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 LT |
40 | LIST_HEAD(super_blocks); |
41 | DEFINE_SPINLOCK(sb_lock); | |
42 | ||
5accdf82 JK |
43 | static char *sb_writers_name[SB_FREEZE_LEVELS] = { |
44 | "sb_writers", | |
45 | "sb_pagefaults", | |
46 | "sb_internal", | |
47 | }; | |
48 | ||
b0d40c92 DC |
49 | /* |
50 | * One thing we have to be careful of with a per-sb shrinker is that we don't | |
51 | * drop the last active reference to the superblock from within the shrinker. | |
52 | * If that happens we could trigger unregistering the shrinker from within the | |
53 | * shrinker path and that leads to deadlock on the shrinker_rwsem. Hence we | |
54 | * take a passive reference to the superblock to avoid this from occurring. | |
55 | */ | |
56 | static int prune_super(struct shrinker *shrink, struct shrink_control *sc) | |
57 | { | |
58 | struct super_block *sb; | |
0e1fdafd DC |
59 | int fs_objects = 0; |
60 | int total_objects; | |
b0d40c92 DC |
61 | |
62 | sb = container_of(shrink, struct super_block, s_shrink); | |
63 | ||
64 | /* | |
65 | * Deadlock avoidance. We may hold various FS locks, and we don't want | |
66 | * to recurse into the FS that called us in clear_inode() and friends.. | |
67 | */ | |
68 | if (sc->nr_to_scan && !(sc->gfp_mask & __GFP_FS)) | |
69 | return -1; | |
70 | ||
71 | if (!grab_super_passive(sb)) | |
8e125cd8 | 72 | return -1; |
b0d40c92 | 73 | |
0e1fdafd DC |
74 | if (sb->s_op && sb->s_op->nr_cached_objects) |
75 | fs_objects = sb->s_op->nr_cached_objects(sb); | |
76 | ||
77 | total_objects = sb->s_nr_dentry_unused + | |
78 | sb->s_nr_inodes_unused + fs_objects + 1; | |
d016a08a TH |
79 | if (!total_objects) |
80 | total_objects = 1; | |
0e1fdafd | 81 | |
b0d40c92 | 82 | if (sc->nr_to_scan) { |
0e1fdafd DC |
83 | int dentries; |
84 | int inodes; | |
b0d40c92 | 85 | |
0e1fdafd DC |
86 | /* proportion the scan between the caches */ |
87 | dentries = (sc->nr_to_scan * sb->s_nr_dentry_unused) / | |
88 | total_objects; | |
89 | inodes = (sc->nr_to_scan * sb->s_nr_inodes_unused) / | |
90 | total_objects; | |
91 | if (fs_objects) | |
92 | fs_objects = (sc->nr_to_scan * fs_objects) / | |
93 | total_objects; | |
94 | /* | |
95 | * prune the dcache first as the icache is pinned by it, then | |
96 | * prune the icache, followed by the filesystem specific caches | |
97 | */ | |
98 | prune_dcache_sb(sb, dentries); | |
99 | prune_icache_sb(sb, inodes); | |
b0d40c92 | 100 | |
0e1fdafd DC |
101 | if (fs_objects && sb->s_op->free_cached_objects) { |
102 | sb->s_op->free_cached_objects(sb, fs_objects); | |
103 | fs_objects = sb->s_op->nr_cached_objects(sb); | |
104 | } | |
105 | total_objects = sb->s_nr_dentry_unused + | |
106 | sb->s_nr_inodes_unused + fs_objects; | |
b0d40c92 DC |
107 | } |
108 | ||
0e1fdafd | 109 | total_objects = (total_objects / 100) * sysctl_vfs_cache_pressure; |
b0d40c92 | 110 | drop_super(sb); |
0e1fdafd | 111 | return total_objects; |
b0d40c92 DC |
112 | } |
113 | ||
5accdf82 JK |
114 | static int init_sb_writers(struct super_block *s, struct file_system_type *type) |
115 | { | |
116 | int err; | |
117 | int i; | |
118 | ||
119 | for (i = 0; i < SB_FREEZE_LEVELS; i++) { | |
120 | err = percpu_counter_init(&s->s_writers.counter[i], 0); | |
121 | if (err < 0) | |
122 | goto err_out; | |
123 | lockdep_init_map(&s->s_writers.lock_map[i], sb_writers_name[i], | |
124 | &type->s_writers_key[i], 0); | |
125 | } | |
126 | init_waitqueue_head(&s->s_writers.wait); | |
127 | init_waitqueue_head(&s->s_writers.wait_unfrozen); | |
128 | return 0; | |
129 | err_out: | |
130 | while (--i >= 0) | |
131 | percpu_counter_destroy(&s->s_writers.counter[i]); | |
132 | return err; | |
133 | } | |
134 | ||
135 | static void destroy_sb_writers(struct super_block *s) | |
136 | { | |
137 | int i; | |
138 | ||
139 | for (i = 0; i < SB_FREEZE_LEVELS; i++) | |
140 | percpu_counter_destroy(&s->s_writers.counter[i]); | |
141 | } | |
142 | ||
1da177e4 LT |
143 | /** |
144 | * alloc_super - create new superblock | |
fe2bbc48 | 145 | * @type: filesystem type superblock should belong to |
9249e17f | 146 | * @flags: the mount flags |
1da177e4 LT |
147 | * |
148 | * Allocates and initializes a new &struct super_block. alloc_super() | |
149 | * returns a pointer new superblock or %NULL if allocation had failed. | |
150 | */ | |
9249e17f | 151 | static struct super_block *alloc_super(struct file_system_type *type, int flags) |
1da177e4 | 152 | { |
11b0b5ab | 153 | struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER); |
b87221de | 154 | static const struct super_operations default_op; |
1da177e4 LT |
155 | |
156 | if (s) { | |
1da177e4 | 157 | if (security_sb_alloc(s)) { |
5accdf82 JK |
158 | /* |
159 | * We cannot call security_sb_free() without | |
160 | * security_sb_alloc() succeeding. So bail out manually | |
161 | */ | |
1da177e4 LT |
162 | kfree(s); |
163 | s = NULL; | |
164 | goto out; | |
165 | } | |
5accdf82 JK |
166 | if (init_sb_writers(s, type)) |
167 | goto err_out; | |
9249e17f | 168 | s->s_flags = flags; |
95f28604 | 169 | s->s_bdi = &default_backing_dev_info; |
a5166169 | 170 | INIT_HLIST_NODE(&s->s_instances); |
ceb5bdc2 | 171 | INIT_HLIST_BL_HEAD(&s->s_anon); |
1da177e4 | 172 | INIT_LIST_HEAD(&s->s_inodes); |
da3bbdd4 | 173 | INIT_LIST_HEAD(&s->s_dentry_lru); |
98b745c6 | 174 | INIT_LIST_HEAD(&s->s_inode_lru); |
09cc9fc7 | 175 | spin_lock_init(&s->s_inode_lru_lock); |
39f7c4db | 176 | INIT_LIST_HEAD(&s->s_mounts); |
1da177e4 | 177 | init_rwsem(&s->s_umount); |
897c6ff9 | 178 | lockdep_set_class(&s->s_umount, &type->s_umount_key); |
ada723dc PZ |
179 | /* |
180 | * sget() can have s_umount recursion. | |
181 | * | |
182 | * When it cannot find a suitable sb, it allocates a new | |
183 | * one (this one), and tries again to find a suitable old | |
184 | * one. | |
185 | * | |
186 | * In case that succeeds, it will acquire the s_umount | |
187 | * lock of the old one. Since these are clearly distrinct | |
188 | * locks, and this object isn't exposed yet, there's no | |
189 | * risk of deadlocks. | |
190 | * | |
191 | * Annotate this by putting this lock in a different | |
192 | * subclass. | |
193 | */ | |
194 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); | |
b20bd1a5 | 195 | s->s_count = 1; |
1da177e4 | 196 | atomic_set(&s->s_active, 1); |
a11f3a05 | 197 | mutex_init(&s->s_vfs_rename_mutex); |
51ee049e | 198 | lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key); |
d3be915f IM |
