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 int prune_super(struct shrinker *shrink, struct shrink_control *sc) |
57 | { |
58 | struct super_block *sb; |
59 | int fs_objects = 0; |
60 | int total_objects; |
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)) |
72 | return -1; |
73 | |
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 | |
80 | if (sc->nr_to_scan) { |
81 | int dentries; |
82 | int inodes; |
83 | |
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); |
98 | |
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; |
105 | } |
106 | |
107 | total_objects = (total_objects / 100) * sysctl_vfs_cache_pressure; |
108 | drop_super(sb); |
109 | return total_objects; |
110 | } |
111 | |
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 | |
141 | /** |
142 | * alloc_super - create new superblock |
143 | * @type: filesystem type superblock should belong to |
144 | * @flags: the mount flags |
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 | */ |
149 | static struct super_block *alloc_super(struct file_system_type *type, int flags) |
150 | { |
151 | struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER); |
152 | static const struct super_operations default_op; |
153 | |
154 | if (s) { |
155 | if (security_sb_alloc(s)) { |
156 | /* |
157 | * We cannot call security_sb_free() without |
158 | * security_sb_alloc() succeeding. So bail out manually |
159 | */ |
160 | kfree(s); |
161 | s = NULL; |
162 | goto out; |
163 | } |
164 | #ifdef CONFIG_SMP |
165 | s->s_files = alloc_percpu(struct list_head); |
166 | if (!s->s_files) |
167 | goto err_out; |
168 | else { |
169 | int i; |
170 | |
171 | for_each_possible_cpu(i) |
172 | INIT_LIST_HEAD(per_cpu_ptr(s->s_files, i)); |
173 | } |
174 | #else |
175 | INIT_LIST_HEAD(&s->s_files); |
176 | #endif |
177 | if (init_sb_writers(s, type)) |
178 | goto err_out; |
179 | s->s_flags = flags; |
180 | s->s_bdi = &default_backing_dev_info; |
181 | INIT_HLIST_NODE(&s->s_instances); |
182 | INIT_HLIST_BL_HEAD(&s->s_anon); |
183 | INIT_LIST_HEAD(&s->s_inodes); |
184 | INIT_LIST_HEAD(&s->s_dentry_lru); |
185 | INIT_LIST_HEAD(&s->s_inode_lru); |
186 | spin_lock_init(&s->s_inode_lru_lock); |
187 | INIT_LIST_HEAD(&s->s_mounts); |
188 | init_rwsem(&s->s_umount); |
189 | lockdep_set_class(&s->s_umount, &type->s_umount_key); |
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); |
206 | s->s_count = 1; |
207 | atomic_set(&s->s_active, 1); |
208 | mutex_init(&s->s_vfs_rename_mutex); |
209 | lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key); |
210 | mutex_init(&s->s_dquot.dqio_mutex); |
211 | mutex_init(&s->s_dquot.dqonoff_mutex); |
212 | init_rwsem(&s->s_dquot.dqptr_sem); |
213 | s->s_maxbytes = MAX_NON_LFS; |
214 | s->s_op = &default_op; |
215 | s->s_time_gran = 1000000000; |
216 | s->cleancache_poolid = -1; |
217 | |
218 | s->s_shrink.seeks = DEFAULT_SEEKS; |
219 | s->s_shrink.shrink = prune_super; |
220 | s->s_shrink.batch = 1024; |
221 | } |
222 | out: |
223 | return s; |
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; |
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 | { |
244 | #ifdef CONFIG_SMP |
245 | free_percpu(s->s_files); |
246 | #endif |
247 | destroy_sb_writers(s); |
248 | security_sb_free(s); |
249 | WARN_ON(!list_empty(&s->s_mounts)); |
250 | kfree(s->s_subtype); |
251 | kfree(s->s_options); |
252 | kfree(s); |
253 | } |
254 | |
255 | /* Superblock refcounting */ |
256 | |
257 | /* |
258 | * Drop a superblock's refcount. The caller must hold sb_lock. |
259 | */ |
260 | static void __put_super(struct super_block *sb) |
261 | { |
262 | if (!--sb->s_count) { |
263 | list_del_init(&sb->s_list); |
264 | destroy_super(sb); |
265 | } |
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 | */ |
275 | static void put_super(struct super_block *sb) |
276 | { |
277 | spin_lock(&sb_lock); |
278 | __put_super(sb); |
279 | spin_unlock(&sb_lock); |
280 | } |
281 | |
282 | |
283 | /** |
284 | * deactivate_locked_super - drop an active reference to superblock |
285 | * @s: superblock to deactivate |
286 | * |
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 |
289 | * tell fs driver to shut it down and drop the temporary reference we |
290 | * had just acquired. |
291 | * |
292 | * Caller holds exclusive lock on superblock; that lock is released. |
293 | */ |
294 | void deactivate_locked_super(struct super_block *s) |
295 | { |
296 | struct file_system_type *fs = s->s_type; |
297 | if (atomic_dec_and_test(&s->s_active)) { |
298 | cleancache_invalidate_fs(s); |
299 | fs->kill_sb(s); |
300 | |
301 | /* caches are now gone, we can safely kill the shrinker now */ |
302 | unregister_shrinker(&s->s_shrink); |
