]>
Commit | Line | Data |
---|---|---|
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/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/mount.h> | |
32 | #include <linux/security.h> | |
33 | #include <linux/syscalls.h> | |
34 | #include <linux/vfs.h> | |
35 | #include <linux/writeback.h> /* for the emergency remount stuff */ | |
36 | #include <linux/idr.h> | |
37 | #include <linux/kobject.h> | |
38 | #include <linux/mutex.h> | |
39 | #include <linux/file.h> | |
40 | #include <linux/backing-dev.h> | |
41 | #include <asm/uaccess.h> | |
42 | #include "internal.h" | |
43 | ||
44 | ||
45 | LIST_HEAD(super_blocks); | |
46 | DEFINE_SPINLOCK(sb_lock); | |
47 | ||
48 | /** | |
49 | * alloc_super - create new superblock | |
50 | * @type: filesystem type superblock should belong to | |
51 | * | |
52 | * Allocates and initializes a new &struct super_block. alloc_super() | |
53 | * returns a pointer new superblock or %NULL if allocation had failed. | |
54 | */ | |
55 | static struct super_block *alloc_super(struct file_system_type *type) | |
56 | { | |
57 | struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER); | |
58 | static const struct super_operations default_op; | |
59 | ||
60 | if (s) { | |
61 | if (security_sb_alloc(s)) { | |
62 | kfree(s); | |
63 | s = NULL; | |
64 | goto out; | |
65 | } | |
66 | INIT_LIST_HEAD(&s->s_files); | |
67 | INIT_LIST_HEAD(&s->s_instances); | |
68 | INIT_HLIST_HEAD(&s->s_anon); | |
69 | INIT_LIST_HEAD(&s->s_inodes); | |
70 | INIT_LIST_HEAD(&s->s_dentry_lru); | |
71 | init_rwsem(&s->s_umount); | |
72 | mutex_init(&s->s_lock); | |
73 | lockdep_set_class(&s->s_umount, &type->s_umount_key); | |
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); | |
80 | /* | |
81 | * sget() can have s_umount recursion. | |
82 | * | |
83 | * When it cannot find a suitable sb, it allocates a new | |
84 | * one (this one), and tries again to find a suitable old | |
85 | * one. | |
86 | * | |
87 | * In case that succeeds, it will acquire the s_umount | |
88 | * lock of the old one. Since these are clearly distrinct | |
89 | * locks, and this object isn't exposed yet, there's no | |
90 | * risk of deadlocks. | |
91 | * | |
92 | * Annotate this by putting this lock in a different | |
93 | * subclass. | |
94 | */ | |
95 | down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); | |
96 | s->s_count = 1; | |
97 | atomic_set(&s->s_active, 1); | |
98 | mutex_init(&s->s_vfs_rename_mutex); | |
99 | mutex_init(&s->s_dquot.dqio_mutex); | |
100 | mutex_init(&s->s_dquot.dqonoff_mutex); | |
101 | init_rwsem(&s->s_dquot.dqptr_sem); | |
102 | init_waitqueue_head(&s->s_wait_unfrozen); | |
103 | s->s_maxbytes = MAX_NON_LFS; | |
104 | s->dq_op = sb_dquot_ops; | |
105 | s->s_qcop = sb_quotactl_ops; | |
106 | s->s_op = &default_op; | |
107 | s->s_time_gran = 1000000000; | |
108 | } | |
109 | out: | |
110 | return s; | |
111 | } | |
112 | ||
113 | /** | |
114 | * destroy_super - frees a superblock | |
115 | * @s: superblock to free | |
116 | * | |
117 | * Frees a superblock. | |
118 | */ | |
119 | static inline void destroy_super(struct super_block *s) | |
120 | { | |
121 | security_sb_free(s); | |
122 | kfree(s->s_subtype); | |
123 | kfree(s->s_options); | |
124 | kfree(s); | |
125 | } | |
126 | ||
127 | /* Superblock refcounting */ | |
128 | ||
129 | /* | |
130 | * Drop a superblock's refcount. The caller must hold sb_lock. | |
131 | */ | |
132 | void __put_super(struct super_block *sb) | |
133 | { | |
134 | if (!--sb->s_count) { | |
135 | list_del_init(&sb->s_list); | |
136 | destroy_super(sb); | |
137 | } | |
138 | } | |
139 | ||
140 | /** | |
141 | * put_super - drop a temporary reference to superblock | |
142 | * @sb: superblock in question | |
143 | * | |
144 | * Drops a temporary reference, frees superblock if there's no | |
145 | * references left. | |
146 | */ | |
