Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/block_dev.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | * Copyright (C) 2001 Andrea Arcangeli <andrea@suse.de> SuSE | |
6 | */ | |
7 | ||
1da177e4 LT |
8 | #include <linux/init.h> |
9 | #include <linux/mm.h> | |
10 | #include <linux/fcntl.h> | |
11 | #include <linux/slab.h> | |
12 | #include <linux/kmod.h> | |
13 | #include <linux/major.h> | |
1da177e4 | 14 | #include <linux/smp_lock.h> |
7db9cfd3 | 15 | #include <linux/device_cgroup.h> |
1da177e4 LT |
16 | #include <linux/highmem.h> |
17 | #include <linux/blkdev.h> | |
18 | #include <linux/module.h> | |
19 | #include <linux/blkpg.h> | |
20 | #include <linux/buffer_head.h> | |
585d3bc0 | 21 | #include <linux/pagevec.h> |
811d736f | 22 | #include <linux/writeback.h> |
1da177e4 LT |
23 | #include <linux/mpage.h> |
24 | #include <linux/mount.h> | |
25 | #include <linux/uio.h> | |
26 | #include <linux/namei.h> | |
1368c4f2 | 27 | #include <linux/log2.h> |
2e1483c9 | 28 | #include <linux/kmemleak.h> |
1da177e4 | 29 | #include <asm/uaccess.h> |
07f3f05c | 30 | #include "internal.h" |
1da177e4 LT |
31 | |
32 | struct bdev_inode { | |
33 | struct block_device bdev; | |
34 | struct inode vfs_inode; | |
35 | }; | |
36 | ||
4c54ac62 AB |
37 | static const struct address_space_operations def_blk_aops; |
38 | ||
1da177e4 LT |
39 | static inline struct bdev_inode *BDEV_I(struct inode *inode) |
40 | { | |
41 | return container_of(inode, struct bdev_inode, vfs_inode); | |
42 | } | |
43 | ||
44 | inline struct block_device *I_BDEV(struct inode *inode) | |
45 | { | |
46 | return &BDEV_I(inode)->bdev; | |
47 | } | |
48 | ||
49 | EXPORT_SYMBOL(I_BDEV); | |
50 | ||
51 | static sector_t max_block(struct block_device *bdev) | |
52 | { | |
53 | sector_t retval = ~((sector_t)0); | |
54 | loff_t sz = i_size_read(bdev->bd_inode); | |
55 | ||
56 | if (sz) { | |
57 | unsigned int size = block_size(bdev); | |
58 | unsigned int sizebits = blksize_bits(size); | |
59 | retval = (sz >> sizebits); | |
60 | } | |
61 | return retval; | |
62 | } | |
63 | ||
f9a14399 | 64 | /* Kill _all_ buffers and pagecache , dirty or not.. */ |
1da177e4 LT |
65 | static void kill_bdev(struct block_device *bdev) |
66 | { | |
f9a14399 PZ |
67 | if (bdev->bd_inode->i_mapping->nrpages == 0) |
68 | return; | |
69 | invalidate_bh_lrus(); | |
1da177e4 LT |
70 | truncate_inode_pages(bdev->bd_inode->i_mapping, 0); |
71 | } | |
72 | ||
73 | int set_blocksize(struct block_device *bdev, int size) | |
74 | { | |
75 | /* Size must be a power of two, and between 512 and PAGE_SIZE */ | |
1368c4f2 | 76 | if (size > PAGE_SIZE || size < 512 || !is_power_of_2(size)) |
1da177e4 LT |
77 | return -EINVAL; |
78 | ||
79 | /* Size cannot be smaller than the size supported by the device */ | |
e1defc4f | 80 | if (size < bdev_logical_block_size(bdev)) |
1da177e4 LT |
81 | return -EINVAL; |
82 | ||
83 | /* Don't change the size if it is same as current */ | |
84 | if (bdev->bd_block_size != size) { | |
85 | sync_blockdev(bdev); | |
86 | bdev->bd_block_size = size; | |
87 | bdev->bd_inode->i_blkbits = blksize_bits(size); | |
88 | kill_bdev(bdev); | |
89 | } | |
90 | return 0; | |
91 | } | |
92 | ||
93 | EXPORT_SYMBOL(set_blocksize); | |
94 | ||
95 | int sb_set_blocksize(struct super_block *sb, int size) | |
96 | { | |
1da177e4 LT |
97 | if (set_blocksize(sb->s_bdev, size)) |
98 | return 0; | |
99 | /* If we get here, we know size is power of two | |
100 | * and it's value is between 512 and PAGE_SIZE */ | |
101 | sb->s_blocksize = size; | |
38885bd4 | 102 | sb->s_blocksize_bits = blksize_bits(size); |
1da177e4 LT |
103 | return sb->s_blocksize; |
104 | } | |
105 | ||
106 | EXPORT_SYMBOL(sb_set_blocksize); | |
107 | ||
108 | int sb_min_blocksize(struct super_block *sb, int size) | |
109 | { | |
e1defc4f | 110 | int minsize = bdev_logical_block_size(sb->s_bdev); |
1da177e4 LT |
111 | if (size < minsize) |
112 | size = minsize; | |
113 | return sb_set_blocksize(sb, size); | |
114 | } | |
115 | ||
116 | EXPORT_SYMBOL(sb_min_blocksize); | |
117 | ||
118 | static int | |
119 | blkdev_get_block(struct inode *inode, sector_t iblock, | |
120 | struct buffer_head *bh, int create) | |
121 | { | |
122 | if (iblock >= max_block(I_BDEV(inode))) { | |
123 | if (create) | |
124 | return -EIO; | |
125 | ||
126 | /* | |
127 | * for reads, we're just trying to fill a partial page. | |
128 | * return a hole, they will have to call get_block again | |
129 | * before they can fill it, and they will get -EIO at that | |
130 | * time | |
131 | */ | |
132 | return 0; | |
133 | } | |
134 | bh->b_bdev = I_BDEV(inode); | |
135 | bh->b_blocknr = iblock; | |
136 | set_buffer_mapped(bh); | |
137 | return 0; | |
138 | } | |
139 | ||
b2e895db AM |
140 | static int |
141 | blkdev_get_blocks(struct inode *inode, sector_t iblock, | |
142 | struct buffer_head *bh, int create) | |
143 | { | |
144 | sector_t end_block = max_block(I_BDEV(inode)); | |
145 | unsigned long max_blocks = bh->b_size >> inode->i_blkbits; | |
146 | ||
147 | if ((iblock + max_blocks) > end_block) { | |
148 | max_blocks = end_block - iblock; | |
149 | if ((long)max_blocks <= 0) { | |
150 | if (create) | |
151 | return -EIO; /* write fully beyond EOF */ | |
152 | /* | |
153 | * It is a read which is fully beyond EOF. We return | |
154 | * a !buffer_mapped buffer | |
155 | */ | |
156 | max_blocks = 0; | |
157 | } | |
158 | } | |
159 | ||
160 | bh->b_bdev = I_BDEV(inode); | |
161 | bh->b_blocknr = iblock; | |
162 | bh->b_size = max_blocks << inode->i_blkbits; | |
