(&sb_lock); |
473 | get_filesystem(type); | |
b0d40c92 | 474 | register_shrinker(&s->s_shrink); |
1da177e4 LT |
475 | return s; |
476 | } | |
477 | ||
478 | EXPORT_SYMBOL(sget); | |
479 | ||
480 | void drop_super(struct super_block *sb) | |
481 | { | |
482 | up_read(&sb->s_umount); | |
483 | put_super(sb); | |
484 | } | |
485 | ||
486 | EXPORT_SYMBOL(drop_super); | |
487 | ||
01a05b33 AV |
488 | /** |
489 | * iterate_supers - call function for all active superblocks | |
490 | * @f: function to call | |
491 | * @arg: argument to pass to it | |
492 | * | |
493 | * Scans the superblock list and calls given function, passing it | |
494 | * locked superblock and given argument. | |
495 | */ | |
496 | void iterate_supers(void (*f)(struct super_block *, void *), void *arg) | |
497 | { | |
dca33252 | 498 | struct super_block *sb, *p = NULL; |
01a05b33 AV |
499 | |
500 | spin_lock(&sb_lock); | |
dca33252 | 501 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 502 | if (hlist_unhashed(&sb->s_instances)) |
01a05b33 AV |
503 | continue; |
504 | sb->s_count++; | |
505 | spin_unlock(&sb_lock); | |
506 | ||
507 | down_read(&sb->s_umount); | |
dabe0dc1 | 508 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
01a05b33 AV |
509 | f(sb, arg); |
510 | up_read(&sb->s_umount); | |
511 | ||
512 | spin_lock(&sb_lock); | |
dca33252 AV |
513 | if (p) |
514 | __put_super(p); | |
515 | p = sb; | |
01a05b33 | 516 | } |
dca33252 AV |
517 | if (p) |
518 | __put_super(p); | |
01a05b33 AV |
519 | spin_unlock(&sb_lock); |
520 | } | |
521 | ||
43e15cdb AV |
522 | /** |
523 | * iterate_supers_type - call function for superblocks of given type | |
524 | * @type: fs type | |
525 | * @f: function to call | |
526 | * @arg: argument to pass to it | |
527 | * | |
528 | * Scans the superblock list and calls given function, passing it | |
529 | * locked superblock and given argument. | |
530 | */ | |
531 | void iterate_supers_type(struct file_system_type *type, | |
532 | void (*f)(struct super_block *, void *), void *arg) | |
533 | { | |
534 | struct super_block *sb, *p = NULL; | |
535 | ||
536 | spin_lock(&sb_lock); | |
b67bfe0d | 537 | hlist_for_each_entry(sb, &type->fs_supers, s_instances) { |
43e15cdb AV |
538 | sb->s_count++; |
539 | spin_unlock(&sb_lock); | |
540 | ||
541 | down_read(&sb->s_umount); | |
dabe0dc1 | 542 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
43e15cdb AV |
543 | f(sb, arg); |
544 | up_read(&sb->s_umount); | |
545 | ||
546 | spin_lock(&sb_lock); | |
547 | if (p) | |
548 | __put_super(p); | |
549 | p = sb; | |
550 | } | |
551 | if (p) | |
552 | __put_super(p); | |
553 | spin_unlock(&sb_lock); | |
554 | } | |
555 | ||
556 | EXPORT_SYMBOL(iterate_supers_type); | |
557 | ||
1da177e4 LT |
558 | /** |
559 | * get_super - get the superblock of a device | |
560 | * @bdev: device to get the superblock for | |
561 | * | |
562 | * Scans the superblock list and finds the superblock of the file system | |
563 | * mounted on the device given. %NULL is returned if no match is found. | |
564 | */ | |
565 | ||
df40c01a | 566 | struct super_block *get_super(struct block_device *bdev) |
1da177e4 | 567 | { |
618f0636 KK |
568 | struct super_block *sb; |
569 | ||
1da177e4 LT |
570 | if (!bdev) |
571 | return NULL; | |
618f0636 | 572 | |
1da177e4 | 573 | spin_lock(&sb_lock); |
618f0636 KK |
574 | rescan: |
575 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 576 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 577 | continue; |
618f0636 KK |
578 | if (sb->s_bdev == bdev) { |
579 | sb->s_count++; | |
1da177e4 | 580 | spin_unlock(&sb_lock); |
618f0636 | 581 | down_read(&sb->s_umount); |
df40c01a | 582 | /* still alive? */ |
dabe0dc1 | 583 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
618f0636 KK |
584 | return sb; |
585 | up_read(&sb->s_umount); | |
df40c01a | 586 | /* nope, got unmounted */ |
618f0636 | 587 | spin_lock(&sb_lock); |
df40c01a AV |
588 | __put_super(sb); |
589 | goto rescan; | |
1da177e4 LT |
590 | } |
591 | } | |
592 | spin_unlock(&sb_lock); | |
593 | return NULL; | |
594 | } | |
595 | ||
596 | EXPORT_SYMBOL(get_super); | |
4504230a | 597 | |
6b6dc836 JK |
598 | /** |
599 | * get_super_thawed - get thawed superblock of a device | |
600 | * @bdev: device to get the superblock for | |
