Root/
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: |
18 | * Torbjörn Lindh (torbjorn.lindh@gopta.se), April 14, 1996. |
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 | |
23 | #include <linux/export.h> |
24 | #include <linux/slab.h> |
25 | #include <linux/acct.h> |
26 | #include <linux/blkdev.h> |
27 | #include <linux/mount.h> |
28 | #include <linux/security.h> |
29 | #include <linux/writeback.h> /* for the emergency remount stuff */ |
30 | #include <linux/idr.h> |
31 | #include <linux/mutex.h> |
32 | #include <linux/backing-dev.h> |
33 | #include <linux/rculist_bl.h> |
34 | #include <linux/cleancache.h> |
35 | #include <linux/fsnotify.h> |
36 | #include <linux/lockdep.h> |
37 | #include "internal.h" |
38 | |
39 | |
40 | LIST_HEAD(super_blocks); |
41 | DEFINE_SPINLOCK(sb_lock); |
42 | |
43 | static char *sb_writers_name[SB_FREEZE_LEVELS] = { |
44 | "sb_writers", |
45 | "sb_pagefaults", |
46 | "sb_internal", |
47 | }; |
48 | |
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 unsigned long super_cache_scan(struct shrinker *shrink, |
57 | struct shrink_control *sc) |
58 | { |
59 | struct super_block *sb; |
60 | long fs_objects = 0; |
61 | long total_objects; |
62 | long freed = 0; |
63 | long dentries; |
64 | long inodes; |
65 | |
66 | sb = container_of(shrink, struct super_block, s_shrink); |
67 | |
68 | /* |
69 | * Deadlock avoidance. We may hold various FS locks, and we don't want |
70 | * to recurse into the FS that called us in clear_inode() and friends.. |
71 | */ |
72 | if (!(sc->gfp_mask & __GFP_FS)) |
73 | return SHRINK_STOP; |
74 | |
75 | if (!grab_super_passive(sb)) |
76 | return SHRINK_STOP; |
77 | |
78 | if (sb->s_op->nr_cached_objects) |
79 | fs_objects = sb->s_op->nr_cached_objects(sb, sc->nid); |
80 | |
81 | inodes = list_lru_count_node(&sb->s_inode_lru, sc->nid); |
82 | dentries = list_lru_count_node(&sb->s_dentry_lru, sc->nid); |
83 | total_objects = dentries + inodes + fs_objects + 1; |
84 | |
85 | /* proportion the scan between the caches */ |
86 | dentries = mult_frac(sc->nr_to_scan, dentries, total_objects); |
87 | inodes = mult_frac(sc->nr_to_scan, inodes, total_objects); |
88 | |
89 | /* |
90 | * prune the dcache first as the icache is pinned by it, then |
91 | * prune the icache, followed by the filesystem specific caches |
92 | */ |
93 | freed = prune_dcache_sb(sb, dentries, sc->nid); |
94 | freed += prune_icache_sb(sb, inodes, sc->nid); |
95 | |
96 | if (fs_objects) { |
97 | fs_objects = mult_frac(sc->nr_to_scan, fs_objects, |
98 | total_objects); |
99 | freed += sb->s_op->free_cached_objects(sb, fs_objects, |
100 | sc->nid); |
101 | } |
102 | |
103 | drop_super(sb); |
104 | return freed; |
105 | } |
106 | |
107 | static unsigned long super_cache_count(struct shrinker *shrink, |
108 | struct shrink_control *sc) |
109 | { |
110 | struct super_block *sb; |
111 | long total_objects = 0; |
112 | |
113 | sb = container_of(shrink, struct super_block, s_shrink); |
114 | |
115 | /* |
116 | * Don't call grab_super_passive as it is a potential |
117 | * scalability bottleneck. The counts could get updated |
118 | * between super_cache_count and super_cache_scan anyway. |
119 | * Call to super_cache_count with shrinker_rwsem held |
120 | * ensures the safety of call to list_lru_count_node() and |
121 | * s_op->nr_cached_objects(). |
122 | */ |
123 | if (sb->s_op && sb->s_op->nr_cached_objects) |
124 | total_objects = sb->s_op->nr_cached_objects(sb, |
125 | sc->nid); |
126 | |
127 | total_objects += list_lru_count_node(&sb->s_dentry_lru, |
128 | sc->nid); |
129 | total_objects += list_lru_count_node(&sb->s_inode_lru, |
130 | sc->nid); |
131 | |
132 | total_objects = vfs_pressure_ratio(total_objects); |
133 | return total_objects; |
134 | } |
135 | |
136 | /** |
137 | * destroy_super - frees a superblock |
138 | * @s: superblock to free |
139 | * |
140 | * Frees a superblock. |
141 | */ |
142 | static void destroy_super(struct super_block *s) |
143 | { |
144 | int i; |
145 | list_lru_destroy(&s->s_dentry_lru); |
146 | list_lru_destroy(&s->s_inode_lru); |
147 | for (i = 0; i < SB_FREEZE_LEVELS; i++) |
148 | percpu_counter_destroy(&s->s_writers.counter[i]); |
149 | security_sb_free(s); |
150 | WARN_ON(!list_empty(&s->s_mounts)); |
151 | kfree(s->s_subtype); |
152 | kfree(s->s_options); |
153 | kfree_rcu(s, rcu); |
154 | } |
155 | |
156 | /** |
157 | * alloc_super - create new superblock |
158 | * @type: filesystem type superblock should belong to |
159 | * @flags: the mount flags |
160 | * |
161 | * Allocates and initializes a new &struct super_block. alloc_super() |
162 | * returns a pointer new superblock or %NULL if allocation had failed. |
163 | */ |
164 | static struct super_block *alloc_super(struct file_system_type *type, int flags) |
165 | { |
166 | struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER); |
167 | static const struct super_operations default_op; |
168 | int i; |
169 | |
170 | if (!s) |
171 | return NULL; |
172 | |
173 | INIT_LIST_HEAD(&s->s_mounts); |
174 | |
175 | if (security_sb_alloc(s)) |
176 | goto fail; |
177 | |
178 | for (i = 0; i < SB_FREEZE_LEVELS; i++) { |
179 | if (percpu_counter_init(&s->s_writers.counter[i], 0) < 0) |
180 | goto fail; |
181 | lockdep_init_map(&s->s_writers.lock_map[i], sb_writers_name[i], |
182 | &type->s_writers_key[i], 0); |
183 | } |
184 | init_waitqueue_head(&s->s_writers.wait); |
185 | init_waitqueue_head(&s->s_writers.wait_unfrozen); |
186 | s->s_flags = flags; |
187 | s->s_bdi = &default_backing_dev_info; |
188 | INIT_HLIST_NODE(&s->s_instances); |