147 | void put_super(struct super_block *sb) | |
148 | { | |
149 | spin_lock(&sb_lock); | |
150 | __put_super(sb); | |
151 | spin_unlock(&sb_lock); | |
152 | } | |
153 | ||
154 | ||
155 | /** | |
156 | * deactivate_locked_super - drop an active reference to superblock | |
157 | * @s: superblock to deactivate | |
158 | * | |
159 | * Drops an active reference to superblock, converting it into a temprory | |
160 | * one if there is no other active references left. In that case we | |
161 | * tell fs driver to shut it down and drop the temporary reference we | |
162 | * had just acquired. | |
163 | * | |
164 | * Caller holds exclusive lock on superblock; that lock is released. | |
165 | */ | |
166 | void deactivate_locked_super(struct super_block *s) | |
167 | { | |
168 | struct file_system_type *fs = s->s_type; | |
169 | if (atomic_dec_and_test(&s->s_active)) { | |
170 | vfs_dq_off(s, 0); | |
171 | fs->kill_sb(s); | |
172 | put_filesystem(fs); | |
173 | put_super(s); | |
174 | } else { | |
175 | up_write(&s->s_umount); | |
176 | } | |
177 | } | |
178 | ||
179 | EXPORT_SYMBOL(deactivate_locked_super); | |
180 | ||
181 | /** | |
182 | * deactivate_super - drop an active reference to superblock | |
183 | * @s: superblock to deactivate | |
184 | * | |
185 | * Variant of deactivate_locked_super(), except that superblock is *not* | |
186 | * locked by caller. If we are going to drop the final active reference, | |
187 | * lock will be acquired prior to that. | |
188 | */ | |
189 | void deactivate_super(struct super_block *s) | |
190 | { | |
191 | if (!atomic_add_unless(&s->s_active, -1, 1)) { | |
192 | down_write(&s->s_umount); | |
193 | deactivate_locked_super(s); | |
194 | } | |
195 | } | |
196 | ||
197 | EXPORT_SYMBOL(deactivate_super); | |
198 | ||
199 | /** | |
200 | * grab_super - acquire an active reference | |
201 | * @s: reference we are trying to make active | |
202 | * | |
203 | * Tries to acquire an active reference. grab_super() is used when we | |
204 | * had just found a superblock in super_blocks or fs_type->fs_supers | |
205 | * and want to turn it into a full-blown active reference. grab_super() | |
206 | * is called with sb_lock held and drops it. Returns 1 in case of | |
207 | * success, 0 if we had failed (superblock contents was already dead or | |
208 | * dying when grab_super() had been called). | |
209 | */ | |
210 | static int grab_super(struct super_block *s) __releases(sb_lock) | |
211 | { | |
212 | if (atomic_inc_not_zero(&s->s_active)) { | |
213 | spin_unlock(&sb_lock); | |
214 | down_write(&s->s_umount); | |
215 | return 1; | |
216 | } | |
217 | /* it's going away */ | |
218 | s->s_count++; | |
219 | spin_unlock(&sb_lock); | |
220 | /* wait for it to die */ | |
221 | down_write(&s->s_umount); | |
222 | up_write(&s->s_umount); | |
223 | put_super(s); | |
224 | return 0; | |
225 | } | |
226 | ||
227 | /* | |
228 | * Superblock locking. We really ought to get rid of these two. | |
229 | */ | |
230 | void lock_super(struct super_block * sb) | |
231 | { | |
232 | get_fs_excl(); | |
233 | mutex_lock(&sb->s_lock); | |
234 | } | |
235 | ||
236 | void unlock_super(struct super_block * sb) | |
237 | { | |
238 | put_fs_excl(); | |
239 | mutex_unlock(&sb->s_lock); | |
240 | } | |
241 | ||
242 | EXPORT_SYMBOL(lock_super); | |
243 | EXPORT_SYMBOL(unlock_super); | |
244 | ||
245 | /** | |
246 | * generic_shutdown_super - common helper for ->kill_sb() | |
247 | * @sb: superblock to kill | |
248 | * | |
249 | * generic_shutdown_super() does all fs-independent work on superblock | |
250 | * shutdown. Typical ->kill_sb() should pick all fs-specific objects | |
251 | * that need destruction out of superblock, call generic_shutdown_super() | |
252 | * and release aforementioned objects. Note: dentries and inodes _are_ | |
253 | * taken care of and do not need specific handling. | |
