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; | |
79 | ||
b0d40c92 | 80 | if (sc->nr_to_scan) { |
0e1fdafd DC |
81 | int dentries; |
82 | int inodes; | |
b0d40c92 | 83 | |
0e1fdafd DC |
84 | /* proportion the scan between the caches */ |
85 | dentries = (sc->nr_to_scan * sb->s_nr_dentry_unused) / | |
86 | total_objects; | |
87 | inodes = (sc->nr_to_scan * sb->s_nr_inodes_unused) / | |
88 | total_objects; | |
89 | if (fs_objects) | |
90 | fs_objects = (sc->nr_to_scan * fs_objects) / | |
91 | total_objects; | |
92 | /* | |
93 | * prune the dcache first as the icache is pinned by it, then | |
94 | * prune the icache, followed by the filesystem specific caches | |
95 | */ | |
96 | prune_dcache_sb(sb, dentries); | |
97 | prune_icache_sb(sb, inodes); | |
b0d40c92 | 98 | |
0e1fdafd DC |
99 | if (fs_objects && sb->s_op->free_cached_objects) { |
100 | sb->s_op->free_cached_objects(sb, fs_objects); | |
101 | fs_objects = sb->s_op->nr_cached_objects(sb); | |
102 | } | |
103 | total_objects = sb->s_nr_dentry_unused + | |
104 | sb->s_nr_inodes_unused + fs_objects; | |
b0d40c92 DC |
105 | } |
106 | ||
0e1fdafd | 107 | total_objects = (total_objects / 100) * sysctl_vfs_cache_pressure; |
b0d40c92 | 108 | drop_super(sb); |
0e1fdafd | 109 | return total_objects; |
b0d40c92 DC |
110 | } |
111 | ||
5accdf82 JK |
112 | static int init_sb_writers(struct super_block *s, struct file_system_type *type) |
113 | { | |
114 | int err; | |
115 | int i; | |
116 | ||
117 | for (i = 0; i < SB_FREEZE_LEVELS; i++) { | |
118 | err = percpu_counter_init(&s->s_writers.counter[i], 0); | |
119 | if (err < 0) | |
120 | goto err_out; | |
121 | lockdep_init_map(&s->s_writers.lock_map[i], sb_writers_name[i], | |
122 | &type->s_writers_key[i], 0); | |
123 | } | |
124 | init_waitqueue_head(&s->s_writers.wait); | |
125 | init_waitqueue_head(&s->s_writers.wait_unfrozen); | |
126 | return 0; | |
127 | err_out: | |
128 | while (--i >= 0) | |
129 | percpu_counter_destroy(&s->s_writers.counter[i]); | |
130 | return err; | |
131 | } | |
132 | ||
133 | static void destroy_sb_writers(struct super_block *s) | |
134 | { | |
135 | int i; | |
136 | ||
137 | for (i = 0; i < SB_FREEZE_LEVELS; i++) | |
138 | percpu_counter_destroy(&s->s_writers.counter[i]); | |
139 | } | |
140 | ||
1da177e4 LT |
141 | /** |
142 | * alloc_super - create new superblock | |
fe2bbc48 | 143 | * @type: filesystem type superblock should belong to |
9249e17f | 144 | * @flags: the mount flags |
1da177e4 LT |
145 | * |
146 | * Allocates and initializes a new &struct super_block. alloc_super() | |
147 | * returns a pointer new superblock or %NULL if allocation had failed. | |
148 | */ | |
9249e17f | 149 | static struct super_block *alloc_super(struct file_system_type *type, int flags) |
1da177e4 | 150 | { |
11b0b5ab | 151 | struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER); |
b87221de | 152 | static const struct super_operations default_op; |
1da177e4 LT |
153 | |
154 | if (s) { | |
1da177e4 | 155 | if (security_sb_alloc(s)) { |
5accdf82 JK |
156 | /* |
157 | * We cannot call security_sb_free() without | |
158 | * security_sb_alloc() succeeding. So bail out manually | |
159 | */ | |
1da177e4 LT |
160 | kfree(s); |
161 | s = NULL; | |
162 | goto out; | |
163 | } | |
6416ccb7 NP |
164 | #ifdef CONFIG_SMP |
165 | s->s_files = alloc_percpu(struct list_head); | |
5accdf82 JK |
166 | if (!s->s_files) |
167 | goto err_out; | |
168 | else { | |
6416ccb7 NP |
169 | int i; |
170 | ||
171 | for_each_possible_cpu(i) | |
172 | INIT_LIST_HEAD(per_cpu_ptr(s->s_files, i)); | |
173 | } | |
174 | #else | |
1da177e4 | 175 | INIT_LIST_HEAD(&s->s_files); |
6416ccb7 | 176 | #endif |
5accdf82 JK |
177 | if (init_sb_writers(s, type)) |
178 | goto err_out; | |
9249e17f | 179 | s->s_flags = flags; |
95f28604 | 180 | s->s_bdi = &default_backing_dev_info; |
a5166169 | 181 | INIT_HLIST_NODE(&s->s_instances); |
ceb5bdc2 | 182 | INIT_HLIST_BL_HEAD(&s->s_anon); |
1da177e4 | 183 | INIT_LIST_HEAD(&s->s_inodes); |
da3bbdd4 | 184 | INIT_LIST_HEAD(&s->s_dentry_lru); |
98b745c6 | 185 | INIT_LIST_HEAD(&s->s_inode_lru); |
09cc9fc7 | 186 | spin_lock_init(&s->s_inode_lru_lock); |
39f7c4db | 187 | INIT_LIST_HEAD(&s->s_mounts); |
1da177e4 | 188 | init_rwsem(&s->s_umount); |
897c6ff9 | 189 | lockdep_set_class(&s->s_umount, &type->s_umount_key); |
ada723dc PZ |
190 | /* |
191 | * sget() can have s_umount recursion. | |
192 | * | |
193 | * When it cannot find a suitable sb, it allocates a new | |
194 | * one (this one), and tries again to find a suitable old | |
195 | * one. | |
196 | * | |
197 | * In case that succeeds, it will acquire the s_umount | |
198 | * lock of the old one. Since these are clearly distrinct | |
199 | * locks, and this object isn't exposed yet, there's no | |
200 | * risk of deadlocks. | |
201 | * | |
202 | * Annotate this by putting this lock in a different | |
203 | * subclass. | |
204 | */ | |
205 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); | |
b20bd1a5 | 206 | s->s_count = 1; |
1da177e4 | 207 | atomic_set(&s->s_active, 1); |
a11f3a05 | 208 | mutex_init(&s->s_vfs_rename_mutex); |
51ee049e | 209 | lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key); |
d3be915f IM |
210 | mutex_init(&s->s_dquot.dqio_mutex); |
211 | mutex_init(&s->s_dquot.dqonoff_mutex); | |
1da177e4 | 212 | init_rwsem(&s->s_dquot.dqptr_sem); |
1da177e4 | 213 | s->s_maxbytes = MAX_NON_LFS; |
1da177e4 LT |
214 | s->s_op = &default_op; |
215 | s->s_time_gran = 1000000000; | |
c515e1fd | 216 | s->cleancache_poolid = -1; |
b0d40c92 DC |
217 | |
218 | s->s_shrink.seeks = DEFAULT_SEEKS; | |
219 | s->s_shrink.shrink = prune_super; | |
8ab47664 | 220 | s->s_shrink.batch = 1024; |
1da177e4 LT |
221 | } |
222 | out: | |
223 | return s; | |
5accdf82 JK |
224 | err_out: |
225 | security_sb_free(s); | |
226 | #ifdef CONFIG_SMP | |
227 | if (s->s_files) | |
228 | free_percpu(s->s_files); | |
229 | #endif | |
230 | destroy_sb_writers(s); | |
231 | kfree(s); | |
232 | s = NULL; | |
233 | goto out; | |
1da177e4 LT |
234 | } |
235 | ||
236 | /** | |
237 | * destroy_super - frees a superblock | |
238 | * @s: superblock to free | |
239 | * | |
240 | * Frees a superblock. | |
241 | */ | |