303 | put_filesystem(fs); |
304 | put_super(s); |
305 | } else { |
306 | up_write(&s->s_umount); |
307 | } |
308 | } |
309 | |
310 | EXPORT_SYMBOL(deactivate_locked_super); |
311 | |
312 | /** |
313 | * deactivate_super - drop an active reference to superblock |
314 | * @s: superblock to deactivate |
315 | * |
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. |
319 | */ |
320 | void deactivate_super(struct super_block *s) |
321 | { |
322 | if (!atomic_add_unless(&s->s_active, -1, 1)) { |
323 | down_write(&s->s_umount); |
324 | deactivate_locked_super(s); |
325 | } |
326 | } |
327 | |
328 | EXPORT_SYMBOL(deactivate_super); |
329 | |
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 |
339 | * dying when grab_super() had been called). |
340 | */ |
341 | static int grab_super(struct super_block *s) __releases(sb_lock) |
342 | { |
343 | if (atomic_inc_not_zero(&s->s_active)) { |
344 | spin_unlock(&sb_lock); |
345 | return 1; |
346 | } |
347 | /* it's going away */ |
348 | s->s_count++; |
349 | spin_unlock(&sb_lock); |
350 | /* wait for it to die */ |
351 | down_write(&s->s_umount); |
352 | up_write(&s->s_umount); |
353 | put_super(s); |
354 | return 0; |
355 | } |
356 | |
357 | /* |
358 | * grab_super_passive - acquire a passive reference |
359 | * @sb: reference we are trying to grab |
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); |
372 | if (hlist_unhashed(&sb->s_instances)) { |
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)) { |
381 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
382 | return true; |
383 | up_read(&sb->s_umount); |
384 | } |
385 | |
386 | put_super(sb); |
387 | return false; |
388 | } |
389 | |
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. |
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. |
403 | */ |
404 | void generic_shutdown_super(struct super_block *sb) |
405 | { |
406 | const struct super_operations *sop = sb->s_op; |
407 | |
408 | if (sb->s_root) { |
409 | shrink_dcache_for_umount(sb); |
410 | sync_filesystem(sb); |
411 | sb->s_flags &= ~MS_ACTIVE; |
412 | |
413 | fsnotify_unmount_inodes(&sb->s_inodes); |
414 | |
415 | evict_inodes(sb); |
416 | |
417 | if (sop->put_super) |
418 | sop->put_super(sb); |
419 | |
420 | if (!list_empty(&sb->s_inodes)) { |
421 | printk("VFS: Busy inodes after unmount of %s. " |
422 | "Self-destruct in 5 seconds. Have a nice day...\n", |
423 | sb->s_id); |
424 | } |
425 | } |
426 | spin_lock(&sb_lock); |
427 | /* should be initialized for __put_super_and_need_restart() */ |
428 | hlist_del_init(&sb->s_instances); |
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 |
440 | * @flags: mount flags |
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 *), |
446 | int flags, |
447 | void *data) |
448 | { |
449 | struct super_block *s = NULL; |
450 | struct super_block *old; |
451 | int err; |
452 | |
453 | retry: |
454 | spin_lock(&sb_lock); |
455 | if (test) { |
456 | hlist_for_each_entry(old, &type->fs_supers, s_instances) { |
457 | if (!test(old, data)) |
458 | continue; |
459 | if (!grab_super(old)) |
460 | goto retry; |
461 | if (s) { |
462 | up_write(&s->s_umount); |
463 | destroy_super(s); |
464 | s = NULL; |
465 | } |
466 | down_write(&old->s_umount); |
467 | if (unlikely(!(old->s_flags & MS_BORN))) { |
468 | deactivate_locked_super(old); |
469 | goto retry; |
470 | } |
471 | return old; |
472 | } |
473 | } |
474 | if (!s) { |
475 | spin_unlock(&sb_lock); |
476 | s = alloc_super(type, flags); |
477 | if (!s) |
478 | return ERR_PTR(-ENOMEM); |
479 | goto retry; |
480 | } |
481 | |
482 | err = set(s, data); |
483 | if (err) { |
484 | spin_unlock(&sb_lock); |
485 | up_write(&s->s_umount); |
486 | destroy_super(s); |
487 | return ERR_PTR(err); |
488 | } |
489 | s->s_type = type; |
490 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); |
491 | list_add_tail(&s->s_list, &super_blocks); |
492 | hlist_add_head(&s->s_instances, &type->fs_supers); |
493 | spin_unlock(&sb_lock); |
494 | get_filesystem(type); |
495 | register_shrinker(&s->s_shrink); |
496 | return s; |
497 | } |
498 | |
499 | EXPORT_SYMBOL(sget); |
500 | |
501 | void drop_super(struct super_block *sb) |
502 | { |
503 | up_read(&sb->s_umount); |
504 | put_super(sb); |
505 | } |
506 | |
507 | EXPORT_SYMBOL(drop_super); |
508 | |
509 | /** |
510 | * iterate_supers - call function for all active superblocks |
511 | * @f: function to call |
512 | * @arg: argument to pass to it |
513 | * |
514 | * Scans the superblock list and calls given function, passing it |
515 | * locked superblock and given argument. |
516 | */ |
517 | void iterate_supers(void (*f)(struct super_block *, void *), void *arg) |
518 | { |