199 | mutex_init(&s->s_dquot.dqio_mutex); |
200 | mutex_init(&s->s_dquot.dqonoff_mutex); | |
1da177e4 | 201 | init_rwsem(&s->s_dquot.dqptr_sem); |
1da177e4 | 202 | s->s_maxbytes = MAX_NON_LFS; |
1da177e4 LT |
203 | s->s_op = &default_op; |
204 | s->s_time_gran = 1000000000; | |
c515e1fd | 205 | s->cleancache_poolid = -1; |
b0d40c92 DC |
206 | |
207 | s->s_shrink.seeks = DEFAULT_SEEKS; | |
208 | s->s_shrink.shrink = prune_super; | |
8ab47664 | 209 | s->s_shrink.batch = 1024; |
1da177e4 LT |
210 | } |
211 | out: | |
212 | return s; | |
5accdf82 JK |
213 | err_out: |
214 | security_sb_free(s); | |
5accdf82 JK |
215 | destroy_sb_writers(s); |
216 | kfree(s); | |
217 | s = NULL; | |
218 | goto out; | |
1da177e4 LT |
219 | } |
220 | ||
221 | /** | |
222 | * destroy_super - frees a superblock | |
223 | * @s: superblock to free | |
224 | * | |
225 | * Frees a superblock. | |
226 | */ | |
227 | static inline void destroy_super(struct super_block *s) | |
228 | { | |
5accdf82 | 229 | destroy_sb_writers(s); |
1da177e4 | 230 | security_sb_free(s); |
39f7c4db | 231 | WARN_ON(!list_empty(&s->s_mounts)); |
79c0b2df | 232 | kfree(s->s_subtype); |
b3b304a2 | 233 | kfree(s->s_options); |
1da177e4 LT |
234 | kfree(s); |
235 | } | |
236 | ||
237 | /* Superblock refcounting */ | |
238 | ||
239 | /* | |
35cf7ba0 | 240 | * Drop a superblock's refcount. The caller must hold sb_lock. |
1da177e4 | 241 | */ |
f47ec3f2 | 242 | static void __put_super(struct super_block *sb) |
1da177e4 | 243 | { |
1da177e4 | 244 | if (!--sb->s_count) { |
551de6f3 | 245 | list_del_init(&sb->s_list); |
1da177e4 | 246 | destroy_super(sb); |
1da177e4 | 247 | } |
1da177e4 LT |
248 | } |
249 | ||
250 | /** | |
251 | * put_super - drop a temporary reference to superblock | |
252 | * @sb: superblock in question | |
253 | * | |
254 | * Drops a temporary reference, frees superblock if there's no | |
255 | * references left. | |
256 | */ | |
f47ec3f2 | 257 | static void put_super(struct super_block *sb) |
1da177e4 LT |
258 | { |
259 | spin_lock(&sb_lock); | |
260 | __put_super(sb); | |
261 | spin_unlock(&sb_lock); | |
262 | } | |
263 | ||
264 | ||
265 | /** | |
1712ac8f | 266 | * deactivate_locked_super - drop an active reference to superblock |
1da177e4 LT |
267 | * @s: superblock to deactivate |
268 | * | |
1712ac8f AV |
269 | * Drops an active reference to superblock, converting it into a temprory |
270 | * one if there is no other active references left. In that case we | |
1da177e4 LT |
271 | * tell fs driver to shut it down and drop the temporary reference we |
272 | * had just acquired. | |
1712ac8f AV |
273 | * |
274 | * Caller holds exclusive lock on superblock; that lock is released. | |
1da177e4 | 275 | */ |
1712ac8f | 276 | void deactivate_locked_super(struct super_block *s) |
1da177e4 LT |
277 | { |
278 | struct file_system_type *fs = s->s_type; | |
b20bd1a5 | 279 | if (atomic_dec_and_test(&s->s_active)) { |
3167760f | 280 | cleancache_invalidate_fs(s); |
1da177e4 | 281 | fs->kill_sb(s); |
b0d40c92 DC |
282 | |
283 | /* caches are now gone, we can safely kill the shrinker now */ | |
284 | unregister_shrinker(&s->s_shrink); | |
1da177e4 LT |
285 | put_filesystem(fs); |
286 | put_super(s); | |
1712ac8f AV |
287 | } else { |
288 | up_write(&s->s_umount); | |
1da177e4 LT |
289 | } |
290 | } | |
291 | ||
1712ac8f | 292 | EXPORT_SYMBOL(deactivate_locked_super); |
1da177e4 | 293 | |
74dbbdd7 | 294 | /** |
1712ac8f | 295 | * deactivate_super - drop an active reference to superblock |
74dbbdd7 AV |
296 | * @s: superblock to deactivate |
297 | * | |
1712ac8f AV |
298 | * Variant of deactivate_locked_super(), except that superblock is *not* |
299 | * locked by caller. If we are going to drop the final active reference, | |
300 | * lock will be acquired prior to that. | |
74dbbdd7 | 301 | */ |
1712ac8f | 302 | void deactivate_super(struct super_block *s) |
74dbbdd7 | 303 | { |
1712ac8f AV |
304 | if (!atomic_add_unless(&s->s_active, -1, 1)) { |
305 | down_write(&s->s_umount); | |
306 | deactivate_locked_super(s); | |
74dbbdd7 AV |
307 | } |
308 | } | |
309 | ||
1712ac8f | 310 | EXPORT_SYMBOL(deactivate_super); |
74dbbdd7 | 311 | |
1da177e4 LT |
312 | /** |
313 | * grab_super - acquire an active reference | |
314 | * @s: reference we are trying to make active | |
315 | * | |
316 | * Tries to acquire an active reference. grab_super() is used when we | |
317 | * had just found a superblock in super_blocks or fs_type->fs_supers | |
318 | * and want to turn it into a full-blown active reference. grab_super() | |
319 | * is called with sb_lock held and drops it. Returns 1 in case of | |
320 | * success, 0 if we had failed (superblock contents was already dead or | |
d0f1a6e5 AV |
321 | * dying when grab_super() had been called). Note that this is only |
322 | * called for superblocks not in rundown mode (== ones still on ->fs_supers | |
323 | * of their type), so increment of ->s_count is OK here. | |
1da177e4 | 324 | */ |
9c4dbee7 | 325 | static int grab_super(struct super_block *s) __releases(sb_lock) |
1da177e4 LT |
326 | { |
327 | s->s_count++; | |
328 | spin_unlock(&sb_lock); | |
329 | down_write(&s->s_umount); | |
d0f1a6e5 AV |
330 | if ((s->s_flags & MS_BORN) && atomic_inc_not_zero(&s->s_active)) { |
331 | put_super(s); | |
332 | return 1; | |
333 | } | |
1da177e4 LT |
334 | up_write(&s->s_umount); |
335 | put_super(s); | |
1da177e4 LT |
336 | return 0; |
337 | } | |
338 | ||
12ad3ab6 DC |
339 | /* |
340 | * grab_super_passive - acquire a passive reference | |
331cbdee | 341 | * @sb: reference we are trying to grab |
12ad3ab6 DC |
342 | * |
343 | * Tries to acquire a passive reference. This is used in places where we | |
344 | * cannot take an active reference but we need to ensure that the | |
345 | * superblock does not go away while we are working on it. It returns | |
346 | * false if a reference was not gained, and returns true with the s_umount | |
347 | * lock held in read mode if a reference is gained. On successful return, | |
348 | * the caller must drop the s_umount lock and the passive reference when | |
349 | * done. | |
350 | */ | |
351 | bool grab_super_passive(struct super_block *sb) | |
352 | { | |
353 | spin_lock(&sb_lock); | |
a5166169 | 354 | if (hlist_unhashed(&sb->s_instances)) { |
12ad3ab6 DC |
355 | spin_unlock(&sb_lock); |
356 | return false; | |
357 | } | |
358 | ||
359 | sb->s_count++; | |
360 | spin_unlock(&sb_lock); | |
361 | ||