242 | static inline void destroy_super(struct super_block *s) | |
243 | { | |
6416ccb7 NP |
244 | #ifdef CONFIG_SMP |
245 | free_percpu(s->s_files); | |
246 | #endif | |
5accdf82 | 247 | destroy_sb_writers(s); |
1da177e4 | 248 | security_sb_free(s); |
39f7c4db | 249 | WARN_ON(!list_empty(&s->s_mounts)); |
79c0b2df | 250 | kfree(s->s_subtype); |
b3b304a2 | 251 | kfree(s->s_options); |
1da177e4 LT |
252 | kfree(s); |
253 | } | |
254 | ||
255 | /* Superblock refcounting */ | |
256 | ||
257 | /* | |
35cf7ba0 | 258 | * Drop a superblock's refcount. The caller must hold sb_lock. |
1da177e4 | 259 | */ |
f47ec3f2 | 260 | static void __put_super(struct super_block *sb) |
1da177e4 | 261 | { |
1da177e4 | 262 | if (!--sb->s_count) { |
551de6f3 | 263 | list_del_init(&sb->s_list); |
1da177e4 | 264 | destroy_super(sb); |
1da177e4 | 265 | } |
1da177e4 LT |
266 | } |
267 | ||
268 | /** | |
269 | * put_super - drop a temporary reference to superblock | |
270 | * @sb: superblock in question | |
271 | * | |
272 | * Drops a temporary reference, frees superblock if there's no | |
273 | * references left. | |
274 | */ | |
f47ec3f2 | 275 | static void put_super(struct super_block *sb) |
1da177e4 LT |
276 | { |
277 | spin_lock(&sb_lock); | |
278 | __put_super(sb); | |
279 | spin_unlock(&sb_lock); | |
280 | } | |
281 | ||
282 | ||
283 | /** | |
1712ac8f | 284 | * deactivate_locked_super - drop an active reference to superblock |
1da177e4 LT |
285 | * @s: superblock to deactivate |
286 | * | |
1712ac8f AV |
287 | * Drops an active reference to superblock, converting it into a temprory |
288 | * one if there is no other active references left. In that case we | |
1da177e4 LT |
289 | * tell fs driver to shut it down and drop the temporary reference we |
290 | * had just acquired. | |
1712ac8f AV |
291 | * |
292 | * Caller holds exclusive lock on superblock; that lock is released. | |
1da177e4 | 293 | */ |
1712ac8f | 294 | void deactivate_locked_super(struct super_block *s) |
1da177e4 LT |
295 | { |
296 | struct file_system_type *fs = s->s_type; | |
b20bd1a5 | 297 | if (atomic_dec_and_test(&s->s_active)) { |
3167760f | 298 | cleancache_invalidate_fs(s); |
1da177e4 | 299 | fs->kill_sb(s); |
b0d40c92 DC |
300 | |
301 | /* caches are now gone, we can safely kill the shrinker now */ | |
302 | unregister_shrinker(&s->s_shrink); | |
1da177e4 LT |
303 | put_filesystem(fs); |
304 | put_super(s); | |
1712ac8f AV |
305 | } else { |
306 | up_write(&s->s_umount); | |
1da177e4 LT |
307 | } |
308 | } | |
309 | ||
1712ac8f | 310 | EXPORT_SYMBOL(deactivate_locked_super); |
1da177e4 | 311 | |
74dbbdd7 | 312 | /** |
1712ac8f | 313 | * deactivate_super - drop an active reference to superblock |
74dbbdd7 AV |
314 | * @s: superblock to deactivate |
315 | * | |
1712ac8f AV |
316 | * Variant of deactivate_locked_super(), except that superblock is *not* |
317 | * locked by caller. If we are going to drop the final active reference, | |
318 | * lock will be acquired prior to that. | |
74dbbdd7 | 319 | */ |
1712ac8f | 320 | void deactivate_super(struct super_block *s) |
74dbbdd7 | 321 | { |
1712ac8f AV |
322 | if (!atomic_add_unless(&s->s_active, -1, 1)) { |
323 | down_write(&s->s_umount); | |
324 | deactivate_locked_super(s); | |
74dbbdd7 AV |
325 | } |
326 | } | |
327 | ||
1712ac8f | 328 | EXPORT_SYMBOL(deactivate_super); |
74dbbdd7 | 329 | |
1da177e4 LT |
330 | /** |
331 | * grab_super - acquire an active reference | |
332 | * @s: reference we are trying to make active | |
333 | * | |
334 | * Tries to acquire an active reference. grab_super() is used when we | |
335 | * had just found a superblock in super_blocks or fs_type->fs_supers | |
336 | * and want to turn it into a full-blown active reference. grab_super() | |
337 | * is called with sb_lock held and drops it. Returns 1 in case of | |
338 | * success, 0 if we had failed (superblock contents was already dead or | |
acfec9a5 AV |
339 | * dying when grab_super() had been called). Note that this is only |
340 | * called for superblocks not in rundown mode (== ones still on ->fs_supers | |
341 | * of their type), so increment of ->s_count is OK here. | |
1da177e4 | 342 | */ |
9c4dbee7 | 343 | static int grab_super(struct super_block *s) __releases(sb_lock) |
1da177e4 LT |
344 | { |
345 | s->s_count++; | |
346 | spin_unlock(&sb_lock); | |
347 | down_write(&s->s_umount); | |
acfec9a5 AV |
348 | if ((s->s_flags & MS_BORN) && atomic_inc_not_zero(&s->s_active)) { |
349 | put_super(s); | |
350 | return 1; | |
351 | } | |
1da177e4 LT |
352 | up_write(&s->s_umount); |
353 | put_super(s); | |
1da177e4 LT |
354 | return 0; |
355 | } | |
356 | ||
12ad3ab6 DC |
357 | /* |
358 | * grab_super_passive - acquire a passive reference | |
331cbdee | 359 | * @sb: reference we are trying to grab |
12ad3ab6 DC |
360 | * |
361 | * Tries to acquire a passive reference. This is used in places where we | |
362 | * cannot take an active reference but we need to ensure that the | |
363 | * superblock does not go away while we are working on it. It returns | |
364 | * false if a reference was not gained, and returns true with the s_umount | |
365 | * lock held in read mode if a reference is gained. On successful return, | |
366 | * the caller must drop the s_umount lock and the passive reference when | |
367 | * done. | |
368 | */ | |
369 | bool grab_super_passive(struct super_block *sb) | |
370 | { | |
371 | spin_lock(&sb_lock); | |
a5166169 | 372 | if (hlist_unhashed(&sb->s_instances)) { |
12ad3ab6 DC |
373 | spin_unlock(&sb_lock); |
374 | return false; | |
375 | } | |
376 | ||
377 | sb->s_count++; | |
378 | spin_unlock(&sb_lock); | |
379 | ||
380 | if (down_read_trylock(&sb->s_umount)) { | |
dabe0dc1 | 381 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
12ad3ab6 DC |
382 | return true; |
383 | up_read(&sb->s_umount); | |
384 | } | |
385 | ||
386 | put_super(sb); | |
387 | return false; | |
388 | } | |
389 | ||
1da177e4 LT |
390 | /** |
391 | * generic_shutdown_super - common helper for ->kill_sb() | |
392 | * @sb: superblock to kill | |
393 | * | |
394 | * generic_shutdown_super() does all fs-independent work on superblock | |
395 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects | |
396 | * that need destruction out of superblock, call generic_shutdown_super() | |
397 | * and release aforementioned objects. Note: dentries and inodes _are_ | |
398 | * taken care of and do not need specific handling. | |
c636ebdb DH |
399 | * |
400 | * Upon calling this function, the filesystem may no longer alter or | |
401 | * rearrange the set of dentries belonging to this super_block, nor may it | |
402 | * change the attachments of dentries to inodes. | |
1da177e4 LT |
403 | */ |
404 | void generic_shutdown_super(struct super_block *sb) | |
405 | { | |
ee9b6d61 | 406 | const struct super_operations *sop = sb->s_op; |
1da177e4 | 407 | |
c636ebdb DH |
408 | if (sb->s_root) { |
409 | shrink_dcache_for_umount(sb); | |
60b0680f | 410 | sync_filesystem(sb); |
1da177e4 | 411 | sb->s_flags &= ~MS_ACTIVE; |
efaee192 | 412 | |
63997e98 AV |
413 | fsnotify_unmount_inodes(&sb->s_inodes); |
414 | ||
415 | evict_inodes(sb); | |
1da177e4 | 416 | |
1da177e4 LT |
417 | if (sop->put_super) |
418 | sop->put_super(sb); | |
419 | ||
63997e98 | 420 | if (!list_empty(&sb->s_inodes)) { |
7b4fe29e DJ |
421 | printk("VFS: Busy inodes after unmount of %s. " |
422 | "Self-destruct in 5 seconds. Have a nice day...\n", | |
423 | sb->s_id); | |
1da177e4 | 424 | } |
1da177e4 LT |
425 | } |
426 | spin_lock(&sb_lock); | |
427 | /* should be initialized for __put_super_and_need_restart() */ | |
a5166169 | 428 | hlist_del_init(&sb->s_instances); |
1da177e4 LT |
429 | spin_unlock(&sb_lock); |
430 | up_write(&sb->s_umount); | |
431 | } | |
432 | ||
433 | EXPORT_SYMBOL(generic_shutdown_super); | |
434 | ||
435 | /** | |
436 | * sget - find or create a superblock | |
437 | * @type: filesystem type superblock should belong to | |
438 | * @test: comparison callback | |
439 | * @set: setup callback | |
9249e17f | 440 | * @flags: mount flags |
1da177e4 LT |
441 | * @data: argument to each of them |
442 | */ | |
443 | struct super_block *sget(struct file_system_type *type, | |
444 | int (*test)(struct super_block *,void *), | |
445 | int (*set)(struct super_block *,void *), | |
9249e17f | 446 | int flags, |
1da177e4 LT |
447 | void *data) |
448 | { | |
449 | struct super_block *s = NULL; | |
d4730127 | 450 | struct super_block *old; |
1da177e4 LT |
451 | int err; |
452 | ||
453 | retry: | |
454 | spin_lock(&sb_lock); | |
d4730127 | 455 | if (test) { |
b67bfe0d | 456 | hlist_for_each_entry(old, &type->fs_supers, s_instances) { |
d4730127 MK |
457 | if (!test(old, data)) |
458 | continue; | |
459 | if (!grab_super(old)) | |
460 | goto retry; | |
a3cfbb53 LZ |
461 | if (s) { |
462 | up_write(&s->s_umount); | |
d4730127 | 463 | destroy_super(s); |
7a4dec53 | 464 | s = NULL; |
a3cfbb53 | 465 | } |
d4730127 MK |
466 | return old; |
467 | } | |
1da177e4 LT |
468 | } |
469 | if (!s) { | |
470 | spin_unlock(&sb_lock); | |
9249e17f | 471 | s = alloc_super(type, flags); |
1da177e4 LT |
472 | if (!s) |
473 | return ERR_PTR(-ENOMEM); | |
474 | goto retry; | |
475 | } | |
476 | ||
477 | err = set(s, data); | |
478 | if (err) { | |
479 | spin_unlock(&sb_lock); | |
a3cfbb53 | 480 | up_write(&s->s_umount); |
1da177e4 LT |
481 | destroy_super(s); |
482 | return ERR_PTR(err); | |
483 | } | |
484 | s->s_type = type; | |
485 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); | |
486 | list_add_tail(&s->s_list, &super_blocks); | |
a5166169 | 487 | hlist_add_head(&s->s_instances, &type->fs_supers); |
1da177e4 LT |
488 | spin_unlock(&sb_lock); |
489 | get_filesystem(type); | |
b0d40c92 | 490 | register_shrinker(&s->s_shrink); |
1da177e4 LT |
491 | return s; |
492 | } | |
493 | ||
494 | EXPORT_SYMBOL(sget); | |
495 | ||
496 | void drop_super(struct super_block *sb) | |
497 | { | |
498 | up_read(&sb->s_umount); | |
499 | put_super(sb); | |
500 | } | |
501 | ||
502 | EXPORT_SYMBOL(drop_super); | |
503 | ||
01a05b33 AV |
504 | /** |
505 | * iterate_supers - call function for all active superblocks | |
506 | * @f: function to call | |
507 | * @arg: argument to pass to it | |
508 | * | |
509 | * Scans the superblock list and calls given function, passing it | |
510 | * locked superblock and given argument. | |
511 | */ | |
512 | void iterate_supers(void (*f)(struct super_block *, void *), void *arg) | |
513 | { | |
dca33252 | 514 | struct super_block *sb, *p = NULL; |
01a05b33 AV |
515 | |
516 | spin_lock(&sb_lock); | |
dca33252 | 517 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 518 | if (hlist_unhashed(&sb->s_instances)) |
01a05b33 AV |
519 | continue; |
520 | sb->s_count++; | |
521 | spin_unlock(&sb_lock); | |
522 | ||
523 | down_read(&sb->s_umount); | |
dabe0dc1 | 524 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
01a05b33 AV |
525 | f(sb, arg); |
526 | up_read(&sb->s_umount); | |
527 | ||
528 | spin_lock(&sb_lock); | |
dca33252 AV |
529 | if (p) |
530 | __put_super(p); | |
531 | p = sb; | |
01a05b33 | 532 | } |
dca33252 AV |
533 | if (p) |
534 | __put_super(p); | |
01a05b33 AV |
535 | spin_unlock(&sb_lock); |
536 | } | |
537 | ||
43e15cdb AV |
538 | /** |
539 | * iterate_supers_type - call function for superblocks of given type | |
540 | * @type: fs type | |
541 | * @f: function to call | |
542 | * @arg: argument to pass to it | |
543 | * | |
544 | * Scans the superblock list and calls given function, passing it | |
545 | * locked superblock and given argument. | |
546 | */ | |
547 | void iterate_supers_type(struct file_system_type *type, | |
548 | void (*f)(struct super_block *, void *), void *arg) | |
549 | { | |
550 | struct super_block *sb, *p = NULL; | |
551 | ||
552 | spin_lock(&sb_lock); | |
b67bfe0d | 553 | hlist_for_each_entry(sb, &type->fs_supers, s_instances) { |
43e15cdb AV |
554 | sb->s_count++; |
555 | spin_unlock(&sb_lock); | |
556 | ||
557 | down_read(&sb->s_umount); | |
dabe0dc1 | 558 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
43e15cdb AV |
559 | f(sb, arg); |
560 | up_read(&sb->s_umount); | |
561 | ||
562 | spin_lock(&sb_lock); | |
563 | if (p) | |
564 | __put_super(p); | |
565 | p = sb; | |
566 | } | |
567 | if (p) | |
568 | __put_super(p); | |
569 | spin_unlock(&sb_lock); | |
570 | } | |
571 | ||
572 | EXPORT_SYMBOL(iterate_supers_type); | |
573 | ||
1da177e4 LT |
574 | /** |
575 | * get_super - get the superblock of a device | |