519 | struct super_block *sb, *p = NULL; |
520 | |
521 | spin_lock(&sb_lock); |
522 | list_for_each_entry(sb, &super_blocks, s_list) { |
523 | if (hlist_unhashed(&sb->s_instances)) |
524 | continue; |
525 | sb->s_count++; |
526 | spin_unlock(&sb_lock); |
527 | |
528 | down_read(&sb->s_umount); |
529 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
530 | f(sb, arg); |
531 | up_read(&sb->s_umount); |
532 | |
533 | spin_lock(&sb_lock); |
534 | if (p) |
535 | __put_super(p); |
536 | p = sb; |
537 | } |
538 | if (p) |
539 | __put_super(p); |
540 | spin_unlock(&sb_lock); |
541 | } |
542 | |
543 | /** |
544 | * iterate_supers_type - call function for superblocks of given type |
545 | * @type: fs type |
546 | * @f: function to call |
547 | * @arg: argument to pass to it |
548 | * |
549 | * Scans the superblock list and calls given function, passing it |
550 | * locked superblock and given argument. |
551 | */ |
552 | void iterate_supers_type(struct file_system_type *type, |
553 | void (*f)(struct super_block *, void *), void *arg) |
554 | { |
555 | struct super_block *sb, *p = NULL; |
556 | |
557 | spin_lock(&sb_lock); |
558 | hlist_for_each_entry(sb, &type->fs_supers, s_instances) { |
559 | sb->s_count++; |
560 | spin_unlock(&sb_lock); |
561 | |
562 | down_read(&sb->s_umount); |
563 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
564 | f(sb, arg); |
565 | up_read(&sb->s_umount); |
566 | |
567 | spin_lock(&sb_lock); |
568 | if (p) |
569 | __put_super(p); |
570 | p = sb; |
571 | } |
572 | if (p) |
573 | __put_super(p); |
574 | spin_unlock(&sb_lock); |
575 | } |
576 | |
577 | EXPORT_SYMBOL(iterate_supers_type); |
578 | |
579 | /** |
580 | * get_super - get the superblock of a device |
581 | * @bdev: device to get the superblock for |
582 | * |
583 | * Scans the superblock list and finds the superblock of the file system |
584 | * mounted on the device given. %NULL is returned if no match is found. |
585 | */ |
586 | |
587 | struct super_block *get_super(struct block_device *bdev) |
588 | { |
589 | struct super_block *sb; |
590 | |
591 | if (!bdev) |
592 | return NULL; |
593 | |
594 | spin_lock(&sb_lock); |
595 | rescan: |
596 | list_for_each_entry(sb, &super_blocks, s_list) { |
597 | if (hlist_unhashed(&sb->s_instances)) |
598 | continue; |
599 | if (sb->s_bdev == bdev) { |
600 | sb->s_count++; |
601 | spin_unlock(&sb_lock); |
602 | down_read(&sb->s_umount); |
603 | /* still alive? */ |
604 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
605 | return sb; |
606 | up_read(&sb->s_umount); |
607 | /* nope, got unmounted */ |
608 | spin_lock(&sb_lock); |
609 | __put_super(sb); |
610 | goto rescan; |
611 | } |
612 | } |
613 | spin_unlock(&sb_lock); |
614 | return NULL; |
615 | } |
616 | |
617 | EXPORT_SYMBOL(get_super); |
618 | |
619 | /** |
620 | * get_super_thawed - get thawed superblock of a device |
621 | * @bdev: device to get the superblock for |
622 | * |
623 | * Scans the superblock list and finds the superblock of the file system |
624 | * mounted on the device. The superblock is returned once it is thawed |
625 | * (or immediately if it was not frozen). %NULL is returned if no match |
626 | * is found. |
627 | */ |
628 | struct super_block *get_super_thawed(struct block_device *bdev) |
629 | { |
630 | while (1) { |
631 | struct super_block *s = get_super(bdev); |
632 | if (!s || s->s_writers.frozen == SB_UNFROZEN) |
633 | return s; |
634 | up_read(&s->s_umount); |
635 | wait_event(s->s_writers.wait_unfrozen, |
636 | s->s_writers.frozen == SB_UNFROZEN); |
637 | put_super(s); |
638 | } |
639 | } |
640 | EXPORT_SYMBOL(get_super_thawed); |
641 | |
642 | /** |
643 | * get_active_super - get an active reference to the superblock of a device |
644 | * @bdev: device to get the superblock for |
645 | * |
646 | * Scans the superblock list and finds the superblock of the file system |
647 | * mounted on the device given. Returns the superblock with an active |
648 | * reference or %NULL if none was found. |
649 | */ |
650 | struct super_block *get_active_super(struct block_device *bdev) |
651 | { |
652 | struct super_block *sb; |
653 | |
654 | if (!bdev) |
655 | return NULL; |
656 | |
657 | restart: |
658 | spin_lock(&sb_lock); |
659 | list_for_each_entry(sb, &super_blocks, s_list) { |
660 | if (hlist_unhashed(&sb->s_instances)) |
661 | continue; |
662 | if (sb->s_bdev == bdev) { |
663 | if (grab_super(sb)) /* drops sb_lock */ |
664 | return sb; |
665 | else |
666 | goto restart; |
667 | } |
668 | } |
669 | spin_unlock(&sb_lock); |
670 | return NULL; |
671 | } |
672 | |
673 | struct super_block *user_get_super(dev_t dev) |
674 | { |
675 | struct super_block *sb; |
676 | |
