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 | ||
1da177e4 LT |
23 | #include <linux/module.h> |
24 | #include <linux/slab.h> | |
25 | #include <linux/init.h> | |
26 | #include <linux/smp_lock.h> | |
27 | #include <linux/acct.h> | |
28 | #include <linux/blkdev.h> | |
29 | #include <linux/quotaops.h> | |
30 | #include <linux/namei.h> | |
31 | #include <linux/buffer_head.h> /* for fsync_super() */ | |
32 | #include <linux/mount.h> | |
33 | #include <linux/security.h> | |
34 | #include <linux/syscalls.h> | |
35 | #include <linux/vfs.h> | |
36 | #include <linux/writeback.h> /* for the emergency remount stuff */ | |
37 | #include <linux/idr.h> | |
38 | #include <linux/kobject.h> | |
353ab6e9 | 39 | #include <linux/mutex.h> |
1da177e4 LT |
40 | #include <asm/uaccess.h> |
41 | ||
42 | ||
1da177e4 LT |
43 | LIST_HEAD(super_blocks); |
44 | DEFINE_SPINLOCK(sb_lock); | |
45 | ||
46 | /** | |
47 | * alloc_super - create new superblock | |
fe2bbc48 | 48 | * @type: filesystem type superblock should belong to |
1da177e4 LT |
49 | * |
50 | * Allocates and initializes a new &struct super_block. alloc_super() | |
51 | * returns a pointer new superblock or %NULL if allocation had failed. | |
52 | */ | |
cf516249 | 53 | static struct super_block *alloc_super(struct file_system_type *type) |
1da177e4 | 54 | { |
11b0b5ab | 55 | struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER); |
1da177e4 LT |
56 | static struct super_operations default_op; |
57 | ||
58 | if (s) { | |
1da177e4 LT |
59 | if (security_sb_alloc(s)) { |
60 | kfree(s); | |
61 | s = NULL; | |
62 | goto out; | |
63 | } | |
64 | INIT_LIST_HEAD(&s->s_dirty); | |
65 | INIT_LIST_HEAD(&s->s_io); | |
0e0f4fc2 | 66 | INIT_LIST_HEAD(&s->s_more_io); |
1da177e4 LT |
67 | INIT_LIST_HEAD(&s->s_files); |
68 | INIT_LIST_HEAD(&s->s_instances); | |
69 | INIT_HLIST_HEAD(&s->s_anon); | |
70 | INIT_LIST_HEAD(&s->s_inodes); | |
71 | init_rwsem(&s->s_umount); | |
7892f2f4 | 72 | mutex_init(&s->s_lock); |
897c6ff9 | 73 | lockdep_set_class(&s->s_umount, &type->s_umount_key); |
cf516249 IM |
74 | /* |
75 | * The locking rules for s_lock are up to the | |
76 | * filesystem. For example ext3fs has different | |
77 | * lock ordering than usbfs: | |
78 | */ | |
79 | lockdep_set_class(&s->s_lock, &type->s_lock_key); | |
1da177e4 LT |
80 | down_write(&s->s_umount); |
81 | s->s_count = S_BIAS; | |
82 | atomic_set(&s->s_active, 1); | |
a11f3a05 | 83 | mutex_init(&s->s_vfs_rename_mutex); |
d3be915f IM |
84 | mutex_init(&s->s_dquot.dqio_mutex); |
85 | mutex_init(&s->s_dquot.dqonoff_mutex); | |
1da177e4 LT |
86 | init_rwsem(&s->s_dquot.dqptr_sem); |
87 | init_waitqueue_head(&s->s_wait_unfrozen); | |
88 | s->s_maxbytes = MAX_NON_LFS; | |
89 | s->dq_op = sb_dquot_ops; | |
90 | s->s_qcop = sb_quotactl_ops; | |
91 | s->s_op = &default_op; | |
92 | s->s_time_gran = 1000000000; | |
93 | } | |
94 | out: | |
95 | return s; | |
96 | } | |
97 | ||
98 | /** | |
99 | * destroy_super - frees a superblock | |
100 | * @s: superblock to free | |
101 | * | |
102 | * Frees a superblock. | |
103 | */ | |
104 | static inline void destroy_super(struct super_block *s) | |
105 | { | |
106 | security_sb_free(s); | |
79c0b2df | 107 | kfree(s->s_subtype); |
b3b304a2 | 108 | kfree(s->s_options); |
1da177e4 LT |
109 | kfree(s); |
110 | } | |
111 | ||
112 | /* Superblock refcounting */ | |
113 | ||
114 | /* | |
115 | * Drop a superblock's refcount. Returns non-zero if the superblock was | |
116 | * destroyed. The caller must hold sb_lock. | |
117 | */ | |
118 | int __put_super(struct super_block *sb) | |
119 | { | |
120 | int ret = 0; | |
121 | ||
122 | if (!--sb->s_count) { | |
123 | destroy_super(sb); | |
124 | ret = 1; | |
125 | } | |
126 | return ret; | |
127 | } | |
128 | ||
129 | /* | |
130 | * Drop a superblock's refcount. | |
131 | * Returns non-zero if the superblock is about to be destroyed and | |
132 | * at least is already removed from super_blocks list, so if we are | |