189 | INIT_HLIST_BL_HEAD(&s->s_anon); |
190 | INIT_LIST_HEAD(&s->s_inodes); |
191 | |
192 | if (list_lru_init(&s->s_dentry_lru)) |
193 | goto fail; |
194 | if (list_lru_init(&s->s_inode_lru)) |
195 | goto fail; |
196 | |
197 | init_rwsem(&s->s_umount); |
198 | lockdep_set_class(&s->s_umount, &type->s_umount_key); |
199 | /* |
200 | * sget() can have s_umount recursion. |
201 | * |
202 | * When it cannot find a suitable sb, it allocates a new |
203 | * one (this one), and tries again to find a suitable old |
204 | * one. |
205 | * |
206 | * In case that succeeds, it will acquire the s_umount |
207 | * lock of the old one. Since these are clearly distrinct |
208 | * locks, and this object isn't exposed yet, there's no |
209 | * risk of deadlocks. |
210 | * |
211 | * Annotate this by putting this lock in a different |
212 | * subclass. |
213 | */ |
214 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); |
215 | s->s_count = 1; |
216 | atomic_set(&s->s_active, 1); |
217 | mutex_init(&s->s_vfs_rename_mutex); |
218 | lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key); |
219 | mutex_init(&s->s_dquot.dqio_mutex); |
220 | mutex_init(&s->s_dquot.dqonoff_mutex); |
221 | init_rwsem(&s->s_dquot.dqptr_sem); |
222 | s->s_maxbytes = MAX_NON_LFS; |
223 | s->s_op = &default_op; |
224 | s->s_time_gran = 1000000000; |
225 | s->cleancache_poolid = -1; |
226 | |
227 | s->s_shrink.seeks = DEFAULT_SEEKS; |
228 | s->s_shrink.scan_objects = super_cache_scan; |
229 | s->s_shrink.count_objects = super_cache_count; |
230 | s->s_shrink.batch = 1024; |
231 | s->s_shrink.flags = SHRINKER_NUMA_AWARE; |
232 | return s; |
233 | |
234 | fail: |
235 | destroy_super(s); |
236 | return NULL; |
237 | } |
238 | |
239 | /* Superblock refcounting */ |
240 | |
241 | /* |
242 | * Drop a superblock's refcount. The caller must hold sb_lock. |
243 | */ |
244 | static void __put_super(struct super_block *sb) |
245 | { |
246 | if (!--sb->s_count) { |
247 | list_del_init(&sb->s_list); |
248 | destroy_super(sb); |
249 | } |
250 | } |
251 | |
252 | /** |
253 | * put_super - drop a temporary reference to superblock |
254 | * @sb: superblock in question |
255 | * |
256 | * Drops a temporary reference, frees superblock if there's no |
257 | * references left. |
258 | */ |
259 | static void put_super(struct super_block *sb) |
260 | { |
261 | spin_lock(&sb_lock); |
262 | __put_super(sb); |
263 | spin_unlock(&sb_lock); |
264 | } |
265 | |
266 | |
267 | /** |
268 | * deactivate_locked_super - drop an active reference to superblock |
269 | * @s: superblock to deactivate |
270 | * |
271 | * Drops an active reference to superblock, converting it into a temprory |
272 | * one if there is no other active references left. In that case we |
273 | * tell fs driver to shut it down and drop the temporary reference we |
274 | * had just acquired. |
275 | * |
276 | * Caller holds exclusive lock on superblock; that lock is released. |
277 | */ |
278 | void deactivate_locked_super(struct super_block *s) |
279 | { |
280 | struct file_system_type *fs = s->s_type; |
281 | if (atomic_dec_and_test(&s->s_active)) { |
282 | cleancache_invalidate_fs(s); |
283 | unregister_shrinker(&s->s_shrink); |
284 | fs->kill_sb(s); |
285 | |
286 | put_filesystem(fs); |
287 | put_super(s); |
288 | } else { |
289 | up_write(&s->s_umount); |
290 | } |
291 | } |
292 | |
293 | EXPORT_SYMBOL(deactivate_locked_super); |
294 | |
295 | /** |
296 | * deactivate_super - drop an active reference to superblock |
297 | * @s: superblock to deactivate |
298 | * |
299 | * Variant of deactivate_locked_super(), except that superblock is *not* |
300 | * locked by caller. If we are going to drop the final active reference, |
301 | * lock will be acquired prior to that. |
302 | */ |
303 | void deactivate_super(struct super_block *s) |
304 | { |
305 | if (!atomic_add_unless(&s->s_active, -1, 1)) { |
306 | down_write(&s->s_umount); |
307 | deactivate_locked_super(s); |
308 | } |
309 | } |
310 | |
311 | EXPORT_SYMBOL(deactivate_super); |
312 | |
313 | /** |
314 | * grab_super - acquire an active reference |
315 | * @s: reference we are trying to make active |
316 | * |
317 | * Tries to acquire an active reference. grab_super() is used when we |
318 | * had just found a superblock in super_blocks or fs_type->fs_supers |
319 | * and want to turn it into a full-blown active reference. grab_super() |
320 | * is called with sb_lock held and drops it. Returns 1 in case of |
321 | * success, 0 if we had failed (superblock contents was already dead or |
322 | * dying when grab_super() had been called). Note that this is only |
323 | * called for superblocks not in rundown mode (== ones still on ->fs_supers |
324 | * of their type), so increment of ->s_count is OK here. |
325 | */ |
326 | static int grab_super(struct super_block *s) __releases(sb_lock) |
327 | { |
328 | s->s_count++; |
329 | spin_unlock(&sb_lock); |
330 | down_write(&s->s_umount); |
331 | if ((s->s_flags & MS_BORN) && atomic_inc_not_zero(&s->s_active)) { |
332 | put_super(s); |
333 | return 1; |
334 | } |
335 | up_write(&s->s_umount); |
336 | put_super(s); |
337 | return 0; |
338 | } |
339 | |
340 | /* |
341 | * grab_super_passive - acquire a passive reference |
342 | * @sb: reference we are trying to grab |
343 | * |
344 | * Tries to acquire a passive reference. This is used in places where we |
345 | * cannot take an active reference but we need to ensure that the |
346 | * superblock does not go away while we are working on it. It returns |