677 | spin_lock(&sb_lock); |
678 | rescan: |
679 | list_for_each_entry(sb, &super_blocks, s_list) { |
680 | if (hlist_unhashed(&sb->s_instances)) |
681 | continue; |
682 | if (sb->s_dev == dev) { |
683 | sb->s_count++; |
684 | spin_unlock(&sb_lock); |
685 | down_read(&sb->s_umount); |
686 | /* still alive? */ |
687 | if (sb->s_root && (sb->s_flags & MS_BORN)) |
688 | return sb; |
689 | up_read(&sb->s_umount); |
690 | /* nope, got unmounted */ |
691 | spin_lock(&sb_lock); |
692 | __put_super(sb); |
693 | goto rescan; |
694 | } |
695 | } |
696 | spin_unlock(&sb_lock); |
697 | return NULL; |
698 | } |
699 | |
700 | /** |
701 | * do_remount_sb - asks filesystem to change mount options. |
702 | * @sb: superblock in question |
703 | * @flags: numeric part of options |
704 | * @data: the rest of options |
705 | * @force: whether or not to force the change |
706 | * |
707 | * Alters the mount options of a mounted file system. |
708 | */ |
709 | int do_remount_sb(struct super_block *sb, int flags, void *data, int force) |
710 | { |
711 | int retval; |
712 | int remount_ro; |
713 | |
714 | if (sb->s_writers.frozen != SB_UNFROZEN) |
715 | return -EBUSY; |
716 | |
717 | #ifdef CONFIG_BLOCK |
718 | if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev)) |
719 | return -EACCES; |
720 | #endif |
721 | |
722 | if (flags & MS_RDONLY) |
723 | acct_auto_close(sb); |
724 | shrink_dcache_sb(sb); |
725 | sync_filesystem(sb); |
726 | |
727 | remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY); |
728 | |
729 | /* If we are remounting RDONLY and current sb is read/write, |
730 | make sure there are no rw files opened */ |
731 | if (remount_ro) { |
732 | if (force) { |
733 | mark_files_ro(sb); |
734 | } else { |
735 | retval = sb_prepare_remount_readonly(sb); |
736 | if (retval) |
737 | return retval; |
738 | } |
739 | } |
740 | |
741 | if (sb->s_op->remount_fs) { |
742 | retval = sb->s_op->remount_fs(sb, &flags, data); |
743 | if (retval) { |
744 | if (!force) |
745 | goto cancel_readonly; |
746 | /* If forced remount, go ahead despite any errors */ |
747 | WARN(1, "forced remount of a %s fs returned %i\n", |
748 | sb->s_type->name, retval); |
749 | } |
750 | } |
751 | sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK); |
752 | /* Needs to be ordered wrt mnt_is_readonly() */ |
753 | smp_wmb(); |
754 | sb->s_readonly_remount = 0; |
755 | |
756 | /* |
757 | * Some filesystems modify their metadata via some other path than the |
758 | * bdev buffer cache (eg. use a private mapping, or directories in |
759 | * pagecache, etc). Also file data modifications go via their own |
760 | * mappings. So If we try to mount readonly then copy the filesystem |
761 | * from bdev, we could get stale data, so invalidate it to give a best |
762 | * effort at coherency. |
763 | */ |
764 | if (remount_ro && sb->s_bdev) |
765 | invalidate_bdev(sb->s_bdev); |
766 | return 0; |
767 | |
768 | cancel_readonly: |
769 | sb->s_readonly_remount = 0; |
770 | return retval; |
771 | } |
772 | |
773 | static void do_emergency_remount(struct work_struct *work) |
774 | { |
775 | struct super_block *sb, *p = NULL; |
776 | |
777 | spin_lock(&sb_lock); |
778 | list_for_each_entry(sb, &super_blocks, s_list) { |
779 | if (hlist_unhashed(&sb->s_instances)) |
780 | continue; |
781 | sb->s_count++; |
782 | spin_unlock(&sb_lock); |
783 | down_write(&sb->s_umount); |
784 | if (sb->s_root && sb->s_bdev && (sb->s_flags & MS_BORN) && |
785 | !(sb->s_flags & MS_RDONLY)) { |
786 | /* |
787 | * What lock protects sb->s_flags?? |
788 | */ |
789 | do_remount_sb(sb, MS_RDONLY, NULL, 1); |
790 | } |
791 | up_write(&sb->s_umount); |
792 | spin_lock(&sb_lock); |
793 | if (p) |
794 | __put_super(p); |
795 | p = sb; |
796 | } |
797 | if (p) |
798 | __put_super(p); |
799 | spin_unlock(&sb_lock); |
800 | kfree(work); |
801 | printk("Emergency Remount complete\n"); |
802 | } |
803 | |
804 | void emergency_remount(void) |
805 | { |
806 | struct work_struct *work; |
807 | |
808 | work = kmalloc(sizeof(*work), GFP_ATOMIC); |
809 | if (work) { |
810 | INIT_WORK(work, do_emergency_remount); |
811 | schedule_work(work); |
812 | } |
813 | } |
814 | |
815 | /* |
816 | * Unnamed block devices are dummy devices used by virtual |
817 | * filesystems which don't use real block-devices. -- jrs |
818 | */ |
819 | |
820 | static DEFINE_IDA(unnamed_dev_ida); |
821 | static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */ |
822 | static int unnamed_dev_start = 0; /* don't bother trying below it */ |
823 | |
824 | int get_anon_bdev(dev_t *p) |
825 | { |
826 | int dev; |
827 | int error; |
828 | |
829 | retry: |
830 | if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0) |