133 | * making a loop through super blocks then we need to restart. | |
134 | * The caller must hold sb_lock. | |
135 | */ | |
136 | int __put_super_and_need_restart(struct super_block *sb) | |
137 | { | |
138 | /* check for race with generic_shutdown_super() */ | |
139 | if (list_empty(&sb->s_list)) { | |
140 | /* super block is removed, need to restart... */ | |
141 | __put_super(sb); | |
142 | return 1; | |
143 | } | |
144 | /* can't be the last, since s_list is still in use */ | |
145 | sb->s_count--; | |
146 | BUG_ON(sb->s_count == 0); | |
147 | return 0; | |
148 | } | |
149 | ||
150 | /** | |
151 | * put_super - drop a temporary reference to superblock | |
152 | * @sb: superblock in question | |
153 | * | |
154 | * Drops a temporary reference, frees superblock if there's no | |
155 | * references left. | |
156 | */ | |
157 | static void put_super(struct super_block *sb) | |
158 | { | |
159 | spin_lock(&sb_lock); | |
160 | __put_super(sb); | |
161 | spin_unlock(&sb_lock); | |
162 | } | |
163 | ||
164 | ||
165 | /** | |
166 | * deactivate_super - drop an active reference to superblock | |
167 | * @s: superblock to deactivate | |
168 | * | |
169 | * Drops an active reference to superblock, acquiring a temprory one if | |
170 | * there is no active references left. In that case we lock superblock, | |
171 | * tell fs driver to shut it down and drop the temporary reference we | |
172 | * had just acquired. | |
173 | */ | |
174 | void deactivate_super(struct super_block *s) | |
175 | { | |
176 | struct file_system_type *fs = s->s_type; | |
177 | if (atomic_dec_and_lock(&s->s_active, &sb_lock)) { | |
178 | s->s_count -= S_BIAS-1; | |
179 | spin_unlock(&sb_lock); | |
7b7b1ace | 180 | DQUOT_OFF(s); |
1da177e4 LT |
181 | down_write(&s->s_umount); |
182 | fs->kill_sb(s); | |
183 | put_filesystem(fs); | |
184 | put_super(s); | |
185 | } | |
186 | } | |
187 | ||
188 | EXPORT_SYMBOL(deactivate_super); | |
189 | ||
190 | /** | |
191 | * grab_super - acquire an active reference | |
192 | * @s: reference we are trying to make active | |
193 | * | |
194 | * Tries to acquire an active reference. grab_super() is used when we | |
195 | * had just found a superblock in super_blocks or fs_type->fs_supers | |
196 | * and want to turn it into a full-blown active reference. grab_super() | |
197 | * is called with sb_lock held and drops it. Returns 1 in case of | |
198 | * success, 0 if we had failed (superblock contents was already dead or | |
199 | * dying when grab_super() had been called). | |
200 | */ | |
9c4dbee7 | 201 | static int grab_super(struct super_block *s) __releases(sb_lock) |
1da177e4 LT |
202 | { |
203 | s->s_count++; | |
204 | spin_unlock(&sb_lock); | |
205 | down_write(&s->s_umount); | |
206 | if (s->s_root) { | |
207 | spin_lock(&sb_lock); | |
208 | if (s->s_count > S_BIAS) { | |
209 | atomic_inc(&s->s_active); | |
210 | s->s_count--; | |
211 | spin_unlock(&sb_lock); | |
212 | return 1; | |
213 | } | |
214 | spin_unlock(&sb_lock); | |
215 | } | |
216 | up_write(&s->s_umount); | |
217 | put_super(s); | |
218 | yield(); | |
219 | return 0; | |
220 | } | |
221 | ||
914e2637 AV |
222 | /* |
223 | * Superblock locking. We really ought to get rid of these two. | |
224 | */ | |
225 | void lock_super(struct super_block * sb) | |
226 | { | |
227 | get_fs_excl(); | |
228 | mutex_lock(&sb->s_lock); | |
229 | } | |
230 | ||
231 | void unlock_super(struct super_block * sb) | |
232 | { | |
233 | put_fs_excl(); | |
234 | mutex_unlock(&sb->s_lock); | |
235 | } | |
236 | ||
237 | EXPORT_SYMBOL(lock_super); | |
238 | EXPORT_SYMBOL(unlock_super); | |
239 | ||
cf9a2ae8 DH |
240 | /* |
241 | * Write out and wait upon all dirty data associated with this | |
242 | * superblock. Filesystem data as well as the underlying block | |
243 | * device. Takes the superblock lock. Requires a second blkdev | |
244 | * flush by the caller to complete the operation. | |
245 | */ | |
246 | void __fsync_super(struct super_block *sb) | |