254 | * | |
255 | * Upon calling this function, the filesystem may no longer alter or | |
256 | * rearrange the set of dentries belonging to this super_block, nor may it | |
257 | * change the attachments of dentries to inodes. | |
258 | */ | |
259 | void generic_shutdown_super(struct super_block *sb) | |
260 | { | |
261 | const struct super_operations *sop = sb->s_op; | |
262 | ||
263 | ||
264 | if (sb->s_root) { | |
265 | shrink_dcache_for_umount(sb); | |
266 | sync_filesystem(sb); | |
267 | get_fs_excl(); | |
268 | sb->s_flags &= ~MS_ACTIVE; | |
269 | ||
270 | /* bad name - it should be evict_inodes() */ | |
271 | invalidate_inodes(sb); | |
272 | ||
273 | if (sop->put_super) | |
274 | sop->put_super(sb); | |
275 | ||
276 | /* Forget any remaining inodes */ | |
277 | if (invalidate_inodes(sb)) { | |
278 | printk("VFS: Busy inodes after unmount of %s. " | |
279 | "Self-destruct in 5 seconds. Have a nice day...\n", | |
280 | sb->s_id); | |
281 | } | |
282 | put_fs_excl(); | |
283 | } | |
284 | spin_lock(&sb_lock); | |
285 | /* should be initialized for __put_super_and_need_restart() */ | |
286 | list_del_init(&sb->s_instances); | |
287 | spin_unlock(&sb_lock); | |
288 | up_write(&sb->s_umount); | |
289 | } | |
290 | ||
291 | EXPORT_SYMBOL(generic_shutdown_super); | |
292 | ||
293 | /** | |
294 | * sget - find or create a superblock | |
295 | * @type: filesystem type superblock should belong to | |
296 | * @test: comparison callback | |
297 | * @set: setup callback | |
298 | * @data: argument to each of them | |
299 | */ | |
300 | struct super_block *sget(struct file_system_type *type, | |
301 | int (*test)(struct super_block *,void *), | |
302 | int (*set)(struct super_block *,void *), | |
303 | void *data) | |
304 | { | |
305 | struct super_block *s = NULL; | |
306 | struct super_block *old; | |
307 | int err; | |
308 | ||
309 | retry: | |
310 | spin_lock(&sb_lock); | |
311 | if (test) { | |
312 | list_for_each_entry(old, &type->fs_supers, s_instances) { | |
313 | if (!test(old, data)) | |
314 | continue; | |
315 | if (!grab_super(old)) | |
316 | goto retry; | |
317 | if (s) { | |
318 | up_write(&s->s_umount); | |
319 | destroy_super(s); | |
320 | } | |
321 | return old; | |
322 | } | |
323 | } | |
324 | if (!s) { | |
325 | spin_unlock(&sb_lock); | |
326 | s = alloc_super(type); | |
327 | if (!s) | |
328 | return ERR_PTR(-ENOMEM); | |
329 | goto retry; | |
330 | } | |
331 | ||
332 | err = set(s, data); | |
333 | if (err) { | |
334 | spin_unlock(&sb_lock); | |
335 | up_write(&s->s_umount); | |
336 | destroy_super(s); | |
337 | return ERR_PTR(err); | |
338 | } | |
339 | s->s_type = type; | |
340 | strlcpy(s->s_id, type->name, sizeof(s->s_id)); | |
341 | list_add_tail(&s->s_list, &super_blocks); | |
342 | list_add(&s->s_instances, &type->fs_supers); | |
343 | spin_unlock(&sb_lock); | |
344 | get_filesystem(type); | |
345 | return s; | |
346 | } | |
347 | ||
348 | EXPORT_SYMBOL(sget); | |
349 | ||
350 | void drop_super(struct super_block *sb) | |
351 | { | |
352 | up_read(&sb->s_umount); | |
353 | put_super(sb); | |
354 | } | |
355 | ||
356 | EXPORT_SYMBOL(drop_super); | |
357 | ||
358 | /** | |
359 | * sync_supers - helper for periodic superblock writeback | |
360 | * | |
361 | * Call the write_super method if present on all dirty superblocks in | |
362 | * the system. This is for the periodic writeback used by most older | |
363 | * filesystems. For data integrity superblock writeback use | |
364 | * sync_filesystems() instead. | |
365 | * | |
366 | * Note: check the dirty flag before waiting, so we don't | |
367 | * hold up the sync while mounting a device. (The newly | |
368 | * mounted device won't need syncing.) | |
369 | */ | |
370 | void sync_supers(void) | |
371 | { | |
372 | struct super_block *sb, *n; | |
373 | ||
374 | spin_lock(&sb_lock); | |