362 | if (down_read_trylock(&sb->s_umount)) { | |
dabe0dc1 | 363 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
12ad3ab6 DC |
364 | return true; |
365 | up_read(&sb->s_umount); | |
366 | } | |
367 | ||
368 | put_super(sb); | |
369 | return false; | |
370 | } | |
371 | ||
1da177e4 LT |
372 | /** |
373 | * generic_shutdown_super - common helper for ->kill_sb() | |
374 | * @sb: superblock to kill | |
375 | * | |
376 | * generic_shutdown_super() does all fs-independent work on superblock | |
377 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects | |
378 | * that need destruction out of superblock, call generic_shutdown_super() | |
379 | * and release aforementioned objects. Note: dentries and inodes _are_ | |
380 | * taken care of and do not need specific handling. | |
c636ebdb DH |
381 | * |
382 | * Upon calling this function, the filesystem may no longer alter or | |
383 | * rearrange the set of dentries belonging to this super_block, nor may it | |
384 | * change the attachments of dentries to inodes. | |
1da177e4 LT |
385 | */ |
386 | void generic_shutdown_super(struct super_block *sb) | |
387 | { | |
ee9b6d61 | 388 | const struct super_operations *sop = sb->s_op; |
1da177e4 | 389 | |
c636ebdb DH |
390 | if (sb->s_root) { |
391 | shrink_dcache_for_umount(sb); | |
60b0680f | 392 | sync_filesystem(sb); |
1da177e4 | 393 | sb->s_flags &= ~MS_ACTIVE; |
efaee192 | 394 | |
63997e98 AV |
395 | fsnotify_unmount_inodes(&sb->s_inodes); |
396 | ||
397 | evict_inodes(sb); | |
1da177e4 | 398 | |
1da177e4 LT |
399 | if (sop->put_super) |
400 | sop->put_super(sb); | |
401 | ||
63997e98 | 402 | if (!list_empty(&sb->s_inodes)) { |
7b4fe29e DJ |
403 | printk("VFS: Busy inodes after unmount of %s. " |
404 | "Self-destruct in 5 seconds. Have a nice day...\n", | |
405 | sb->s_id); | |
1da177e4 | 406 | } |
1da177e4 LT |
407 | } |
408 | spin_lock(&sb_lock); | |
409 | /* should be initialized for __put_super_and_need_restart() */ | |
a5166169 | 410 | hlist_del_init(&sb->s_instances); |
1da177e4 LT |
411 | spin_unlock(&sb_lock); |
412 | up_write(&sb->s_umount); | |
413 | } | |
414 | ||
415 | EXPORT_SYMBOL(generic_shutdown_super); | |
416 | ||
417 | /** | |
418 | * sget - find or create a superblock | |
419 | * @type: filesystem type superblock should belong to | |
420 | * @test: comparison callback | |
421 | * @set: setup callback | |
9249e17f | 422 | * @flags: mount flags |
1da177e4 LT |
423 | * @data: argument to each of them |
424 | */ | |
425 | struct super_block *sget(struct file_system_type *type, | |
426 | int (*test)(struct super_block *,void *), | |
427 | int (*set)(struct super_block *,void *), | |
9249e17f | 428 | int flags, |
1da177e4 LT |
429 | void *data) |
430 | { | |
431 | struct super_block *s = NULL; | |
d4730127 | 432 | struct super_block *old; |
1da177e4 LT |
433 | int err; |
434 | ||
435 | retry: | |
436 | spin_lock(&sb_lock); | |
d4730127 | 437 | if (test) { |
b67bfe0d | 438 | hlist_for_each_entry(old, &type->fs_supers, s_instances) { |
d4730127 MK |
439 | if (!test(old, data)) |
440 | continue; | |
441 | if (!grab_super(old)) | |
442 | goto retry; | |
a3cfbb53 LZ |
443 | if (s) { |
444 | up_write(&s->s_umount); | |
d4730127 | 445 | destroy_super(s); |
7a4dec53 | 446 | s = NULL; |
a3cfbb53 | 447 | } |
d4730127 MK |
448 | return old; |
449 | } | |
1da177e4 LT |
450 | } |
451 | if (!s) { | |
452 | spin_unlock(&sb_lock); | |
9249e17f | 453 | s = alloc_super(type, flags); |
1da177e4 LT |
454 | if (!s) |
455 | return ERR_PTR(-ENOMEM); | |
456 | goto retry; | |
457 | } | |
458 | ||
459 | err = set(s, data); | |
460 | if (err) { | |
461 | spin_unlock(&sb_lock); | |
a3cfbb53 | 462 | up_write(&s->s_umount); |
1da177e4 LT |
463 | destroy_super(s); |
464 | return ERR_PTR(err); | |
465 | } | |
466 | s->s_type = type; | |
467 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); | |
468 | list_add_tail(&s->s_list, &super_blocks); | |
a5166169 | 469 | hlist_add_head(&s->s_instances, &type->fs_supers); |
1da177e4 LT |
470 | spin_unlock(&sb_lock); |
471 | get_filesystem(type); | |
b0d40c92 | 472 | register_shrinker(&s->s_shrink); |
1da177e4 LT |
473 | return s; |
474 | } | |
475 | ||
476 | EXPORT_SYMBOL(sget); | |
477 | ||
478 | void drop_super(struct super_block *sb) | |
479 | { | |
480 | up_read(&sb->s_umount); | |
481 | put_super(sb); | |
482 | } | |
483 | ||
484 | EXPORT_SYMBOL(drop_super); | |
485 | ||
01a05b33 AV |
486 | /** |
487 | * iterate_supers - call function for all active superblocks | |
488 | * @f: function to call | |
489 | * @arg: argument to pass to it | |
490 | * | |
491 | * Scans the superblock list and calls given function, passing it | |
492 | * locked superblock and given argument. | |
493 | */ | |
494 | void iterate_supers(void (*f)(struct super_block *, void *), void *arg) | |
495 | { | |
dca33252 | 496 | struct super_block *sb, *p = NULL; |
01a05b33 AV |
497 | |
498 | spin_lock(&sb_lock); | |
dca33252 | 499 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 500 | if (hlist_unhashed(&sb->s_instances)) |
01a05b33 AV |
501 | continue; |
502 | sb->s_count++; | |
503 | spin_unlock(&sb_lock); | |
504 | ||
505 | down_read(&sb->s_umount); | |
dabe0dc1 | 506 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
01a05b33 AV |
507 | f(sb, arg); |
508 | up_read(&sb->s_umount); | |
509 | ||
510 | spin_lock(&sb_lock); | |
dca33252 AV |
511 | if (p) |
512 | __put_super(p); | |
513 | p = sb; | |
01a05b33 | 514 | } |
dca33252 AV |
515 | if (p) |
516 | __put_super(p); | |
01a05b33 AV |
517 | spin_unlock(&sb_lock); |
518 | } | |
519 | ||
43e15cdb AV |
520 | /** |
521 | * iterate_supers_type - call function for superblocks of given type | |
522 | * @type: fs type | |
523 | * @f: function to call | |
524 | * @arg: argument to pass to it | |
525 | * | |
526 | * Scans the superblock list and calls given function, passing it | |
527 | * locked superblock and given argument. | |
528 | */ | |
529 | void iterate_supers_type(struct file_system_type *type, | |
530 | void (*f)(struct super_block *, void *), void *arg) | |
531 | { | |
532 | struct super_block *sb, *p = NULL; | |
533 | ||
534 | spin_lock(&sb_lock); | |
b67bfe0d | 535 | hlist_for_each_entry(sb, &type->fs_supers, s_instances) { |
43e15cdb AV |
536 | sb->s_count++; |
537 | spin_unlock(&sb_lock); | |
538 | ||
539 | down_read(&sb->s_umount); | |
dabe0dc1 | 540 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
43e15cdb AV |
541 | f(sb, arg); |
542 | up_read(&sb->s_umount); | |