163 | if (max_blocks) | |
164 | set_buffer_mapped(bh); | |
165 | return 0; | |
166 | } | |
167 | ||
168 | static ssize_t | |
169 | blkdev_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, | |
170 | loff_t offset, unsigned long nr_segs) | |
171 | { | |
172 | struct file *file = iocb->ki_filp; | |
173 | struct inode *inode = file->f_mapping->host; | |
174 | ||
175 | return blockdev_direct_IO_no_locking(rw, iocb, inode, I_BDEV(inode), | |
176 | iov, offset, nr_segs, blkdev_get_blocks, NULL); | |
177 | } | |
178 | ||
5cee5815 JK |
179 | int __sync_blockdev(struct block_device *bdev, int wait) |
180 | { | |
181 | if (!bdev) | |
182 | return 0; | |
183 | if (!wait) | |
184 | return filemap_flush(bdev->bd_inode->i_mapping); | |
185 | return filemap_write_and_wait(bdev->bd_inode->i_mapping); | |
186 | } | |
187 | ||
585d3bc0 NP |
188 | /* |
189 | * Write out and wait upon all the dirty data associated with a block | |
190 | * device via its mapping. Does not take the superblock lock. | |
191 | */ | |
192 | int sync_blockdev(struct block_device *bdev) | |
193 | { | |
5cee5815 | 194 | return __sync_blockdev(bdev, 1); |
585d3bc0 NP |
195 | } |
196 | EXPORT_SYMBOL(sync_blockdev); | |
197 | ||
198 | /* | |
199 | * Write out and wait upon all dirty data associated with this | |
200 | * device. Filesystem data as well as the underlying block | |
201 | * device. Takes the superblock lock. | |
202 | */ | |
203 | int fsync_bdev(struct block_device *bdev) | |
204 | { | |
205 | struct super_block *sb = get_super(bdev); | |
206 | if (sb) { | |
60b0680f | 207 | int res = sync_filesystem(sb); |
585d3bc0 NP |
208 | drop_super(sb); |
209 | return res; | |
210 | } | |
211 | return sync_blockdev(bdev); | |
212 | } | |
47e4491b | 213 | EXPORT_SYMBOL(fsync_bdev); |
585d3bc0 NP |
214 | |
215 | /** | |
216 | * freeze_bdev -- lock a filesystem and force it into a consistent state | |
217 | * @bdev: blockdevice to lock | |
218 | * | |
585d3bc0 NP |
219 | * If a superblock is found on this device, we take the s_umount semaphore |
220 | * on it to make sure nobody unmounts until the snapshot creation is done. | |
221 | * The reference counter (bd_fsfreeze_count) guarantees that only the last | |
222 | * unfreeze process can unfreeze the frozen filesystem actually when multiple | |
223 | * freeze requests arrive simultaneously. It counts up in freeze_bdev() and | |
224 | * count down in thaw_bdev(). When it becomes 0, thaw_bdev() will unfreeze | |
225 | * actually. | |
226 | */ | |
227 | struct super_block *freeze_bdev(struct block_device *bdev) | |
228 | { | |
229 | struct super_block *sb; | |
230 | int error = 0; | |
231 | ||
232 | mutex_lock(&bdev->bd_fsfreeze_mutex); | |
4504230a CH |
233 | if (++bdev->bd_fsfreeze_count > 1) { |
234 | /* | |
235 | * We don't even need to grab a reference - the first call | |
236 | * to freeze_bdev grab an active reference and only the last | |
237 | * thaw_bdev drops it. | |
238 | */ | |
585d3bc0 | 239 | sb = get_super(bdev); |
4504230a CH |
240 | drop_super(sb); |
241 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
242 | return sb; | |
243 | } | |
244 | ||
245 | sb = get_active_super(bdev); | |
246 | if (!sb) | |
247 | goto out; | |
248 | if (sb->s_flags & MS_RDONLY) { | |
4b06e5b9 JN |
249 | sb->s_frozen = SB_FREEZE_TRANS; |
250 | up_write(&sb->s_umount); | |
585d3bc0 NP |
251 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
252 | return sb; | |
253 | } | |
4504230a CH |
254 | |
255 | sb->s_frozen = SB_FREEZE_WRITE; | |
256 | smp_wmb(); | |
257 | ||
258 | sync_filesystem(sb); | |
259 | ||
260 | sb->s_frozen = SB_FREEZE_TRANS; | |
261 | smp_wmb(); | |
262 | ||
263 | sync_blockdev(sb->s_bdev); | |
264 | ||
265 | if (sb->s_op->freeze_fs) { | |
266 | error = sb->s_op->freeze_fs(sb); | |
267 | if (error) { | |
268 | printk(KERN_ERR | |
269 | "VFS:Filesystem freeze failed\n"); | |
270 | sb->s_frozen = SB_UNFROZEN; | |
271 | deactivate_locked_super(sb); | |
272 | bdev->bd_fsfreeze_count--; | |
273 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
274 | return ERR_PTR(error); | |
585d3bc0 NP |
275 | } |
276 | } | |
4504230a | 277 | up_write(&sb->s_umount); |
585d3bc0 | 278 | |
4504230a | 279 | out: |
585d3bc0 NP |
280 | sync_blockdev(bdev); |
281 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
4fadd7bb | 282 | return sb; /* thaw_bdev releases s->s_umount */ |
585d3bc0 NP |
283 | } |
284 | EXPORT_SYMBOL(freeze_bdev); | |
285 | ||
286 | /** | |
287 | * thaw_bdev -- unlock filesystem | |
288 | * @bdev: blockdevice to unlock | |
289 | * @sb: associated superblock | |
290 | * | |
291 | * Unlocks the filesystem and marks it writeable again after freeze_bdev(). | |
292 | */ | |
293 | int thaw_bdev(struct block_device *bdev, struct super_block *sb) | |
294 | { | |
4504230a | 295 | int error = -EINVAL; |
585d3bc0 NP |
296 | |
297 | mutex_lock(&bdev->bd_fsfreeze_mutex); | |
4504230a CH |
298 | if (!bdev->bd_fsfreeze_count) |
299 | goto out_unlock; | |
300 | ||
301 | error = 0; | |
302 | if (--bdev->bd_fsfreeze_count > 0) | |
303 | goto out_unlock; | |
304 | ||
305 | if (!sb) | |
306 | goto out_unlock; | |
307 | ||
308 | BUG_ON(sb->s_bdev != bdev); | |
309 | down_write(&sb->s_umount); | |
310 | if (sb->s_flags & MS_RDONLY) | |
4b06e5b9 | 311 | goto out_unfrozen; |
4504230a CH |
312 | |
313 | if (sb->s_op->unfreeze_fs) { | |
314 | error = sb->s_op->unfreeze_fs(sb); | |
315 | if (error) { | |
316 | printk(KERN_ERR | |
317 | "VFS:Filesystem thaw failed\n"); | |
318 | sb->s_frozen = SB_FREEZE_TRANS; | |