375 | list_for_each_entry_safe(sb, n, &super_blocks, s_list) { | |
376 | if (list_empty(&sb->s_instances)) | |
377 | continue; | |
378 | if (sb->s_op->write_super && sb->s_dirt) { | |
379 | sb->s_count++; | |
380 | spin_unlock(&sb_lock); | |
381 | ||
382 | down_read(&sb->s_umount); | |
383 | if (sb->s_root && sb->s_dirt) | |
384 | sb->s_op->write_super(sb); | |
385 | up_read(&sb->s_umount); | |
386 | ||
387 | spin_lock(&sb_lock); | |
388 | __put_super(sb); | |
389 | } | |
390 | } | |
391 | spin_unlock(&sb_lock); | |
392 | } | |
393 | ||
394 | /** | |
395 | * get_super - get the superblock of a device | |
396 | * @bdev: device to get the superblock for | |
397 | * | |
398 | * Scans the superblock list and finds the superblock of the file system | |
399 | * mounted on the device given. %NULL is returned if no match is found. | |
400 | */ | |
401 | ||
402 | struct super_block *get_super(struct block_device *bdev) | |
403 | { | |
404 | struct super_block *sb; | |
405 | ||
406 | if (!bdev) | |
407 | return NULL; | |
408 | ||
409 | spin_lock(&sb_lock); | |
410 | rescan: | |
411 | list_for_each_entry(sb, &super_blocks, s_list) { | |
412 | if (list_empty(&sb->s_instances)) | |
413 | continue; | |
414 | if (sb->s_bdev == bdev) { | |
415 | sb->s_count++; | |
416 | spin_unlock(&sb_lock); | |
417 | down_read(&sb->s_umount); | |
418 | /* still alive? */ | |
419 | if (sb->s_root) | |
420 | return sb; | |
421 | up_read(&sb->s_umount); | |
422 | /* nope, got unmounted */ | |
423 | spin_lock(&sb_lock); | |
424 | __put_super(sb); | |
425 | goto rescan; | |
426 | } | |
427 | } | |
428 | spin_unlock(&sb_lock); | |
429 | return NULL; | |
430 | } | |
431 | ||
432 | EXPORT_SYMBOL(get_super); | |
433 | ||
434 | /** | |
435 | * get_active_super - get an active reference to the superblock of a device | |
436 | * @bdev: device to get the superblock for | |
437 | * | |
438 | * Scans the superblock list and finds the superblock of the file system | |
439 | * mounted on the device given. Returns the superblock with an active | |
440 | * reference and s_umount held exclusively or %NULL if none was found. | |
441 | */ | |
442 | struct super_block *get_active_super(struct block_device *bdev) | |
443 | { | |
444 | struct super_block *sb; | |
445 | ||
446 | if (!bdev) | |
447 | return NULL; | |
448 | ||
449 | restart: | |
450 | spin_lock(&sb_lock); | |
451 | list_for_each_entry(sb, &super_blocks, s_list) { | |
452 | if (list_empty(&sb->s_instances)) | |
453 | continue; | |
454 | if (sb->s_bdev == bdev) { | |
455 | if (grab_super(sb)) /* drops sb_lock */ | |
456 | return sb; | |
457 | else | |
458 | goto restart; | |
459 | } | |
460 | } | |
461 | spin_unlock(&sb_lock); | |
462 | return NULL; | |
463 | } | |
464 | ||
465 | struct super_block *user_get_super(dev_t dev) | |
466 | { | |
467 | struct super_block *sb; | |
468 | ||
469 | spin_lock(&sb_lock); | |
470 | rescan: | |
471 | list_for_each_entry(sb, &super_blocks, s_list) { | |
472 | if (list_empty(&sb->s_instances)) | |
473 | continue; | |
474 | if (sb->s_dev == dev) { | |
475 | sb->s_count++; | |
476 | spin_unlock(&sb_lock); | |
477 | down_read(&sb->s_umount); | |
478 | /* still alive? */ | |
479 | if (sb->s_root) | |
480 | return sb; | |
481 | up_read(&sb->s_umount); | |
482 | /* nope, got unmounted */ | |
483 | spin_lock(&sb_lock); | |
484 | __put_super(sb); | |
485 | goto rescan; | |
486 | } | |
487 | } | |
488 | spin_unlock(&sb_lock); | |
489 | return NULL; | |
490 | } | |
491 | ||
492 | SYSCALL_DEFINE2(ustat, unsigned, dev, struct ustat __user *, ubuf) | |
493 | { | |
494 | struct super_block *s; | |
495 | struct ustat tmp; | |
496 | struct kstatfs sbuf; | |
497 | int err = -EINVAL; | |
498 | ||
499 | s = user_get_super(new_decode_dev(dev)); | |
500 | if (s == NULL) | |
501 | goto out; | |
502 | err = vfs_statfs(s->s_root, &sbuf); | |
503 | drop_super(s); | |