601 | * | |
602 | * Scans the superblock list and finds the superblock of the file system | |
603 | * mounted on the device. The superblock is returned once it is thawed | |
604 | * (or immediately if it was not frozen). %NULL is returned if no match | |
605 | * is found. | |
606 | */ | |
607 | struct super_block *get_super_thawed(struct block_device *bdev) | |
608 | { | |
609 | while (1) { | |
610 | struct super_block *s = get_super(bdev); | |
5accdf82 | 611 | if (!s || s->s_writers.frozen == SB_UNFROZEN) |
6b6dc836 JK |
612 | return s; |
613 | up_read(&s->s_umount); | |
5accdf82 JK |
614 | wait_event(s->s_writers.wait_unfrozen, |
615 | s->s_writers.frozen == SB_UNFROZEN); | |
6b6dc836 JK |
616 | put_super(s); |
617 | } | |
618 | } | |
619 | EXPORT_SYMBOL(get_super_thawed); | |
620 | ||
4504230a CH |
621 | /** |
622 | * get_active_super - get an active reference to the superblock of a device | |
623 | * @bdev: device to get the superblock for | |
624 | * | |
625 | * Scans the superblock list and finds the superblock of the file system | |
626 | * mounted on the device given. Returns the superblock with an active | |
d3f21473 | 627 | * reference or %NULL if none was found. |
4504230a CH |
628 | */ |
629 | struct super_block *get_active_super(struct block_device *bdev) | |
630 | { | |
631 | struct super_block *sb; | |
632 | ||
633 | if (!bdev) | |
634 | return NULL; | |
635 | ||
1494583d | 636 | restart: |
4504230a CH |
637 | spin_lock(&sb_lock); |
638 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 639 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 640 | continue; |
1494583d | 641 | if (sb->s_bdev == bdev) { |
d0f1a6e5 | 642 | if (!grab_super(sb)) |
1494583d | 643 | goto restart; |
d0f1a6e5 AV |
644 | up_write(&sb->s_umount); |
645 | return sb; | |
1494583d | 646 | } |
4504230a CH |
647 | } |
648 | spin_unlock(&sb_lock); | |
649 | return NULL; | |
650 | } | |
1da177e4 | 651 | |
df40c01a | 652 | struct super_block *user_get_super(dev_t dev) |
1da177e4 | 653 | { |
618f0636 | 654 | struct super_block *sb; |
1da177e4 | 655 | |
1da177e4 | 656 | spin_lock(&sb_lock); |
618f0636 KK |
657 | rescan: |
658 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 659 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 660 | continue; |
618f0636 KK |
661 | if (sb->s_dev == dev) { |
662 | sb->s_count++; | |
1da177e4 | 663 | spin_unlock(&sb_lock); |
618f0636 | 664 | down_read(&sb->s_umount); |
df40c01a | 665 | /* still alive? */ |
dabe0dc1 | 666 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
618f0636 KK |
667 | return sb; |
668 | up_read(&sb->s_umount); | |
df40c01a | 669 | /* nope, got unmounted */ |
618f0636 | 670 | spin_lock(&sb_lock); |
df40c01a AV |
671 | __put_super(sb); |
672 | goto rescan; | |
1da177e4 LT |
673 | } |
674 | } | |
675 | spin_unlock(&sb_lock); | |
676 | return NULL; | |
677 | } | |
678 | ||
1da177e4 LT |
679 | /** |
680 | * do_remount_sb - asks filesystem to change mount options. | |
681 | * @sb: superblock in question | |
682 | * @flags: numeric part of options | |
683 | * @data: the rest of options | |
684 | * @force: whether or not to force the change | |
685 | * | |
686 | * Alters the mount options of a mounted file system. | |
687 | */ | |
688 | int do_remount_sb(struct super_block *sb, int flags, void *data, int force) | |
689 | { | |
690 | int retval; | |
c79d967d | 691 | int remount_ro; |
4504230a | 692 | |
5accdf82 | 693 | if (sb->s_writers.frozen != SB_UNFROZEN) |
4504230a CH |
694 | return -EBUSY; |
695 | ||
9361401e | 696 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
697 | if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev)) |
698 | return -EACCES; | |
9361401e | 699 | #endif |
4504230a | 700 | |
1da177e4 LT |
701 | if (flags & MS_RDONLY) |
702 | acct_auto_close(sb); | |
703 | shrink_dcache_sb(sb); | |
60b0680f | 704 | sync_filesystem(sb); |
1da177e4 | 705 | |
d208bbdd | 706 | remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY); |
d208bbdd | 707 | |
1da177e4 LT |
708 | /* If we are remounting RDONLY and current sb is read/write, |
709 | make sure there are no rw files opened */ | |
d208bbdd | 710 | if (remount_ro) { |
4ed5e82f | 711 | if (force) { |
68c8a7ae AV |
712 | sb->s_readonly_remount = 1; |