543 | ||
544 | spin_lock(&sb_lock); | |
545 | if (p) | |
546 | __put_super(p); | |
547 | p = sb; | |
548 | } | |
549 | if (p) | |
550 | __put_super(p); | |
551 | spin_unlock(&sb_lock); | |
552 | } | |
553 | ||
554 | EXPORT_SYMBOL(iterate_supers_type); | |
555 | ||
1da177e4 LT |
556 | /** |
557 | * get_super - get the superblock of a device | |
558 | * @bdev: device to get the superblock for | |
559 | * | |
560 | * Scans the superblock list and finds the superblock of the file system | |
561 | * mounted on the device given. %NULL is returned if no match is found. | |
562 | */ | |
563 | ||
df40c01a | 564 | struct super_block *get_super(struct block_device *bdev) |
1da177e4 | 565 | { |
618f0636 KK |
566 | struct super_block *sb; |
567 | ||
1da177e4 LT |
568 | if (!bdev) |
569 | return NULL; | |
618f0636 | 570 | |
1da177e4 | 571 | spin_lock(&sb_lock); |
618f0636 KK |
572 | rescan: |
573 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 574 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 575 | continue; |
618f0636 KK |
576 | if (sb->s_bdev == bdev) { |
577 | sb->s_count++; | |
1da177e4 | 578 | spin_unlock(&sb_lock); |
618f0636 | 579 | down_read(&sb->s_umount); |
df40c01a | 580 | /* still alive? */ |
dabe0dc1 | 581 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
618f0636 KK |
582 | return sb; |
583 | up_read(&sb->s_umount); | |
df40c01a | 584 | /* nope, got unmounted */ |
618f0636 | 585 | spin_lock(&sb_lock); |
df40c01a AV |
586 | __put_super(sb); |
587 | goto rescan; | |
1da177e4 LT |
588 | } |
589 | } | |
590 | spin_unlock(&sb_lock); | |
591 | return NULL; | |
592 | } | |
593 | ||
594 | EXPORT_SYMBOL(get_super); | |
4504230a | 595 | |
6b6dc836 JK |
596 | /** |
597 | * get_super_thawed - get thawed superblock of a device | |
598 | * @bdev: device to get the superblock for | |
599 | * | |
600 | * Scans the superblock list and finds the superblock of the file system | |
601 | * mounted on the device. The superblock is returned once it is thawed | |
602 | * (or immediately if it was not frozen). %NULL is returned if no match | |
603 | * is found. | |
604 | */ | |
605 | struct super_block *get_super_thawed(struct block_device *bdev) | |
606 | { | |
607 | while (1) { | |
608 | struct super_block *s = get_super(bdev); | |
5accdf82 | 609 | if (!s || s->s_writers.frozen == SB_UNFROZEN) |
6b6dc836 JK |
610 | return s; |
611 | up_read(&s->s_umount); | |
5accdf82 JK |
612 | wait_event(s->s_writers.wait_unfrozen, |
613 | s->s_writers.frozen == SB_UNFROZEN); | |
6b6dc836 JK |
614 | put_super(s); |
615 | } | |
616 | } | |
617 | EXPORT_SYMBOL(get_super_thawed); | |
618 | ||
4504230a CH |
619 | /** |
620 | * get_active_super - get an active reference to the superblock of a device | |
621 | * @bdev: device to get the superblock for | |
622 | * | |
623 | * Scans the superblock list and finds the superblock of the file system | |
624 | * mounted on the device given. Returns the superblock with an active | |
d3f21473 | 625 | * reference or %NULL if none was found. |
4504230a CH |
626 | */ |
627 | struct super_block *get_active_super(struct block_device *bdev) | |
628 | { | |
629 | struct super_block *sb; | |
630 | ||
631 | if (!bdev) | |
632 | return NULL; | |
633 | ||
1494583d | 634 | restart: |
4504230a CH |
635 | spin_lock(&sb_lock); |
636 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 637 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 638 | continue; |
1494583d | 639 | if (sb->s_bdev == bdev) { |
d0f1a6e5 | 640 | if (!grab_super(sb)) |
1494583d | 641 | goto restart; |
d0f1a6e5 AV |
642 | up_write(&sb->s_umount); |
643 | return sb; | |
1494583d | 644 | } |
4504230a CH |
645 | } |
646 | spin_unlock(&sb_lock); | |
647 | return NULL; | |
648 | } | |
1da177e4 | 649 | |
df40c01a | 650 | struct super_block *user_get_super(dev_t dev) |
1da177e4 | 651 | { |
618f0636 | 652 | struct super_block *sb; |
1da177e4 | 653 | |
1da177e4 | 654 | spin_lock(&sb_lock); |
618f0636 KK |
655 | rescan: |
656 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 657 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 658 | continue; |
618f0636 KK |
659 | if (sb->s_dev == dev) { |
660 | sb->s_count++; | |
1da177e4 | 661 | spin_unlock(&sb_lock); |
618f0636 | 662 | down_read(&sb->s_umount); |
df40c01a | 663 | /* still alive? */ |
dabe0dc1 | 664 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
618f0636 KK |
665 | return sb; |
666 | up_read(&sb->s_umount); | |
df40c01a | 667 | /* nope, got unmounted */ |
618f0636 | 668 | spin_lock(&sb_lock); |
df40c01a AV |
669 | __put_super(sb); |
670 | goto rescan; | |
1da177e4 LT |
671 | } |
672 | } | |
673 | spin_unlock(&sb_lock); | |
674 | return NULL; | |
675 | } | |
676 | ||
1da177e4 | 677 | /** |
27d99302 DR |
678 | * do_remount_sb2 - asks filesystem to change mount options. |
679 | * @mnt: mount we are looking at | |
1da177e4 LT |
680 | * @sb: superblock in question |
681 | * @flags: numeric part of options | |
682 | * @data: the rest of options | |
683 | * @force: whether or not to force the change | |
684 | * | |
685 | * Alters the mount options of a mounted file system. | |
686 | */ | |
27d99302 | 687 | int do_remount_sb2(struct vfsmount *mnt, struct super_block *sb, int flags, void *data, int force) |
1da177e4 LT |
688 | { |
689 | int retval; | |
c79d967d | 690 | int remount_ro; |
4504230a | 691 | |
5accdf82 | 692 | if (sb->s_writers.frozen != SB_UNFROZEN) |
4504230a CH |
693 | return -EBUSY; |
694 | ||
9361401e | 695 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
696 | if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev)) |
697 | return -EACCES; | |
9361401e | 698 | #endif |
4504230a | 699 | |
1da177e4 LT |
700 | if (flags & MS_RDONLY) |
701 | acct_auto_close(sb); | |
702 | shrink_dcache_sb(sb); | |
60b0680f | 703 | sync_filesystem(sb); |
1da177e4 | 704 | |
d208bbdd | 705 | remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY); |
d208bbdd | 706 | |
1da177e4 LT |
707 | /* If we are remounting RDONLY and current sb is read/write, |
708 | make sure there are no rw files opened */ | |
d208bbdd | 709 | if (remount_ro) { |
4ed5e82f | 710 | if (force) { |
e7551ffb AV |
711 | sb->s_readonly_remount = 1; |
712 | smp_wmb(); | |
4ed5e82f MS |
713 | } else { |
714 | retval = sb_prepare_remount_readonly(sb); | |
715 | if (retval) | |
716 | return retval; | |
4ed5e82f | 717 | } |
1da177e4 LT |
718 | } |
719 | ||
27d99302 DR |
720 | if (mnt && sb->s_op->remount_fs2) { |