576 | * @bdev: device to get the superblock for | |
577 | * | |
578 | * Scans the superblock list and finds the superblock of the file system | |
579 | * mounted on the device given. %NULL is returned if no match is found. | |
580 | */ | |
581 | ||
df40c01a | 582 | struct super_block *get_super(struct block_device *bdev) |
1da177e4 | 583 | { |
618f0636 KK |
584 | struct super_block *sb; |
585 | ||
1da177e4 LT |
586 | if (!bdev) |
587 | return NULL; | |
618f0636 | 588 | |
1da177e4 | 589 | spin_lock(&sb_lock); |
618f0636 KK |
590 | rescan: |
591 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 592 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 593 | continue; |
618f0636 KK |
594 | if (sb->s_bdev == bdev) { |
595 | sb->s_count++; | |
1da177e4 | 596 | spin_unlock(&sb_lock); |
618f0636 | 597 | down_read(&sb->s_umount); |
df40c01a | 598 | /* still alive? */ |
dabe0dc1 | 599 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
618f0636 KK |
600 | return sb; |
601 | up_read(&sb->s_umount); | |
df40c01a | 602 | /* nope, got unmounted */ |
618f0636 | 603 | spin_lock(&sb_lock); |
df40c01a AV |
604 | __put_super(sb); |
605 | goto rescan; | |
1da177e4 LT |
606 | } |
607 | } | |
608 | spin_unlock(&sb_lock); | |
609 | return NULL; | |
610 | } | |
611 | ||
612 | EXPORT_SYMBOL(get_super); | |
4504230a | 613 | |
6b6dc836 JK |
614 | /** |
615 | * get_super_thawed - get thawed superblock of a device | |
616 | * @bdev: device to get the superblock for | |
617 | * | |
618 | * Scans the superblock list and finds the superblock of the file system | |
619 | * mounted on the device. The superblock is returned once it is thawed | |
620 | * (or immediately if it was not frozen). %NULL is returned if no match | |
621 | * is found. | |
622 | */ | |
623 | struct super_block *get_super_thawed(struct block_device *bdev) | |
624 | { | |
625 | while (1) { | |
626 | struct super_block *s = get_super(bdev); | |
5accdf82 | 627 | if (!s || s->s_writers.frozen == SB_UNFROZEN) |
6b6dc836 JK |
628 | return s; |
629 | up_read(&s->s_umount); | |
5accdf82 JK |
630 | wait_event(s->s_writers.wait_unfrozen, |
631 | s->s_writers.frozen == SB_UNFROZEN); | |
6b6dc836 JK |
632 | put_super(s); |
633 | } | |
634 | } | |
635 | EXPORT_SYMBOL(get_super_thawed); | |
636 | ||
4504230a CH |
637 | /** |
638 | * get_active_super - get an active reference to the superblock of a device | |
639 | * @bdev: device to get the superblock for | |
640 | * | |
641 | * Scans the superblock list and finds the superblock of the file system | |
642 | * mounted on the device given. Returns the superblock with an active | |
d3f21473 | 643 | * reference or %NULL if none was found. |
4504230a CH |
644 | */ |
645 | struct super_block *get_active_super(struct block_device *bdev) | |
646 | { | |
647 | struct super_block *sb; | |
648 | ||
649 | if (!bdev) | |
650 | return NULL; | |
651 | ||
1494583d | 652 | restart: |
4504230a CH |
653 | spin_lock(&sb_lock); |
654 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 655 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 656 | continue; |
1494583d | 657 | if (sb->s_bdev == bdev) { |
acfec9a5 | 658 | if (!grab_super(sb)) |
1494583d | 659 | goto restart; |
acfec9a5 AV |
660 | up_write(&sb->s_umount); |
661 | return sb; | |
1494583d | 662 | } |
4504230a CH |
663 | } |
664 | spin_unlock(&sb_lock); | |
665 | return NULL; | |
666 | } | |
1da177e4 | 667 | |
df40c01a | 668 | struct super_block *user_get_super(dev_t dev) |
1da177e4 | 669 | { |
618f0636 | 670 | struct super_block *sb; |
1da177e4 | 671 | |
1da177e4 | 672 | spin_lock(&sb_lock); |
618f0636 KK |
673 | rescan: |
674 | list_for_each_entry(sb, &super_blocks, s_list) { | |
a5166169 | 675 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 676 | continue; |
618f0636 KK |
677 | if (sb->s_dev == dev) { |
678 | sb->s_count++; | |
1da177e4 | 679 | spin_unlock(&sb_lock); |
618f0636 | 680 | down_read(&sb->s_umount); |
df40c01a | 681 | /* still alive? */ |
dabe0dc1 | 682 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
618f0636 KK |
683 | return sb; |
684 | up_read(&sb->s_umount); | |
df40c01a | 685 | /* nope, got unmounted */ |
618f0636 | 686 | spin_lock(&sb_lock); |
df40c01a AV |
687 | __put_super(sb); |
688 | goto rescan; | |
1da177e4 LT |
689 | } |
690 | } | |
691 | spin_unlock(&sb_lock); | |
692 | return NULL; | |
693 | } | |
694 | ||
1da177e4 LT |
695 | /** |
696 | * do_remount_sb - asks filesystem to change mount options. | |
697 | * @sb: superblock in question | |
698 | * @flags: numeric part of options | |
699 | * @data: the rest of options | |
700 | * @force: whether or not to force the change | |
701 | * | |
702 | * Alters the mount options of a mounted file system. | |
703 | */ | |
704 | int do_remount_sb(struct super_block *sb, int flags, void *data, int force) | |
705 | { | |
706 | int retval; | |
c79d967d | 707 | int remount_ro; |
4504230a | 708 | |
5accdf82 | 709 | if (sb->s_writers.frozen != SB_UNFROZEN) |
4504230a CH |
710 | return -EBUSY; |
711 | ||
9361401e | 712 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
713 | if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev)) |
714 | return -EACCES; | |
9361401e | 715 | #endif |
4504230a | 716 | |
1da177e4 LT |
717 | if (flags & MS_RDONLY) |
718 | acct_auto_close(sb); | |
719 | shrink_dcache_sb(sb); | |
60b0680f | 720 | sync_filesystem(sb); |
1da177e4 | 721 | |
d208bbdd | 722 | remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY); |
d208bbdd | 723 | |
1da177e4 LT |
724 | /* If we are remounting RDONLY and current sb is read/write, |
725 | make sure there are no rw files opened */ | |
d208bbdd | 726 | if (remount_ro) { |
4ed5e82f | 727 | if (force) { |
1da177e4 | 728 | mark_files_ro(sb); |
4ed5e82f MS |
729 | } else { |
730 | retval = sb_prepare_remount_readonly(sb); | |
731 | if (retval) | |
732 | return retval; | |
4ed5e82f | 733 | } |
1da177e4 LT |
734 | } |
735 | ||
736 | if (sb->s_op->remount_fs) { | |
1da177e4 | 737 | retval = sb->s_op->remount_fs(sb, &flags, data); |
2833eb2b MS |
738 | if (retval) { |
739 | if (!force) | |
4ed5e82f | 740 | goto cancel_readonly; |
2833eb2b MS |
741 | /* If forced remount, go ahead despite any errors */ |
742 | WARN(1, "forced remount of a %s fs returned %i\n", | |