504 | if (err) | |
505 | goto out; | |
506 | ||
507 | memset(&tmp,0,sizeof(struct ustat)); | |
508 | tmp.f_tfree = sbuf.f_bfree; | |
509 | tmp.f_tinode = sbuf.f_ffree; | |
510 | ||
511 | err = copy_to_user(ubuf,&tmp,sizeof(struct ustat)) ? -EFAULT : 0; | |
512 | out: | |
513 | return err; | |
514 | } | |
515 | ||
516 | /** | |
517 | * do_remount_sb - asks filesystem to change mount options. | |
518 | * @sb: superblock in question | |
519 | * @flags: numeric part of options | |
520 | * @data: the rest of options | |
521 | * @force: whether or not to force the change | |
522 | * | |
523 | * Alters the mount options of a mounted file system. | |
524 | */ | |
525 | int do_remount_sb(struct super_block *sb, int flags, void *data, int force) | |
526 | { | |
527 | int retval; | |
528 | int remount_rw, remount_ro; | |
529 | ||
530 | if (sb->s_frozen != SB_UNFROZEN) | |
531 | return -EBUSY; | |
532 | ||
533 | #ifdef CONFIG_BLOCK | |
534 | if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev)) | |
535 | return -EACCES; | |
536 | #endif | |
537 | ||
538 | if (flags & MS_RDONLY) | |
539 | acct_auto_close(sb); | |
540 | shrink_dcache_sb(sb); | |
541 | sync_filesystem(sb); | |
542 | ||
543 | remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY); | |
544 | remount_rw = !(flags & MS_RDONLY) && (sb->s_flags & MS_RDONLY); | |
545 | ||
546 | /* If we are remounting RDONLY and current sb is read/write, | |
547 | make sure there are no rw files opened */ | |
548 | if (remount_ro) { | |
549 | if (force) | |
550 | mark_files_ro(sb); | |
551 | else if (!fs_may_remount_ro(sb)) | |
552 | return -EBUSY; | |
553 | retval = vfs_dq_off(sb, 1); | |
554 | if (retval < 0 && retval != -ENOSYS) | |
555 | return -EBUSY; | |
556 | } | |
557 | ||
558 | if (sb->s_op->remount_fs) { | |
559 | retval = sb->s_op->remount_fs(sb, &flags, data); | |
560 | if (retval) | |
561 | return retval; | |
562 | } | |
563 | sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK); | |
564 | if (remount_rw) | |
565 | vfs_dq_quota_on_remount(sb); | |
566 | /* | |
567 | * Some filesystems modify their metadata via some other path than the | |
568 | * bdev buffer cache (eg. use a private mapping, or directories in | |
569 | * pagecache, etc). Also file data modifications go via their own | |
570 | * mappings. So If we try to mount readonly then copy the filesystem | |
571 | * from bdev, we could get stale data, so invalidate it to give a best | |
572 | * effort at coherency. | |
573 | */ | |
574 | if (remount_ro && sb->s_bdev) | |
575 | invalidate_bdev(sb->s_bdev); | |
576 | return 0; | |
577 | } | |
578 | ||
579 | static void do_emergency_remount(struct work_struct *work) | |
580 | { | |
581 | struct super_block *sb, *n; | |
582 | ||
583 | spin_lock(&sb_lock); | |
584 | list_for_each_entry_safe(sb, n, &super_blocks, s_list) { | |
585 | if (list_empty(&sb->s_instances)) | |
586 | continue; | |
587 | sb->s_count++; | |
588 | spin_unlock(&sb_lock); | |
589 | down_write(&sb->s_umount); | |
590 | if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) { | |
591 | /* | |
592 | * What lock protects sb->s_flags?? | |
593 | */ | |
594 | do_remount_sb(sb, MS_RDONLY, NULL, 1); | |
595 | } | |
596 | up_write(&sb->s_umount); | |
597 | spin_lock(&sb_lock); | |
598 | __put_super(sb); | |
599 | } | |
600 | spin_unlock(&sb_lock); | |
601 | kfree(work); | |
602 | printk("Emergency Remount complete\n"); | |
603 | } | |
604 | ||
605 | void emergency_remount(void) | |
606 | { | |
607 | struct work_struct *work; | |
608 | ||
609 | work = kmalloc(sizeof(*work), GFP_ATOMIC); | |
610 | if (work) { | |
611 | INIT_WORK(work, do_emergency_remount); | |
612 | schedule_work(work); | |
613 | } | |
614 | } | |
615 | ||
616 | /* | |
617 | * Unnamed block devices are dummy devices used by virtual | |
618 | * filesystems which don't use real block-devices. -- jrs | |