347 | * false if a reference was not gained, and returns true with the s_umount |
348 | * lock held in read mode if a reference is gained. On successful return, |
349 | * the caller must drop the s_umount lock and the passive reference when |
350 | * done. |
351 | */ |
352 | bool grab_super_passive(struct super_block *sb) |
353 | { |
354 | spin_lock(&sb_lock); |
355 | if (hlist_unhashed(&sb->s_instances)) { |
356 | spin_unlock(&sb_lock); |
357 | return false; |
358 | } |
359 | |
360 | sb->s_count++; |
361 | spin_unlock(&sb_lock); |
362 | |
363 | if (down_read_trylock(&sb->s_umount)) { |
364 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
365 | return true; |
366 | up_read(&sb->s_umount); |
367 | } |
368 | |
369 | put_super(sb); |
370 | return false; |
371 | } |
372 | |
373 | /** |
374 | * generic_shutdown_super - common helper for ->kill_sb() |
375 | * @sb: superblock to kill |
376 | * |
377 | * generic_shutdown_super() does all fs-independent work on superblock |
378 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects |
379 | * that need destruction out of superblock, call generic_shutdown_super() |
380 | * and release aforementioned objects. Note: dentries and inodes _are_ |
381 | * taken care of and do not need specific handling. |
382 | * |
383 | * Upon calling this function, the filesystem may no longer alter or |
384 | * rearrange the set of dentries belonging to this super_block, nor may it |
385 | * change the attachments of dentries to inodes. |
386 | */ |
387 | void generic_shutdown_super(struct super_block *sb) |
388 | { |
389 | const struct super_operations *sop = sb->s_op; |
390 | |
391 | if (sb->s_root) { |
392 | shrink_dcache_for_umount(sb); |
393 | sync_filesystem(sb); |
394 | sb->s_flags &= ~MS_ACTIVE; |
395 | |
396 | fsnotify_unmount_inodes(&sb->s_inodes); |
397 | |
398 | evict_inodes(sb); |
399 | |
400 | if (sb->s_dio_done_wq) { |
401 | destroy_workqueue(sb->s_dio_done_wq); |
402 | sb->s_dio_done_wq = NULL; |
403 | } |
404 | |
405 | if (sop->put_super) |
406 | sop->put_super(sb); |
407 | |
408 | if (!list_empty(&sb->s_inodes)) { |
409 | printk("VFS: Busy inodes after unmount of %s. " |
410 | "Self-destruct in 5 seconds. Have a nice day...\n", |
411 | sb->s_id); |
412 | } |
413 | } |
414 | spin_lock(&sb_lock); |
415 | /* should be initialized for __put_super_and_need_restart() */ |
416 | hlist_del_init(&sb->s_instances); |
417 | spin_unlock(&sb_lock); |
418 | up_write(&sb->s_umount); |
419 | } |
420 | |
421 | EXPORT_SYMBOL(generic_shutdown_super); |
422 | |
423 | /** |
424 | * sget - find or create a superblock |
425 | * @type: filesystem type superblock should belong to |
426 | * @test: comparison callback |
427 | * @set: setup callback |
428 | * @flags: mount flags |
429 | * @data: argument to each of them |
430 | */ |
431 | struct super_block *sget(struct file_system_type *type, |
432 | int (*test)(struct super_block *,void *), |
433 | int (*set)(struct super_block *,void *), |
434 | int flags, |
435 | void *data) |
436 | { |
437 | struct super_block *s = NULL; |
438 | struct super_block *old; |
439 | int err; |
440 | |
441 | retry: |
442 | spin_lock(&sb_lock); |
443 | if (test) { |
444 | hlist_for_each_entry(old, &type->fs_supers, s_instances) { |
445 | if (!test(old, data)) |
446 | continue; |
447 | if (!grab_super(old)) |
448 | goto retry; |
449 | if (s) { |
450 | up_write(&s->s_umount); |
451 | destroy_super(s); |
452 | s = NULL; |
453 | } |
454 | return old; |
455 | } |
456 | } |
457 | if (!s) { |
458 | spin_unlock(&sb_lock); |
459 | s = alloc_super(type, flags); |
460 | if (!s) |
461 | return ERR_PTR(-ENOMEM); |
462 | goto retry; |
463 | } |
464 | |
465 | err = set(s, data); |
466 | if (err) { |
467 | spin_unlock(&sb_lock); |
468 | up_write(&s->s_umount); |
469 | destroy_super(s); |
470 | return ERR_PTR(err); |
471 | } |
472 | s->s_type = type; |
473 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); |
474 | list_add_tail(&s->s_list, &super_blocks); |
475 | hlist_add_head(&s->s_instances, &type->fs_supers); |
476 | spin_unlock(&sb_lock); |
477 | get_filesystem(type); |
478 | register_shrinker(&s->s_shrink); |
479 | return s; |
480 | } |
481 | |
482 | EXPORT_SYMBOL(sget); |
483 | |
484 | void drop_super(struct super_block *sb) |
485 | { |
486 | up_read(&sb->s_umount); |
487 | put_super(sb); |
488 | } |
489 | |
490 | EXPORT_SYMBOL(drop_super); |
491 | |
492 | /** |
493 | * iterate_supers - call function for all active superblocks |
494 | * @f: function to call |
495 | * @arg: argument to pass to it |
496 | * |
497 | * Scans the superblock list and calls given function, passing it |
498 | * locked superblock and given argument. |
499 | */ |
500 | void iterate_supers(void (*f)(struct super_block *, void *), void *arg) |
501 | { |
502 | struct super_block *sb, *p = NULL; |
503 | |
504 | spin_lock(&sb_lock); |
505 | list_for_each_entry(sb, &super_blocks, s_list) { |
506 | if (hlist_unhashed(&sb->s_instances)) |
507 | continue; |
508 | sb->s_count++; |
509 | spin_unlock(&sb_lock); |
510 | |
511 | down_read(&sb->s_umount); |
512 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
513 | f(sb, arg); |
514 | up_read(&sb->s_umount); |
515 | |
516 | spin_lock(&sb_lock); |
517 | if (p) |
518 | __put_super(p); |
519 | p = sb; |
520 | } |
521 | if (p) |
522 | __put_super(p); |
523 | spin_unlock(&sb_lock); |
524 | } |
525 | |
526 | /** |
527 | * iterate_supers_type - call function for superblocks of given type |
528 | * @type: fs type |
529 | * @f: function to call |
530 | * @arg: argument to pass to it |
531 | * |