319 | bdev->bd_fsfreeze_count++; | |
320 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
321 | return error; | |
585d3bc0 | 322 | } |
585d3bc0 NP |
323 | } |
324 | ||
4b06e5b9 | 325 | out_unfrozen: |
4504230a CH |
326 | sb->s_frozen = SB_UNFROZEN; |
327 | smp_wmb(); | |
328 | wake_up(&sb->s_wait_unfrozen); | |
329 | ||
4504230a CH |
330 | if (sb) |
331 | deactivate_locked_super(sb); | |
332 | out_unlock: | |
585d3bc0 NP |
333 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
334 | return 0; | |
335 | } | |
336 | EXPORT_SYMBOL(thaw_bdev); | |
337 | ||
1da177e4 LT |
338 | static int blkdev_writepage(struct page *page, struct writeback_control *wbc) |
339 | { | |
340 | return block_write_full_page(page, blkdev_get_block, wbc); | |
341 | } | |
342 | ||
343 | static int blkdev_readpage(struct file * file, struct page * page) | |
344 | { | |
345 | return block_read_full_page(page, blkdev_get_block); | |
346 | } | |
347 | ||
6272b5a5 NP |
348 | static int blkdev_write_begin(struct file *file, struct address_space *mapping, |
349 | loff_t pos, unsigned len, unsigned flags, | |
350 | struct page **pagep, void **fsdata) | |
1da177e4 | 351 | { |
6272b5a5 NP |
352 | *pagep = NULL; |
353 | return block_write_begin(file, mapping, pos, len, flags, pagep, fsdata, | |
354 | blkdev_get_block); | |
1da177e4 LT |
355 | } |
356 | ||
6272b5a5 NP |
357 | static int blkdev_write_end(struct file *file, struct address_space *mapping, |
358 | loff_t pos, unsigned len, unsigned copied, | |
359 | struct page *page, void *fsdata) | |
1da177e4 | 360 | { |
6272b5a5 NP |
361 | int ret; |
362 | ret = block_write_end(file, mapping, pos, len, copied, page, fsdata); | |
363 | ||
364 | unlock_page(page); | |
365 | page_cache_release(page); | |
366 | ||
367 | return ret; | |
1da177e4 LT |
368 | } |
369 | ||
370 | /* | |
371 | * private llseek: | |
0f7fc9e4 | 372 | * for a block special file file->f_path.dentry->d_inode->i_size is zero |
1da177e4 LT |
373 | * so we compute the size by hand (just as in block_read/write above) |
374 | */ | |
375 | static loff_t block_llseek(struct file *file, loff_t offset, int origin) | |
376 | { | |
377 | struct inode *bd_inode = file->f_mapping->host; | |
378 | loff_t size; | |
379 | loff_t retval; | |
380 | ||
1b1dcc1b | 381 | mutex_lock(&bd_inode->i_mutex); |
1da177e4 LT |
382 | size = i_size_read(bd_inode); |
383 | ||
384 | switch (origin) { | |
385 | case 2: | |
386 | offset += size; | |
387 | break; | |
388 | case 1: | |
389 | offset += file->f_pos; | |
390 | } | |
391 | retval = -EINVAL; | |
392 | if (offset >= 0 && offset <= size) { | |
393 | if (offset != file->f_pos) { | |
394 | file->f_pos = offset; | |
395 | } | |
396 | retval = offset; | |
397 | } | |
1b1dcc1b | 398 | mutex_unlock(&bd_inode->i_mutex); |
1da177e4 LT |
399 | return retval; |
400 | } | |
401 | ||
402 | /* | |
403 | * Filp is never NULL; the only case when ->fsync() is called with | |
404 | * NULL first argument is nfsd_sync_dir() and that's not a directory. | |
405 | */ | |
406 | ||
b1dd3b28 | 407 | int blkdev_fsync(struct file *filp, struct dentry *dentry, int datasync) |
1da177e4 | 408 | { |
ab0a9735 CH |
409 | struct block_device *bdev = I_BDEV(filp->f_mapping->host); |
410 | int error; | |
411 | ||
412 | error = sync_blockdev(bdev); | |
413 | if (error) | |
414 | return error; | |
415 | ||
416 | error = blkdev_issue_flush(bdev, NULL); | |
417 | if (error == -EOPNOTSUPP) | |
418 | error = 0; | |
419 | return error; | |
1da177e4 | 420 | } |
b1dd3b28 | 421 | EXPORT_SYMBOL(blkdev_fsync); |
1da177e4 LT |
422 | |
423 | /* | |
424 | * pseudo-fs | |
425 | */ | |
426 | ||
427 | static __cacheline_aligned_in_smp DEFINE_SPINLOCK(bdev_lock); | |
e18b890b | 428 | static struct kmem_cache * bdev_cachep __read_mostly; |
1da177e4 LT |
429 | |
430 | static struct inode *bdev_alloc_inode(struct super_block *sb) | |
431 | { | |
e94b1766 | 432 | struct bdev_inode *ei = kmem_cache_alloc(bdev_cachep, GFP_KERNEL); |
1da177e4 LT |
433 | if (!ei) |
434 | return NULL; | |
435 | return &ei->vfs_inode; | |
436 | } | |
437 | ||
438 | static void bdev_destroy_inode(struct inode *inode) | |
439 | { | |
440 | struct bdev_inode *bdi = BDEV_I(inode); | |
441 | ||
1da177e4 LT |
442 | kmem_cache_free(bdev_cachep, bdi); |
443 | } | |
444 | ||
51cc5068 | 445 | static void init_once(void *foo) |
1da177e4 LT |
446 | { |
447 | struct bdev_inode *ei = (struct bdev_inode *) foo; | |
448 | struct block_device *bdev = &ei->bdev; | |
449 | ||
a35afb83 CL |
450 | memset(bdev, 0, sizeof(*bdev)); |
451 | mutex_init(&bdev->bd_mutex); | |
a35afb83 CL |
452 | INIT_LIST_HEAD(&bdev->bd_inodes); |
453 | INIT_LIST_HEAD(&bdev->bd_list); | |
641dc636 | 454 | #ifdef CONFIG_SYSFS |
a35afb83 | 455 | INIT_LIST_HEAD(&bdev->bd_holder_list); |
641dc636 | 456 | #endif |
a35afb83 | 457 | inode_init_once(&ei->vfs_inode); |
fcccf502 TS |
458 | /* Initialize mutex for freeze. */ |
459 | mutex_init(&bdev->bd_fsfreeze_mutex); | |
1da177e4 LT |
460 | } |
461 | ||
462 | static inline void __bd_forget(struct inode *inode) | |
463 | { | |
464 | list_del_init(&inode->i_devices); | |
465 | inode->i_bdev = NULL; | |
466 | inode->i_mapping = &inode->i_data; | |
467 | } | |
468 | ||
469 | static void bdev_clear_inode(struct inode *inode) | |
470 | { | |
471 | struct block_device *bdev = &BDEV_I(inode)->bdev; | |
472 | struct list_head *p; | |
473 | spin_lock(&bdev_lock); | |
474 | while ( (p = bdev->bd_inodes.next) != &bdev->bd_inodes ) { | |
475 | __bd_forget(list_entry(p, struct inode, i_devices)); | |