721 | retval = sb->s_op->remount_fs2(mnt, sb, &flags, data); | |
722 | if (retval) { | |
723 | if (!force) | |
724 | goto cancel_readonly; | |
725 | /* If forced remount, go ahead despite any errors */ | |
726 | WARN(1, "forced remount of a %s fs returned %i\n", | |
727 | sb->s_type->name, retval); | |
728 | } | |
729 | } else if (sb->s_op->remount_fs) { | |
1da177e4 | 730 | retval = sb->s_op->remount_fs(sb, &flags, data); |
2833eb2b MS |
731 | if (retval) { |
732 | if (!force) | |
4ed5e82f | 733 | goto cancel_readonly; |
2833eb2b MS |
734 | /* If forced remount, go ahead despite any errors */ |
735 | WARN(1, "forced remount of a %s fs returned %i\n", | |
736 | sb->s_type->name, retval); | |
737 | } | |
1da177e4 LT |
738 | } |
739 | sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK); | |
4ed5e82f MS |
740 | /* Needs to be ordered wrt mnt_is_readonly() */ |
741 | smp_wmb(); | |
742 | sb->s_readonly_remount = 0; | |
c79d967d | 743 | |
d208bbdd NP |
744 | /* |
745 | * Some filesystems modify their metadata via some other path than the | |
746 | * bdev buffer cache (eg. use a private mapping, or directories in | |
747 | * pagecache, etc). Also file data modifications go via their own | |
748 | * mappings. So If we try to mount readonly then copy the filesystem | |
749 | * from bdev, we could get stale data, so invalidate it to give a best | |
750 | * effort at coherency. | |
751 | */ | |
752 | if (remount_ro && sb->s_bdev) | |
753 | invalidate_bdev(sb->s_bdev); | |
1da177e4 | 754 | return 0; |
4ed5e82f MS |
755 | |
756 | cancel_readonly: | |
757 | sb->s_readonly_remount = 0; | |
758 | return retval; | |
1da177e4 LT |
759 | } |
760 | ||
27d99302 DR |
761 | int do_remount_sb(struct super_block *sb, int flags, void *data, int force) |
762 | { | |
763 | return do_remount_sb2(NULL, sb, flags, data, force); | |
764 | } | |
765 | ||
3c2a0909 | 766 | void do_emergency_remount(struct work_struct *work) |
1da177e4 | 767 | { |
dca33252 | 768 | struct super_block *sb, *p = NULL; |
1da177e4 LT |
769 | |
770 | spin_lock(&sb_lock); | |
b6afee17 | 771 | list_for_each_entry_reverse(sb, &super_blocks, s_list) { |
a5166169 | 772 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 773 | continue; |
1da177e4 LT |
774 | sb->s_count++; |
775 | spin_unlock(&sb_lock); | |
443b94ba | 776 | down_write(&sb->s_umount); |
dabe0dc1 AV |
777 | if (sb->s_root && sb->s_bdev && (sb->s_flags & MS_BORN) && |
778 | !(sb->s_flags & MS_RDONLY)) { | |
1da177e4 | 779 | /* |
1da177e4 LT |
780 | * What lock protects sb->s_flags?? |
781 | */ | |
1da177e4 | 782 | do_remount_sb(sb, MS_RDONLY, NULL, 1); |
1da177e4 | 783 | } |
443b94ba | 784 | up_write(&sb->s_umount); |
1da177e4 | 785 | spin_lock(&sb_lock); |
dca33252 AV |
786 | if (p) |
787 | __put_super(p); | |
788 | p = sb; | |
1da177e4 | 789 | } |
dca33252 AV |
790 | if (p) |
791 | __put_super(p); | |
1da177e4 | 792 | spin_unlock(&sb_lock); |
a2a9537a | 793 | kfree(work); |
1da177e4 LT |
794 | printk("Emergency Remount complete\n"); |
795 | } | |
796 | ||
797 | void emergency_remount(void) | |
798 | { | |
a2a9537a JA |
799 | struct work_struct *work; |
800 | ||
801 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | |
802 | if (work) { | |
803 | INIT_WORK(work, do_emergency_remount); | |
804 | schedule_work(work); | |
805 | } | |
1da177e4 LT |
806 | } |
807 | ||
808 | /* | |
809 | * Unnamed block devices are dummy devices used by virtual | |
810 | * filesystems which don't use real block-devices. -- jrs | |
811 | */ | |
812 | ||
ad76cbc6 | 813 | static DEFINE_IDA(unnamed_dev_ida); |
1da177e4 | 814 | static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */ |
c63e09ec | 815 | static int unnamed_dev_start = 0; /* don't bother trying below it */ |
1da177e4 | 816 | |
0ee5dc67 | 817 | int get_anon_bdev(dev_t *p) |
1da177e4 LT |
818 | { |
819 | int dev; | |
820 | int error; | |
821 | ||
822 | retry: | |
ad76cbc6 | 823 | if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0) |
1da177e4 LT |
824 | return -ENOMEM; |
825 | spin_lock(&unnamed_dev_lock); | |
c63e09ec | 826 | error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev); |
f21f6220 AV |
827 | if (!error) |
828 | unnamed_dev_start = dev + 1; | |
1da177e4 LT |
829 | spin_unlock(&unnamed_dev_lock); |
830 | if (error == -EAGAIN) | |
831 | /* We raced and lost with another CPU. */ | |
832 | goto retry; | |
833 | else if (error) | |
834 | return -EAGAIN; | |
835 | ||
e8c8d1bc | 836 | if (dev == (1 << MINORBITS)) { |
1da177e4 | 837 | spin_lock(&unnamed_dev_lock); |
ad76cbc6 | 838 | ida_remove(&unnamed_dev_ida, dev); |
f21f6220 AV |
839 | if (unnamed_dev_start > dev) |
840 | unnamed_dev_start = dev; | |
1da177e4 LT |
841 | spin_unlock(&unnamed_dev_lock); |
842 | return -EMFILE; | |
843 | } | |
0ee5dc67 | 844 | *p = MKDEV(0, dev & MINORMASK); |
1da177e4 LT |
845 | return 0; |
846 | } | |
0ee5dc67 | 847 | EXPORT_SYMBOL(get_anon_bdev); |
1da177e4 | 848 | |
0ee5dc67 | 849 | void free_anon_bdev(dev_t dev) |
1da177e4 | 850 | { |
0ee5dc67 | 851 | int slot = MINOR(dev); |
1da177e4 | 852 | spin_lock(&unnamed_dev_lock); |
ad76cbc6 | 853 | ida_remove(&unnamed_dev_ida, slot); |
c63e09ec AV |
854 | if (slot < unnamed_dev_start) |
855 | unnamed_dev_start = slot; | |
1da177e4 LT |
856 | spin_unlock(&unnamed_dev_lock); |
857 | } | |
0ee5dc67 AV |
858 | EXPORT_SYMBOL(free_anon_bdev); |
859 | ||
860 | int set_anon_super(struct super_block *s, void *data) | |
861 | { | |
862 | int error = get_anon_bdev(&s->s_dev); | |
863 | if (!error) | |
864 | s->s_bdi = &noop_backing_dev_info; | |
865 | return error; | |
866 | } | |
867 | ||
868 | EXPORT_SYMBOL(set_anon_super); | |
869 | ||
870 | void kill_anon_super(struct super_block *sb) | |
871 | { | |
872 | dev_t dev = sb->s_dev; | |
873 | generic_shutdown_super(sb); | |
874 | free_anon_bdev(dev); | |
875 | } | |
1da177e4 LT |
876 | |
877 | EXPORT_SYMBOL(kill_anon_super); | |
878 | ||
1da177e4 LT |
879 | void kill_litter_super(struct super_block *sb) |
880 | { | |
881 | if (sb->s_root) | |
882 | d_genocide(sb->s_root); | |
883 | kill_anon_super(sb); | |
884 | } | |
885 | ||
886 | EXPORT_SYMBOL(kill_litter_super); | |
887 | ||
909e6d94 SH |
888 | static int ns_test_super(struct super_block *sb, void *data) |
889 | { | |
890 | return sb->s_fs_info == data; | |
891 | } | |
892 | ||
893 | static int ns_set_super(struct super_block *sb, void *data) | |
894 | { | |
895 | sb->s_fs_info = data; | |
896 | return set_anon_super(sb, NULL); | |
897 | } | |
898 | ||