619 | */ | |
620 | ||
621 | static DEFINE_IDA(unnamed_dev_ida); | |
622 | static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */ | |
623 | static int unnamed_dev_start = 0; /* don't bother trying below it */ | |
624 | ||
625 | int set_anon_super(struct super_block *s, void *data) | |
626 | { | |
627 | int dev; | |
628 | int error; | |
629 | ||
630 | retry: | |
631 | if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0) | |
632 | return -ENOMEM; | |
633 | spin_lock(&unnamed_dev_lock); | |
634 | error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev); | |
635 | if (!error) | |
636 | unnamed_dev_start = dev + 1; | |
637 | spin_unlock(&unnamed_dev_lock); | |
638 | if (error == -EAGAIN) | |
639 | /* We raced and lost with another CPU. */ | |
640 | goto retry; | |
641 | else if (error) | |
642 | return -EAGAIN; | |
643 | ||
644 | if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) { | |
645 | spin_lock(&unnamed_dev_lock); | |
646 | ida_remove(&unnamed_dev_ida, dev); | |
647 | if (unnamed_dev_start > dev) | |
648 | unnamed_dev_start = dev; | |
649 | spin_unlock(&unnamed_dev_lock); | |
650 | return -EMFILE; | |
651 | } | |
652 | s->s_dev = MKDEV(0, dev & MINORMASK); | |
653 | s->s_bdi = &noop_backing_dev_info; | |
654 | return 0; | |
655 | } | |
656 | ||
657 | EXPORT_SYMBOL(set_anon_super); | |
658 | ||
659 | void kill_anon_super(struct super_block *sb) | |
660 | { | |
661 | int slot = MINOR(sb->s_dev); | |
662 | ||
663 | generic_shutdown_super(sb); | |
664 | spin_lock(&unnamed_dev_lock); | |
665 | ida_remove(&unnamed_dev_ida, slot); | |
666 | if (slot < unnamed_dev_start) | |
667 | unnamed_dev_start = slot; | |
668 | spin_unlock(&unnamed_dev_lock); | |
669 | } | |
670 | ||
671 | EXPORT_SYMBOL(kill_anon_super); | |
672 | ||
673 | void kill_litter_super(struct super_block *sb) | |
674 | { | |
675 | if (sb->s_root) | |
676 | d_genocide(sb->s_root); | |
677 | kill_anon_super(sb); | |
678 | } | |
679 | ||
680 | EXPORT_SYMBOL(kill_litter_super); | |
681 | ||
682 | static int ns_test_super(struct super_block *sb, void *data) | |
683 | { | |
684 | return sb->s_fs_info == data; | |
685 | } | |
686 | ||
687 | static int ns_set_super(struct super_block *sb, void *data) | |
688 | { | |
689 | sb->s_fs_info = data; | |
690 | return set_anon_super(sb, NULL); | |
691 | } | |
692 | ||
693 | int get_sb_ns(struct file_system_type *fs_type, int flags, void *data, | |
694 | int (*fill_super)(struct super_block *, void *, int), | |
695 | struct vfsmount *mnt) | |
696 | { | |
697 | struct super_block *sb; | |
698 | ||
699 | sb = sget(fs_type, ns_test_super, ns_set_super, data); | |
700 | if (IS_ERR(sb)) | |
701 | return PTR_ERR(sb); | |
702 | ||
703 | if (!sb->s_root) { | |
704 | int err; | |
705 | sb->s_flags = flags; | |
706 | err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0); | |
707 | if (err) { | |
708 | deactivate_locked_super(sb); | |
709 | return err; | |
710 | } | |
711 | ||
712 | sb->s_flags |= MS_ACTIVE; | |
713 | } | |
714 | ||
715 | simple_set_mnt(mnt, sb); | |
716 | return 0; | |
717 | } | |
718 | ||
719 | EXPORT_SYMBOL(get_sb_ns); | |
720 | ||
721 | #ifdef CONFIG_BLOCK | |
722 | static int set_bdev_super(struct super_block *s, void *data) | |
723 | { | |
724 | s->s_bdev = data; | |
725 | s->s_dev = s->s_bdev->bd_dev; | |
726 | ||
727 | /* | |
728 | * We set the bdi here to the queue backing, file systems can | |
729 | * overwrite this in ->fill_super() | |
730 | */ | |
731 | s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info; | |
732 | return 0; | |
733 | } | |
734 | ||
735 | static int test_bdev_super(struct super_block *s, void *data) | |
736 | { | |
737 | return (void *)s->s_bdev == data; | |
738 | } | |
739 | ||
740 | int get_sb_bdev(struct file_system_type *fs_type, | |
741 | int flags, const char *dev_name, void *data, | |