743 | sb->s_type->name, retval); | |
744 | } | |
1da177e4 LT |
745 | } |
746 | sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK); | |
4ed5e82f MS |
747 | /* Needs to be ordered wrt mnt_is_readonly() */ |
748 | smp_wmb(); | |
749 | sb->s_readonly_remount = 0; | |
c79d967d | 750 | |
d208bbdd NP |
751 | /* |
752 | * Some filesystems modify their metadata via some other path than the | |
753 | * bdev buffer cache (eg. use a private mapping, or directories in | |
754 | * pagecache, etc). Also file data modifications go via their own | |
755 | * mappings. So If we try to mount readonly then copy the filesystem | |
756 | * from bdev, we could get stale data, so invalidate it to give a best | |
757 | * effort at coherency. | |
758 | */ | |
759 | if (remount_ro && sb->s_bdev) | |
760 | invalidate_bdev(sb->s_bdev); | |
1da177e4 | 761 | return 0; |
4ed5e82f MS |
762 | |
763 | cancel_readonly: | |
764 | sb->s_readonly_remount = 0; | |
765 | return retval; | |
1da177e4 LT |
766 | } |
767 | ||
a2a9537a | 768 | static void do_emergency_remount(struct work_struct *work) |
1da177e4 | 769 | { |
dca33252 | 770 | struct super_block *sb, *p = NULL; |
1da177e4 LT |
771 | |
772 | spin_lock(&sb_lock); | |
dca33252 | 773 | list_for_each_entry(sb, &super_blocks, s_list) { |
a5166169 | 774 | if (hlist_unhashed(&sb->s_instances)) |
551de6f3 | 775 | continue; |
1da177e4 LT |
776 | sb->s_count++; |
777 | spin_unlock(&sb_lock); | |
443b94ba | 778 | down_write(&sb->s_umount); |
dabe0dc1 AV |
779 | if (sb->s_root && sb->s_bdev && (sb->s_flags & MS_BORN) && |
780 | !(sb->s_flags & MS_RDONLY)) { | |
1da177e4 | 781 | /* |
1da177e4 LT |
782 | * What lock protects sb->s_flags?? |
783 | */ | |
1da177e4 | 784 | do_remount_sb(sb, MS_RDONLY, NULL, 1); |
1da177e4 | 785 | } |
443b94ba | 786 | up_write(&sb->s_umount); |
1da177e4 | 787 | spin_lock(&sb_lock); |
dca33252 AV |
788 | if (p) |
789 | __put_super(p); | |
790 | p = sb; | |
1da177e4 | 791 | } |
dca33252 AV |
792 | if (p) |
793 | __put_super(p); | |
1da177e4 | 794 | spin_unlock(&sb_lock); |
a2a9537a | 795 | kfree(work); |
1da177e4 LT |
796 | printk("Emergency Remount complete\n"); |
797 | } | |
798 | ||
799 | void emergency_remount(void) | |
800 | { | |
a2a9537a JA |
801 | struct work_struct *work; |
802 | ||
803 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | |
804 | if (work) { | |
805 | INIT_WORK(work, do_emergency_remount); | |
806 | schedule_work(work); | |
807 | } | |
1da177e4 LT |
808 | } |
809 | ||
810 | /* | |
811 | * Unnamed block devices are dummy devices used by virtual | |
812 | * filesystems which don't use real block-devices. -- jrs | |
813 | */ | |
814 | ||
ad76cbc6 | 815 | static DEFINE_IDA(unnamed_dev_ida); |
1da177e4 | 816 | static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */ |
c63e09ec | 817 | static int unnamed_dev_start = 0; /* don't bother trying below it */ |
1da177e4 | 818 | |
0ee5dc67 | 819 | int get_anon_bdev(dev_t *p) |
1da177e4 LT |
820 | { |
821 | int dev; | |
822 | int error; | |
823 | ||
824 | retry: | |
ad76cbc6 | 825 | if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0) |
1da177e4 LT |
826 | return -ENOMEM; |
827 | spin_lock(&unnamed_dev_lock); | |
c63e09ec | 828 | error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev); |
f21f6220 AV |
829 | if (!error) |
830 | unnamed_dev_start = dev + 1; | |
1da177e4 LT |
831 | spin_unlock(&unnamed_dev_lock); |
832 | if (error == -EAGAIN) | |
833 | /* We raced and lost with another CPU. */ | |
834 | goto retry; | |
835 | else if (error) | |
836 | return -EAGAIN; | |
837 | ||
e8c8d1bc | 838 | if (dev == (1 << MINORBITS)) { |
1da177e4 | 839 | spin_lock(&unnamed_dev_lock); |
ad76cbc6 | 840 | ida_remove(&unnamed_dev_ida, dev); |
f21f6220 AV |
841 | if (unnamed_dev_start > dev) |
842 | unnamed_dev_start = dev; | |
1da177e4 LT |
843 | spin_unlock(&unnamed_dev_lock); |
844 | return -EMFILE; | |
845 | } | |
0ee5dc67 | 846 | *p = MKDEV(0, dev & MINORMASK); |
1da177e4 LT |
847 | return 0; |
848 | } | |
0ee5dc67 | 849 | EXPORT_SYMBOL(get_anon_bdev); |
1da177e4 | 850 | |
0ee5dc67 | 851 | void free_anon_bdev(dev_t dev) |
1da177e4 | 852 | { |
0ee5dc67 | 853 | int slot = MINOR(dev); |
1da177e4 | 854 | spin_lock(&unnamed_dev_lock); |
ad76cbc6 | 855 | ida_remove(&unnamed_dev_ida, slot); |
c63e09ec AV |
856 | if (slot < unnamed_dev_start) |
857 | unnamed_dev_start = slot; | |
1da177e4 LT |
858 | spin_unlock(&unnamed_dev_lock); |
859 | } | |
0ee5dc67 AV |
860 | EXPORT_SYMBOL(free_anon_bdev); |
861 | ||
862 | int set_anon_super(struct super_block *s, void *data) | |
863 | { | |
864 | int error = get_anon_bdev(&s->s_dev); | |
865 | if (!error) | |
866 | s->s_bdi = &noop_backing_dev_info; | |
867 | return error; | |
868 | } | |
869 | ||
870 | EXPORT_SYMBOL(set_anon_super); | |
871 | ||
872 | void kill_anon_super(struct super_block *sb) | |
873 | { | |
874 | dev_t dev = sb->s_dev; | |
875 | generic_shutdown_super(sb); | |
876 | free_anon_bdev(dev); | |
877 | } | |
1da177e4 LT |
878 | |
879 | EXPORT_SYMBOL(kill_anon_super); | |
880 | ||
1da177e4 LT |
881 | void kill_litter_super(struct super_block *sb) |
882 | { | |
883 | if (sb->s_root) | |
884 | d_genocide(sb->s_root); | |
885 | kill_anon_super(sb); | |
886 | } | |
887 | ||
888 | EXPORT_SYMBOL(kill_litter_super); | |
889 | ||
909e6d94 SH |
890 | static int ns_test_super(struct super_block *sb, void *data) |
891 | { | |
892 | return sb->s_fs_info == data; | |
893 | } | |
894 | ||
895 | static int ns_set_super(struct super_block *sb, void *data) | |
896 | { | |
897 | sb->s_fs_info = data; | |
898 | return set_anon_super(sb, NULL); | |
899 | } | |
900 | ||
ceefda69 AV |
901 | struct dentry *mount_ns(struct file_system_type *fs_type, int flags, |
902 | void *data, int (*fill_super)(struct super_block *, void *, int)) | |
909e6d94 SH |
903 | { |
904 | struct super_block *sb; | |
905 | ||
9249e17f | 906 | sb = sget(fs_type, ns_test_super, ns_set_super, flags, data); |
909e6d94 | 907 | if (IS_ERR(sb)) |
ceefda69 | 908 | return ERR_CAST(sb); |
909e6d94 SH |
909 | |
910 | if (!sb->s_root) { | |
911 | int err; | |
909e6d94 SH |
912 | err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0); |
913 | if (err) { | |