ceefda69 AV |
899 | struct dentry *mount_ns(struct file_system_type *fs_type, int flags, |
900 | void *data, int (*fill_super)(struct super_block *, void *, int)) | |
909e6d94 SH |
901 | { |
902 | struct super_block *sb; | |
903 | ||
9249e17f | 904 | sb = sget(fs_type, ns_test_super, ns_set_super, flags, data); |
909e6d94 | 905 | if (IS_ERR(sb)) |
ceefda69 | 906 | return ERR_CAST(sb); |
909e6d94 SH |
907 | |
908 | if (!sb->s_root) { | |
909 | int err; | |
909e6d94 SH |
910 | err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0); |
911 | if (err) { | |
74dbbdd7 | 912 | deactivate_locked_super(sb); |
ceefda69 | 913 | return ERR_PTR(err); |
909e6d94 SH |
914 | } |
915 | ||
916 | sb->s_flags |= MS_ACTIVE; | |
917 | } | |
918 | ||
ceefda69 | 919 | return dget(sb->s_root); |
909e6d94 SH |
920 | } |
921 | ||
ceefda69 | 922 | EXPORT_SYMBOL(mount_ns); |
909e6d94 | 923 | |
9361401e | 924 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
925 | static int set_bdev_super(struct super_block *s, void *data) |
926 | { | |
927 | s->s_bdev = data; | |
928 | s->s_dev = s->s_bdev->bd_dev; | |
32a88aa1 JA |
929 | |
930 | /* | |
931 | * We set the bdi here to the queue backing, file systems can | |
932 | * overwrite this in ->fill_super() | |
933 | */ | |
934 | s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info; | |
1da177e4 LT |
935 | return 0; |
936 | } | |
937 | ||
938 | static int test_bdev_super(struct super_block *s, void *data) | |
939 | { | |
940 | return (void *)s->s_bdev == data; | |
941 | } | |
942 | ||
152a0836 | 943 | struct dentry *mount_bdev(struct file_system_type *fs_type, |
1da177e4 | 944 | int flags, const char *dev_name, void *data, |
152a0836 | 945 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
946 | { |
947 | struct block_device *bdev; | |
948 | struct super_block *s; | |
d4d77629 | 949 | fmode_t mode = FMODE_READ | FMODE_EXCL; |
1da177e4 LT |
950 | int error = 0; |
951 | ||
30c40d2c AV |
952 | if (!(flags & MS_RDONLY)) |
953 | mode |= FMODE_WRITE; | |
954 | ||
d4d77629 | 955 | bdev = blkdev_get_by_path(dev_name, mode, fs_type); |
1da177e4 | 956 | if (IS_ERR(bdev)) |
152a0836 | 957 | return ERR_CAST(bdev); |
1da177e4 LT |
958 | |
959 | /* | |
960 | * once the super is inserted into the list by sget, s_umount | |
961 | * will protect the lockfs code from trying to start a snapshot | |
962 | * while we are mounting | |
963 | */ | |
4fadd7bb CH |
964 | mutex_lock(&bdev->bd_fsfreeze_mutex); |
965 | if (bdev->bd_fsfreeze_count > 0) { | |
966 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
967 | error = -EBUSY; | |
968 | goto error_bdev; | |
969 | } | |
9249e17f DH |
970 | s = sget(fs_type, test_bdev_super, set_bdev_super, flags | MS_NOSEC, |
971 | bdev); | |
4fadd7bb | 972 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
1da177e4 | 973 | if (IS_ERR(s)) |
454e2398 | 974 | goto error_s; |
1da177e4 LT |
975 | |
976 | if (s->s_root) { | |
977 | if ((flags ^ s->s_flags) & MS_RDONLY) { | |
74dbbdd7 | 978 | deactivate_locked_super(s); |
454e2398 DH |
979 | error = -EBUSY; |
980 | goto error_bdev; | |
1da177e4 | 981 | } |
454e2398 | 982 | |
4f331f01 TH |
983 | /* |
984 | * s_umount nests inside bd_mutex during | |
e525fd89 TH |
985 | * __invalidate_device(). blkdev_put() acquires |
986 | * bd_mutex and can't be called under s_umount. Drop | |
987 | * s_umount temporarily. This is safe as we're | |
988 | * holding an active reference. | |
4f331f01 TH |
989 | */ |
990 | up_write(&s->s_umount); | |
d4d77629 | 991 | blkdev_put(bdev, mode); |
4f331f01 | 992 | down_write(&s->s_umount); |
1da177e4 LT |
993 | } else { |
994 | char b[BDEVNAME_SIZE]; | |
995 | ||
30c40d2c | 996 | s->s_mode = mode; |
1da177e4 | 997 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); |
e78c9a00 | 998 | sb_set_blocksize(s, block_size(bdev)); |
9b04c997 | 999 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1000 | if (error) { |
74dbbdd7 | 1001 | deactivate_locked_super(s); |
454e2398 | 1002 | goto error; |
fa675765 | 1003 | } |
454e2398 DH |
1004 | |
1005 | s->s_flags |= MS_ACTIVE; | |
87d8fe1e | 1006 | bdev->bd_super = s; |
1da177e4 LT |
1007 | } |
1008 | ||
152a0836 | 1009 | return dget(s->s_root); |
1da177e4 | 1010 | |
454e2398 DH |
1011 | error_s: |
1012 | error = PTR_ERR(s); | |
1013 | error_bdev: | |
d4d77629 | 1014 | blkdev_put(bdev, mode); |
454e2398 | 1015 | error: |
152a0836 AV |
1016 | return ERR_PTR(error); |
1017 | } | |
1018 | EXPORT_SYMBOL(mount_bdev); | |
1019 | ||
1da177e4 LT |
1020 | void kill_block_super(struct super_block *sb) |
1021 | { | |
1022 | struct block_device *bdev = sb->s_bdev; | |
30c40d2c | 1023 | fmode_t mode = sb->s_mode; |
1da177e4 | 1024 | |
ddbaaf30 | 1025 | bdev->bd_super = NULL; |
1da177e4 LT |
1026 | generic_shutdown_super(sb); |
1027 | sync_blockdev(bdev); | |
d4d77629 | 1028 | WARN_ON_ONCE(!(mode & FMODE_EXCL)); |
e525fd89 | 1029 | blkdev_put(bdev, mode | FMODE_EXCL); |
1da177e4 LT |
1030 | } |
1031 | ||
1032 | EXPORT_SYMBOL(kill_block_super); | |
9361401e | 1033 | #endif |
1da177e4 | 1034 | |
3c26ff6e | 1035 | struct dentry *mount_nodev(struct file_system_type *fs_type, |
1da177e4 | 1036 | int flags, void *data, |
3c26ff6e | 1037 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1038 | { |
1039 | int error; | |
9249e17f | 1040 | struct super_block *s = sget(fs_type, NULL, set_anon_super, flags, NULL); |
1da177e4 LT |
1041 | |
1042 | if (IS_ERR(s)) | |
3c26ff6e | 1043 | return ERR_CAST(s); |
1da177e4 | 1044 | |
9b04c997 | 1045 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1046 | if (error) { |
74dbbdd7 | 1047 | deactivate_locked_super(s); |
3c26ff6e | 1048 | return ERR_PTR(error); |
1da177e4 LT |
1049 | } |
1050 | s->s_flags |= MS_ACTIVE; | |
3c26ff6e | 1051 | return dget(s->s_root); |
1da177e4 | 1052 | } |
3c26ff6e AV |
1053 | EXPORT_SYMBOL(mount_nodev); |
1054 | ||
1da177e4 LT |
1055 | static int compare_single(struct super_block *s, void *p) |
1056 | { | |
1057 | return 1; | |
1058 | } | |
1059 | ||
fc14f2fe | 1060 | struct dentry *mount_single(struct file_system_type *fs_type, |
1da177e4 | 1061 | int flags, void *data, |
fc14f2fe | 1062 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1063 | { |
1064 | struct super_block *s; | |
1065 | int error; | |
1066 | ||
9249e17f | 1067 | s = sget(fs_type, compare_single, set_anon_super, flags, NULL); |
1da177e4 | 1068 | if (IS_ERR(s)) |
fc14f2fe | 1069 | return ERR_CAST(s); |
1da177e4 | 1070 | if (!s->s_root) { |