742 | int (*fill_super)(struct super_block *, void *, int), | |
743 | struct vfsmount *mnt) | |
744 | { | |
745 | struct block_device *bdev; | |
746 | struct super_block *s; | |
747 | fmode_t mode = FMODE_READ; | |
748 | int error = 0; | |
749 | ||
750 | if (!(flags & MS_RDONLY)) | |
751 | mode |= FMODE_WRITE; | |
752 | ||
753 | bdev = open_bdev_exclusive(dev_name, mode, fs_type); | |
754 | if (IS_ERR(bdev)) | |
755 | return PTR_ERR(bdev); | |
756 | ||
757 | /* | |
758 | * once the super is inserted into the list by sget, s_umount | |
759 | * will protect the lockfs code from trying to start a snapshot | |
760 | * while we are mounting | |
761 | */ | |
762 | mutex_lock(&bdev->bd_fsfreeze_mutex); | |
763 | if (bdev->bd_fsfreeze_count > 0) { | |
764 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
765 | error = -EBUSY; | |
766 | goto error_bdev; | |
767 | } | |
768 | s = sget(fs_type, test_bdev_super, set_bdev_super, bdev); | |
769 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
770 | if (IS_ERR(s)) | |
771 | goto error_s; | |
772 | ||
773 | if (s->s_root) { | |
774 | if ((flags ^ s->s_flags) & MS_RDONLY) { | |
775 | deactivate_locked_super(s); | |
776 | error = -EBUSY; | |
777 | goto error_bdev; | |
778 | } | |
779 | ||
780 | close_bdev_exclusive(bdev, mode); | |
781 | } else { | |
782 | char b[BDEVNAME_SIZE]; | |
783 | ||
784 | s->s_flags = flags; | |
785 | s->s_mode = mode; | |
786 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); | |
787 | sb_set_blocksize(s, block_size(bdev)); | |
788 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); | |
789 | if (error) { | |
790 | deactivate_locked_super(s); | |
791 | goto error; | |
792 | } | |
793 | ||
794 | s->s_flags |= MS_ACTIVE; | |
795 | bdev->bd_super = s; | |
796 | } | |
797 | ||
798 | simple_set_mnt(mnt, s); | |
799 | return 0; | |
800 | ||
801 | error_s: | |
802 | error = PTR_ERR(s); | |
803 | error_bdev: | |
804 | close_bdev_exclusive(bdev, mode); | |
805 | error: | |
806 | return error; | |
807 | } | |
808 | ||
809 | EXPORT_SYMBOL(get_sb_bdev); | |
810 | ||
811 | void kill_block_super(struct super_block *sb) | |
812 | { | |
813 | struct block_device *bdev = sb->s_bdev; | |
814 | fmode_t mode = sb->s_mode; | |
815 | ||
816 | bdev->bd_super = NULL; | |
817 | generic_shutdown_super(sb); | |
818 | sync_blockdev(bdev); | |
819 | close_bdev_exclusive(bdev, mode); | |
820 | } | |
821 | ||
822 | EXPORT_SYMBOL(kill_block_super); | |
823 | #endif | |
824 | ||
825 | int get_sb_nodev(struct file_system_type *fs_type, | |
826 | int flags, void *data, | |
827 | int (*fill_super)(struct super_block *, void *, int), | |
828 | struct vfsmount *mnt) | |
829 | { | |
830 | int error; | |
831 | struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL); | |
832 | ||
833 | if (IS_ERR(s)) | |
834 | return PTR_ERR(s); | |
835 | ||
836 | s->s_flags = flags; | |
837 | ||
838 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); | |
839 | if (error) { | |
840 | deactivate_locked_super(s); | |
841 | return error; | |
842 | } | |
843 | s->s_flags |= MS_ACTIVE; | |
844 | simple_set_mnt(mnt, s); | |
845 | return 0; | |
846 | } | |
847 | ||
848 | EXPORT_SYMBOL(get_sb_nodev); | |
849 | ||
850 | static int compare_single(struct super_block *s, void *p) | |
851 | { | |
852 | return 1; | |
853 | } | |
854 | ||
855 | int get_sb_single(struct file_system_type *fs_type, | |
856 | int flags, void *data, | |
857 | int (*fill_super)(struct super_block *, void *, int), | |
858 | struct vfsmount *mnt) | |
859 | { | |
860 | struct super_block *s; | |
861 | int error; | |
862 | ||
863 | s = sget(fs_type, compare_single, set_anon_super, NULL); | |
864 | if (IS_ERR(s)) | |
865 | return PTR_ERR(s); | |
866 | if (!s->s_root) { | |
867 | s->s_flags = flags; | |
868 | error = fill_super(s, data, flags & MS_SILENT ? 1 : 0); | |
869 | if (error) { | |