247 | { | |
248 | sync_inodes_sb(sb, 0); | |
249 | DQUOT_SYNC(sb); | |
250 | lock_super(sb); | |
251 | if (sb->s_dirt && sb->s_op->write_super) | |
252 | sb->s_op->write_super(sb); | |
253 | unlock_super(sb); | |
254 | if (sb->s_op->sync_fs) | |
255 | sb->s_op->sync_fs(sb, 1); | |
256 | sync_blockdev(sb->s_bdev); | |
257 | sync_inodes_sb(sb, 1); | |
258 | } | |
259 | ||
260 | /* | |
261 | * Write out and wait upon all dirty data associated with this | |
262 | * superblock. Filesystem data as well as the underlying block | |
263 | * device. Takes the superblock lock. | |
264 | */ | |
265 | int fsync_super(struct super_block *sb) | |
266 | { | |
267 | __fsync_super(sb); | |
268 | return sync_blockdev(sb->s_bdev); | |
269 | } | |
270 | ||
1da177e4 LT |
271 | /** |
272 | * generic_shutdown_super - common helper for ->kill_sb() | |
273 | * @sb: superblock to kill | |
274 | * | |
275 | * generic_shutdown_super() does all fs-independent work on superblock | |
276 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects | |
277 | * that need destruction out of superblock, call generic_shutdown_super() | |
278 | * and release aforementioned objects. Note: dentries and inodes _are_ | |
279 | * taken care of and do not need specific handling. | |
c636ebdb DH |
280 | * |
281 | * Upon calling this function, the filesystem may no longer alter or | |
282 | * rearrange the set of dentries belonging to this super_block, nor may it | |
283 | * change the attachments of dentries to inodes. | |
1da177e4 LT |
284 | */ |
285 | void generic_shutdown_super(struct super_block *sb) | |
286 | { | |
ee9b6d61 | 287 | const struct super_operations *sop = sb->s_op; |
1da177e4 | 288 | |
c636ebdb DH |
289 | if (sb->s_root) { |
290 | shrink_dcache_for_umount(sb); | |
1da177e4 LT |
291 | fsync_super(sb); |
292 | lock_super(sb); | |
293 | sb->s_flags &= ~MS_ACTIVE; | |
294 | /* bad name - it should be evict_inodes() */ | |
295 | invalidate_inodes(sb); | |
296 | lock_kernel(); | |
297 | ||
298 | if (sop->write_super && sb->s_dirt) | |
299 | sop->write_super(sb); | |
300 | if (sop->put_super) | |
301 | sop->put_super(sb); | |
302 | ||
303 | /* Forget any remaining inodes */ | |
304 | if (invalidate_inodes(sb)) { | |
7b4fe29e DJ |
305 | printk("VFS: Busy inodes after unmount of %s. " |
306 | "Self-destruct in 5 seconds. Have a nice day...\n", | |
307 | sb->s_id); | |
1da177e4 LT |
308 | } |
309 | ||
310 | unlock_kernel(); | |
311 | unlock_super(sb); | |
312 | } | |
313 | spin_lock(&sb_lock); | |
314 | /* should be initialized for __put_super_and_need_restart() */ | |
315 | list_del_init(&sb->s_list); | |
316 | list_del(&sb->s_instances); | |
317 | spin_unlock(&sb_lock); | |
318 | up_write(&sb->s_umount); | |
319 | } | |
320 | ||
321 | EXPORT_SYMBOL(generic_shutdown_super); | |
322 | ||
323 | /** | |
324 | * sget - find or create a superblock | |
325 | * @type: filesystem type superblock should belong to | |
326 | * @test: comparison callback | |
327 | * @set: setup callback | |
328 | * @data: argument to each of them | |
329 | */ | |
330 | struct super_block *sget(struct file_system_type *type, | |
331 | int (*test)(struct super_block *,void *), | |
332 | int (*set)(struct super_block *,void *), | |
333 | void *data) | |
334 | { | |
335 | struct super_block *s = NULL; | |
d4730127 | 336 | struct super_block *old; |
1da177e4 LT |
337 | int err; |
338 | ||
339 | retry: | |
340 | spin_lock(&sb_lock); | |
d4730127 MK |
341 | if (test) { |
342 | list_for_each_entry(old, &type->fs_supers, s_instances) { | |
343 | if (!test(old, data)) | |
344 | continue; | |
345 | if (!grab_super(old)) | |
346 | goto retry; | |
347 | if (s) | |
348 | destroy_super(s); | |
349 | return old; | |
350 | } | |
1da177e4 LT |
351 | } |
352 | if (!s) { | |
353 | spin_unlock(&sb_lock); | |
cf516249 | 354 | s = alloc_super(type); |
1da177e4 LT |
355 | if (!s) |
356 | return ERR_PTR(-ENOMEM); | |
357 | goto retry; | |
358 | } | |
359 | ||