831 | return -ENOMEM; |
832 | spin_lock(&unnamed_dev_lock); |
833 | error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev); |
834 | if (!error) |
835 | unnamed_dev_start = dev + 1; |
836 | spin_unlock(&unnamed_dev_lock); |
837 | if (error == -EAGAIN) |
838 | /* We raced and lost with another CPU. */ |
839 | goto retry; |
840 | else if (error) |
841 | return -EAGAIN; |
842 | |
843 | if (dev == (1 << MINORBITS)) { |
844 | spin_lock(&unnamed_dev_lock); |
845 | ida_remove(&unnamed_dev_ida, dev); |
846 | if (unnamed_dev_start > dev) |
847 | unnamed_dev_start = dev; |
848 | spin_unlock(&unnamed_dev_lock); |
849 | return -EMFILE; |
850 | } |
851 | *p = MKDEV(0, dev & MINORMASK); |
852 | return 0; |
853 | } |
854 | EXPORT_SYMBOL(get_anon_bdev); |
855 | |
856 | void free_anon_bdev(dev_t dev) |
857 | { |
858 | int slot = MINOR(dev); |
859 | spin_lock(&unnamed_dev_lock); |
860 | ida_remove(&unnamed_dev_ida, slot); |
861 | if (slot < unnamed_dev_start) |
862 | unnamed_dev_start = slot; |
863 | spin_unlock(&unnamed_dev_lock); |
864 | } |
865 | EXPORT_SYMBOL(free_anon_bdev); |
866 | |
867 | int set_anon_super(struct super_block *s, void *data) |
868 | { |
869 | int error = get_anon_bdev(&s->s_dev); |
870 | if (!error) |
871 | s->s_bdi = &noop_backing_dev_info; |
872 | return error; |
873 | } |
874 | |
875 | EXPORT_SYMBOL(set_anon_super); |
876 | |
877 | void kill_anon_super(struct super_block *sb) |
878 | { |
879 | dev_t dev = sb->s_dev; |
880 | generic_shutdown_super(sb); |
881 | free_anon_bdev(dev); |
882 | } |
883 | |
884 | EXPORT_SYMBOL(kill_anon_super); |
885 | |
886 | void kill_litter_super(struct super_block *sb) |
887 | { |
888 | if (sb->s_root) |
889 | d_genocide(sb->s_root); |
890 | kill_anon_super(sb); |
891 | } |
892 | |
893 | EXPORT_SYMBOL(kill_litter_super); |
894 | |
895 | static int ns_test_super(struct super_block *sb, void *data) |
896 | { |
897 | return sb->s_fs_info == data; |
898 | } |
899 | |
900 | static int ns_set_super(struct super_block *sb, void *data) |
901 | { |
902 | sb->s_fs_info = data; |
903 | return set_anon_super(sb, NULL); |
904 | } |
905 | |
906 | struct dentry *mount_ns(struct file_system_type *fs_type, int flags, |
907 | void *data, int (*fill_super)(struct super_block *, void *, int)) |
908 | { |
909 | struct super_block *sb; |
910 | |
911 | sb = sget(fs_type, ns_test_super, ns_set_super, flags, data); |
912 | if (IS_ERR(sb)) |
913 | return ERR_CAST(sb); |
914 | |
915 | if (!sb->s_root) { |
916 | int err; |
917 | err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0); |
918 | if (err) { |
919 | deactivate_locked_super(sb); |
920 | return ERR_PTR(err); |
921 | } |
922 | |
923 | sb->s_flags |= MS_ACTIVE; |
924 | } |
925 | |
926 | return dget(sb->s_root); |
927 | } |
928 | |
929 | EXPORT_SYMBOL(mount_ns); |
930 | |
931 | #ifdef CONFIG_BLOCK |
932 | static int set_bdev_super(struct super_block *s, void *data) |
933 | { |
934 | s->s_bdev = data; |
935 | s->s_dev = s->s_bdev->bd_dev; |
936 | |
937 | /* |
938 | * We set the bdi here to the queue backing, file systems can |
939 | * overwrite this in ->fill_super() |
940 | */ |
941 | s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info; |
942 | return 0; |
943 | } |
944 | |
945 | static int test_bdev_super(struct super_block *s, void *data) |
946 | { |
947 | return (void *)s->s_bdev == data; |
948 | } |
949 | |
950 | struct dentry *mount_bdev(struct file_system_type *fs_type, |
951 | int flags, const char *dev_name, void *data, |
952 | int (*fill_super)(struct super_block *, void *, int)) |
953 | { |
954 | struct block_device *bdev; |
955 | struct super_block *s; |
956 | fmode_t mode = FMODE_READ | FMODE_EXCL; |
957 | int error = 0; |
958 | |
959 | if (!(flags & MS_RDONLY)) |
960 | mode |= FMODE_WRITE; |
961 | |
962 | bdev = blkdev_get_by_path(dev_name, mode, fs_type); |
963 | if (IS_ERR(bdev)) |
964 | return ERR_CAST(bdev); |
965 | |
966 | /* |
967 | * once the super is inserted into the list by sget, s_umount |
968 | * will protect the lockfs code from trying to start a snapshot |
969 | * while we are mounting |
970 | */ |
971 | mutex_lock(&bdev->bd_fsfreeze_mutex); |
972 | if (bdev->bd_fsfreeze_count > 0) { |
973 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
974 | error = -EBUSY; |
975 | goto error_bdev; |
976 | } |
977 | s = sget(fs_type, test_bdev_super, set_bdev_super, flags | MS_NOSEC, |
978 | bdev); |
979 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
980 | if (IS_ERR(s)) |
981 | goto error_s; |
982 | |
983 | if (s->s_root) { |
984 | if ((flags ^ s->s_flags) & MS_RDONLY) { |
985 | deactivate_locked_super(s); |
986 | error = -EBUSY; |
987 | goto error_bdev; |
988 | } |