713 | smp_wmb(); | |
4ed5e82f MS |
714 | } else { |
715 | retval = sb_prepare_remount_readonly(sb); | |
716 | if (retval) | |
717 | return retval; | |
4ed5e82f | 718 | } |
1da177e4 LT |
719 | } |
720 | ||
721 | if (sb->s_op->remount_fs) { | |
1da177e4 | 722 | retval = sb->s_op->remount_fs(sb, &flags, data); |
2833eb2b MS |
723 | if (retval) { |
724 | if (!force) | |
4ed5e82f | 725 | goto cancel_readonly; |
2833eb2b MS |
726 | /* If forced remount, go ahead despite any errors */ |
727 | WARN(1, "forced remount of a %s fs returned %i\n", | |
728 | sb->s_type->name, retval); | |
729 | } | |
1da177e4 LT |
730 | } |
731 | sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK); | |
4ed5e82f MS |
732 | /* Needs to be ordered wrt mnt_is_readonly() */ |
733 | smp_wmb(); | |
734 | sb->s_readonly_remount = 0; | |
c79d967d | 735 | |
d208bbdd NP |
736 | /* |
737 | * Some filesystems modify their metadata via some other path than the | |
738 | * bdev buffer cache (eg. use a private mapping, or directories in | |
739 | * pagecache, etc). Also file data modifications go via their own | |
740 | * mappings. So If we try to mount readonly then copy the filesystem | |
741 | * from bdev, we could get stale data, so invalidate it to give a best | |
742 | * effort at coherency. | |
743 | */ | |
744 | if (remount_ro && sb->s_bdev) | |
745 | invalidate_bdev(sb->s_bdev); | |
1da177e4 | 746 | return 0; |
4ed5e82f MS |
747 | |
748 | cancel_readonly: | |
749 | sb->s_readonly_remount = 0; | |
750 | return retval; | |
1da177e4 LT |
751 | } |
752 | ||
a2a9537a | 753 | static void do_emergency_remount(struct work_struct *work) |
1da177e4 | 754 | { |
dca33252 | 755 | struct super_block *sb, *p = NULL; |
1da177e4 LT |
756 | |
757 | spin_lock(&sb_lock); | |
dca33252 | 758 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 759 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 760 | continue; |
1da177e4 LT |
761 | sb->s_count++; |
762 | spin_unlock(&sb_lock); | |
443b94ba | 763 | down_write(&sb->s_umount); |
dabe0dc1 AV |
764 | if (sb->s_root && sb->s_bdev && (sb->s_flags & MS_BORN) && |
765 | !(sb->s_flags & MS_RDONLY)) { | |
1da177e4 | 766 | /* |
1da177e4 LT |
767 | * What lock protects sb->s_flags?? |
768 | */ | |
1da177e4 | 769 | do_remount_sb(sb, MS_RDONLY, NULL, 1); |
1da177e4 | 770 | } |
443b94ba | 771 | up_write(&sb->s_umount); |
1da177e4 | 772 | spin_lock(&sb_lock); |
dca33252 AV |
773 | if (p) |
774 | __put_super(p); | |
775 | p = sb; | |
1da177e4 | 776 | } |
dca33252 AV |
777 | if (p) |
778 | __put_super(p); | |
1da177e4 | 779 | spin_unlock(&sb_lock); |
a2a9537a | 780 | kfree(work); |
1da177e4 LT |
781 | printk("Emergency Remount complete\n"); |
782 | } | |
783 | ||
784 | void emergency_remount(void) | |
785 | { | |
a2a9537a JA |
786 | struct work_struct *work; |
787 | ||
788 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | |
789 | if (work) { | |
790 | INIT_WORK(work, do_emergency_remount); | |
791 | schedule_work(work); | |
792 | } | |
1da177e4 LT |
793 | } |
794 | ||
795 | /* | |
796 | * Unnamed block devices are dummy devices used by virtual | |
797 | * filesystems which don't use real block-devices. -- jrs | |
798 | */ | |
799 | ||
ad76cbc6 | 800 | static DEFINE_IDA(unnamed_dev_ida); |
1da177e4 | 801 | static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */ |
c63e09ec | 802 | static int unnamed_dev_start = 0; /* don't bother trying below it */ |
1da177e4 | 803 | |
0ee5dc67 | 804 | int get_anon_bdev(dev_t *p) |
1da177e4 LT |
805 | { |
806 | int dev; | |
807 | int error; | |
808 | ||
809 | retry: | |
ad76cbc6 | 810 | if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0) |
1da177e4 LT |
811 | return -ENOMEM; |
812 | spin_lock(&unnamed_dev_lock); | |
c63e09ec | 813 | error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev); |
f21f6220 AV |
814 | if (!error) |
815 | unnamed_dev_start = dev + 1; | |
1da177e4 LT |
816 | spin_unlock(&unnamed_dev_lock); |
817 | if (error == -EAGAIN) | |
818 | /* We raced and lost with another CPU. */ | |
819 | goto retry; | |
820 | else if (error) | |
821 | return -EAGAIN; | |
822 | ||
e8c8d1bc | 823 | if (dev == (1 << MINORBITS)) { |
1da177e4 | 824 | spin_lock(&unnamed_dev_lock); |
ad76cbc6 | 825 | ida_remove(&unnamed_dev_ida, dev); |