360 | err = set(s, data); | |
361 | if (err) { | |
362 | spin_unlock(&sb_lock); | |
363 | destroy_super(s); | |
364 | return ERR_PTR(err); | |
365 | } | |
366 | s->s_type = type; | |
367 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); | |
368 | list_add_tail(&s->s_list, &super_blocks); | |
369 | list_add(&s->s_instances, &type->fs_supers); | |
370 | spin_unlock(&sb_lock); | |
371 | get_filesystem(type); | |
372 | return s; | |
373 | } | |
374 | ||
375 | EXPORT_SYMBOL(sget); | |
376 | ||
377 | void drop_super(struct super_block *sb) | |
378 | { | |
379 | up_read(&sb->s_umount); | |
380 | put_super(sb); | |
381 | } | |
382 | ||
383 | EXPORT_SYMBOL(drop_super); | |
384 | ||
385 | static inline void write_super(struct super_block *sb) | |
386 | { | |
387 | lock_super(sb); | |
388 | if (sb->s_root && sb->s_dirt) | |
389 | if (sb->s_op->write_super) | |
390 | sb->s_op->write_super(sb); | |
391 | unlock_super(sb); | |
392 | } | |
393 | ||
394 | /* | |
395 | * Note: check the dirty flag before waiting, so we don't | |
396 | * hold up the sync while mounting a device. (The newly | |
397 | * mounted device won't need syncing.) | |
398 | */ | |
399 | void sync_supers(void) | |
400 | { | |
618f0636 KK |
401 | struct super_block *sb; |
402 | ||
1da177e4 | 403 | spin_lock(&sb_lock); |
618f0636 KK |
404 | restart: |
405 | list_for_each_entry(sb, &super_blocks, s_list) { | |
1da177e4 LT |
406 | if (sb->s_dirt) { |
407 | sb->s_count++; | |
408 | spin_unlock(&sb_lock); | |
409 | down_read(&sb->s_umount); | |
410 | write_super(sb); | |
618f0636 KK |
411 | up_read(&sb->s_umount); |
412 | spin_lock(&sb_lock); | |
413 | if (__put_super_and_need_restart(sb)) | |
414 | goto restart; | |
415 | } | |
416 | } | |
1da177e4 LT |
417 | spin_unlock(&sb_lock); |
418 | } | |
419 | ||
420 | /* | |
3a4fa0a2 | 421 | * Call the ->sync_fs super_op against all filesystems which are r/w and |
1da177e4 LT |
422 | * which implement it. |
423 | * | |
424 | * This operation is careful to avoid the livelock which could easily happen | |
425 | * if two or more filesystems are being continuously dirtied. s_need_sync_fs | |
426 | * is used only here. We set it against all filesystems and then clear it as | |
427 | * we sync them. So redirtied filesystems are skipped. | |
428 | * | |
3a4fa0a2 | 429 | * But if process A is currently running sync_filesystems and then process B |
1da177e4 LT |
430 | * calls sync_filesystems as well, process B will set all the s_need_sync_fs |
431 | * flags again, which will cause process A to resync everything. Fix that with | |
432 | * a local mutex. | |
433 | * | |
434 | * (Fabian) Avoid sync_fs with clean fs & wait mode 0 | |
435 | */ | |
436 | void sync_filesystems(int wait) | |
437 | { | |
438 | struct super_block *sb; | |
353ab6e9 | 439 | static DEFINE_MUTEX(mutex); |
1da177e4 | 440 | |
353ab6e9 | 441 | mutex_lock(&mutex); /* Could be down_interruptible */ |
1da177e4 | 442 | spin_lock(&sb_lock); |
618f0636 | 443 | list_for_each_entry(sb, &super_blocks, s_list) { |
1da177e4 LT |
444 | if (!sb->s_op->sync_fs) |
445 | continue; | |
446 | if (sb->s_flags & MS_RDONLY) | |
447 | continue; | |
448 | sb->s_need_sync_fs = 1; | |
449 | } | |
1da177e4 LT |
450 | |
451 | restart: | |
618f0636 | 452 | list_for_each_entry(sb, &super_blocks, s_list) { |
1da177e4 LT |
453 | if (!sb->s_need_sync_fs) |
454 | continue; | |
455 | sb->s_need_sync_fs = 0; | |
456 | if (sb->s_flags & MS_RDONLY) | |
457 | continue; /* hm. Was remounted r/o meanwhile */ | |
458 | sb->s_count++; | |
459 | spin_unlock(&sb_lock); | |
460 | down_read(&sb->s_umount); | |
461 | if (sb->s_root && (wait || sb->s_dirt)) | |
462 | sb->s_op->sync_fs(sb, wait); | |
618f0636 KK |
463 | up_read(&sb->s_umount); |
464 | /* restart only when sb is no longer on the list */ | |
465 | spin_lock(&sb_lock); | |
466 | if (__put_super_and_need_restart(sb)) | |
467 | goto restart; | |
1da177e4 LT |
468 | } |
469 | spin_unlock(&sb_lock); | |