989 | |
990 | /* |
991 | * s_umount nests inside bd_mutex during |
992 | * __invalidate_device(). blkdev_put() acquires |
993 | * bd_mutex and can't be called under s_umount. Drop |
994 | * s_umount temporarily. This is safe as we're |
995 | * holding an active reference. |
996 | */ |
997 | up_write(&s->s_umount); |
998 | blkdev_put(bdev, mode); |
999 | down_write(&s->s_umount); |
1000 | } else { |
1001 | char b[BDEVNAME_SIZE]; |
1002 | |
1003 | s->s_mode = mode; |
1004 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); |
1005 | sb_set_blocksize(s, block_size(bdev)); |
1006 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1007 | if (error) { |
1008 | deactivate_locked_super(s); |
1009 | goto error; |
1010 | } |
1011 | |
1012 | s->s_flags |= MS_ACTIVE; |
1013 | bdev->bd_super = s; |
1014 | } |
1015 | |
1016 | return dget(s->s_root); |
1017 | |
1018 | error_s: |
1019 | error = PTR_ERR(s); |
1020 | error_bdev: |
1021 | blkdev_put(bdev, mode); |
1022 | error: |
1023 | return ERR_PTR(error); |
1024 | } |
1025 | EXPORT_SYMBOL(mount_bdev); |
1026 | |
1027 | void kill_block_super(struct super_block *sb) |
1028 | { |
1029 | struct block_device *bdev = sb->s_bdev; |
1030 | fmode_t mode = sb->s_mode; |
1031 | |
1032 | bdev->bd_super = NULL; |
1033 | generic_shutdown_super(sb); |
1034 | sync_blockdev(bdev); |
1035 | WARN_ON_ONCE(!(mode & FMODE_EXCL)); |
1036 | blkdev_put(bdev, mode | FMODE_EXCL); |
1037 | } |
1038 | |
1039 | EXPORT_SYMBOL(kill_block_super); |
1040 | #endif |
1041 | |
1042 | struct dentry *mount_nodev(struct file_system_type *fs_type, |
1043 | int flags, void *data, |
1044 | int (*fill_super)(struct super_block *, void *, int)) |
1045 | { |
1046 | int error; |
1047 | struct super_block *s = sget(fs_type, NULL, set_anon_super, flags, NULL); |
1048 | |
1049 | if (IS_ERR(s)) |
1050 | return ERR_CAST(s); |
1051 | |
1052 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1053 | if (error) { |
1054 | deactivate_locked_super(s); |
1055 | return ERR_PTR(error); |
1056 | } |
1057 | s->s_flags |= MS_ACTIVE; |
1058 | return dget(s->s_root); |
1059 | } |
1060 | EXPORT_SYMBOL(mount_nodev); |
1061 | |
1062 | static int compare_single(struct super_block *s, void *p) |
1063 | { |
1064 | return 1; |
1065 | } |
1066 | |
1067 | struct dentry *mount_single(struct file_system_type *fs_type, |
1068 | int flags, void *data, |
1069 | int (*fill_super)(struct super_block *, void *, int)) |
1070 | { |
1071 | struct super_block *s; |
1072 | int error; |
1073 | |
1074 | s = sget(fs_type, compare_single, set_anon_super, flags, NULL); |
1075 | if (IS_ERR(s)) |
1076 | return ERR_CAST(s); |
1077 | if (!s->s_root) { |
1078 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1079 | if (error) { |
1080 | deactivate_locked_super(s); |
1081 | return ERR_PTR(error); |
1082 | } |
1083 | s->s_flags |= MS_ACTIVE; |
1084 | } else { |
1085 | do_remount_sb(s, flags, data, 0); |
1086 | } |
1087 | return dget(s->s_root); |
1088 | } |
1089 | EXPORT_SYMBOL(mount_single); |
1090 | |
1091 | struct dentry * |
1092 | mount_fs(struct file_system_type *type, int flags, const char *name, void *data) |
1093 | { |
1094 | struct dentry *root; |
1095 | struct super_block *sb; |
1096 | char *secdata = NULL; |
1097 | int error = -ENOMEM; |
1098 | |
1099 | if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) { |
1100 | secdata = alloc_secdata(); |
1101 | if (!secdata) |
1102 | goto out; |
1103 | |
1104 | error = security_sb_copy_data(data, secdata); |
1105 | if (error) |
1106 | goto out_free_secdata; |
1107 | } |
1108 | |
1109 | root = type->mount(type, flags, name, data); |
1110 | if (IS_ERR(root)) { |
1111 | error = PTR_ERR(root); |
1112 | goto out_free_secdata; |
1113 | } |
1114 | sb = root->d_sb; |
1115 | BUG_ON(!sb); |
1116 | WARN_ON(!sb->s_bdi); |
1117 | WARN_ON(sb->s_bdi == &default_backing_dev_info); |
1118 | sb->s_flags |= MS_BORN; |
1119 | |
1120 | error = security_sb_kern_mount(sb, flags, secdata); |
1121 | if (error) |
1122 | goto out_sb; |
1123 | |
1124 | /* |
1125 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE |
1126 | * but s_maxbytes was an unsigned long long for many releases. Throw |
1127 | * this warning for a little while to try and catch filesystems that |
1128 | * violate this rule. |
1129 | */ |
1130 | WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " |
1131 | "negative value (%lld)\n", type->name, sb->s_maxbytes); |
1132 | |
1133 | up_write(&sb->s_umount); |
1134 | free_secdata(secdata); |
1135 | return root; |
1136 | out_sb: |
1137 | dput(root); |
1138 | deactivate_locked_super(sb); |
1139 | out_free_secdata: |
1140 | free_secdata(secdata); |
1141 | out: |
1142 | return ERR_PTR(error); |