f21f6220 AV |
826 | if (unnamed_dev_start > dev) |
827 | unnamed_dev_start = dev; | |
1da177e4 LT |
828 | spin_unlock(&unnamed_dev_lock); |
829 | return -EMFILE; | |
830 | } | |
0ee5dc67 | 831 | *p = MKDEV(0, dev & MINORMASK); |
1da177e4 LT |
832 | return 0; |
833 | } | |
0ee5dc67 | 834 | EXPORT_SYMBOL(get_anon_bdev); |
1da177e4 | 835 | |
0ee5dc67 | 836 | void free_anon_bdev(dev_t dev) |
1da177e4 | 837 | { |
0ee5dc67 | 838 | int slot = MINOR(dev); |
1da177e4 | 839 | spin_lock(&unnamed_dev_lock); |
ad76cbc6 | 840 | ida_remove(&unnamed_dev_ida, slot); |
c63e09ec AV |
841 | if (slot < unnamed_dev_start) |
842 | unnamed_dev_start = slot; | |
1da177e4 LT |
843 | spin_unlock(&unnamed_dev_lock); |
844 | } | |
0ee5dc67 AV |
845 | EXPORT_SYMBOL(free_anon_bdev); |
846 | ||
847 | int set_anon_super(struct super_block *s, void *data) | |
848 | { | |
849 | int error = get_anon_bdev(&s->s_dev); | |
850 | if (!error) | |
851 | s->s_bdi = &noop_backing_dev_info; | |
852 | return error; | |
853 | } | |
854 | ||
855 | EXPORT_SYMBOL(set_anon_super); | |
856 | ||
857 | void kill_anon_super(struct super_block *sb) | |
858 | { | |
859 | dev_t dev = sb->s_dev; | |
860 | generic_shutdown_super(sb); | |
861 | free_anon_bdev(dev); | |
862 | } | |
1da177e4 LT |
863 | |
864 | EXPORT_SYMBOL(kill_anon_super); | |
865 | ||
1da177e4 LT |
866 | void kill_litter_super(struct super_block *sb) |
867 | { | |
868 | if (sb->s_root) | |
869 | d_genocide(sb->s_root); | |
870 | kill_anon_super(sb); | |
871 | } | |
872 | ||
873 | EXPORT_SYMBOL(kill_litter_super); | |
874 | ||
909e6d94 SH |
875 | static int ns_test_super(struct super_block *sb, void *data) |
876 | { | |
877 | return sb->s_fs_info == data; | |
878 | } | |
879 | ||
880 | static int ns_set_super(struct super_block *sb, void *data) | |
881 | { | |
882 | sb->s_fs_info = data; | |
883 | return set_anon_super(sb, NULL); | |
884 | } | |
885 | ||
ceefda69 AV |
886 | struct dentry *mount_ns(struct file_system_type *fs_type, int flags, |
887 | void *data, int (*fill_super)(struct super_block *, void *, int)) | |
909e6d94 SH |
888 | { |
889 | struct super_block *sb; | |
890 | ||
9249e17f | 891 | sb = sget(fs_type, ns_test_super, ns_set_super, flags, data); |
909e6d94 | 892 | if (IS_ERR(sb)) |
ceefda69 | 893 | return ERR_CAST(sb); |
909e6d94 SH |
894 | |
895 | if (!sb->s_root) { | |
896 | int err; | |
909e6d94 SH |
897 | err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0); |
898 | if (err) { | |
74dbbdd7 | 899 | deactivate_locked_super(sb); |
ceefda69 | 900 | return ERR_PTR(err); |
909e6d94 SH |
901 | } |
902 | ||
903 | sb->s_flags |= MS_ACTIVE; | |
904 | } | |
905 | ||
ceefda69 | 906 | return dget(sb->s_root); |
909e6d94 SH |
907 | } |
908 | ||
ceefda69 | 909 | EXPORT_SYMBOL(mount_ns); |
909e6d94 | 910 | |
9361401e | 911 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
912 | static int set_bdev_super(struct super_block *s, void *data) |
913 | { | |
914 | s->s_bdev = data; | |
915 | s->s_dev = s->s_bdev->bd_dev; | |
32a88aa1 JA |
916 | |
917 | /* | |
918 | * We set the bdi here to the queue backing, file systems can | |
919 | * overwrite this in ->fill_super() | |
920 | */ | |
921 | s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info; | |
1da177e4 LT |
922 | return 0; |
923 | } | |
924 | ||
925 | static int test_bdev_super(struct super_block *s, void *data) | |
926 | { | |
927 | return (void *)s->s_bdev == data; | |
928 | } | |
929 | ||
152a0836 | 930 | struct dentry *mount_bdev(struct file_system_type *fs_type, |
1da177e4 | 931 | int flags, const char *dev_name, void *data, |
152a0836 | 932 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
933 | { |
934 | struct block_device *bdev; | |
935 | struct super_block *s; | |
d4d77629 | 936 | fmode_t mode = FMODE_READ | FMODE_EXCL; |
1da177e4 LT |
937 | int error = 0; |
938 | ||
30c40d2c AV |
939 | if (!(flags & MS_RDONLY)) |
940 | mode |= FMODE_WRITE; | |
941 | ||
d4d77629 | 942 | bdev = blkdev_get_by_path(dev_name, mode, fs_type); |
1da177e4 | 943 | if (IS_ERR(bdev)) |
152a0836 | 944 | return ERR_CAST(bdev); |
1da177e4 LT |
945 | |
946 | /* | |
947 | * once the super is inserted into the list by sget, s_umount | |
948 | * will protect the lockfs code from trying to start a snapshot | |