353ab6e9 | 470 | mutex_unlock(&mutex); |
1da177e4 LT |
471 | } |
472 | ||
473 | /** | |
474 | * get_super - get the superblock of a device | |
475 | * @bdev: device to get the superblock for | |
476 | * | |
477 | * Scans the superblock list and finds the superblock of the file system | |
478 | * mounted on the device given. %NULL is returned if no match is found. | |
479 | */ | |
480 | ||
481 | struct super_block * get_super(struct block_device *bdev) | |
482 | { | |
618f0636 KK |
483 | struct super_block *sb; |
484 | ||
1da177e4 LT |
485 | if (!bdev) |
486 | return NULL; | |
618f0636 | 487 | |
1da177e4 | 488 | spin_lock(&sb_lock); |
618f0636 KK |
489 | rescan: |
490 | list_for_each_entry(sb, &super_blocks, s_list) { | |
491 | if (sb->s_bdev == bdev) { | |
492 | sb->s_count++; | |
1da177e4 | 493 | spin_unlock(&sb_lock); |
618f0636 KK |
494 | down_read(&sb->s_umount); |
495 | if (sb->s_root) | |
496 | return sb; | |
497 | up_read(&sb->s_umount); | |
498 | /* restart only when sb is no longer on the list */ | |
499 | spin_lock(&sb_lock); | |
500 | if (__put_super_and_need_restart(sb)) | |
501 | goto rescan; | |
1da177e4 LT |
502 | } |
503 | } | |
504 | spin_unlock(&sb_lock); | |
505 | return NULL; | |
506 | } | |
507 | ||
508 | EXPORT_SYMBOL(get_super); | |
509 | ||
510 | struct super_block * user_get_super(dev_t dev) | |
511 | { | |
618f0636 | 512 | struct super_block *sb; |
1da177e4 | 513 | |
1da177e4 | 514 | spin_lock(&sb_lock); |
618f0636 KK |
515 | rescan: |
516 | list_for_each_entry(sb, &super_blocks, s_list) { | |
517 | if (sb->s_dev == dev) { | |
518 | sb->s_count++; | |
1da177e4 | 519 | spin_unlock(&sb_lock); |
618f0636 KK |
520 | down_read(&sb->s_umount); |
521 | if (sb->s_root) | |
522 | return sb; | |
523 | up_read(&sb->s_umount); | |
524 | /* restart only when sb is no longer on the list */ | |
525 | spin_lock(&sb_lock); | |
526 | if (__put_super_and_need_restart(sb)) | |
527 | goto rescan; | |
1da177e4 LT |
528 | } |
529 | } | |
530 | spin_unlock(&sb_lock); | |
531 | return NULL; | |
532 | } | |
533 | ||
1da177e4 LT |
534 | asmlinkage long sys_ustat(unsigned dev, struct ustat __user * ubuf) |
535 | { | |
536 | struct super_block *s; | |
537 | struct ustat tmp; | |
538 | struct kstatfs sbuf; | |
539 | int err = -EINVAL; | |
540 | ||
541 | s = user_get_super(new_decode_dev(dev)); | |
542 | if (s == NULL) | |
543 | goto out; | |
726c3342 | 544 | err = vfs_statfs(s->s_root, &sbuf); |
1da177e4 LT |
545 | drop_super(s); |
546 | if (err) | |
547 | goto out; | |
548 | ||
549 | memset(&tmp,0,sizeof(struct ustat)); | |
550 | tmp.f_tfree = sbuf.f_bfree; | |
551 | tmp.f_tinode = sbuf.f_ffree; | |
552 | ||
553 | err = copy_to_user(ubuf,&tmp,sizeof(struct ustat)) ? -EFAULT : 0; | |
554 | out: | |
555 | return err; | |
556 | } | |
557 | ||
558 | /** | |
559 | * mark_files_ro | |
560 | * @sb: superblock in question | |
561 | * | |
562 | * All files are marked read/only. We don't care about pending | |
563 | * delete files so this should be used in 'force' mode only | |
564 | */ | |
565 | ||
566 | static void mark_files_ro(struct super_block *sb) | |
567 | { | |
568 | struct file *f; | |
569 | ||
570 | file_list_lock(); | |
2f512016 | 571 | list_for_each_entry(f, &sb->s_files, f_u.fu_list) { |
0f7fc9e4 | 572 | if (S_ISREG(f->f_path.dentry->d_inode->i_mode) && file_count(f)) |
1da177e4 LT |
573 | f->f_mode &= ~FMODE_WRITE; |
574 | } | |
575 | file_list_unlock(); | |
576 | } | |
577 | ||
578 | /** | |
579 | * do_remount_sb - asks filesystem to change mount options. | |
580 | * @sb: superblock in question | |
581 | * @flags: numeric part of options | |
582 | * @data: the rest of options | |
583 | * @force: whether or not to force the change | |
584 | * | |
585 | * Alters the mount options of a mounted file system. | |
586 | */ | |
587 | int do_remount_sb(struct super_block *sb, int flags, void *data, int force) | |
588 | { | |
589 | int retval; | |
590 | ||
9361401e | 591 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