476 | } | |
477 | list_del_init(&bdev->bd_list); | |
478 | spin_unlock(&bdev_lock); | |
479 | } | |
480 | ||
ee9b6d61 | 481 | static const struct super_operations bdev_sops = { |
1da177e4 LT |
482 | .statfs = simple_statfs, |
483 | .alloc_inode = bdev_alloc_inode, | |
484 | .destroy_inode = bdev_destroy_inode, | |
485 | .drop_inode = generic_delete_inode, | |
486 | .clear_inode = bdev_clear_inode, | |
487 | }; | |
488 | ||
454e2398 DH |
489 | static int bd_get_sb(struct file_system_type *fs_type, |
490 | int flags, const char *dev_name, void *data, struct vfsmount *mnt) | |
1da177e4 | 491 | { |
454e2398 | 492 | return get_sb_pseudo(fs_type, "bdev:", &bdev_sops, 0x62646576, mnt); |
1da177e4 LT |
493 | } |
494 | ||
495 | static struct file_system_type bd_type = { | |
496 | .name = "bdev", | |
497 | .get_sb = bd_get_sb, | |
498 | .kill_sb = kill_anon_super, | |
499 | }; | |
500 | ||
c2acf7b9 | 501 | struct super_block *blockdev_superblock __read_mostly; |
1da177e4 LT |
502 | |
503 | void __init bdev_cache_init(void) | |
504 | { | |
505 | int err; | |
c2acf7b9 DC |
506 | struct vfsmount *bd_mnt; |
507 | ||
1da177e4 | 508 | bdev_cachep = kmem_cache_create("bdev_cache", sizeof(struct bdev_inode), |
fffb60f9 PJ |
509 | 0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| |
510 | SLAB_MEM_SPREAD|SLAB_PANIC), | |
20c2df83 | 511 | init_once); |
1da177e4 LT |
512 | err = register_filesystem(&bd_type); |
513 | if (err) | |
514 | panic("Cannot register bdev pseudo-fs"); | |
515 | bd_mnt = kern_mount(&bd_type); | |
1da177e4 LT |
516 | if (IS_ERR(bd_mnt)) |
517 | panic("Cannot create bdev pseudo-fs"); | |
2e1483c9 CM |
518 | /* |
519 | * This vfsmount structure is only used to obtain the | |
520 | * blockdev_superblock, so tell kmemleak not to report it. | |
521 | */ | |
522 | kmemleak_not_leak(bd_mnt); | |
1da177e4 LT |
523 | blockdev_superblock = bd_mnt->mnt_sb; /* For writeback */ |
524 | } | |
525 | ||
526 | /* | |
527 | * Most likely _very_ bad one - but then it's hardly critical for small | |
528 | * /dev and can be fixed when somebody will need really large one. | |
529 | * Keep in mind that it will be fed through icache hash function too. | |
530 | */ | |
531 | static inline unsigned long hash(dev_t dev) | |
532 | { | |
533 | return MAJOR(dev)+MINOR(dev); | |
534 | } | |
535 | ||
536 | static int bdev_test(struct inode *inode, void *data) | |
537 | { | |
538 | return BDEV_I(inode)->bdev.bd_dev == *(dev_t *)data; | |
539 | } | |
540 | ||
541 | static int bdev_set(struct inode *inode, void *data) | |
542 | { | |
543 | BDEV_I(inode)->bdev.bd_dev = *(dev_t *)data; | |
544 | return 0; | |
545 | } | |
546 | ||
547 | static LIST_HEAD(all_bdevs); | |
548 | ||
549 | struct block_device *bdget(dev_t dev) | |
550 | { | |
551 | struct block_device *bdev; | |
552 | struct inode *inode; | |
553 | ||
c2acf7b9 | 554 | inode = iget5_locked(blockdev_superblock, hash(dev), |
1da177e4 LT |
555 | bdev_test, bdev_set, &dev); |
556 | ||
557 | if (!inode) | |
558 | return NULL; | |
559 | ||
560 | bdev = &BDEV_I(inode)->bdev; | |
561 | ||
562 | if (inode->i_state & I_NEW) { | |
563 | bdev->bd_contains = NULL; | |
564 | bdev->bd_inode = inode; | |
565 | bdev->bd_block_size = (1 << inode->i_blkbits); | |
566 | bdev->bd_part_count = 0; | |
567 | bdev->bd_invalidated = 0; | |
568 | inode->i_mode = S_IFBLK; | |
569 | inode->i_rdev = dev; | |
570 | inode->i_bdev = bdev; | |
571 | inode->i_data.a_ops = &def_blk_aops; | |
572 | mapping_set_gfp_mask(&inode->i_data, GFP_USER); | |
573 | inode->i_data.backing_dev_info = &default_backing_dev_info; | |
574 | spin_lock(&bdev_lock); | |
575 | list_add(&bdev->bd_list, &all_bdevs); | |
576 | spin_unlock(&bdev_lock); | |
577 | unlock_new_inode(inode); | |
578 | } | |
579 | return bdev; | |
580 | } | |
581 | ||
582 | EXPORT_SYMBOL(bdget); | |
583 | ||
dddac6a7 AJ |
584 | /** |
585 | * bdgrab -- Grab a reference to an already referenced block device | |
586 | * @bdev: Block device to grab a reference to. | |
587 | */ | |
588 | struct block_device *bdgrab(struct block_device *bdev) | |
589 | { | |
590 | atomic_inc(&bdev->bd_inode->i_count); | |
591 | return bdev; | |
592 | } | |
593 | ||
1da177e4 LT |
594 | long nr_blockdev_pages(void) |
595 | { | |
203a2935 | 596 | struct block_device *bdev; |
1da177e4 LT |
597 | long ret = 0; |
598 | spin_lock(&bdev_lock); | |
203a2935 | 599 | list_for_each_entry(bdev, &all_bdevs, bd_list) { |
1da177e4 LT |
600 | ret += bdev->bd_inode->i_mapping->nrpages; |
601 | } | |
602 | spin_unlock(&bdev_lock); | |
603 | return ret; | |
604 | } | |
605 | ||
606 | void bdput(struct block_device *bdev) | |
607 | { | |
608 | iput(bdev->bd_inode); | |
609 | } | |
610 | ||
611 | EXPORT_SYMBOL(bdput); | |
612 | ||
613 | static struct block_device *bd_acquire(struct inode *inode) | |
614 | { | |
615 | struct block_device *bdev; | |
09d967c6 | 616 | |
1da177e4 LT |
617 | spin_lock(&bdev_lock); |
618 | bdev = inode->i_bdev; | |
09d967c6 OH |
619 | if (bdev) { |
620 | atomic_inc(&bdev->bd_inode->i_count); | |
1da177e4 LT |
621 | spin_unlock(&bdev_lock); |
622 | return bdev; | |
623 | } | |
624 | spin_unlock(&bdev_lock); | |
09d967c6 | 625 | |
1da177e4 LT |
626 | bdev = bdget(inode->i_rdev); |
627 | if (bdev) { | |
628 | spin_lock(&bdev_lock); | |
09d967c6 OH |
629 | if (!inode->i_bdev) { |
630 | /* | |
631 | * We take an additional bd_inode->i_count for inode, | |
632 | * and it's released in clear_inode() of inode. | |
633 | * So, we can access it via ->i_mapping always | |