949 | * while we are mounting | |
950 | */ | |
4fadd7bb CH |
951 | mutex_lock(&bdev->bd_fsfreeze_mutex); |
952 | if (bdev->bd_fsfreeze_count > 0) { | |
953 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
954 | error = -EBUSY; | |
955 | goto error_bdev; | |
956 | } | |
9249e17f DH |
957 | s = sget(fs_type, test_bdev_super, set_bdev_super, flags | MS_NOSEC, |
958 | bdev); | |
4fadd7bb | 959 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
1da177e4 | 960 | if (IS_ERR(s)) |
454e2398 | 961 | goto error_s; |
1da177e4 LT |
962 | |
963 | if (s->s_root) { | |
964 | if ((flags ^ s->s_flags) & MS_RDONLY) { | |
74dbbdd7 | 965 | deactivate_locked_super(s); |
454e2398 DH |
966 | error = -EBUSY; |
967 | goto error_bdev; | |
1da177e4 | 968 | } |
454e2398 | 969 | |
4f331f01 TH |
970 | /* |
971 | * s_umount nests inside bd_mutex during | |
e525fd89 TH |
972 | * __invalidate_device(). blkdev_put() acquires |
973 | * bd_mutex and can't be called under s_umount. Drop | |
974 | * s_umount temporarily. This is safe as we're | |
975 | * holding an active reference. | |
4f331f01 TH |
976 | */ |
977 | up_write(&s->s_umount); | |
d4d77629 | 978 | blkdev_put(bdev, mode); |
4f331f01 | 979 | down_write(&s->s_umount); |
1da177e4 LT |
980 | } else { |
981 | char b[BDEVNAME_SIZE]; | |
982 | ||
30c40d2c | 983 | s->s_mode = mode; |
1da177e4 | 984 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); |
e78c9a00 | 985 | sb_set_blocksize(s, block_size(bdev)); |
9b04c997 | 986 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 987 | if (error) { |
74dbbdd7 | 988 | deactivate_locked_super(s); |
454e2398 | 989 | goto error; |
fa675765 | 990 | } |
454e2398 DH |
991 | |
992 | s->s_flags |= MS_ACTIVE; | |
87d8fe1e | 993 | bdev->bd_super = s; |
1da177e4 LT |
994 | } |
995 | ||
152a0836 | 996 | return dget(s->s_root); |
1da177e4 | 997 | |
454e2398 DH |
998 | error_s: |
999 | error = PTR_ERR(s); | |
1000 | error_bdev: | |
d4d77629 | 1001 | blkdev_put(bdev, mode); |
454e2398 | 1002 | error: |
152a0836 AV |
1003 | return ERR_PTR(error); |
1004 | } | |
1005 | EXPORT_SYMBOL(mount_bdev); | |
1006 | ||
1da177e4 LT |
1007 | void kill_block_super(struct super_block *sb) |
1008 | { | |
1009 | struct block_device *bdev = sb->s_bdev; | |
30c40d2c | 1010 | fmode_t mode = sb->s_mode; |
1da177e4 | 1011 | |
ddbaaf30 | 1012 | bdev->bd_super = NULL; |
1da177e4 LT |
1013 | generic_shutdown_super(sb); |
1014 | sync_blockdev(bdev); | |
d4d77629 | 1015 | WARN_ON_ONCE(!(mode & FMODE_EXCL)); |
e525fd89 | 1016 | blkdev_put(bdev, mode | FMODE_EXCL); |
1da177e4 LT |
1017 | } |
1018 | ||
1019 | EXPORT_SYMBOL(kill_block_super); | |
9361401e | 1020 | #endif |
1da177e4 | 1021 | |
3c26ff6e | 1022 | struct dentry *mount_nodev(struct file_system_type *fs_type, |
1da177e4 | 1023 | int flags, void *data, |
3c26ff6e | 1024 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1025 | { |
1026 | int error; | |
9249e17f | 1027 | struct super_block *s = sget(fs_type, NULL, set_anon_super, flags, NULL); |
1da177e4 LT |
1028 | |
1029 | if (IS_ERR(s)) | |
3c26ff6e | 1030 | return ERR_CAST(s); |
1da177e4 | 1031 | |
9b04c997 | 1032 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1033 | if (error) { |
74dbbdd7 | 1034 | deactivate_locked_super(s); |
3c26ff6e | 1035 | return ERR_PTR(error); |
1da177e4 LT |
1036 | } |
1037 | s->s_flags |= MS_ACTIVE; | |
3c26ff6e | 1038 | return dget(s->s_root); |
1da177e4 | 1039 | } |
3c26ff6e AV |
1040 | EXPORT_SYMBOL(mount_nodev); |
1041 | ||
1da177e4 LT |
1042 | static int compare_single(struct super_block *s, void *p) |
1043 | { | |
1044 | return 1; | |
1045 | } | |
1046 | ||
fc14f2fe | 1047 | struct dentry *mount_single(struct file_system_type *fs_type, |
1da177e4 | 1048 | int flags, void *data, |
fc14f2fe | 1049 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1050 | { |
1051 | struct super_block *s; | |
1052 | int error; | |
1053 | ||
9249e17f | 1054 | s = sget(fs_type, compare_single, set_anon_super, flags, NULL); |
1da177e4 | 1055 | if (IS_ERR(s)) |
fc14f2fe | 1056 | return ERR_CAST(s); |
1da177e4 | 1057 | if (!s->s_root) { |
9b04c997 | 1058 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1059 | if (error) { |
74dbbdd7 | 1060 | deactivate_locked_super(s); |
fc14f2fe | 1061 | return ERR_PTR(error); |