592 | if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev)) |
593 | return -EACCES; | |
9361401e | 594 | #endif |
1da177e4 LT |
595 | if (flags & MS_RDONLY) |
596 | acct_auto_close(sb); | |
597 | shrink_dcache_sb(sb); | |
598 | fsync_super(sb); | |
599 | ||
600 | /* If we are remounting RDONLY and current sb is read/write, | |
601 | make sure there are no rw files opened */ | |
602 | if ((flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY)) { | |
603 | if (force) | |
604 | mark_files_ro(sb); | |
605 | else if (!fs_may_remount_ro(sb)) | |
606 | return -EBUSY; | |
66191dc6 | 607 | DQUOT_OFF(sb); |
1da177e4 LT |
608 | } |
609 | ||
610 | if (sb->s_op->remount_fs) { | |
611 | lock_super(sb); | |
612 | retval = sb->s_op->remount_fs(sb, &flags, data); | |
613 | unlock_super(sb); | |
614 | if (retval) | |
615 | return retval; | |
616 | } | |
617 | sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK); | |
618 | return 0; | |
619 | } | |
620 | ||
621 | static void do_emergency_remount(unsigned long foo) | |
622 | { | |
623 | struct super_block *sb; | |
624 | ||
625 | spin_lock(&sb_lock); | |
626 | list_for_each_entry(sb, &super_blocks, s_list) { | |
627 | sb->s_count++; | |
628 | spin_unlock(&sb_lock); | |
629 | down_read(&sb->s_umount); | |
630 | if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) { | |
631 | /* | |
632 | * ->remount_fs needs lock_kernel(). | |
633 | * | |
634 | * What lock protects sb->s_flags?? | |
635 | */ | |
636 | lock_kernel(); | |
637 | do_remount_sb(sb, MS_RDONLY, NULL, 1); | |
638 | unlock_kernel(); | |
639 | } | |
640 | drop_super(sb); | |
641 | spin_lock(&sb_lock); | |
642 | } | |
643 | spin_unlock(&sb_lock); | |
644 | printk("Emergency Remount complete\n"); | |
645 | } | |
646 | ||
647 | void emergency_remount(void) | |
648 | { | |
649 | pdflush_operation(do_emergency_remount, 0); | |
650 | } | |
651 | ||
652 | /* | |
653 | * Unnamed block devices are dummy devices used by virtual | |
654 | * filesystems which don't use real block-devices. -- jrs | |
655 | */ | |
656 | ||
657 | static struct idr unnamed_dev_idr; | |
658 | static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */ | |
659 | ||
660 | int set_anon_super(struct super_block *s, void *data) | |
661 | { | |
662 | int dev; | |
663 | int error; | |
664 | ||
665 | retry: | |
666 | if (idr_pre_get(&unnamed_dev_idr, GFP_ATOMIC) == 0) | |
667 | return -ENOMEM; | |
668 | spin_lock(&unnamed_dev_lock); | |
669 | error = idr_get_new(&unnamed_dev_idr, NULL, &dev); | |
670 | spin_unlock(&unnamed_dev_lock); | |
671 | if (error == -EAGAIN) | |
672 | /* We raced and lost with another CPU. */ | |
673 | goto retry; | |
674 | else if (error) | |
675 | return -EAGAIN; | |
676 | ||
677 | if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) { | |
678 | spin_lock(&unnamed_dev_lock); | |
679 | idr_remove(&unnamed_dev_idr, dev); | |
680 | spin_unlock(&unnamed_dev_lock); | |
681 | return -EMFILE; | |
682 | } | |
683 | s->s_dev = MKDEV(0, dev & MINORMASK); | |
684 | return 0; | |
685 | } | |
686 | ||
687 | EXPORT_SYMBOL(set_anon_super); | |
688 | ||
689 | void kill_anon_super(struct super_block *sb) | |
690 | { | |
691 | int slot = MINOR(sb->s_dev); | |
692 | ||
693 | generic_shutdown_super(sb); | |
694 | spin_lock(&unnamed_dev_lock); | |
695 | idr_remove(&unnamed_dev_idr, slot); | |
696 | spin_unlock(&unnamed_dev_lock); | |
697 | } | |
698 | ||
699 | EXPORT_SYMBOL(kill_anon_super); | |
700 | ||
701 | void __init unnamed_dev_init(void) | |
702 | { | |
703 | idr_init(&unnamed_dev_idr); | |
704 | } | |
705 | ||
706 | void kill_litter_super(struct super_block *sb) | |
707 | { | |
708 | if (sb->s_root) | |
709 | d_genocide(sb->s_root); | |
710 | kill_anon_super(sb); | |
711 | } | |
712 | ||
713 | EXPORT_SYMBOL(kill_litter_super); | |
714 | ||
9361401e | 715 | #ifdef CONFIG_BLOCK |
1da177e4 LT |
716 | static int set_bdev_super(struct super_block *s, void *data) |
717 | { | |