634 | * without igrab(). | |
635 | */ | |
636 | atomic_inc(&bdev->bd_inode->i_count); | |
637 | inode->i_bdev = bdev; | |
638 | inode->i_mapping = bdev->bd_inode->i_mapping; | |
639 | list_add(&inode->i_devices, &bdev->bd_inodes); | |
640 | } | |
1da177e4 LT |
641 | spin_unlock(&bdev_lock); |
642 | } | |
643 | return bdev; | |
644 | } | |
645 | ||
646 | /* Call when you free inode */ | |
647 | ||
648 | void bd_forget(struct inode *inode) | |
649 | { | |
09d967c6 OH |
650 | struct block_device *bdev = NULL; |
651 | ||
1da177e4 | 652 | spin_lock(&bdev_lock); |
09d967c6 | 653 | if (inode->i_bdev) { |
c2acf7b9 | 654 | if (!sb_is_blkdev_sb(inode->i_sb)) |
09d967c6 | 655 | bdev = inode->i_bdev; |
1da177e4 | 656 | __bd_forget(inode); |
09d967c6 | 657 | } |
1da177e4 | 658 | spin_unlock(&bdev_lock); |
09d967c6 OH |
659 | |
660 | if (bdev) | |
661 | iput(bdev->bd_inode); | |
1da177e4 LT |
662 | } |
663 | ||
1a3cbbc5 TH |
664 | /** |
665 | * bd_may_claim - test whether a block device can be claimed | |
666 | * @bdev: block device of interest | |
667 | * @whole: whole block device containing @bdev, may equal @bdev | |
668 | * @holder: holder trying to claim @bdev | |
669 | * | |
670 | * Test whther @bdev can be claimed by @holder. | |
671 | * | |
672 | * CONTEXT: | |
673 | * spin_lock(&bdev_lock). | |
674 | * | |
675 | * RETURNS: | |
676 | * %true if @bdev can be claimed, %false otherwise. | |
677 | */ | |
678 | static bool bd_may_claim(struct block_device *bdev, struct block_device *whole, | |
679 | void *holder) | |
1da177e4 | 680 | { |
1da177e4 | 681 | if (bdev->bd_holder == holder) |
1a3cbbc5 | 682 | return true; /* already a holder */ |
1da177e4 | 683 | else if (bdev->bd_holder != NULL) |
1a3cbbc5 | 684 | return false; /* held by someone else */ |
1da177e4 | 685 | else if (bdev->bd_contains == bdev) |
1a3cbbc5 | 686 | return true; /* is a whole device which isn't held */ |
1da177e4 | 687 | |
1a3cbbc5 TH |
688 | else if (whole->bd_holder == bd_claim) |
689 | return true; /* is a partition of a device that is being partitioned */ | |
690 | else if (whole->bd_holder != NULL) | |
691 | return false; /* is a partition of a held device */ | |
1da177e4 | 692 | else |
1a3cbbc5 TH |
693 | return true; /* is a partition of an un-held device */ |
694 | } | |
695 | ||
696 | /** | |
697 | * bd_claim - claim a block device | |
698 | * @bdev: block device to claim | |
699 | * @holder: holder trying to claim @bdev | |
700 | * | |
701 | * Try to claim @bdev. | |
702 | * | |
703 | * RETURNS: | |
704 | * 0 if successful, -EBUSY if @bdev is already claimed. | |
705 | */ | |
706 | int bd_claim(struct block_device *bdev, void *holder) | |
707 | { | |
708 | struct block_device *whole = bdev->bd_contains; | |
709 | int res = -EBUSY; | |
710 | ||
711 | spin_lock(&bdev_lock); | |
1da177e4 | 712 | |
1a3cbbc5 | 713 | if (bd_may_claim(bdev, whole, holder)) { |
1da177e4 LT |
714 | /* note that for a whole device bd_holders |
715 | * will be incremented twice, and bd_holder will | |
716 | * be set to bd_claim before being set to holder | |
717 | */ | |
1a3cbbc5 TH |
718 | whole->bd_holders++; |
719 | whole->bd_holder = bd_claim; | |
1da177e4 LT |
720 | bdev->bd_holders++; |
721 | bdev->bd_holder = holder; | |
1a3cbbc5 | 722 | res = 0; |
1da177e4 | 723 | } |
1a3cbbc5 | 724 | |
1da177e4 LT |
725 | spin_unlock(&bdev_lock); |
726 | return res; | |
727 | } | |
1da177e4 LT |
728 | EXPORT_SYMBOL(bd_claim); |
729 | ||
730 | void bd_release(struct block_device *bdev) | |
731 | { | |
732 | spin_lock(&bdev_lock); | |
733 | if (!--bdev->bd_contains->bd_holders) | |
734 | bdev->bd_contains->bd_holder = NULL; | |
735 | if (!--bdev->bd_holders) | |
736 | bdev->bd_holder = NULL; | |
737 | spin_unlock(&bdev_lock); | |
738 | } | |
739 | ||
740 | EXPORT_SYMBOL(bd_release); | |
741 | ||
641dc636 JN |
742 | #ifdef CONFIG_SYSFS |
743 | /* | |
744 | * Functions for bd_claim_by_kobject / bd_release_from_kobject | |
745 | * | |
746 | * If a kobject is passed to bd_claim_by_kobject() | |
747 | * and the kobject has a parent directory, | |
748 | * following symlinks are created: | |
749 | * o from the kobject to the claimed bdev | |
750 | * o from "holders" directory of the bdev to the parent of the kobject | |
751 | * bd_release_from_kobject() removes these symlinks. | |
752 | * | |
753 | * Example: | |
754 | * If /dev/dm-0 maps to /dev/sda, kobject corresponding to | |
755 | * /sys/block/dm-0/slaves is passed to bd_claim_by_kobject(), then: | |
756 | * /sys/block/dm-0/slaves/sda --> /sys/block/sda | |
757 | * /sys/block/sda/holders/dm-0 --> /sys/block/dm-0 | |
758 | */ | |
759 | ||
4d7dd8fd | 760 | static int add_symlink(struct kobject *from, struct kobject *to) |
641dc636 JN |
761 | { |
762 | if (!from || !to) | |
4d7dd8fd AM |
763 | return 0; |
764 | return sysfs_create_link(from, to, kobject_name(to)); | |
641dc636 JN |
765 | } |
766 | ||
767 | static void del_symlink(struct kobject *from, struct kobject *to) | |
768 | { | |
769 | if (!from || !to) | |
770 | return; | |
771 | sysfs_remove_link(from, kobject_name(to)); | |
772 | } | |
773 | ||
774 | /* | |
775 | * 'struct bd_holder' contains pointers to kobjects symlinked by | |
776 | * bd_claim_by_kobject. | |
777 | * It's connected to bd_holder_list which is protected by bdev->bd_sem. | |
778 | */ | |
779 | struct bd_holder { | |
780 | struct list_head list; /* chain of holders of the bdev */ | |
781 | int count; /* references from the holder */ | |
782 | struct kobject *sdir; /* holder object, e.g. "/block/dm-0/slaves" */ | |