1143 | } |
1144 | |
1145 | /* |
1146 | * This is an internal function, please use sb_end_{write,pagefault,intwrite} |
1147 | * instead. |
1148 | */ |
1149 | void __sb_end_write(struct super_block *sb, int level) |
1150 | { |
1151 | percpu_counter_dec(&sb->s_writers.counter[level-1]); |
1152 | /* |
1153 | * Make sure s_writers are updated before we wake up waiters in |
1154 | * freeze_super(). |
1155 | */ |
1156 | smp_mb(); |
1157 | if (waitqueue_active(&sb->s_writers.wait)) |
1158 | wake_up(&sb->s_writers.wait); |
1159 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _RET_IP_); |
1160 | } |
1161 | EXPORT_SYMBOL(__sb_end_write); |
1162 | |
1163 | #ifdef CONFIG_LOCKDEP |
1164 | /* |
1165 | * We want lockdep to tell us about possible deadlocks with freezing but |
1166 | * it's it bit tricky to properly instrument it. Getting a freeze protection |
1167 | * works as getting a read lock but there are subtle problems. XFS for example |
1168 | * gets freeze protection on internal level twice in some cases, which is OK |
1169 | * only because we already hold a freeze protection also on higher level. Due |
1170 | * to these cases we have to tell lockdep we are doing trylock when we |
1171 | * already hold a freeze protection for a higher freeze level. |
1172 | */ |
1173 | static void acquire_freeze_lock(struct super_block *sb, int level, bool trylock, |
1174 | unsigned long ip) |
1175 | { |
1176 | int i; |
1177 | |
1178 | if (!trylock) { |
1179 | for (i = 0; i < level - 1; i++) |
1180 | if (lock_is_held(&sb->s_writers.lock_map[i])) { |
1181 | trylock = true; |
1182 | break; |
1183 | } |
1184 | } |
1185 | rwsem_acquire_read(&sb->s_writers.lock_map[level-1], 0, trylock, ip); |
1186 | } |
1187 | #endif |
1188 | |
1189 | /* |
1190 | * This is an internal function, please use sb_start_{write,pagefault,intwrite} |
1191 | * instead. |
1192 | */ |
1193 | int __sb_start_write(struct super_block *sb, int level, bool wait) |
1194 | { |
1195 | retry: |
1196 | if (unlikely(sb->s_writers.frozen >= level)) { |
1197 | if (!wait) |
1198 | return 0; |
1199 | wait_event(sb->s_writers.wait_unfrozen, |
1200 | sb->s_writers.frozen < level); |
1201 | } |
1202 | |
1203 | #ifdef CONFIG_LOCKDEP |
1204 | acquire_freeze_lock(sb, level, !wait, _RET_IP_); |
1205 | #endif |
1206 | percpu_counter_inc(&sb->s_writers.counter[level-1]); |
1207 | /* |
1208 | * Make sure counter is updated before we check for frozen. |
1209 | * freeze_super() first sets frozen and then checks the counter. |
1210 | */ |
1211 | smp_mb(); |
1212 | if (unlikely(sb->s_writers.frozen >= level)) { |
1213 | __sb_end_write(sb, level); |
1214 | goto retry; |
1215 | } |
1216 | return 1; |
1217 | } |
1218 | EXPORT_SYMBOL(__sb_start_write); |
1219 | |
1220 | /** |
1221 | * sb_wait_write - wait until all writers to given file system finish |
1222 | * @sb: the super for which we wait |
1223 | * @level: type of writers we wait for (normal vs page fault) |
1224 | * |
1225 | * This function waits until there are no writers of given type to given file |
1226 | * system. Caller of this function should make sure there can be no new writers |
1227 | * of type @level before calling this function. Otherwise this function can |
1228 | * livelock. |
1229 | */ |
1230 | static void sb_wait_write(struct super_block *sb, int level) |
1231 | { |
1232 | s64 writers; |
1233 | |
1234 | /* |
1235 | * We just cycle-through lockdep here so that it does not complain |
1236 | * about returning with lock to userspace |
1237 | */ |
1238 | rwsem_acquire(&sb->s_writers.lock_map[level-1], 0, 0, _THIS_IP_); |
1239 | rwsem_release(&sb->s_writers.lock_map[level-1], 1, _THIS_IP_); |
1240 | |
1241 | do { |
1242 | DEFINE_WAIT(wait); |
1243 | |
1244 | /* |
1245 | * We use a barrier in prepare_to_wait() to separate setting |
1246 | * of frozen and checking of the counter |
1247 | */ |
1248 | prepare_to_wait(&sb->s_writers.wait, &wait, |
1249 | TASK_UNINTERRUPTIBLE); |
1250 | |
1251 | writers = percpu_counter_sum(&sb->s_writers.counter[level-1]); |
1252 | if (writers) |
1253 | schedule(); |
1254 | |
1255 | finish_wait(&sb->s_writers.wait, &wait); |
1256 | } while (writers); |
1257 | } |
1258 | |
1259 | /** |
1260 | * freeze_super - lock the filesystem and force it into a consistent state |
1261 | * @sb: the super to lock |
1262 | * |
1263 | * Syncs the super to make sure the filesystem is consistent and calls the fs's |
1264 | * freeze_fs. Subsequent calls to this without first thawing the fs will return |
1265 | * -EBUSY. |
1266 | * |
1267 | * During this function, sb->s_writers.frozen goes through these values: |
1268 | * |
1269 | * SB_UNFROZEN: File system is normal, all writes progress as usual. |