532 | * Scans the superblock list and calls given function, passing it |
533 | * locked superblock and given argument. |
534 | */ |
535 | void iterate_supers_type(struct file_system_type *type, |
536 | void (*f)(struct super_block *, void *), void *arg) |
537 | { |
538 | struct super_block *sb, *p = NULL; |
539 | |
540 | spin_lock(&sb_lock); |
541 | hlist_for_each_entry(sb, &type->fs_supers, s_instances) { |
542 | sb->s_count++; |
543 | spin_unlock(&sb_lock); |
544 | |
545 | down_read(&sb->s_umount); |
546 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
547 | f(sb, arg); |
548 | up_read(&sb->s_umount); |
549 | |
550 | spin_lock(&sb_lock); |
551 | if (p) |
552 | __put_super(p); |
553 | p = sb; |
554 | } |
555 | if (p) |
556 | __put_super(p); |
557 | spin_unlock(&sb_lock); |
558 | } |
559 | |
560 | EXPORT_SYMBOL(iterate_supers_type); |
561 | |
562 | /** |
563 | * get_super - get the superblock of a device |
564 | * @bdev: device to get the superblock for |
565 | * |
566 | * Scans the superblock list and finds the superblock of the file system |
567 | * mounted on the device given. %NULL is returned if no match is found. |
568 | */ |
569 | |
570 | struct super_block *get_super(struct block_device *bdev) |
571 | { |
572 | struct super_block *sb; |
573 | |
574 | if (!bdev) |
575 | return NULL; |
576 | |
577 | spin_lock(&sb_lock); |
578 | rescan: |
579 | list_for_each_entry(sb, &super_blocks, s_list) { |
580 | if (hlist_unhashed(&sb->s_instances)) |
581 | continue; |
582 | if (sb->s_bdev == bdev) { |
583 | sb->s_count++; |
584 | spin_unlock(&sb_lock); |
585 | down_read(&sb->s_umount); |
586 | /* still alive? */ |
587 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
588 | return sb; |
589 | up_read(&sb->s_umount); |
590 | /* nope, got unmounted */ |
591 | spin_lock(&sb_lock); |
592 | __put_super(sb); |
593 | goto rescan; |
594 | } |
595 | } |
596 | spin_unlock(&sb_lock); |
597 | return NULL; |
598 | } |
599 | |
600 | EXPORT_SYMBOL(get_super); |
601 | |
602 | /** |
603 | * get_super_thawed - get thawed superblock of a device |
604 | * @bdev: device to get the superblock for |
605 | * |
606 | * Scans the superblock list and finds the superblock of the file system |
607 | * mounted on the device. The superblock is returned once it is thawed |
608 | * (or immediately if it was not frozen). %NULL is returned if no match |
609 | * is found. |
610 | */ |
611 | struct super_block *get_super_thawed(struct block_device *bdev) |
612 | { |
613 | while (1) { |
614 | struct super_block *s = get_super(bdev); |
615 | if (!s || s->s_writers.frozen == SB_UNFROZEN) |
616 | return s; |
617 | up_read(&s->s_umount); |
618 | wait_event(s->s_writers.wait_unfrozen, |
619 | s->s_writers.frozen == SB_UNFROZEN); |
620 | put_super(s); |
621 | } |
622 | } |
623 | EXPORT_SYMBOL(get_super_thawed); |
624 | |
625 | /** |
626 | * get_active_super - get an active reference to the superblock of a device |
627 | * @bdev: device to get the superblock for |
628 | * |
629 | * Scans the superblock list and finds the superblock of the file system |
630 | * mounted on the device given. Returns the superblock with an active |
631 | * reference or %NULL if none was found. |
632 | */ |
633 | struct super_block *get_active_super(struct block_device *bdev) |
634 | { |
635 | struct super_block *sb; |
636 | |
637 | if (!bdev) |
638 | return NULL; |
639 | |
640 | restart: |
641 | spin_lock(&sb_lock); |
642 | list_for_each_entry(sb, &super_blocks, s_list) { |
643 | if (hlist_unhashed(&sb->s_instances)) |
644 | continue; |
645 | if (sb->s_bdev == bdev) { |
646 | if (!grab_super(sb)) |
647 | goto restart; |
648 | up_write(&sb->s_umount); |
649 | return sb; |
650 | } |
651 | } |
652 | spin_unlock(&sb_lock); |
653 | return NULL; |
654 | } |
655 | |
656 | struct super_block *user_get_super(dev_t dev) |
657 | { |
658 | struct super_block *sb; |
659 | |
660 | spin_lock(&sb_lock); |
661 | rescan: |
662 | list_for_each_entry(sb, &super_blocks, s_list) { |
663 | if (hlist_unhashed(&sb->s_instances)) |
664 | continue; |
665 | if (sb->s_dev == dev) { |
666 | sb->s_count++; |
667 | spin_unlock(&sb_lock); |
668 | down_read(&sb->s_umount); |
669 | /* still alive? */ |
670 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
671 | return sb; |
672 | up_read(&sb->s_umount); |
673 | /* nope, got unmounted */ |
674 | spin_lock(&sb_lock); |
675 | __put_super(sb); |
676 | goto rescan; |
677 | } |
678 | } |
679 | spin_unlock(&sb_lock); |
680 | return NULL; |
681 | } |
682 | |
683 | /** |
684 | * do_remount_sb - asks filesystem to change mount options. |
685 | * @sb: superblock in question |
686 | * @flags: numeric part of options |
687 | * @data: the rest of options |
688 | * @force: whether or not to force the change |
689 | * |
690 | * Alters the mount options of a mounted file system. |
691 | */ |
692 | int do_remount_sb(struct super_block *sb, int flags, void *data, int force) |
693 | { |
694 | int retval; |
695 | int remount_ro; |
696 | |
697 | if (sb->s_writers.frozen != SB_UNFROZEN) |
698 | return -EBUSY; |
699 | |
700 | #ifdef CONFIG_BLOCK |
701 | if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev)) |
702 | return -EACCES; |
703 | #endif |
704 | |
705 | if (flags & MS_RDONLY) |
706 | acct_auto_close(sb); |
707 | shrink_dcache_sb(sb); |
708 | |
709 | remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY); |
710 | |
711 | /* If we are remounting RDONLY and current sb is read/write, |
712 | make sure there are no rw files opened */ |
713 | if (remount_ro) { |
714 | if (force) { |