783 | struct kobject *hdev; /* e.g. "/block/dm-0" */ | |
784 | struct kobject *hdir; /* e.g. "/block/sda/holders" */ | |
785 | struct kobject *sdev; /* e.g. "/block/sda" */ | |
786 | }; | |
787 | ||
788 | /* | |
789 | * Get references of related kobjects at once. | |
790 | * Returns 1 on success. 0 on failure. | |
791 | * | |
792 | * Should call bd_holder_release_dirs() after successful use. | |
793 | */ | |
794 | static int bd_holder_grab_dirs(struct block_device *bdev, | |
795 | struct bd_holder *bo) | |
796 | { | |
797 | if (!bdev || !bo) | |
798 | return 0; | |
799 | ||
800 | bo->sdir = kobject_get(bo->sdir); | |
801 | if (!bo->sdir) | |
802 | return 0; | |
803 | ||
804 | bo->hdev = kobject_get(bo->sdir->parent); | |
805 | if (!bo->hdev) | |
806 | goto fail_put_sdir; | |
807 | ||
0762b8bd | 808 | bo->sdev = kobject_get(&part_to_dev(bdev->bd_part)->kobj); |
641dc636 JN |
809 | if (!bo->sdev) |
810 | goto fail_put_hdev; | |
811 | ||
4c46501d | 812 | bo->hdir = kobject_get(bdev->bd_part->holder_dir); |
641dc636 JN |
813 | if (!bo->hdir) |
814 | goto fail_put_sdev; | |
815 | ||
816 | return 1; | |
817 | ||
818 | fail_put_sdev: | |
819 | kobject_put(bo->sdev); | |
820 | fail_put_hdev: | |
821 | kobject_put(bo->hdev); | |
822 | fail_put_sdir: | |
823 | kobject_put(bo->sdir); | |
824 | ||
825 | return 0; | |
826 | } | |
827 | ||
828 | /* Put references of related kobjects at once. */ | |
829 | static void bd_holder_release_dirs(struct bd_holder *bo) | |
830 | { | |
831 | kobject_put(bo->hdir); | |
832 | kobject_put(bo->sdev); | |
833 | kobject_put(bo->hdev); | |
834 | kobject_put(bo->sdir); | |
835 | } | |
836 | ||
837 | static struct bd_holder *alloc_bd_holder(struct kobject *kobj) | |
838 | { | |
839 | struct bd_holder *bo; | |
840 | ||
841 | bo = kzalloc(sizeof(*bo), GFP_KERNEL); | |
842 | if (!bo) | |
843 | return NULL; | |
844 | ||
845 | bo->count = 1; | |
846 | bo->sdir = kobj; | |
847 | ||
848 | return bo; | |
849 | } | |
850 | ||
851 | static void free_bd_holder(struct bd_holder *bo) | |
852 | { | |
853 | kfree(bo); | |
854 | } | |
855 | ||
df6c0cd9 JN |
856 | /** |
857 | * find_bd_holder - find matching struct bd_holder from the block device | |
858 | * | |
859 | * @bdev: struct block device to be searched | |
860 | * @bo: target struct bd_holder | |
861 | * | |
862 | * Returns matching entry with @bo in @bdev->bd_holder_list. | |
863 | * If found, increment the reference count and return the pointer. | |
864 | * If not found, returns NULL. | |
865 | */ | |
36a561d6 AM |
866 | static struct bd_holder *find_bd_holder(struct block_device *bdev, |
867 | struct bd_holder *bo) | |
df6c0cd9 JN |
868 | { |
869 | struct bd_holder *tmp; | |
870 | ||
871 | list_for_each_entry(tmp, &bdev->bd_holder_list, list) | |
872 | if (tmp->sdir == bo->sdir) { | |
873 | tmp->count++; | |
874 | return tmp; | |
875 | } | |
876 | ||
877 | return NULL; | |
878 | } | |
879 | ||
641dc636 JN |
880 | /** |
881 | * add_bd_holder - create sysfs symlinks for bd_claim() relationship | |
882 | * | |
883 | * @bdev: block device to be bd_claimed | |
884 | * @bo: preallocated and initialized by alloc_bd_holder() | |
885 | * | |
df6c0cd9 | 886 | * Add @bo to @bdev->bd_holder_list, create symlinks. |
641dc636 | 887 | * |
df6c0cd9 JN |
888 | * Returns 0 if symlinks are created. |
889 | * Returns -ve if something fails. | |
641dc636 JN |
890 | */ |
891 | static int add_bd_holder(struct block_device *bdev, struct bd_holder *bo) | |
892 | { | |
4e91672c | 893 | int err; |
641dc636 JN |
894 | |
895 | if (!bo) | |
4d7dd8fd | 896 | return -EINVAL; |
641dc636 | 897 | |
641dc636 | 898 | if (!bd_holder_grab_dirs(bdev, bo)) |
4d7dd8fd | 899 | return -EBUSY; |
641dc636 | 900 | |
4e91672c JW |
901 | err = add_symlink(bo->sdir, bo->sdev); |
902 | if (err) | |
903 | return err; | |
904 | ||
905 | err = add_symlink(bo->hdir, bo->hdev); | |
906 | if (err) { | |
907 | del_symlink(bo->sdir, bo->sdev); | |
908 | return err; | |
4d7dd8fd | 909 | } |
4e91672c JW |
910 | |
911 | list_add_tail(&bo->list, &bdev->bd_holder_list); | |
912 | return 0; | |
641dc636 JN |
913 | } |
914 | ||
915 | /** | |
916 | * del_bd_holder - delete sysfs symlinks for bd_claim() relationship | |
917 | * | |
918 | * @bdev: block device to be bd_claimed | |
919 | * @kobj: holder's kobject | |
920 | * | |
921 | * If there is matching entry with @kobj in @bdev->bd_holder_list | |
922 | * and no other bd_claim() from the same kobject, | |
923 | * remove the struct bd_holder from the list, delete symlinks for it. | |
924 | * | |
925 | * Returns a pointer to the struct bd_holder when it's removed from the list | |
926 | * and ready to be freed. | |
927 | * Returns NULL if matching claim isn't found or there is other bd_claim() | |
928 | * by the same kobject. | |
929 | */ | |
930 | static struct bd_holder *del_bd_holder(struct block_device *bdev, | |
931 | struct kobject *kobj) | |
932 | { | |
933 | struct bd_holder *bo; | |
934 | ||
935 | list_for_each_entry(bo, &bdev->bd_holder_list, list) { | |
936 | if (bo->sdir == kobj) { | |
937 | bo->count--; | |
938 | BUG_ON(bo->count < 0); | |
939 | if (!bo->count) { | |
940 | list_del(&bo->list); | |
941 | del_symlink(bo->sdir, bo->sdev); | |
942 | del_symlink(bo->hdir, bo->hdev); | |
943 | bd_holder_release_dirs(bo); | |
944 | return bo; | |
945 | } | |
946 | break; | |
947 | } | |
948 | } | |
949 | ||
950 | return NULL; | |
951 | } | |
952 | ||
953 | /** | |
954 | * bd_claim_by_kobject - bd_claim() with additional kobject signature | |
955 | * | |