718 | s->s_bdev = data; | |
719 | s->s_dev = s->s_bdev->bd_dev; | |
720 | return 0; | |
721 | } | |
722 | ||
723 | static int test_bdev_super(struct super_block *s, void *data) | |
724 | { | |
725 | return (void *)s->s_bdev == data; | |
726 | } | |
727 | ||
454e2398 | 728 | int get_sb_bdev(struct file_system_type *fs_type, |
1da177e4 | 729 | int flags, const char *dev_name, void *data, |
454e2398 DH |
730 | int (*fill_super)(struct super_block *, void *, int), |
731 | struct vfsmount *mnt) | |
1da177e4 LT |
732 | { |
733 | struct block_device *bdev; | |
734 | struct super_block *s; | |
735 | int error = 0; | |
736 | ||
737 | bdev = open_bdev_excl(dev_name, flags, fs_type); | |
738 | if (IS_ERR(bdev)) | |
454e2398 | 739 | return PTR_ERR(bdev); |
1da177e4 LT |
740 | |
741 | /* | |
742 | * once the super is inserted into the list by sget, s_umount | |
743 | * will protect the lockfs code from trying to start a snapshot | |
744 | * while we are mounting | |
745 | */ | |
f73ca1b7 | 746 | down(&bdev->bd_mount_sem); |
1da177e4 | 747 | s = sget(fs_type, test_bdev_super, set_bdev_super, bdev); |
f73ca1b7 | 748 | up(&bdev->bd_mount_sem); |
1da177e4 | 749 | if (IS_ERR(s)) |
454e2398 | 750 | goto error_s; |
1da177e4 LT |
751 | |
752 | if (s->s_root) { | |
753 | if ((flags ^ s->s_flags) & MS_RDONLY) { | |
754 | up_write(&s->s_umount); | |
755 | deactivate_super(s); | |
454e2398 DH |
756 | error = -EBUSY; |
757 | goto error_bdev; | |
1da177e4 | 758 | } |
454e2398 DH |
759 | |
760 | close_bdev_excl(bdev); | |
1da177e4 LT |
761 | } else { |
762 | char b[BDEVNAME_SIZE]; | |
763 | ||
764 | s->s_flags = flags; | |
765 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); | |
e78c9a00 | 766 | sb_set_blocksize(s, block_size(bdev)); |
9b04c997 | 767 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 LT |
768 | if (error) { |
769 | up_write(&s->s_umount); | |
770 | deactivate_super(s); | |
454e2398 | 771 | goto error; |
fa675765 | 772 | } |
454e2398 DH |
773 | |
774 | s->s_flags |= MS_ACTIVE; | |
1da177e4 LT |
775 | } |
776 | ||
454e2398 | 777 | return simple_set_mnt(mnt, s); |
1da177e4 | 778 | |
454e2398 DH |
779 | error_s: |
780 | error = PTR_ERR(s); | |
781 | error_bdev: | |
1da177e4 | 782 | close_bdev_excl(bdev); |
454e2398 DH |
783 | error: |
784 | return error; | |
1da177e4 LT |
785 | } |
786 | ||
787 | EXPORT_SYMBOL(get_sb_bdev); | |
788 | ||
789 | void kill_block_super(struct super_block *sb) | |
790 | { | |
791 | struct block_device *bdev = sb->s_bdev; | |
792 | ||
1da177e4 LT |
793 | generic_shutdown_super(sb); |
794 | sync_blockdev(bdev); | |
795 | close_bdev_excl(bdev); | |
796 | } | |
797 | ||
798 | EXPORT_SYMBOL(kill_block_super); | |
9361401e | 799 | #endif |
1da177e4 | 800 | |
454e2398 | 801 | int get_sb_nodev(struct file_system_type *fs_type, |
1da177e4 | 802 | int flags, void *data, |
454e2398 DH |
803 | int (*fill_super)(struct super_block *, void *, int), |
804 | struct vfsmount *mnt) | |
1da177e4 LT |
805 | { |
806 | int error; | |
807 | struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL); | |
808 | ||
809 | if (IS_ERR(s)) | |
454e2398 | 810 | return PTR_ERR(s); |
1da177e4 LT |
811 | |
812 | s->s_flags = flags; | |
813 | ||
9b04c997 | 814 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 LT |
815 | if (error) { |
816 | up_write(&s->s_umount); | |
817 | deactivate_super(s); | |
454e2398 | 818 | return error; |
1da177e4 LT |
819 | } |
820 | s->s_flags |= MS_ACTIVE; | |
454e2398 | 821 | return simple_set_mnt(mnt, s); |
1da177e4 LT |
822 | } |
823 | ||
824 | EXPORT_SYMBOL(get_sb_nodev); | |
825 | ||
826 | static int compare_single(struct super_block *s, void *p) | |
827 | { | |
828 | return 1; | |
829 | } | |
830 | ||
454e2398 | 831 | int get_sb_single(struct file_system_type *fs_type, |
1da177e4 | 832 | int flags, void *data, |
454e2398 DH |
833 | int (*fill_super)(struct super_block *, void *, int), |
834 | struct vfsmount *mnt) | |
1da177e4 LT |
835 | { |
836 | struct super_block *s; | |