74dbbdd7 | 914 | deactivate_locked_super(sb); |
ceefda69 | 915 | return ERR_PTR(err); |
909e6d94 SH |
916 | } |
917 | ||
918 | sb->s_flags |= MS_ACTIVE; | |
919 | } | |
920 | ||
ceefda69 | 921 | return dget(sb->s_root); |
909e6d94 SH |
922 | } |
923 | ||
ceefda69 | 924 | EXPORT_SYMBOL(mount_ns); |
909e6d94 | 925 | |
9361401e | 926 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
927 | static int set_bdev_super(struct super_block *s, void *data) |
928 | { | |
929 | s->s_bdev = data; | |
930 | s->s_dev = s->s_bdev->bd_dev; | |
32a88aa1 JA |
931 | |
932 | /* | |
933 | * We set the bdi here to the queue backing, file systems can | |
934 | * overwrite this in ->fill_super() | |
935 | */ | |
936 | s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info; | |
1da177e4 LT |
937 | return 0; |
938 | } | |
939 | ||
940 | static int test_bdev_super(struct super_block *s, void *data) | |
941 | { | |
942 | return (void *)s->s_bdev == data; | |
943 | } | |
944 | ||
152a0836 | 945 | struct dentry *mount_bdev(struct file_system_type *fs_type, |
1da177e4 | 946 | int flags, const char *dev_name, void *data, |
152a0836 | 947 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
948 | { |
949 | struct block_device *bdev; | |
950 | struct super_block *s; | |
d4d77629 | 951 | fmode_t mode = FMODE_READ | FMODE_EXCL; |
1da177e4 LT |
952 | int error = 0; |
953 | ||
30c40d2c AV |
954 | if (!(flags & MS_RDONLY)) |
955 | mode |= FMODE_WRITE; | |
956 | ||
d4d77629 | 957 | bdev = blkdev_get_by_path(dev_name, mode, fs_type); |
1da177e4 | 958 | if (IS_ERR(bdev)) |
152a0836 | 959 | return ERR_CAST(bdev); |
1da177e4 LT |
960 | |
961 | /* | |
962 | * once the super is inserted into the list by sget, s_umount | |
963 | * will protect the lockfs code from trying to start a snapshot | |
964 | * while we are mounting | |
965 | */ | |
4fadd7bb CH |
966 | mutex_lock(&bdev->bd_fsfreeze_mutex); |
967 | if (bdev->bd_fsfreeze_count > 0) { | |
968 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
969 | error = -EBUSY; | |
970 | goto error_bdev; | |
971 | } | |
9249e17f DH |
972 | s = sget(fs_type, test_bdev_super, set_bdev_super, flags | MS_NOSEC, |
973 | bdev); | |
4fadd7bb | 974 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
1da177e4 | 975 | if (IS_ERR(s)) |
454e2398 | 976 | goto error_s; |
1da177e4 LT |
977 | |
978 | if (s->s_root) { | |
979 | if ((flags ^ s->s_flags) & MS_RDONLY) { | |
74dbbdd7 | 980 | deactivate_locked_super(s); |
454e2398 DH |
981 | error = -EBUSY; |
982 | goto error_bdev; | |
1da177e4 | 983 | } |
454e2398 | 984 | |
4f331f01 TH |
985 | /* |
986 | * s_umount nests inside bd_mutex during | |
e525fd89 TH |
987 | * __invalidate_device(). blkdev_put() acquires |
988 | * bd_mutex and can't be called under s_umount. Drop | |
989 | * s_umount temporarily. This is safe as we're | |
990 | * holding an active reference. | |
4f331f01 TH |
991 | */ |
992 | up_write(&s->s_umount); | |
d4d77629 | 993 | blkdev_put(bdev, mode); |
4f331f01 | 994 | down_write(&s->s_umount); |
1da177e4 LT |
995 | } else { |
996 | char b[BDEVNAME_SIZE]; | |
997 | ||
30c40d2c | 998 | s->s_mode = mode; |
1da177e4 | 999 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); |
e78c9a00 | 1000 | sb_set_blocksize(s, block_size(bdev)); |
9b04c997 | 1001 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1002 | if (error) { |
74dbbdd7 | 1003 | deactivate_locked_super(s); |
454e2398 | 1004 | goto error; |
fa675765 | 1005 | } |
454e2398 DH |
1006 | |
1007 | s->s_flags |= MS_ACTIVE; | |
87d8fe1e | 1008 | bdev->bd_super = s; |
1da177e4 LT |
1009 | } |
1010 | ||
152a0836 | 1011 | return dget(s->s_root); |
1da177e4 | 1012 | |
454e2398 DH |
1013 | error_s: |
1014 | error = PTR_ERR(s); | |
1015 | error_bdev: | |
d4d77629 | 1016 | blkdev_put(bdev, mode); |
454e2398 | 1017 | error: |
152a0836 AV |
1018 | return ERR_PTR(error); |
1019 | } | |
1020 | EXPORT_SYMBOL(mount_bdev); | |
1021 | ||
1da177e4 LT |
1022 | void kill_block_super(struct super_block *sb) |
1023 | { | |
1024 | struct block_device *bdev = sb->s_bdev; | |
30c40d2c | 1025 | fmode_t mode = sb->s_mode; |
1da177e4 | 1026 | |
ddbaaf30 | 1027 | bdev->bd_super = NULL; |
1da177e4 LT |
1028 | generic_shutdown_super(sb); |
1029 | sync_blockdev(bdev); | |
d4d77629 | 1030 | WARN_ON_ONCE(!(mode & FMODE_EXCL)); |
e525fd89 | 1031 | blkdev_put(bdev, mode | FMODE_EXCL); |
1da177e4 LT |
1032 | } |
1033 | ||
1034 | EXPORT_SYMBOL(kill_block_super); | |
9361401e | 1035 | #endif |
1da177e4 | 1036 | |
3c26ff6e | 1037 | struct dentry *mount_nodev(struct file_system_type *fs_type, |
1da177e4 | 1038 | int flags, void *data, |
3c26ff6e | 1039 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1040 | { |
1041 | int error; | |
9249e17f | 1042 | struct super_block *s = sget(fs_type, NULL, set_anon_super, flags, NULL); |
1da177e4 LT |
1043 | |
1044 | if (IS_ERR(s)) | |
3c26ff6e | 1045 | return ERR_CAST(s); |
1da177e4 | 1046 | |
9b04c997 | 1047 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1048 | if (error) { |
74dbbdd7 | 1049 | deactivate_locked_super(s); |
3c26ff6e | 1050 | return ERR_PTR(error); |
1da177e4 LT |
1051 | } |
1052 | s->s_flags |= MS_ACTIVE; | |
3c26ff6e | 1053 | return dget(s->s_root); |
1da177e4 | 1054 | } |
3c26ff6e AV |
1055 | EXPORT_SYMBOL(mount_nodev); |
1056 | ||
1da177e4 LT |
1057 | static int compare_single(struct super_block *s, void *p) |
1058 | { | |
1059 | return 1; | |
1060 | } | |
1061 | ||
fc14f2fe | 1062 | struct dentry *mount_single(struct file_system_type *fs_type, |
1da177e4 | 1063 | int flags, void *data, |
fc14f2fe | 1064 | int (*fill_super)(struct super_block *, void *, int)) |
1da177e4 LT |
1065 | { |
1066 | struct super_block *s; | |
1067 | int error; | |
1068 | ||
9249e17f | 1069 | s = sget(fs_type, compare_single, set_anon_super, flags, NULL); |
1da177e4 | 1070 | if (IS_ERR(s)) |
fc14f2fe | 1071 | return ERR_CAST(s); |
1da177e4 | 1072 | if (!s->s_root) { |
9b04c997 | 1073 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 | 1074 | if (error) { |
74dbbdd7 | 1075 | deactivate_locked_super(s); |
fc14f2fe | 1076 | return ERR_PTR(error); |
1da177e4 LT |
1077 | } |
1078 | s->s_flags |= MS_ACTIVE; | |
9329d1be KS |
1079 | } else { |
1080 | do_remount_sb(s, flags, data, 0); | |