956 | * @bdev: block device to be claimed | |
957 | * @holder: holder's signature | |
958 | * @kobj: holder's kobject | |
959 | * | |
960 | * Do bd_claim() and if it succeeds, create sysfs symlinks between | |
961 | * the bdev and the holder's kobject. | |
962 | * Use bd_release_from_kobject() when relesing the claimed bdev. | |
963 | * | |
964 | * Returns 0 on success. (same as bd_claim()) | |
965 | * Returns errno on failure. | |
966 | */ | |
967 | static int bd_claim_by_kobject(struct block_device *bdev, void *holder, | |
968 | struct kobject *kobj) | |
969 | { | |
4e91672c | 970 | int err; |
df6c0cd9 | 971 | struct bd_holder *bo, *found; |
641dc636 JN |
972 | |
973 | if (!kobj) | |
974 | return -EINVAL; | |
975 | ||
976 | bo = alloc_bd_holder(kobj); | |
977 | if (!bo) | |
978 | return -ENOMEM; | |
979 | ||
2e7b651d | 980 | mutex_lock(&bdev->bd_mutex); |
df6c0cd9 | 981 | |
4e91672c JW |
982 | err = bd_claim(bdev, holder); |
983 | if (err) | |
4210df28 | 984 | goto fail; |
4e91672c JW |
985 | |
986 | found = find_bd_holder(bdev, bo); | |
987 | if (found) | |
4210df28 | 988 | goto fail; |
4e91672c JW |
989 | |
990 | err = add_bd_holder(bdev, bo); | |
991 | if (err) | |
992 | bd_release(bdev); | |
4210df28 AM |
993 | else |
994 | bo = NULL; | |
995 | fail: | |
b4cf1b72 | 996 | mutex_unlock(&bdev->bd_mutex); |
4210df28 | 997 | free_bd_holder(bo); |
4e91672c | 998 | return err; |
641dc636 JN |
999 | } |
1000 | ||
1001 | /** | |
1002 | * bd_release_from_kobject - bd_release() with additional kobject signature | |
1003 | * | |
1004 | * @bdev: block device to be released | |
1005 | * @kobj: holder's kobject | |
1006 | * | |
1007 | * Do bd_release() and remove sysfs symlinks created by bd_claim_by_kobject(). | |
1008 | */ | |
1009 | static void bd_release_from_kobject(struct block_device *bdev, | |
1010 | struct kobject *kobj) | |
1011 | { | |
641dc636 JN |
1012 | if (!kobj) |
1013 | return; | |
1014 | ||
2e7b651d | 1015 | mutex_lock(&bdev->bd_mutex); |
641dc636 | 1016 | bd_release(bdev); |
4210df28 | 1017 | free_bd_holder(del_bd_holder(bdev, kobj)); |
b4cf1b72 | 1018 | mutex_unlock(&bdev->bd_mutex); |
641dc636 JN |
1019 | } |
1020 | ||
1021 | /** | |
1022 | * bd_claim_by_disk - wrapper function for bd_claim_by_kobject() | |
1023 | * | |
1024 | * @bdev: block device to be claimed | |
1025 | * @holder: holder's signature | |
1026 | * @disk: holder's gendisk | |
1027 | * | |
1028 | * Call bd_claim_by_kobject() with getting @disk->slave_dir. | |
1029 | */ | |
1030 | int bd_claim_by_disk(struct block_device *bdev, void *holder, | |
1031 | struct gendisk *disk) | |
1032 | { | |
1033 | return bd_claim_by_kobject(bdev, holder, kobject_get(disk->slave_dir)); | |
1034 | } | |
1035 | EXPORT_SYMBOL_GPL(bd_claim_by_disk); | |
1036 | ||
1037 | /** | |
1038 | * bd_release_from_disk - wrapper function for bd_release_from_kobject() | |
1039 | * | |
1040 | * @bdev: block device to be claimed | |
1041 | * @disk: holder's gendisk | |
1042 | * | |
1043 | * Call bd_release_from_kobject() and put @disk->slave_dir. | |
1044 | */ | |
1045 | void bd_release_from_disk(struct block_device *bdev, struct gendisk *disk) | |
1046 | { | |
1047 | bd_release_from_kobject(bdev, disk->slave_dir); | |
1048 | kobject_put(disk->slave_dir); | |
1049 | } | |
1050 | EXPORT_SYMBOL_GPL(bd_release_from_disk); | |
1051 | #endif | |
1052 | ||
1da177e4 LT |
1053 | /* |
1054 | * Tries to open block device by device number. Use it ONLY if you | |
1055 | * really do not have anything better - i.e. when you are behind a | |
1056 | * truly sucky interface and all you are given is a device number. _Never_ | |
1057 | * to be used for internal purposes. If you ever need it - reconsider | |
1058 | * your API. | |
1059 | */ | |
aeb5d727 | 1060 | struct block_device *open_by_devnum(dev_t dev, fmode_t mode) |
1da177e4 LT |
1061 | { |
1062 | struct block_device *bdev = bdget(dev); | |
1063 | int err = -ENOMEM; | |
1da177e4 | 1064 | if (bdev) |
572c4892 | 1065 | err = blkdev_get(bdev, mode); |
1da177e4 LT |
1066 | return err ? ERR_PTR(err) : bdev; |
1067 | } | |
1068 | ||
1069 | EXPORT_SYMBOL(open_by_devnum); | |
1070 | ||
56ade44b AP |
1071 | /** |
1072 | * flush_disk - invalidates all buffer-cache entries on a disk | |
1073 | * | |
1074 | * @bdev: struct block device to be flushed | |
1075 | * | |
1076 | * Invalidates all buffer-cache entries on a disk. It should be called | |
1077 | * when a disk has been changed -- either by a media change or online | |
1078 | * resize. | |
1079 | */ | |
1080 | static void flush_disk(struct block_device *bdev) | |
1081 | { | |
1082 | if (__invalidate_device(bdev)) { | |
1083 | char name[BDEVNAME_SIZE] = ""; | |
1084 | ||
1085 | if (bdev->bd_disk) | |
1086 | disk_name(bdev->bd_disk, 0, name); | |
1087 | printk(KERN_WARNING "VFS: busy inodes on changed media or " | |
1088 | "resized disk %s\n", name); | |
1089 | } | |
1090 | ||
1091 | if (!bdev->bd_disk) | |
1092 | return; | |
1093 | if (disk_partitionable(bdev->bd_disk)) | |
1094 | bdev->bd_invalidated = 1; | |
1095 | } | |
1096 | ||
c3279d14 | 1097 | /** |
57d1b536 | 1098 | * check_disk_size_change - checks for disk size change and adjusts bdev size. |
c3279d14 AP |
1099 | * @disk: struct gendisk to check |
1100 | * @bdev: struct bdev to adjust. | |
1101 | * | |
1102 | * This routine checks to see if the bdev size does not match the disk size | |
1103 | * and adjusts it if it differs. | |
1104 | */ | |
1105 | void check_disk_size_change(struct gendisk *disk, struct block_device *bdev) | |
1106 | { | |