715 | sb->s_readonly_remount = 1; |
716 | smp_wmb(); |
717 | } else { |
718 | retval = sb_prepare_remount_readonly(sb); |
719 | if (retval) |
720 | return retval; |
721 | } |
722 | } |
723 | |
724 | if (sb->s_op->remount_fs) { |
725 | retval = sb->s_op->remount_fs(sb, &flags, data); |
726 | if (retval) { |
727 | if (!force) |
728 | goto cancel_readonly; |
729 | /* If forced remount, go ahead despite any errors */ |
730 | WARN(1, "forced remount of a %s fs returned %i\n", |
731 | sb->s_type->name, retval); |
732 | } |
733 | } |
734 | sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK); |
735 | /* Needs to be ordered wrt mnt_is_readonly() */ |
736 | smp_wmb(); |
737 | sb->s_readonly_remount = 0; |
738 | |
739 | /* |
740 | * Some filesystems modify their metadata via some other path than the |
741 | * bdev buffer cache (eg. use a private mapping, or directories in |
742 | * pagecache, etc). Also file data modifications go via their own |
743 | * mappings. So If we try to mount readonly then copy the filesystem |
744 | * from bdev, we could get stale data, so invalidate it to give a best |
745 | * effort at coherency. |
746 | */ |
747 | if (remount_ro && sb->s_bdev) |
748 | invalidate_bdev(sb->s_bdev); |
749 | return 0; |
750 | |
751 | cancel_readonly: |
752 | sb->s_readonly_remount = 0; |
753 | return retval; |
754 | } |
755 | |
756 | static void do_emergency_remount(struct work_struct *work) |
757 | { |
758 | struct super_block *sb, *p = NULL; |
759 | |
760 | spin_lock(&sb_lock); |
761 | list_for_each_entry(sb, &super_blocks, s_list) { |
762 | if (hlist_unhashed(&sb->s_instances)) |
763 | continue; |
764 | sb->s_count++; |
765 | spin_unlock(&sb_lock); |
766 | down_write(&sb->s_umount); |
767 | if (sb->s_root && sb->s_bdev && (sb->s_flags & MS_BORN) && |
768 | !(sb->s_flags & MS_RDONLY)) { |
769 | /* |
770 | * What lock protects sb->s_flags?? |
771 | */ |
772 | do_remount_sb(sb, MS_RDONLY, NULL, 1); |
773 | } |
774 | up_write(&sb->s_umount); |
775 | spin_lock(&sb_lock); |
776 | if (p) |
777 | __put_super(p); |
778 | p = sb; |
779 | } |
780 | if (p) |
781 | __put_super(p); |
782 | spin_unlock(&sb_lock); |
783 | kfree(work); |
784 | printk("Emergency Remount complete\n"); |
785 | } |
786 | |
787 | void emergency_remount(void) |
788 | { |
789 | struct work_struct *work; |
790 | |
791 | work = kmalloc(sizeof(*work), GFP_ATOMIC); |
792 | if (work) { |
793 | INIT_WORK(work, do_emergency_remount); |
794 | schedule_work(work); |
795 | } |
796 | } |
797 | |
798 | /* |
799 | * Unnamed block devices are dummy devices used by virtual |
800 | * filesystems which don't use real block-devices. -- jrs |
801 | */ |
802 | |
803 | static DEFINE_IDA(unnamed_dev_ida); |
804 | static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */ |
805 | /* Many userspace utilities consider an FSID of 0 invalid. |
806 | * Always return at least 1 from get_anon_bdev. |
807 | */ |
808 | static int unnamed_dev_start = 1; |
809 | |
810 | int get_anon_bdev(dev_t *p) |
811 | { |
812 | int dev; |
813 | int error; |
814 | |
815 | retry: |
816 | if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0) |
817 | return -ENOMEM; |
818 | spin_lock(&unnamed_dev_lock); |
819 | error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev); |
820 | if (!error) |
821 | unnamed_dev_start = dev + 1; |
822 | spin_unlock(&unnamed_dev_lock); |
823 | if (error == -EAGAIN) |
824 | /* We raced and lost with another CPU. */ |
825 | goto retry; |
826 | else if (error) |
827 | return -EAGAIN; |
828 | |
829 | if (dev == (1 << MINORBITS)) { |
830 | spin_lock(&unnamed_dev_lock); |
831 | ida_remove(&unnamed_dev_ida, dev); |
832 | if (unnamed_dev_start > dev) |
833 | unnamed_dev_start = dev; |
834 | spin_unlock(&unnamed_dev_lock); |
835 | return -EMFILE; |
836 | } |
837 | *p = MKDEV(0, dev & MINORMASK); |
838 | return 0; |
839 | } |
840 | EXPORT_SYMBOL(get_anon_bdev); |
841 | |
842 | void free_anon_bdev(dev_t dev) |
843 | { |
844 | int slot = MINOR(dev); |
845 | spin_lock(&unnamed_dev_lock); |
846 | ida_remove(&unnamed_dev_ida, slot); |
847 | if (slot < unnamed_dev_start) |
848 | unnamed_dev_start = slot; |
849 | spin_unlock(&unnamed_dev_lock); |
850 | } |
851 | EXPORT_SYMBOL(free_anon_bdev); |
852 | |
853 | int set_anon_super(struct super_block *s, void *data) |
854 | { |
855 | int error = get_anon_bdev(&s->s_dev); |
856 | if (!error) |
857 | s->s_bdi = &noop_backing_dev_info; |
858 | return error; |
859 | } |
860 | |
861 | EXPORT_SYMBOL(set_anon_super); |
862 | |
863 | void kill_anon_super(struct super_block *sb) |
864 | { |
865 | dev_t dev = sb->s_dev; |
866 | generic_shutdown_super(sb); |
867 | free_anon_bdev(dev); |
868 | } |
869 | |
870 | EXPORT_SYMBOL(kill_anon_super); |
871 | |
872 | void kill_litter_super(struct super_block *sb) |
873 | { |
874 | if (sb->s_root) |
875 | d_genocide(sb->s_root); |
876 | kill_anon_super(sb); |
877 | } |
878 | |
879 | EXPORT_SYMBOL(kill_litter_super); |
880 | |
881 | static int ns_test_super(struct super_block *sb, void *data) |
882 | { |
883 | return sb->s_fs_info == data; |
884 | } |
885 | |
886 | static int ns_set_super(struct super_block *sb, void *data) |
887 | { |
888 | sb->s_fs_info = data; |
889 | return set_anon_super(sb, NULL); |
890 | } |
891 | |
892 | struct dentry *mount_ns(struct file_system_type *fs_type, int flags, |
893 | void *data, int (*fill_super)(struct super_block *, void *, int)) |
894 | { |
895 | struct super_block *sb; |
896 | |
897 | sb = sget(fs_type, ns_test_super, ns_set_super, flags, data); |
898 | if (IS_ERR(sb)) |