1da177e4 | 1081 | } |
fc14f2fe AV |
1082 | return dget(s->s_root); |
1083 | } | |
1084 | EXPORT_SYMBOL(mount_single); | |
1085 | ||
9d412a43 AV |
1086 | struct dentry * |
1087 | mount_fs(struct file_system_type *type, int flags, const char *name, void *data) | |
1da177e4 | 1088 | { |
c96e41e9 | 1089 | struct dentry *root; |
9d412a43 | 1090 | struct super_block *sb; |
1da177e4 | 1091 | char *secdata = NULL; |
9d412a43 | 1092 | int error = -ENOMEM; |
8089352a | 1093 | |
e0007529 | 1094 | if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) { |
1da177e4 | 1095 | secdata = alloc_secdata(); |
454e2398 | 1096 | if (!secdata) |
9d412a43 | 1097 | goto out; |
1da177e4 | 1098 | |
e0007529 | 1099 | error = security_sb_copy_data(data, secdata); |
454e2398 | 1100 | if (error) |
1da177e4 | 1101 | goto out_free_secdata; |
1da177e4 LT |
1102 | } |
1103 | ||
1a102ff9 AV |
1104 | root = type->mount(type, flags, name, data); |
1105 | if (IS_ERR(root)) { | |
1106 | error = PTR_ERR(root); | |
1107 | goto out_free_secdata; | |
c96e41e9 | 1108 | } |
9d412a43 AV |
1109 | sb = root->d_sb; |
1110 | BUG_ON(!sb); | |
1111 | WARN_ON(!sb->s_bdi); | |
6c510389 | 1112 | WARN_ON(sb->s_bdi == &default_backing_dev_info); |
9d412a43 | 1113 | sb->s_flags |= MS_BORN; |
454e2398 | 1114 | |
9d412a43 | 1115 | error = security_sb_kern_mount(sb, flags, secdata); |
5129a469 JE |
1116 | if (error) |
1117 | goto out_sb; | |
454e2398 | 1118 | |
42cb56ae JL |
1119 | /* |
1120 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE | |
1121 | * but s_maxbytes was an unsigned long long for many releases. Throw | |
1122 | * this warning for a little while to try and catch filesystems that | |
4358b567 | 1123 | * violate this rule. |
42cb56ae | 1124 | */ |
9d412a43 AV |
1125 | WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " |
1126 | "negative value (%lld)\n", type->name, sb->s_maxbytes); | |
42cb56ae | 1127 | |
9d412a43 | 1128 | up_write(&sb->s_umount); |
8680e22f | 1129 | free_secdata(secdata); |
9d412a43 | 1130 | return root; |
1da177e4 | 1131 | out_sb: |
9d412a43 AV |
1132 | dput(root); |
1133 | deactivate_locked_super(sb); | |
1da177e4 LT |
1134 | out_free_secdata: |
1135 | free_secdata(secdata); | |
1da177e4 | 1136 | out: |
454e2398 | 1137 | return ERR_PTR(error); |
1da177e4 LT |
1138 | } |
1139 | ||
5accdf82 JK |
1140 | /* |
1141 | * This is an internal function, please use sb_end_{write,pagefault,intwrite} | |
1142 | * instead. | |
1143 | */ | |
1144 | void __sb_end_write(struct super_block *sb, int level) | |
1145 | { | |
1146 | percpu_counter_dec(&sb->s_writers.counter[level-1]); | |
1147 | /* | |
1148 | * Make sure s_writers are updated before we wake up waiters in | |
1149 | * freeze_super(). | |
1150 | */ | |
1151 | smp_mb(); | |
1152 | if (waitqueue_active(&sb->s_writers.wait)) | |
1153 | wake_up(&sb->s_writers.wait); | |
1154 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _RET_IP_); | |
1155 | } | |
1156 | EXPORT_SYMBOL(__sb_end_write); | |
1157 | ||
1158 | #ifdef CONFIG_LOCKDEP | |
1159 | /* | |
1160 | * We want lockdep to tell us about possible deadlocks with freezing but | |
1161 | * it's it bit tricky to properly instrument it. Getting a freeze protection | |
1162 | * works as getting a read lock but there are subtle problems. XFS for example | |
1163 | * gets freeze protection on internal level twice in some cases, which is OK | |
1164 | * only because we already hold a freeze protection also on higher level. Due | |
1165 | * to these cases we have to tell lockdep we are doing trylock when we | |
1166 | * already hold a freeze protection for a higher freeze level. | |
1167 | */ | |
1168 | static void acquire_freeze_lock(struct super_block *sb, int level, bool trylock, | |
1169 | unsigned long ip) | |
1170 | { | |
1171 | int i; | |
1172 | ||
1173 | if (!trylock) { | |
1174 | for (i = 0; i < level - 1; i++) | |
1175 | if (lock_is_held(&sb->s_writers.lock_map[i])) { | |
1176 | trylock = true; | |
1177 | break; | |
1178 | } | |
1179 | } | |
1180 | rwsem_acquire_read(&sb->s_writers.lock_map[level-1], 0, trylock, ip); | |
1181 | } | |
1182 | #endif | |
1183 | ||
1184 | /* | |
1185 | * This is an internal function, please use sb_start_{write,pagefault,intwrite} | |
1186 | * instead. | |
1187 | */ | |
1188 | int __sb_start_write(struct super_block *sb, int level, bool wait) | |
1189 | { | |
1190 | retry: | |
1191 | if (unlikely(sb->s_writers.frozen >= level)) { | |
1192 | if (!wait) | |
1193 | return 0; | |
1194 | wait_event(sb->s_writers.wait_unfrozen, | |
1195 | sb->s_writers.frozen < level); | |
1196 | } | |
1197 | ||
1198 | #ifdef CONFIG_LOCKDEP | |
1199 | acquire_freeze_lock(sb, level, !wait, _RET_IP_); | |
1200 | #endif | |
1201 | percpu_counter_inc(&sb->s_writers.counter[level-1]); | |
1202 | /* | |
1203 | * Make sure counter is updated before we check for frozen. | |
1204 | * freeze_super() first sets frozen and then checks the counter. | |
1205 | */ | |
1206 | smp_mb(); | |
1207 | if (unlikely(sb->s_writers.frozen >= level)) { | |
1208 | __sb_end_write(sb, level); | |
1209 | goto retry; | |
1210 | } | |
1211 | return 1; | |
1212 | } | |
1213 | EXPORT_SYMBOL(__sb_start_write); | |
1214 | ||
1215 | /** | |
1216 | * sb_wait_write - wait until all writers to given file system finish | |
1217 | * @sb: the super for which we wait | |
1218 | * @level: type of writers we wait for (normal vs page fault) | |
1219 | * | |
1220 | * This function waits until there are no writers of given type to given file | |
1221 | * system. Caller of this function should make sure there can be no new writers | |
1222 | * of type @level before calling this function. Otherwise this function can | |
1223 | * livelock. | |
1224 | */ | |
1225 | static void sb_wait_write(struct super_block *sb, int level) | |
1226 | { | |
1227 | s64 writers; | |
1228 | ||
1229 | /* | |
1230 | * We just cycle-through lockdep here so that it does not complain | |
1231 | * about returning with lock to userspace | |
1232 | */ | |
1233 | rwsem_acquire(&sb->s_writers.lock_map[level-1], 0, 0, _THIS_IP_); | |
1234 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _THIS_IP_); | |
1235 | ||
1236 | do { | |
1237 | DEFINE_WAIT(wait); | |
1238 | ||