9b04c997 | 1071 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1072 | if (error) { |
74dbbdd7 | 1073 | deactivate_locked_super(s); |
fc14f2fe | 1074 | return ERR_PTR(error); |
1da177e4 LT |
1075 | } |
1076 | s->s_flags |= MS_ACTIVE; | |
9329d1be KS |
1077 | } else { |
1078 | do_remount_sb(s, flags, data, 0); | |
1da177e4 | 1079 | } |
fc14f2fe AV |
1080 | return dget(s->s_root); |
1081 | } | |
1082 | EXPORT_SYMBOL(mount_single); | |
1083 | ||
9d412a43 | 1084 | struct dentry * |
27d99302 | 1085 | mount_fs(struct file_system_type *type, int flags, const char *name, struct vfsmount *mnt, void *data) |
1da177e4 | 1086 | { |
c96e41e9 | 1087 | struct dentry *root; |
9d412a43 | 1088 | struct super_block *sb; |
1da177e4 | 1089 | char *secdata = NULL; |
9d412a43 | 1090 | int error = -ENOMEM; |
8089352a | 1091 | |
e0007529 | 1092 | if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) { |
1da177e4 | 1093 | secdata = alloc_secdata(); |
454e2398 | 1094 | if (!secdata) |
9d412a43 | 1095 | goto out; |
1da177e4 | 1096 | |
e0007529 | 1097 | error = security_sb_copy_data(data, secdata); |
454e2398 | 1098 | if (error) |
1da177e4 | 1099 | goto out_free_secdata; |
1da177e4 LT |
1100 | } |
1101 | ||
27d99302 DR |
1102 | if (type->mount2) |
1103 | root = type->mount2(mnt, type, flags, name, data); | |
1104 | else | |
1105 | root = type->mount(type, flags, name, data); | |
1a102ff9 AV |
1106 | if (IS_ERR(root)) { |
1107 | error = PTR_ERR(root); | |
1108 | goto out_free_secdata; | |
c96e41e9 | 1109 | } |
9d412a43 AV |
1110 | sb = root->d_sb; |
1111 | BUG_ON(!sb); | |
1112 | WARN_ON(!sb->s_bdi); | |
6c510389 | 1113 | WARN_ON(sb->s_bdi == &default_backing_dev_info); |
9d412a43 | 1114 | sb->s_flags |= MS_BORN; |
454e2398 | 1115 | |
9d412a43 | 1116 | error = security_sb_kern_mount(sb, flags, secdata); |
5129a469 JE |
1117 | if (error) |
1118 | goto out_sb; | |
454e2398 | 1119 | |
42cb56ae JL |
1120 | /* |
1121 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE | |
1122 | * but s_maxbytes was an unsigned long long for many releases. Throw | |
1123 | * this warning for a little while to try and catch filesystems that | |
4358b567 | 1124 | * violate this rule. |
42cb56ae | 1125 | */ |
9d412a43 AV |
1126 | WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " |
1127 | "negative value (%lld)\n", type->name, sb->s_maxbytes); | |
42cb56ae | 1128 | |
9d412a43 | 1129 | up_write(&sb->s_umount); |
8680e22f | 1130 | free_secdata(secdata); |
9d412a43 | 1131 | return root; |
1da177e4 | 1132 | out_sb: |
9d412a43 AV |
1133 | dput(root); |
1134 | deactivate_locked_super(sb); | |
1da177e4 LT |
1135 | out_free_secdata: |
1136 | free_secdata(secdata); | |
1da177e4 | 1137 | out: |
454e2398 | 1138 | return ERR_PTR(error); |
1da177e4 LT |
1139 | } |
1140 | ||
5accdf82 JK |
1141 | /* |
1142 | * This is an internal function, please use sb_end_{write,pagefault,intwrite} | |
1143 | * instead. | |
1144 | */ | |
1145 | void __sb_end_write(struct super_block *sb, int level) | |
1146 | { | |
1147 | percpu_counter_dec(&sb->s_writers.counter[level-1]); | |
1148 | /* | |
1149 | * Make sure s_writers are updated before we wake up waiters in | |
1150 | * freeze_super(). | |
1151 | */ | |
1152 | smp_mb(); | |
1153 | if (waitqueue_active(&sb->s_writers.wait)) | |
1154 | wake_up(&sb->s_writers.wait); | |
1155 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _RET_IP_); | |
1156 | } | |
1157 | EXPORT_SYMBOL(__sb_end_write); | |
1158 | ||
1159 | #ifdef CONFIG_LOCKDEP | |
1160 | /* | |
1161 | * We want lockdep to tell us about possible deadlocks with freezing but | |
1162 | * it's it bit tricky to properly instrument it. Getting a freeze protection | |
1163 | * works as getting a read lock but there are subtle problems. XFS for example | |
1164 | * gets freeze protection on internal level twice in some cases, which is OK | |
1165 | * only because we already hold a freeze protection also on higher level. Due | |
1166 | * to these cases we have to tell lockdep we are doing trylock when we | |
1167 | * already hold a freeze protection for a higher freeze level. | |
1168 | */ | |
1169 | static void acquire_freeze_lock(struct super_block *sb, int level, bool trylock, | |
1170 | unsigned long ip) | |
1171 | { | |
1172 | int i; | |
1173 | ||
1174 | if (!trylock) { | |
1175 | for (i = 0; i < level - 1; i++) | |
1176 | if (lock_is_held(&sb->s_writers.lock_map[i])) { | |
1177 | trylock = true; | |
1178 | break; | |
1179 | } | |
1180 | } | |
1181 | rwsem_acquire_read(&sb->s_writers.lock_map[level-1], 0, trylock, ip); | |
1182 | } | |
1183 | #endif | |
1184 | ||
1185 | /* | |
1186 | * This is an internal function, please use sb_start_{write,pagefault,intwrite} | |
1187 | * instead. | |
1188 | */ | |
1189 | int __sb_start_write(struct super_block *sb, int level, bool wait) | |
1190 | { | |
1191 | retry: | |
1192 | if (unlikely(sb->s_writers.frozen >= level)) { | |
1193 | if (!wait) | |
1194 | return 0; | |
1195 | wait_event(sb->s_writers.wait_unfrozen, | |
1196 | sb->s_writers.frozen < level); | |
1197 | } | |
1198 | ||
1199 | #ifdef CONFIG_LOCKDEP | |
1200 | acquire_freeze_lock(sb, level, !wait, _RET_IP_); | |
1201 | #endif | |
1202 | percpu_counter_inc(&sb->s_writers.counter[level-1]); | |
1203 | /* | |
1204 | * Make sure counter is updated before we check for frozen. | |
1205 | * freeze_super() first sets frozen and then checks the counter. | |
1206 | */ | |
1207 | smp_mb(); | |
1208 | if (unlikely(sb->s_writers.frozen >= level)) { | |
1209 | __sb_end_write(sb, level); | |
1210 | goto retry; | |
1211 | } | |
1212 | return 1; | |
1213 | } | |
1214 | EXPORT_SYMBOL(__sb_start_write); | |
1215 | ||
1216 | /** | |
1217 | * sb_wait_write - wait until all writers to given file system finish | |
1218 | * @sb: the super for which we wait | |
1219 | * @level: type of writers we wait for (normal vs page fault) | |
1220 | * | |
1221 | * This function waits until there are no writers of given type to given file | |
1222 | * system. Caller of this function should make sure there can be no new writers | |
1223 | * of type @level before calling this function. Otherwise this function can | |
1224 | * livelock. | |
1225 | */ | |
1226 | static void sb_wait_write(struct super_block *sb, int level) | |
1227 | { | |
1228 | s64 writers; | |
1229 | ||
1230 | /* | |
1231 | * We just cycle-through lockdep here so that it does not complain | |
1232 | * about returning with lock to userspace | |
1233 | */ | |