870 | deactivate_locked_super(s); | |
871 | return error; | |
872 | } | |
873 | s->s_flags |= MS_ACTIVE; | |
874 | } else { | |
875 | do_remount_sb(s, flags, data, 0); | |
876 | } | |
877 | simple_set_mnt(mnt, s); | |
878 | return 0; | |
879 | } | |
880 | ||
881 | EXPORT_SYMBOL(get_sb_single); | |
882 | ||
883 | struct vfsmount * | |
884 | vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data) | |
885 | { | |
886 | struct vfsmount *mnt; | |
887 | char *secdata = NULL; | |
888 | int error; | |
889 | ||
890 | if (!type) | |
891 | return ERR_PTR(-ENODEV); | |
892 | ||
893 | error = -ENOMEM; | |
894 | mnt = alloc_vfsmnt(name); | |
895 | if (!mnt) | |
896 | goto out; | |
897 | ||
898 | if (flags & MS_KERNMOUNT) | |
899 | mnt->mnt_flags = MNT_INTERNAL; | |
900 | ||
901 | if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) { | |
902 | secdata = alloc_secdata(); | |
903 | if (!secdata) | |
904 | goto out_mnt; | |
905 | ||
906 | error = security_sb_copy_data(data, secdata); | |
907 | if (error) | |
908 | goto out_free_secdata; | |
909 | } | |
910 | ||
911 | error = type->get_sb(type, flags, name, data, mnt); | |
912 | if (error < 0) | |
913 | goto out_free_secdata; | |
914 | BUG_ON(!mnt->mnt_sb); | |
915 | WARN_ON(!mnt->mnt_sb->s_bdi); | |
916 | ||
917 | error = security_sb_kern_mount(mnt->mnt_sb, flags, secdata); | |
918 | if (error) | |
919 | goto out_sb; | |
920 | ||
921 | /* | |
922 | * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE | |
923 | * but s_maxbytes was an unsigned long long for many releases. Throw | |
924 | * this warning for a little while to try and catch filesystems that | |
925 | * violate this rule. This warning should be either removed or | |
926 | * converted to a BUG() in 2.6.34. | |
927 | */ | |
928 | WARN((mnt->mnt_sb->s_maxbytes < 0), "%s set sb->s_maxbytes to " | |
929 | "negative value (%lld)\n", type->name, mnt->mnt_sb->s_maxbytes); | |
930 | ||
931 | mnt->mnt_mountpoint = mnt->mnt_root; | |
932 | mnt->mnt_parent = mnt; | |
933 | up_write(&mnt->mnt_sb->s_umount); | |
934 | free_secdata(secdata); | |
935 | return mnt; | |
936 | out_sb: | |
937 | dput(mnt->mnt_root); | |
938 | deactivate_locked_super(mnt->mnt_sb); | |
939 | out_free_secdata: | |
940 | free_secdata(secdata); | |
941 | out_mnt: | |
942 | free_vfsmnt(mnt); | |
943 | out: | |
944 | return ERR_PTR(error); | |
945 | } | |
946 | ||
947 | EXPORT_SYMBOL_GPL(vfs_kern_mount); | |
948 | ||
949 | static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype) | |
950 | { | |
951 | int err; | |
952 | const char *subtype = strchr(fstype, '.'); | |
953 | if (subtype) { | |
954 | subtype++; | |
955 | err = -EINVAL; | |
956 | if (!subtype[0]) | |
957 | goto err; | |
958 | } else | |
959 | subtype = ""; | |
960 | ||
961 | mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL); | |
962 | err = -ENOMEM; | |
963 | if (!mnt->mnt_sb->s_subtype) | |
964 | goto err; | |
965 | return mnt; | |
966 | ||
967 | err: | |
968 | mntput(mnt); | |
969 | return ERR_PTR(err); | |
970 | } | |
971 | ||
972 | struct vfsmount * | |
973 | do_kern_mount(const char *fstype, int flags, const char *name, void *data) | |
974 | { | |
975 | struct file_system_type *type = get_fs_type(fstype); | |
976 | struct vfsmount *mnt; | |
977 | if (!type) | |
978 | return ERR_PTR(-ENODEV); | |
979 | mnt = vfs_kern_mount(type, flags, name, data); | |
980 | if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) && | |
981 | !mnt->mnt_sb->s_subtype) | |
982 | mnt = fs_set_subtype(mnt, fstype); | |
983 | put_filesystem(type); | |
984 | return mnt; | |
985 | } | |
986 | EXPORT_SYMBOL_GPL(do_kern_mount); | |
987 | ||
988 | struct vfsmount *kern_mount_data(struct file_system_type *type, void *data) | |
989 | { | |
990 | return vfs_kern_mount(type, MS_KERNMOUNT, type->name, data); | |
991 | } | |
992 | ||
993 | EXPORT_SYMBOL_GPL(kern_mount_data); |