1da177e4 LT |
1062 | } |
1063 | s->s_flags |= MS_ACTIVE; | |
9329d1be KS |
1064 | } else { |
1065 | do_remount_sb(s, flags, data, 0); | |
1da177e4 | 1066 | } |
fc14f2fe AV |
1067 | return dget(s->s_root); |
1068 | } | |
1069 | EXPORT_SYMBOL(mount_single); | |
1070 | ||
9d412a43 AV |
1071 | struct dentry * |
1072 | mount_fs(struct file_system_type *type, int flags, const char *name, void *data) | |
1da177e4 | 1073 | { |
c96e41e9 | 1074 | struct dentry *root; |
9d412a43 | 1075 | struct super_block *sb; |
1da177e4 | 1076 | char *secdata = NULL; |
9d412a43 | 1077 | int error = -ENOMEM; |
8089352a | 1078 | |
e0007529 | 1079 | if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) { |
1da177e4 | 1080 | secdata = alloc_secdata(); |
454e2398 | 1081 | if (!secdata) |
9d412a43 | 1082 | goto out; |
1da177e4 | 1083 | |
e0007529 | 1084 | error = security_sb_copy_data(data, secdata); |
454e2398 | 1085 | if (error) |
1da177e4 | 1086 | goto out_free_secdata; |
1da177e4 LT |
1087 | } |
1088 | ||
1a102ff9 AV |
1089 | root = type->mount(type, flags, name, data); |
1090 | if (IS_ERR(root)) { | |
1091 | error = PTR_ERR(root); | |
1092 | goto out_free_secdata; | |
c96e41e9 | 1093 | } |
9d412a43 AV |
1094 | sb = root->d_sb; |
1095 | BUG_ON(!sb); | |
1096 | WARN_ON(!sb->s_bdi); | |
6c510389 | 1097 | WARN_ON(sb->s_bdi == &default_backing_dev_info); |
9d412a43 | 1098 | sb->s_flags |= MS_BORN; |
454e2398 | 1099 | |
9d412a43 | 1100 | error = security_sb_kern_mount(sb, flags, secdata); |
5129a469 JE |
1101 | if (error) |
1102 | goto out_sb; | |
454e2398 | 1103 | |
42cb56ae JL |
1104 | /* |
1105 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE | |
1106 | * but s_maxbytes was an unsigned long long for many releases. Throw | |
1107 | * this warning for a little while to try and catch filesystems that | |
4358b567 | 1108 | * violate this rule. |
42cb56ae | 1109 | */ |
9d412a43 AV |
1110 | WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " |
1111 | "negative value (%lld)\n", type->name, sb->s_maxbytes); | |
42cb56ae | 1112 | |
9d412a43 | 1113 | up_write(&sb->s_umount); |
8680e22f | 1114 | free_secdata(secdata); |
9d412a43 | 1115 | return root; |
1da177e4 | 1116 | out_sb: |
9d412a43 AV |
1117 | dput(root); |
1118 | deactivate_locked_super(sb); | |
1da177e4 LT |
1119 | out_free_secdata: |
1120 | free_secdata(secdata); | |
1da177e4 | 1121 | out: |
454e2398 | 1122 | return ERR_PTR(error); |
1da177e4 LT |
1123 | } |
1124 | ||
5accdf82 JK |
1125 | /* |
1126 | * This is an internal function, please use sb_end_{write,pagefault,intwrite} | |
1127 | * instead. | |
1128 | */ | |
1129 | void __sb_end_write(struct super_block *sb, int level) | |
1130 | { | |
1131 | percpu_counter_dec(&sb->s_writers.counter[level-1]); | |
1132 | /* | |
1133 | * Make sure s_writers are updated before we wake up waiters in | |
1134 | * freeze_super(). | |
1135 | */ | |
1136 | smp_mb(); | |
1137 | if (waitqueue_active(&sb->s_writers.wait)) | |
1138 | wake_up(&sb->s_writers.wait); | |
6fa3eb70 | 1139 | lockdep_off(); |
5accdf82 | 1140 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _RET_IP_); |
6fa3eb70 | 1141 | lockdep_on(); |
5accdf82 JK |
1142 | } |
1143 | EXPORT_SYMBOL(__sb_end_write); | |
1144 | ||
1145 | #ifdef CONFIG_LOCKDEP | |
1146 | /* | |
1147 | * We want lockdep to tell us about possible deadlocks with freezing but | |
1148 | * it's it bit tricky to properly instrument it. Getting a freeze protection | |
1149 | * works as getting a read lock but there are subtle problems. XFS for example | |
1150 | * gets freeze protection on internal level twice in some cases, which is OK | |
1151 | * only because we already hold a freeze protection also on higher level. Due | |
1152 | * to these cases we have to tell lockdep we are doing trylock when we | |
1153 | * already hold a freeze protection for a higher freeze level. | |
1154 | */ | |
1155 | static void acquire_freeze_lock(struct super_block *sb, int level, bool trylock, | |
1156 | unsigned long ip) | |
1157 | { | |
1158 | int i; | |
1159 | ||
1160 | if (!trylock) { | |
1161 | for (i = 0; i < level - 1; i++) | |
1162 | if (lock_is_held(&sb->s_writers.lock_map[i])) { | |
1163 | trylock = true; | |
1164 | break; | |
1165 | } | |
1166 | } | |
6fa3eb70 | 1167 | lockdep_off(); |