899 | return ERR_CAST(sb); |
900 | |
901 | if (!sb->s_root) { |
902 | int err; |
903 | err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0); |
904 | if (err) { |
905 | deactivate_locked_super(sb); |
906 | return ERR_PTR(err); |
907 | } |
908 | |
909 | sb->s_flags |= MS_ACTIVE; |
910 | } |
911 | |
912 | return dget(sb->s_root); |
913 | } |
914 | |
915 | EXPORT_SYMBOL(mount_ns); |
916 | |
917 | #ifdef CONFIG_BLOCK |
918 | static int set_bdev_super(struct super_block *s, void *data) |
919 | { |
920 | s->s_bdev = data; |
921 | s->s_dev = s->s_bdev->bd_dev; |
922 | |
923 | /* |
924 | * We set the bdi here to the queue backing, file systems can |
925 | * overwrite this in ->fill_super() |
926 | */ |
927 | s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info; |
928 | return 0; |
929 | } |
930 | |
931 | static int test_bdev_super(struct super_block *s, void *data) |
932 | { |
933 | return (void *)s->s_bdev == data; |
934 | } |
935 | |
936 | struct dentry *mount_bdev(struct file_system_type *fs_type, |
937 | int flags, const char *dev_name, void *data, |
938 | int (*fill_super)(struct super_block *, void *, int)) |
939 | { |
940 | struct block_device *bdev; |
941 | struct super_block *s; |
942 | fmode_t mode = FMODE_READ | FMODE_EXCL; |
943 | int error = 0; |
944 | |
945 | if (!(flags & MS_RDONLY)) |
946 | mode |= FMODE_WRITE; |
947 | |
948 | bdev = blkdev_get_by_path(dev_name, mode, fs_type); |
949 | if (IS_ERR(bdev)) |
950 | return ERR_CAST(bdev); |
951 | |
952 | /* |
953 | * once the super is inserted into the list by sget, s_umount |
954 | * will protect the lockfs code from trying to start a snapshot |
955 | * while we are mounting |
956 | */ |
957 | mutex_lock(&bdev->bd_fsfreeze_mutex); |
958 | if (bdev->bd_fsfreeze_count > 0) { |
959 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
960 | error = -EBUSY; |
961 | goto error_bdev; |
962 | } |
963 | s = sget(fs_type, test_bdev_super, set_bdev_super, flags | MS_NOSEC, |
964 | bdev); |
965 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
966 | if (IS_ERR(s)) |
967 | goto error_s; |
968 | |
969 | if (s->s_root) { |
970 | if ((flags ^ s->s_flags) & MS_RDONLY) { |
971 | deactivate_locked_super(s); |
972 | error = -EBUSY; |
973 | goto error_bdev; |
974 | } |
975 | |
976 | /* |
977 | * s_umount nests inside bd_mutex during |
978 | * __invalidate_device(). blkdev_put() acquires |
979 | * bd_mutex and can't be called under s_umount. Drop |
980 | * s_umount temporarily. This is safe as we're |
981 | * holding an active reference. |
982 | */ |
983 | up_write(&s->s_umount); |
984 | blkdev_put(bdev, mode); |
985 | down_write(&s->s_umount); |
986 | } else { |
987 | char b[BDEVNAME_SIZE]; |
988 | |
989 | s->s_mode = mode; |
990 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); |
991 | sb_set_blocksize(s, block_size(bdev)); |
992 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
993 | if (error) { |
994 | deactivate_locked_super(s); |
995 | goto error; |
996 | } |
997 | |
998 | s->s_flags |= MS_ACTIVE; |
999 | bdev->bd_super = s; |
1000 | } |
1001 | |
1002 | return dget(s->s_root); |
1003 | |
1004 | error_s: |
1005 | error = PTR_ERR(s); |
1006 | error_bdev: |
1007 | blkdev_put(bdev, mode); |
1008 | error: |
1009 | return ERR_PTR(error); |
1010 | } |
1011 | EXPORT_SYMBOL(mount_bdev); |
1012 | |
1013 | void kill_block_super(struct super_block *sb) |
1014 | { |
1015 | struct block_device *bdev = sb->s_bdev; |
1016 | fmode_t mode = sb->s_mode; |
1017 | |
1018 | bdev->bd_super = NULL; |
1019 | generic_shutdown_super(sb); |
1020 | sync_blockdev(bdev); |
1021 | WARN_ON_ONCE(!(mode & FMODE_EXCL)); |
1022 | blkdev_put(bdev, mode | FMODE_EXCL); |
1023 | } |
1024 | |
1025 | EXPORT_SYMBOL(kill_block_super); |
1026 | #endif |
1027 | |
1028 | struct dentry *mount_nodev(struct file_system_type *fs_type, |
1029 | int flags, void *data, |
1030 | int (*fill_super)(struct super_block *, void *, int)) |
1031 | { |
1032 | int error; |
1033 | struct super_block *s = sget(fs_type, NULL, set_anon_super, flags, NULL); |
1034 | |
1035 | if (IS_ERR(s)) |
1036 | return ERR_CAST(s); |
1037 | |
1038 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1039 | if (error) { |
1040 | deactivate_locked_super(s); |
1041 | return ERR_PTR(error); |
1042 | } |
1043 | s->s_flags |= MS_ACTIVE; |
1044 | return dget(s->s_root); |
1045 | } |
1046 | EXPORT_SYMBOL(mount_nodev); |
1047 | |
1048 | static int compare_single(struct super_block *s, void *p) |
1049 | { |
1050 | return 1; |
1051 | } |
1052 | |
1053 | struct dentry *mount_single(struct file_system_type *fs_type, |
1054 | int flags, void *data, |
1055 | int (*fill_super)(struct super_block *, void *, int)) |
1056 | { |
1057 | struct super_block *s; |
1058 | int error; |
1059 | |
1060 | s = sget(fs_type, compare_single, set_anon_super, flags, NULL); |
1061 | if (IS_ERR(s)) |
1062 | return ERR_CAST(s); |
1063 | if (!s->s_root) { |
1064 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1065 | if (error) { |
1066 | deactivate_locked_super(s); |
1067 | return ERR_PTR(error); |
1068 | } |
1069 | s->s_flags |= MS_ACTIVE; |
1070 | } else { |
1071 | do_remount_sb(s, flags, data, 0); |
1072 | } |
1073 | return dget(s->s_root); |
1074 | } |
1075 | EXPORT_SYMBOL(mount_single); |
1076 | |
1077 | struct dentry * |
1078 | mount_fs(struct file_system_type *type, int flags, const char *name, void *data) |
1079 | { |
1080 | struct dentry *root; |
1081 | struct super_block *sb; |
1082 | char *secdata = NULL; |
1083 | int error = -ENOMEM; |
1084 | |