1270 | * |
1271 | * SB_FREEZE_WRITE: The file system is in the process of being frozen. New |
1272 | * writes should be blocked, though page faults are still allowed. We wait for |
1273 | * all writes to complete and then proceed to the next stage. |
1274 | * |
1275 | * SB_FREEZE_PAGEFAULT: Freezing continues. Now also page faults are blocked |
1276 | * but internal fs threads can still modify the filesystem (although they |
1277 | * should not dirty new pages or inodes), writeback can run etc. After waiting |
1278 | * for all running page faults we sync the filesystem which will clean all |
1279 | * dirty pages and inodes (no new dirty pages or inodes can be created when |
1280 | * sync is running). |
1281 | * |
1282 | * SB_FREEZE_FS: The file system is frozen. Now all internal sources of fs |
1283 | * modification are blocked (e.g. XFS preallocation truncation on inode |
1284 | * reclaim). This is usually implemented by blocking new transactions for |
1285 | * filesystems that have them and need this additional guard. After all |
1286 | * internal writers are finished we call ->freeze_fs() to finish filesystem |
1287 | * freezing. Then we transition to SB_FREEZE_COMPLETE state. This state is |
1288 | * mostly auxiliary for filesystems to verify they do not modify frozen fs. |
1289 | * |
1290 | * sb->s_writers.frozen is protected by sb->s_umount. |
1291 | */ |
1292 | int freeze_super(struct super_block *sb) |
1293 | { |
1294 | int ret; |
1295 | |
1296 | atomic_inc(&sb->s_active); |
1297 | down_write(&sb->s_umount); |
1298 | if (sb->s_writers.frozen != SB_UNFROZEN) { |
1299 | deactivate_locked_super(sb); |
1300 | return -EBUSY; |
1301 | } |
1302 | |
1303 | if (!(sb->s_flags & MS_BORN)) { |
1304 | up_write(&sb->s_umount); |
1305 | return 0; /* sic - it's "nothing to do" */ |
1306 | } |
1307 | |
1308 | if (sb->s_flags & MS_RDONLY) { |
1309 | /* Nothing to do really... */ |
1310 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; |
1311 | up_write(&sb->s_umount); |
1312 | return 0; |
1313 | } |
1314 | |
1315 | /* From now on, no new normal writers can start */ |
1316 | sb->s_writers.frozen = SB_FREEZE_WRITE; |
1317 | smp_wmb(); |
1318 | |
1319 | /* Release s_umount to preserve sb_start_write -> s_umount ordering */ |
1320 | up_write(&sb->s_umount); |
1321 | |
1322 | sb_wait_write(sb, SB_FREEZE_WRITE); |
1323 | |
1324 | /* Now we go and block page faults... */ |
1325 | down_write(&sb->s_umount); |
1326 | sb->s_writers.frozen = SB_FREEZE_PAGEFAULT; |
1327 | smp_wmb(); |
1328 | |
1329 | sb_wait_write(sb, SB_FREEZE_PAGEFAULT); |
1330 | |
1331 | /* All writers are done so after syncing there won't be dirty data */ |
1332 | sync_filesystem(sb); |
1333 | |
1334 | /* Now wait for internal filesystem counter */ |
1335 | sb->s_writers.frozen = SB_FREEZE_FS; |
1336 | smp_wmb(); |
1337 | sb_wait_write(sb, SB_FREEZE_FS); |
1338 | |
1339 | if (sb->s_op->freeze_fs) { |
1340 | ret = sb->s_op->freeze_fs(sb); |
1341 | if (ret) { |
1342 | printk(KERN_ERR |
1343 | "VFS:Filesystem freeze failed\n"); |
1344 | sb->s_writers.frozen = SB_UNFROZEN; |
1345 | smp_wmb(); |
1346 | wake_up(&sb->s_writers.wait_unfrozen); |
1347 | deactivate_locked_super(sb); |
1348 | return ret; |
1349 | } |
1350 | } |
1351 | /* |
1352 | * This is just for debugging purposes so that fs can warn if it |
1353 | * sees write activity when frozen is set to SB_FREEZE_COMPLETE. |
1354 | */ |
1355 | sb->s_writers.frozen = SB_FREEZE_COMPLETE; |
1356 | up_write(&sb->s_umount); |
1357 | return 0; |
1358 | } |
1359 | EXPORT_SYMBOL(freeze_super); |
1360 | |
1361 | /** |
1362 | * thaw_super -- unlock filesystem |
1363 | * @sb: the super to thaw |
1364 | * |
1365 | * Unlocks the filesystem and marks it writeable again after freeze_super(). |
1366 | */ |
1367 | int thaw_super(struct super_block *sb) |
1368 | { |
1369 | int error; |
1370 | |
1371 | down_write(&sb->s_umount); |
1372 | if (sb->s_writers.frozen == SB_UNFROZEN) { |
1373 | up_write(&sb->s_umount); |
1374 | return -EINVAL; |
1375 | } |
1376 | |
1377 | if (sb->s_flags & MS_RDONLY) |
1378 | goto out; |
1379 | |
1380 | if (sb->s_op->unfreeze_fs) { |
1381 | error = sb->s_op->unfreeze_fs(sb); |
1382 | if (error) { |
1383 | printk(KERN_ERR |
1384 | "VFS:Filesystem thaw failed\n"); |
1385 | up_write(&sb->s_umount); |
1386 | return error; |
1387 | } |
1388 | } |
1389 | |
1390 | out: |
1391 | sb->s_writers.frozen = SB_UNFROZEN; |
1392 | smp_wmb(); |
1393 | wake_up(&sb->s_writers.wait_unfrozen); |
1394 | deactivate_locked_super(sb); |
1395 | |
1396 | return 0; |
1397 | } |
1398 | EXPORT_SYMBOL(thaw_super); |
1399 |
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