5accdf82 | 1168 | rwsem_acquire_read(&sb->s_writers.lock_map[level-1], 0, trylock, ip); |
6fa3eb70 | 1169 | lockdep_on(); |
5accdf82 JK |
1170 | } |
1171 | #endif | |
1172 | ||
1173 | /* | |
1174 | * This is an internal function, please use sb_start_{write,pagefault,intwrite} | |
1175 | * instead. | |
1176 | */ | |
1177 | int __sb_start_write(struct super_block *sb, int level, bool wait) | |
1178 | { | |
1179 | retry: | |
1180 | if (unlikely(sb->s_writers.frozen >= level)) { | |
1181 | if (!wait) | |
1182 | return 0; | |
1183 | wait_event(sb->s_writers.wait_unfrozen, | |
1184 | sb->s_writers.frozen < level); | |
1185 | } | |
1186 | ||
1187 | #ifdef CONFIG_LOCKDEP | |
1188 | acquire_freeze_lock(sb, level, !wait, _RET_IP_); | |
1189 | #endif | |
1190 | percpu_counter_inc(&sb->s_writers.counter[level-1]); | |
1191 | /* | |
1192 | * Make sure counter is updated before we check for frozen. | |
1193 | * freeze_super() first sets frozen and then checks the counter. | |
1194 | */ | |
1195 | smp_mb(); | |
1196 | if (unlikely(sb->s_writers.frozen >= level)) { | |
1197 | __sb_end_write(sb, level); | |
1198 | goto retry; | |
1199 | } | |
1200 | return 1; | |
1201 | } | |
1202 | EXPORT_SYMBOL(__sb_start_write); | |
1203 | ||
1204 | /** | |
1205 | * sb_wait_write - wait until all writers to given file system finish | |
1206 | * @sb: the super for which we wait | |
1207 | * @level: type of writers we wait for (normal vs page fault) | |
1208 | * | |
1209 | * This function waits until there are no writers of given type to given file | |
1210 | * system. Caller of this function should make sure there can be no new writers | |
1211 | * of type @level before calling this function. Otherwise this function can | |
1212 | * livelock. | |
1213 | */ | |
1214 | static void sb_wait_write(struct super_block *sb, int level) | |
1215 | { | |
1216 | s64 writers; | |
1217 | ||
1218 | /* | |
1219 | * We just cycle-through lockdep here so that it does not complain | |
1220 | * about returning with lock to userspace | |
1221 | */ | |
1222 | rwsem_acquire(&sb->s_writers.lock_map[level-1], 0, 0, _THIS_IP_); | |
1223 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _THIS_IP_); | |
1224 | ||
1225 | do { | |
1226 | DEFINE_WAIT(wait); | |
1227 | ||
1228 | /* | |
1229 | * We use a barrier in prepare_to_wait() to separate setting | |
1230 | * of frozen and checking of the counter | |
1231 | */ | |
1232 | prepare_to_wait(&sb->s_writers.wait, &wait, | |
1233 | TASK_UNINTERRUPTIBLE); | |
1234 | ||
1235 | writers = percpu_counter_sum(&sb->s_writers.counter[level-1]); | |
1236 | if (writers) | |
1237 | schedule(); | |
1238 | ||
1239 | finish_wait(&sb->s_writers.wait, &wait); | |
1240 | } while (writers); | |
1241 | } | |
1242 | ||
18e9e510 | 1243 | /** |
7000d3c4 RD |
1244 | * freeze_super - lock the filesystem and force it into a consistent state |
1245 | * @sb: the super to lock | |
18e9e510 JB |
1246 | * |
1247 | * Syncs the super to make sure the filesystem is consistent and calls the fs's | |
1248 | * freeze_fs. Subsequent calls to this without first thawing the fs will return | |
1249 | * -EBUSY. | |
5accdf82 JK |
1250 | * |
1251 | * During this function, sb->s_writers.frozen goes through these values: | |
1252 | * | |
1253 | * SB_UNFROZEN: File system is normal, all writes progress as usual. | |
1254 | * | |
1255 | * SB_FREEZE_WRITE: The file system is in the process of being frozen. New | |
1256 | * writes should be blocked, though page faults are still allowed. We wait for | |
1257 | * all writes to complete and then proceed to the next stage. | |
1258 | * | |
1259 | * SB_FREEZE_PAGEFAULT: Freezing continues. Now also page faults are blocked | |
1260 | * but internal fs threads can still modify the filesystem (although they | |
1261 | * should not dirty new pages or inodes), writeback can run etc. After waiting | |
1262 | * for all running page faults we sync the filesystem which will clean all | |
1263 | * dirty pages and inodes (no new dirty pages or inodes can be created when | |
1264 | * sync is running). | |
1265 | * | |
1266 | * SB_FREEZE_FS: The file system is frozen. Now all internal sources of fs | |
1267 | * modification are blocked (e.g. XFS preallocation truncation on inode | |