1085 | if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) { |
1086 | secdata = alloc_secdata(); |
1087 | if (!secdata) |
1088 | goto out; |
1089 | |
1090 | error = security_sb_copy_data(data, secdata); |
1091 | if (error) |
1092 | goto out_free_secdata; |
1093 | } |
1094 | |
1095 | root = type->mount(type, flags, name, data); |
1096 | if (IS_ERR(root)) { |
1097 | error = PTR_ERR(root); |
1098 | goto out_free_secdata; |
1099 | } |
1100 | sb = root->d_sb; |
1101 | BUG_ON(!sb); |
1102 | WARN_ON(!sb->s_bdi); |
1103 | WARN_ON(sb->s_bdi == &default_backing_dev_info); |
1104 | sb->s_flags |= MS_BORN; |
1105 | |
1106 | error = security_sb_kern_mount(sb, flags, secdata); |
1107 | if (error) |
1108 | goto out_sb; |
1109 | |
1110 | /* |
1111 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE |
1112 | * but s_maxbytes was an unsigned long long for many releases. Throw |
1113 | * this warning for a little while to try and catch filesystems that |
1114 | * violate this rule. |
1115 | */ |
1116 | WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " |
1117 | "negative value (%lld)\n", type->name, sb->s_maxbytes); |
1118 | |
1119 | up_write(&sb->s_umount); |
1120 | free_secdata(secdata); |
1121 | return root; |
1122 | out_sb: |
1123 | dput(root); |
1124 | deactivate_locked_super(sb); |
1125 | out_free_secdata: |
1126 | free_secdata(secdata); |
1127 | out: |
1128 | return ERR_PTR(error); |
1129 | } |
1130 | |
1131 | /* |
1132 | * This is an internal function, please use sb_end_{write,pagefault,intwrite} |
1133 | * instead. |
1134 | */ |
1135 | void __sb_end_write(struct super_block *sb, int level) |
1136 | { |
1137 | percpu_counter_dec(&sb->s_writers.counter[level-1]); |
1138 | /* |
1139 | * Make sure s_writers are updated before we wake up waiters in |
1140 | * freeze_super(). |
1141 | */ |
1142 | smp_mb(); |
1143 | if (waitqueue_active(&sb->s_writers.wait)) |
1144 | wake_up(&sb->s_writers.wait); |
1145 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _RET_IP_); |
1146 | } |
1147 | EXPORT_SYMBOL(__sb_end_write); |
1148 | |
1149 | #ifdef CONFIG_LOCKDEP |
1150 | /* |
1151 | * We want lockdep to tell us about possible deadlocks with freezing but |
1152 | * it's it bit tricky to properly instrument it. Getting a freeze protection |
1153 | * works as getting a read lock but there are subtle problems. XFS for example |
1154 | * gets freeze protection on internal level twice in some cases, which is OK |
1155 | * only because we already hold a freeze protection also on higher level. Due |
1156 | * to these cases we have to tell lockdep we are doing trylock when we |
1157 | * already hold a freeze protection for a higher freeze level. |
1158 | */ |
1159 | static void acquire_freeze_lock(struct super_block *sb, int level, bool trylock, |
1160 | unsigned long ip) |
1161 | { |
1162 | int i; |
1163 | |
1164 | if (!trylock) { |
1165 | for (i = 0; i < level - 1; i++) |
1166 | if (lock_is_held(&sb->s_writers.lock_map[i])) { |
1167 | trylock = true; |
1168 | break; |
1169 | } |
1170 | } |
1171 | rwsem_acquire_read(&sb->s_writers.lock_map[level-1], 0, trylock, ip); |
1172 | } |
1173 | #endif |
1174 | |
1175 | /* |
1176 | * This is an internal function, please use sb_start_{write,pagefault,intwrite} |
1177 | * instead. |
1178 | */ |
1179 | int __sb_start_write(struct super_block *sb, int level, bool wait) |
1180 | { |
1181 | retry: |
1182 | if (unlikely(sb->s_writers.frozen >= level)) { |
1183 | if (!wait) |
1184 | return 0; |
1185 | wait_event(sb->s_writers.wait_unfrozen, |
1186 | sb->s_writers.frozen < level); |
1187 | } |
1188 | |
1189 | #ifdef CONFIG_LOCKDEP |
1190 | acquire_freeze_lock(sb, level, !wait, _RET_IP_); |
1191 | #endif |
1192 | percpu_counter_inc(&sb->s_writers.counter[level-1]); |
1193 | /* |
1194 | * Make sure counter is updated before we check for frozen. |
1195 | * freeze_super() first sets frozen and then checks the counter. |
1196 | */ |
1197 | smp_mb(); |
1198 | if (unlikely(sb->s_writers.frozen >= level)) { |
1199 | __sb_end_write(sb, level); |
1200 | goto retry; |
1201 | } |
1202 | return 1; |
1203 | } |
1204 | EXPORT_SYMBOL(__sb_start_write); |
1205 | |
1206 | /** |
1207 | * sb_wait_write - wait until all writers to given file system finish |
1208 | * @sb: the super for which we wait |
1209 | * @level: type of writers we wait for (normal vs page fault) |
1210 | * |
1211 | * This function waits until there are no writers of given type to given file |
1212 | * system. Caller of this function should make sure there can be no new writers |
1213 | * of type @level before calling this function. Otherwise this function can |
1214 | * livelock. |
1215 | */ |
1216 | static void sb_wait_write(struct super_block *sb, int level) |
1217 | { |
1218 | s64 writers; |
1219 | |
1220 | /* |
1221 | * We just cycle-through lockdep here so that it does not complain |
1222 | * about returning with lock to userspace |
1223 | */ |
1224 | rwsem_acquire(&sb->s_writers.lock_map[level-1], 0, 0, _THIS_IP_); |
1225 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _THIS_IP_); |
1226 | |
1227 | do { |
1228 | DEFINE_WAIT(wait); |
1229 | |
1230 | /* |
1231 | * We use a barrier in prepare_to_wait() to separate setting |
1232 | * of frozen and checking of the counter |
1233 | */ |
1234 | prepare_to_wait(&sb->s_writers.wait, &wait, |
1235 | TASK_UNINTERRUPTIBLE); |
1236 | |
1237 | writers = percpu_counter_sum(&sb->s_writers.counter[level-1]); |
1238 | if (writers) |
1239 | schedule(); |
1240 | |
1241 | finish_wait(&sb->s_writers.wait, &wait); |
1242 | } while (writers); |