1239 | /* | |
1240 | * We use a barrier in prepare_to_wait() to separate setting | |
1241 | * of frozen and checking of the counter | |
1242 | */ | |
1243 | prepare_to_wait(&sb->s_writers.wait, &wait, | |
1244 | TASK_UNINTERRUPTIBLE); | |
1245 | ||
1246 | writers = percpu_counter_sum(&sb->s_writers.counter[level-1]); | |
1247 | if (writers) | |
1248 | schedule(); | |
1249 | ||
1250 | finish_wait(&sb->s_writers.wait, &wait); | |
1251 | } while (writers); | |
1252 | } | |
1253 | ||
18e9e510 | 1254 | /** |
7000d3c4 RD |
1255 | * freeze_super - lock the filesystem and force it into a consistent state |
1256 | * @sb: the super to lock | |
18e9e510 JB |
1257 | * |
1258 | * Syncs the super to make sure the filesystem is consistent and calls the fs's | |
1259 | * freeze_fs. Subsequent calls to this without first thawing the fs will return | |
1260 | * -EBUSY. | |
5accdf82 JK |
1261 | * |
1262 | * During this function, sb->s_writers.frozen goes through these values: | |
1263 | * | |
1264 | * SB_UNFROZEN: File system is normal, all writes progress as usual. | |
1265 | * | |
1266 | * SB_FREEZE_WRITE: The file system is in the process of being frozen. New | |
1267 | * writes should be blocked, though page faults are still allowed. We wait for | |
1268 | * all writes to complete and then proceed to the next stage. | |
1269 | * | |
1270 | * SB_FREEZE_PAGEFAULT: Freezing continues. Now also page faults are blocked | |
1271 | * but internal fs threads can still modify the filesystem (although they | |
1272 | * should not dirty new pages or inodes), writeback can run etc. After waiting | |
1273 | * for all running page faults we sync the filesystem which will clean all | |
1274 | * dirty pages and inodes (no new dirty pages or inodes can be created when | |
1275 | * sync is running). | |
1276 | * | |
1277 | * SB_FREEZE_FS: The file system is frozen. Now all internal sources of fs | |
1278 | * modification are blocked (e.g. XFS preallocation truncation on inode | |
1279 | * reclaim). This is usually implemented by blocking new transactions for | |
1280 | * filesystems that have them and need this additional guard. After all | |
1281 | * internal writers are finished we call ->freeze_fs() to finish filesystem | |
1282 | * freezing. Then we transition to SB_FREEZE_COMPLETE state. This state is | |
1283 | * mostly auxiliary for filesystems to verify they do not modify frozen fs. | |
1284 | * | |
1285 | * sb->s_writers.frozen is protected by sb->s_umount. | |
18e9e510 JB |
1286 | */ |
1287 | int freeze_super(struct super_block *sb) | |
1288 | { | |
1289 | int ret; | |
1290 | ||
1291 | atomic_inc(&sb->s_active); | |
1292 | down_write(&sb->s_umount); | |
5accdf82 | 1293 | if (sb->s_writers.frozen != SB_UNFROZEN) { |
18e9e510 JB |
1294 | deactivate_locked_super(sb); |
1295 | return -EBUSY; | |
1296 | } | |
1297 | ||
dabe0dc1 AV |
1298 | if (!(sb->s_flags & MS_BORN)) { |
1299 | up_write(&sb->s_umount); | |
1300 | return 0; /* sic - it's "nothing to do" */ | |
1301 | } | |
1302 | ||
18e9e510 | 1303 | if (sb->s_flags & MS_RDONLY) { |
5accdf82 JK |
1304 | /* Nothing to do really... */ |
1305 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
18e9e510 JB |
1306 | up_write(&sb->s_umount); |
1307 | return 0; | |
1308 | } | |
1309 | ||
5accdf82 JK |
1310 | /* From now on, no new normal writers can start */ |
1311 | sb->s_writers.frozen = SB_FREEZE_WRITE; | |
1312 | smp_wmb(); | |
1313 | ||
1314 | /* Release s_umount to preserve sb_start_write -> s_umount ordering */ | |
1315 | up_write(&sb->s_umount); | |
1316 | ||
1317 | sb_wait_write(sb, SB_FREEZE_WRITE); | |
1318 | ||
1319 | /* Now we go and block page faults... */ | |
1320 | down_write(&sb->s_umount); | |
1321 | sb->s_writers.frozen = SB_FREEZE_PAGEFAULT; | |
18e9e510 JB |
1322 | smp_wmb(); |
1323 | ||
5accdf82 JK |
1324 | sb_wait_write(sb, SB_FREEZE_PAGEFAULT); |
1325 | ||
1326 | /* All writers are done so after syncing there won't be dirty data */ | |
18e9e510 JB |
1327 | sync_filesystem(sb); |
1328 | ||
5accdf82 JK |
1329 | /* Now wait for internal filesystem counter */ |
1330 | sb->s_writers.frozen = SB_FREEZE_FS; | |
18e9e510 | 1331 | smp_wmb(); |
5accdf82 | 1332 | sb_wait_write(sb, SB_FREEZE_FS); |
18e9e510 | 1333 | |
18e9e510 JB |
1334 | if (sb->s_op->freeze_fs) { |
1335 | ret = sb->s_op->freeze_fs(sb); | |
1336 | if (ret) { | |
1337 | printk(KERN_ERR | |
1338 | "VFS:Filesystem freeze failed\n"); | |
5accdf82 | 1339 | sb->s_writers.frozen = SB_UNFROZEN; |
e1616300 | 1340 | smp_wmb(); |
5accdf82 | 1341 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 JB |
1342 | deactivate_locked_super(sb); |
1343 | return ret; | |
1344 | } | |
1345 | } | |
5accdf82 JK |
1346 | /* |
1347 | * This is just for debugging purposes so that fs can warn if it | |
1348 | * sees write activity when frozen is set to SB_FREEZE_COMPLETE. | |
1349 | */ | |
1350 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; | |
18e9e510 JB |
1351 | up_write(&sb->s_umount); |
1352 | return 0; | |
1353 | } | |
1354 | EXPORT_SYMBOL(freeze_super); | |
1355 | ||
1356 | /** | |
1357 | * thaw_super -- unlock filesystem | |
1358 | * @sb: the super to thaw | |
1359 | * | |
1360 | * Unlocks the filesystem and marks it writeable again after freeze_super(). | |
1361 | */ | |
1362 | int thaw_super(struct super_block *sb) | |
1363 | { | |
1364 | int error; | |
1365 | ||
1366 | down_write(&sb->s_umount); | |
5accdf82 | 1367 | if (sb->s_writers.frozen == SB_UNFROZEN) { |
18e9e510 JB |
1368 | up_write(&sb->s_umount); |
1369 | return -EINVAL; | |
1370 | } | |
1371 | ||
1372 | if (sb->s_flags & MS_RDONLY) | |
1373 | goto out; | |
1374 | ||
1375 | if (sb->s_op->unfreeze_fs) { | |
1376 | error = sb->s_op->unfreeze_fs(sb); | |
1377 | if (error) { | |
1378 | printk(KERN_ERR | |
1379 | "VFS:Filesystem thaw failed\n"); | |
18e9e510 JB |
1380 | up_write(&sb->s_umount); |
1381 | return error; | |
1382 | } | |
1383 | } | |
1384 | ||
1385 | out: | |
5accdf82 | 1386 | sb->s_writers.frozen = SB_UNFROZEN; |
18e9e510 | 1387 | smp_wmb(); |
5accdf82 | 1388 | wake_up(&sb->s_writers.wait_unfrozen); |
18e9e510 JB |
1389 | deactivate_locked_super(sb); |
1390 | ||
1391 | return 0; | |
1392 | } | |
1393 | EXPORT_SYMBOL(thaw_super); |