1234 | rwsem_acquire(&sb->s_writers.lock_map[level-1], 0, 0, _THIS_IP_); | |
1235 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _THIS_IP_); | |
1236 | ||
1237 | do { | |
1238 | DEFINE_WAIT(wait); | |
1239 | ||
1240 | /* | |
1241 | * We use a barrier in prepare_to_wait() to separate setting | |
1242 | * of frozen and checking of the counter | |
1243 | */ | |
1244 | prepare_to_wait(&sb->s_writers.wait, &wait, | |
1245 | TASK_UNINTERRUPTIBLE); | |
1246 | ||
1247 | writers = percpu_counter_sum(&sb->s_writers.counter[level-1]); | |
1248 | if (writers) | |
1249 | schedule(); | |
1250 | ||
1251 | finish_wait(&sb->s_writers.wait, &wait); | |
1252 | } while (writers); | |
1253 | } | |
1254 | ||
18e9e510 | 1255 | /** |
7000d3c4 RD |
1256 | * freeze_super - lock the filesystem and force it into a consistent state |
1257 | * @sb: the super to lock | |
18e9e510 JB |
1258 | * |
1259 | * Syncs the super to make sure the filesystem is consistent and calls the fs's | |
1260 | * freeze_fs. Subsequent calls to this without first thawing the fs will return | |
1261 | * -EBUSY. | |
5accdf82 JK |
1262 | * |
1263 | * During this function, sb->s_writers.frozen goes through these values: | |
1264 | * | |
1265 | * SB_UNFROZEN: File system is normal, all writes progress as usual. | |
1266 | * | |
1267 | * SB_FREEZE_WRITE: The file system is in the process of being frozen. New | |
1268 | * writes should be blocked, though page faults are still allowed. We wait for | |
1269 | * all writes to complete and then proceed to the next stage. | |
1270 | * | |
1271 | * SB_FREEZE_PAGEFAULT: Freezing continues. Now also page faults are blocked | |
1272 | * but internal fs threads can still modify the filesystem (although they | |
1273 | * should not dirty new pages or inodes), writeback can run etc. After waiting | |
1274 | * for all running page faults we sync the filesystem which will clean all | |
1275 | * dirty pages and inodes (no new dirty pages or inodes can be created when | |
1276 | * sync is running). | |
1277 | * | |
1278 | * SB_FREEZE_FS: The file system is frozen. Now all internal sources of fs | |
1279 | * modification are blocked (e.g. XFS preallocation truncation on inode | |
1280 | * reclaim). This is usually implemented by blocking new transactions for | |
1281 | * filesystems that have them and need this additional guard. After all | |
1282 | * internal writers are finished we call ->freeze_fs() to finish filesystem | |
1283 | * freezing. Then we transition to SB_FREEZE_COMPLETE state. This state is | |
1284 | * mostly auxiliary for filesystems to verify they do not modify frozen fs. | |
1285 | * | |
1286 | * sb->s_writers.frozen is protected by sb->s_umount. | |
18e9e510 JB |
1287 | */ |
1288 | int freeze_super(struct super_block *sb) | |
1289 | { | |
1290 | int ret; | |
1291 | ||
1292 | atomic_inc(&sb->s_active); | |
1293 | down_write(&sb->s_umount); | |
5accdf82 | 1294 | if (sb->s_writers.frozen != SB_UNFROZEN) { |
18e9e510 JB |
1295 | deactivate_locked_super(sb); |
1296 | return -EBUSY; | |
1297 | } | |
1298 | ||
dabe0dc1 AV |
1299 | if (!(sb->s_flags & MS_BORN)) { |
1300 | up_write(&sb->s_umount); | |
1301 | return 0; /* sic - it's "nothing to do" */ | |
1302 | } | |
1303 | ||
18e9e510 | 1304 | if (sb->s_flags & MS_RDONLY) { |
5accdf82 JK |
1305 | /* Nothing to do really... */ |
1306 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
18e9e510 JB |
1307 | up_write(&sb->s_umount); |
1308 | return 0; | |
1309 | } | |
1310 | ||
5accdf82 JK |
1311 | /* From now on, no new normal writers can start */ |
1312 | sb->s_writers.frozen = SB_FREEZE_WRITE; | |
1313 | smp_wmb(); | |
1314 | ||
1315 | /* Release s_umount to preserve sb_start_write -> s_umount ordering */ | |
1316 | up_write(&sb->s_umount); | |
1317 | ||
1318 | sb_wait_write(sb, SB_FREEZE_WRITE); | |
1319 | ||
1320 | /* Now we go and block page faults... */ | |
1321 | down_write(&sb->s_umount); | |
1322 | sb->s_writers.frozen = SB_FREEZE_PAGEFAULT; | |
18e9e510 JB |
1323 | smp_wmb(); |
1324 | ||
5accdf82 JK |
1325 | sb_wait_write(sb, SB_FREEZE_PAGEFAULT); |
1326 | ||
1327 | /* All writers are done so after syncing there won't be dirty data */ | |
18e9e510 JB |
1328 | sync_filesystem(sb); |
1329 | ||
5accdf82 JK |
1330 | /* Now wait for internal filesystem counter */ |
1331 | sb->s_writers.frozen = SB_FREEZE_FS; | |
18e9e510 | 1332 | smp_wmb(); |
5accdf82 | 1333 | sb_wait_write(sb, SB_FREEZE_FS); |
18e9e510 | 1334 | |
18e9e510 JB |
1335 | if (sb->s_op->freeze_fs) { |
1336 | ret = sb->s_op->freeze_fs(sb); | |
1337 | if (ret) { | |
1338 | printk(KERN_ERR | |
1339 | "VFS:Filesystem freeze failed\n"); | |
5accdf82 | 1340 | sb->s_writers.frozen = SB_UNFROZEN; |
e1616300 | 1341 | smp_wmb(); |
5accdf82 | 1342 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 JB |
1343 | deactivate_locked_super(sb); |
1344 | return ret; | |
1345 | } | |
1346 | } | |
5accdf82 | 1347 | /* |
f6324bcc ON |
1348 | * For debugging purposes so that fs can warn if it sees write activity |
1349 | * when frozen is set to SB_FREEZE_COMPLETE, and for thaw_super(). | |
5accdf82 JK |
1350 | */ |
1351 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
18e9e510 JB |
1352 | up_write(&sb->s_umount); |
1353 | return 0; | |
1354 | } | |
1355 | EXPORT_SYMBOL(freeze_super); | |
1356 | ||
1357 | /** | |
1358 | * thaw_super -- unlock filesystem | |
1359 | * @sb: the super to thaw | |
1360 | * | |
1361 | * Unlocks the filesystem and marks it writeable again after freeze_super(). | |
1362 | */ | |
1363 | int thaw_super(struct super_block *sb) | |
1364 | { | |
1365 | int error; | |
1366 | ||
1367 | down_write(&sb->s_umount); | |
f6324bcc | 1368 | if (sb->s_writers.frozen != SB_FREEZE_COMPLETE) { |
18e9e510 JB |
1369 | up_write(&sb->s_umount); |
1370 | return -EINVAL; | |
1371 | } | |
1372 | ||
1373 | if (sb->s_flags & MS_RDONLY) | |
1374 | goto out; | |
1375 | ||
1376 | if (sb->s_op->unfreeze_fs) { | |
1377 | error = sb->s_op->unfreeze_fs(sb); | |
1378 | if (error) { | |
1379 | printk(KERN_ERR | |
1380 | "VFS:Filesystem thaw failed\n"); | |
18e9e510 JB |
1381 | up_write(&sb->s_umount); |
1382 | return error; | |
1383 | } | |
1384 | } | |
1385 | ||
1386 | out: | |
5accdf82 | 1387 | sb->s_writers.frozen = SB_UNFROZEN; |
18e9e510 | 1388 | smp_wmb(); |
5accdf82 | 1389 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 JB |
1390 | deactivate_locked_super(sb); |
1391 | ||
1392 | return 0; | |
1393 | } | |
1394 | EXPORT_SYMBOL(thaw_super); |