837 | int error; | |
838 | ||
839 | s = sget(fs_type, compare_single, set_anon_super, NULL); | |
840 | if (IS_ERR(s)) | |
454e2398 | 841 | return PTR_ERR(s); |
1da177e4 LT |
842 | if (!s->s_root) { |
843 | s->s_flags = flags; | |
9b04c997 | 844 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); |
1da177e4 LT |
845 | if (error) { |
846 | up_write(&s->s_umount); | |
847 | deactivate_super(s); | |
454e2398 | 848 | return error; |
1da177e4 LT |
849 | } |
850 | s->s_flags |= MS_ACTIVE; | |
851 | } | |
852 | do_remount_sb(s, flags, data, 0); | |
454e2398 | 853 | return simple_set_mnt(mnt, s); |
1da177e4 LT |
854 | } |
855 | ||
856 | EXPORT_SYMBOL(get_sb_single); | |
857 | ||
858 | struct vfsmount * | |
bb4a58bf | 859 | vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data) |
1da177e4 | 860 | { |
1da177e4 | 861 | struct vfsmount *mnt; |
1da177e4 | 862 | char *secdata = NULL; |
454e2398 | 863 | int error; |
1da177e4 LT |
864 | |
865 | if (!type) | |
866 | return ERR_PTR(-ENODEV); | |
1da177e4 | 867 | |
454e2398 | 868 | error = -ENOMEM; |
1da177e4 LT |
869 | mnt = alloc_vfsmnt(name); |
870 | if (!mnt) | |
871 | goto out; | |
872 | ||
873 | if (data) { | |
874 | secdata = alloc_secdata(); | |
454e2398 | 875 | if (!secdata) |
1da177e4 | 876 | goto out_mnt; |
1da177e4 LT |
877 | |
878 | error = security_sb_copy_data(type, data, secdata); | |
454e2398 | 879 | if (error) |
1da177e4 | 880 | goto out_free_secdata; |
1da177e4 LT |
881 | } |
882 | ||
454e2398 DH |
883 | error = type->get_sb(type, flags, name, data, mnt); |
884 | if (error < 0) | |
1da177e4 | 885 | goto out_free_secdata; |
b4c07bce | 886 | BUG_ON(!mnt->mnt_sb); |
454e2398 DH |
887 | |
888 | error = security_sb_kern_mount(mnt->mnt_sb, secdata); | |
1da177e4 LT |
889 | if (error) |
890 | goto out_sb; | |
454e2398 DH |
891 | |
892 | mnt->mnt_mountpoint = mnt->mnt_root; | |
1da177e4 | 893 | mnt->mnt_parent = mnt; |
454e2398 | 894 | up_write(&mnt->mnt_sb->s_umount); |
8680e22f | 895 | free_secdata(secdata); |
1da177e4 LT |
896 | return mnt; |
897 | out_sb: | |
454e2398 DH |
898 | dput(mnt->mnt_root); |
899 | up_write(&mnt->mnt_sb->s_umount); | |
900 | deactivate_super(mnt->mnt_sb); | |
1da177e4 LT |
901 | out_free_secdata: |
902 | free_secdata(secdata); | |
903 | out_mnt: | |
904 | free_vfsmnt(mnt); | |
905 | out: | |
454e2398 | 906 | return ERR_PTR(error); |
1da177e4 LT |
907 | } |
908 | ||
bb4a58bf TM |
909 | EXPORT_SYMBOL_GPL(vfs_kern_mount); |
910 | ||
79c0b2df MS |
911 | static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype) |
912 | { | |
913 | int err; | |
914 | const char *subtype = strchr(fstype, '.'); | |
915 | if (subtype) { | |
916 | subtype++; | |
917 | err = -EINVAL; | |
918 | if (!subtype[0]) | |
919 | goto err; | |
920 | } else | |
921 | subtype = ""; | |
922 | ||
923 | mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL); | |
924 | err = -ENOMEM; | |
925 | if (!mnt->mnt_sb->s_subtype) | |
926 | goto err; | |
927 | return mnt; | |
928 | ||
929 | err: | |
930 | mntput(mnt); | |
931 | return ERR_PTR(err); | |
932 | } | |
933 | ||
bb4a58bf TM |
934 | struct vfsmount * |
935 | do_kern_mount(const char *fstype, int flags, const char *name, void *data) | |
936 | { | |
937 | struct file_system_type *type = get_fs_type(fstype); | |
938 | struct vfsmount *mnt; | |
939 | if (!type) | |
940 | return ERR_PTR(-ENODEV); | |
941 | mnt = vfs_kern_mount(type, flags, name, data); | |
79c0b2df MS |
942 | if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) && |
943 | !mnt->mnt_sb->s_subtype) | |
944 | mnt = fs_set_subtype(mnt, fstype); | |
bb4a58bf TM |
945 | put_filesystem(type); |
946 | return mnt; | |
947 | } | |
948 | ||
8bf9725c | 949 | struct vfsmount *kern_mount_data(struct file_system_type *type, void *data) |
1da177e4 | 950 | { |
8bf9725c | 951 | return vfs_kern_mount(type, MS_KERNMOUNT, type->name, data); |
1da177e4 LT |
952 | } |
953 | ||
8bf9725c | 954 | EXPORT_SYMBOL_GPL(kern_mount_data); |