1268 | * reclaim). This is usually implemented by blocking new transactions for | |
1269 | * filesystems that have them and need this additional guard. After all | |
1270 | * internal writers are finished we call ->freeze_fs() to finish filesystem | |
1271 | * freezing. Then we transition to SB_FREEZE_COMPLETE state. This state is | |
1272 | * mostly auxiliary for filesystems to verify they do not modify frozen fs. | |
1273 | * | |
1274 | * sb->s_writers.frozen is protected by sb->s_umount. | |
18e9e510 JB |
1275 | */ |
1276 | int freeze_super(struct super_block *sb) | |
1277 | { | |
1278 | int ret; | |
1279 | ||
1280 | atomic_inc(&sb->s_active); | |
1281 | down_write(&sb->s_umount); | |
5accdf82 | 1282 | if (sb->s_writers.frozen != SB_UNFROZEN) { |
18e9e510 JB |
1283 | deactivate_locked_super(sb); |
1284 | return -EBUSY; | |
1285 | } | |
1286 | ||
dabe0dc1 AV |
1287 | if (!(sb->s_flags & MS_BORN)) { |
1288 | up_write(&sb->s_umount); | |
1289 | return 0; /* sic - it's "nothing to do" */ | |
1290 | } | |
1291 | ||
18e9e510 | 1292 | if (sb->s_flags & MS_RDONLY) { |
5accdf82 JK |
1293 | /* Nothing to do really... */ |
1294 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
18e9e510 JB |
1295 | up_write(&sb->s_umount); |
1296 | return 0; | |
1297 | } | |
1298 | ||
5accdf82 JK |
1299 | /* From now on, no new normal writers can start */ |
1300 | sb->s_writers.frozen = SB_FREEZE_WRITE; | |
1301 | smp_wmb(); | |
1302 | ||
1303 | /* Release s_umount to preserve sb_start_write -> s_umount ordering */ | |
1304 | up_write(&sb->s_umount); | |
1305 | ||
1306 | sb_wait_write(sb, SB_FREEZE_WRITE); | |
1307 | ||
1308 | /* Now we go and block page faults... */ | |
1309 | down_write(&sb->s_umount); | |
1310 | sb->s_writers.frozen = SB_FREEZE_PAGEFAULT; | |
18e9e510 JB |
1311 | smp_wmb(); |
1312 | ||
5accdf82 JK |
1313 | sb_wait_write(sb, SB_FREEZE_PAGEFAULT); |
1314 | ||
1315 | /* All writers are done so after syncing there won't be dirty data */ | |
18e9e510 JB |
1316 | sync_filesystem(sb); |
1317 | ||
5accdf82 JK |
1318 | /* Now wait for internal filesystem counter */ |
1319 | sb->s_writers.frozen = SB_FREEZE_FS; | |
18e9e510 | 1320 | smp_wmb(); |
5accdf82 | 1321 | sb_wait_write(sb, SB_FREEZE_FS); |
18e9e510 | 1322 | |
18e9e510 JB |
1323 | if (sb->s_op->freeze_fs) { |
1324 | ret = sb->s_op->freeze_fs(sb); | |
1325 | if (ret) { | |
1326 | printk(KERN_ERR | |
1327 | "VFS:Filesystem freeze failed\n"); | |
5accdf82 | 1328 | sb->s_writers.frozen = SB_UNFROZEN; |
e1616300 | 1329 | smp_wmb(); |
5accdf82 | 1330 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 JB |
1331 | deactivate_locked_super(sb); |
1332 | return ret; | |
1333 | } | |
1334 | } | |
5accdf82 | 1335 | /* |
b24eac86 ON |
1336 | * For debugging purposes so that fs can warn if it sees write activity |
1337 | * when frozen is set to SB_FREEZE_COMPLETE, and for thaw_super(). | |
5accdf82 JK |
1338 | */ |
1339 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
18e9e510 JB |
1340 | up_write(&sb->s_umount); |
1341 | return 0; | |
1342 | } | |
1343 | EXPORT_SYMBOL(freeze_super); | |
1344 | ||
1345 | /** | |
1346 | * thaw_super -- unlock filesystem | |
1347 | * @sb: the super to thaw | |
1348 | * | |
1349 | * Unlocks the filesystem and marks it writeable again after freeze_super(). | |
1350 | */ | |
1351 | int thaw_super(struct super_block *sb) | |
1352 | { | |
1353 | int error; | |
1354 | ||
1355 | down_write(&sb->s_umount); | |
b24eac86 | 1356 | if (sb->s_writers.frozen != SB_FREEZE_COMPLETE) { |
18e9e510 JB |
1357 | up_write(&sb->s_umount); |
1358 | return -EINVAL; | |
1359 | } | |
1360 | ||
1361 | if (sb->s_flags & MS_RDONLY) | |
1362 | goto out; | |
1363 | ||
1364 | if (sb->s_op->unfreeze_fs) { | |
1365 | error = sb->s_op->unfreeze_fs(sb); | |
1366 | if (error) { | |
1367 | printk(KERN_ERR | |
1368 | "VFS:Filesystem thaw failed\n"); | |
18e9e510 JB |
1369 | up_write(&sb->s_umount); |
1370 | return error; | |
1371 | } | |
1372 | } | |
1373 | ||
1374 | out: | |
5accdf82 | 1375 | sb->s_writers.frozen = SB_UNFROZEN; |
18e9e510 | 1376 | smp_wmb(); |
5accdf82 | 1377 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 JB |
1378 | deactivate_locked_super(sb); |
1379 | ||
1380 | return 0; | |
1381 | } | |
1382 | EXPORT_SYMBOL(thaw_super); |