1243 | } |
1244 | |
1245 | /** |
1246 | * freeze_super - lock the filesystem and force it into a consistent state |
1247 | * @sb: the super to lock |
1248 | * |
1249 | * Syncs the super to make sure the filesystem is consistent and calls the fs's |
1250 | * freeze_fs. Subsequent calls to this without first thawing the fs will return |
1251 | * -EBUSY. |
1252 | * |
1253 | * During this function, sb->s_writers.frozen goes through these values: |
1254 | * |
1255 | * SB_UNFROZEN: File system is normal, all writes progress as usual. |
1256 | * |
1257 | * SB_FREEZE_WRITE: The file system is in the process of being frozen. New |
1258 | * writes should be blocked, though page faults are still allowed. We wait for |
1259 | * all writes to complete and then proceed to the next stage. |
1260 | * |
1261 | * SB_FREEZE_PAGEFAULT: Freezing continues. Now also page faults are blocked |
1262 | * but internal fs threads can still modify the filesystem (although they |
1263 | * should not dirty new pages or inodes), writeback can run etc. After waiting |
1264 | * for all running page faults we sync the filesystem which will clean all |
1265 | * dirty pages and inodes (no new dirty pages or inodes can be created when |
1266 | * sync is running). |
1267 | * |
1268 | * SB_FREEZE_FS: The file system is frozen. Now all internal sources of fs |
1269 | * modification are blocked (e.g. XFS preallocation truncation on inode |
1270 | * reclaim). This is usually implemented by blocking new transactions for |
1271 | * filesystems that have them and need this additional guard. After all |
1272 | * internal writers are finished we call ->freeze_fs() to finish filesystem |
1273 | * freezing. Then we transition to SB_FREEZE_COMPLETE state. This state is |
1274 | * mostly auxiliary for filesystems to verify they do not modify frozen fs. |
1275 | * |
1276 | * sb->s_writers.frozen is protected by sb->s_umount. |
1277 | */ |
1278 | int freeze_super(struct super_block *sb) |
1279 | { |
1280 | int ret; |
1281 | |
1282 | atomic_inc(&sb->s_active); |
1283 | down_write(&sb->s_umount); |
1284 | if (sb->s_writers.frozen != SB_UNFROZEN) { |
1285 | deactivate_locked_super(sb); |
1286 | return -EBUSY; |
1287 | } |
1288 | |
1289 | if (!(sb->s_flags & MS_BORN)) { |
1290 | up_write(&sb->s_umount); |
1291 | return 0; /* sic - it's "nothing to do" */ |
1292 | } |
1293 | |
1294 | if (sb->s_flags & MS_RDONLY) { |
1295 | /* Nothing to do really... */ |
1296 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; |
1297 | up_write(&sb->s_umount); |
1298 | return 0; |
1299 | } |
1300 | |
1301 | /* From now on, no new normal writers can start */ |
1302 | sb->s_writers.frozen = SB_FREEZE_WRITE; |
1303 | smp_wmb(); |
1304 | |
1305 | /* Release s_umount to preserve sb_start_write -> s_umount ordering */ |
1306 | up_write(&sb->s_umount); |
1307 | |
1308 | sb_wait_write(sb, SB_FREEZE_WRITE); |
1309 | |
1310 | /* Now we go and block page faults... */ |
1311 | down_write(&sb->s_umount); |
1312 | sb->s_writers.frozen = SB_FREEZE_PAGEFAULT; |
1313 | smp_wmb(); |
1314 | |
1315 | sb_wait_write(sb, SB_FREEZE_PAGEFAULT); |
1316 | |
1317 | /* All writers are done so after syncing there won't be dirty data */ |
1318 | sync_filesystem(sb); |
1319 | |
1320 | /* Now wait for internal filesystem counter */ |
1321 | sb->s_writers.frozen = SB_FREEZE_FS; |
1322 | smp_wmb(); |
1323 | sb_wait_write(sb, SB_FREEZE_FS); |
1324 | |
1325 | if (sb->s_op->freeze_fs) { |
1326 | ret = sb->s_op->freeze_fs(sb); |
1327 | if (ret) { |
1328 | printk(KERN_ERR |
1329 | "VFS:Filesystem freeze failed\n"); |
1330 | sb->s_writers.frozen = SB_UNFROZEN; |
1331 | smp_wmb(); |
1332 | wake_up(&sb->s_writers.wait_unfrozen); |
1333 | deactivate_locked_super(sb); |
1334 | return ret; |
1335 | } |
1336 | } |
1337 | /* |
1338 | * This is just for debugging purposes so that fs can warn if it |
1339 | * sees write activity when frozen is set to SB_FREEZE_COMPLETE. |
1340 | */ |
1341 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; |
1342 | up_write(&sb->s_umount); |
1343 | return 0; |
1344 | } |
1345 | EXPORT_SYMBOL(freeze_super); |
1346 | |
1347 | /** |
1348 | * thaw_super -- unlock filesystem |
1349 | * @sb: the super to thaw |
1350 | * |
1351 | * Unlocks the filesystem and marks it writeable again after freeze_super(). |
1352 | */ |
1353 | int thaw_super(struct super_block *sb) |
1354 | { |
1355 | int error; |
1356 | |
1357 | down_write(&sb->s_umount); |
1358 | if (sb->s_writers.frozen == SB_UNFROZEN) { |
1359 | up_write(&sb->s_umount); |
1360 | return -EINVAL; |
1361 | } |
1362 | |
1363 | if (sb->s_flags & MS_RDONLY) |
1364 | goto out; |
1365 | |
1366 | if (sb->s_op->unfreeze_fs) { |
1367 | error = sb->s_op->unfreeze_fs(sb); |
1368 | if (error) { |
1369 | printk(KERN_ERR |
1370 | "VFS:Filesystem thaw failed\n"); |
1371 | up_write(&sb->s_umount); |
1372 | return error; |
1373 | } |
1374 | } |
1375 | |
1376 | out: |
1377 | sb->s_writers.frozen = SB_UNFROZEN; |
1378 | smp_wmb(); |
1379 | wake_up(&sb->s_writers.wait_unfrozen); |
1380 | deactivate_locked_super(sb); |
1381 | |
1382 | return 0; |
1383 | } |
1384 | EXPORT_SYMBOL(thaw_super); |
1385 |
Branches:
ben-wpan
ben-wpan-stefan
javiroman/ks7010
jz-2.6.34
jz-2.6.34-rc5
jz-2.6.34-rc6
jz-2.6.34-rc7
jz-2.6.35
jz-2.6.36
jz-2.6.37
jz-2.6.38
jz-2.6.39
jz-3.0
jz-3.1
jz-3.11
jz-3.12
jz-3.13
jz-3.15
jz-3.16
jz-3.18-dt
jz-3.2
jz-3.3
jz-3.4
jz-3.5
jz-3.6
jz-3.6-rc2-pwm
jz-3.9
jz-3.9-clk
jz-3.9-rc8
jz47xx
jz47xx-2.6.38
master
Tags:
od-2011-09-04
od-2011-09-18
v2.6.34-rc5
v2.6.34-rc6
v2.6.34-rc7
v3.9