1107 | loff_t disk_size, bdev_size; | |
1108 | ||
1109 | disk_size = (loff_t)get_capacity(disk) << 9; | |
1110 | bdev_size = i_size_read(bdev->bd_inode); | |
1111 | if (disk_size != bdev_size) { | |
1112 | char name[BDEVNAME_SIZE]; | |
1113 | ||
1114 | disk_name(disk, 0, name); | |
1115 | printk(KERN_INFO | |
1116 | "%s: detected capacity change from %lld to %lld\n", | |
1117 | name, bdev_size, disk_size); | |
1118 | i_size_write(bdev->bd_inode, disk_size); | |
608aeef1 | 1119 | flush_disk(bdev); |
c3279d14 AP |
1120 | } |
1121 | } | |
1122 | EXPORT_SYMBOL(check_disk_size_change); | |
1123 | ||
0c002c2f | 1124 | /** |
57d1b536 | 1125 | * revalidate_disk - wrapper for lower-level driver's revalidate_disk call-back |
0c002c2f AP |
1126 | * @disk: struct gendisk to be revalidated |
1127 | * | |
1128 | * This routine is a wrapper for lower-level driver's revalidate_disk | |
1129 | * call-backs. It is used to do common pre and post operations needed | |
1130 | * for all revalidate_disk operations. | |
1131 | */ | |
1132 | int revalidate_disk(struct gendisk *disk) | |
1133 | { | |
c3279d14 | 1134 | struct block_device *bdev; |
0c002c2f AP |
1135 | int ret = 0; |
1136 | ||
1137 | if (disk->fops->revalidate_disk) | |
1138 | ret = disk->fops->revalidate_disk(disk); | |
1139 | ||
c3279d14 AP |
1140 | bdev = bdget_disk(disk, 0); |
1141 | if (!bdev) | |
1142 | return ret; | |
1143 | ||
1144 | mutex_lock(&bdev->bd_mutex); | |
1145 | check_disk_size_change(disk, bdev); | |
1146 | mutex_unlock(&bdev->bd_mutex); | |
1147 | bdput(bdev); | |
0c002c2f AP |
1148 | return ret; |
1149 | } | |
1150 | EXPORT_SYMBOL(revalidate_disk); | |
1151 | ||
1da177e4 LT |
1152 | /* |
1153 | * This routine checks whether a removable media has been changed, | |
1154 | * and invalidates all buffer-cache-entries in that case. This | |
1155 | * is a relatively slow routine, so we have to try to minimize using | |
1156 | * it. Thus it is called only upon a 'mount' or 'open'. This | |
1157 | * is the best way of combining speed and utility, I think. | |
1158 | * People changing diskettes in the middle of an operation deserve | |
1159 | * to lose :-) | |
1160 | */ | |
1161 | int check_disk_change(struct block_device *bdev) | |
1162 | { | |
1163 | struct gendisk *disk = bdev->bd_disk; | |
83d5cde4 | 1164 | const struct block_device_operations *bdops = disk->fops; |
1da177e4 LT |
1165 | |
1166 | if (!bdops->media_changed) | |
1167 | return 0; | |
1168 | if (!bdops->media_changed(bdev->bd_disk)) | |
1169 | return 0; | |
1170 | ||
56ade44b | 1171 | flush_disk(bdev); |
1da177e4 LT |
1172 | if (bdops->revalidate_disk) |
1173 | bdops->revalidate_disk(bdev->bd_disk); | |
1da177e4 LT |
1174 | return 1; |
1175 | } | |
1176 | ||
1177 | EXPORT_SYMBOL(check_disk_change); | |
1178 | ||
1179 | void bd_set_size(struct block_device *bdev, loff_t size) | |
1180 | { | |
e1defc4f | 1181 | unsigned bsize = bdev_logical_block_size(bdev); |
1da177e4 LT |
1182 | |
1183 | bdev->bd_inode->i_size = size; | |
1184 | while (bsize < PAGE_CACHE_SIZE) { | |
1185 | if (size & bsize) | |
1186 | break; | |
1187 | bsize <<= 1; | |
1188 | } | |
1189 | bdev->bd_block_size = bsize; | |
1190 | bdev->bd_inode->i_blkbits = blksize_bits(bsize); | |
1191 | } | |
1192 | EXPORT_SYMBOL(bd_set_size); | |
1193 | ||
9a1c3542 | 1194 | static int __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part); |
37be4124 | 1195 | |
6d740cd5 PZ |
1196 | /* |
1197 | * bd_mutex locking: | |
1198 | * | |
1199 | * mutex_lock(part->bd_mutex) | |
1200 | * mutex_lock_nested(whole->bd_mutex, 1) | |
1201 | */ | |
1202 | ||
572c4892 | 1203 | static int __blkdev_get(struct block_device *bdev, fmode_t mode, int for_part) |
1da177e4 | 1204 | { |
1da177e4 | 1205 | struct gendisk *disk; |
7db9cfd3 | 1206 | int ret; |
cf771cb5 | 1207 | int partno; |
fe6e9c1f AV |
1208 | int perm = 0; |
1209 | ||
572c4892 | 1210 | if (mode & FMODE_READ) |
fe6e9c1f | 1211 | perm |= MAY_READ; |
572c4892 | 1212 | if (mode & FMODE_WRITE) |
fe6e9c1f AV |
1213 | perm |= MAY_WRITE; |
1214 | /* | |
1215 | * hooks: /n/, see "layering violations". | |
1216 | */ | |
1217 | ret = devcgroup_inode_permission(bdev->bd_inode, perm); | |
82666020 AV |
1218 | if (ret != 0) { |
1219 | bdput(bdev); | |
7db9cfd3 | 1220 | return ret; |
82666020 | 1221 | } |
7db9cfd3 | 1222 | |
1da177e4 | 1223 | lock_kernel(); |
d3374825 | 1224 | restart: |
0762b8bd | 1225 | |
89f97496 | 1226 | ret = -ENXIO; |
cf771cb5 | 1227 | disk = get_gendisk(bdev->bd_dev, &partno); |
0762b8bd TH |
1228 | if (!disk) |
1229 | goto out_unlock_kernel; | |
1da177e4 | 1230 | |
6796bf54 | 1231 | mutex_lock_nested(&bdev->bd_mutex, for_part); |
1da177e4 LT |
1232 | if (!bdev->bd_openers) { |
1233 | bdev->bd_disk = disk; | |
1234 | bdev->bd_contains = bdev; | |
cf771cb5 | 1235 | if (!partno) { |
1da177e4 | 1236 | struct backing_dev_info *bdi; |
89f97496 TH |
1237 | |
1238 | ret = -ENXIO; | |
1239 | bdev->bd_part = disk_get_part(disk, partno); | |
1240 | if (!bdev->bd_part) | |
1241 | goto out_clear; | |
1242 | ||
1da177e4 | 1243 | if (disk->fops->open) { |
572c4892 | 1244 | ret = disk->fops->open(bdev, mode); |
d3374825 N |
1245 | if (ret == -ERESTARTSYS) { |
1246 | /* Lost a race with 'disk' being | |
1247 | * deleted, try again. | |
1248 | * See md.c | |
1249 | */ | |
1250 | disk_put_part(bdev->bd_part); | |
1251 | bdev->bd_part = NULL; | |
1252 | module_put(disk->fops->owner); | |
1253 | put_disk(disk); | |
1254 | bdev->bd_disk = NULL; | |
1255 | mutex_unlock(&bdev->bd_mutex); | |
1256 | goto restart; | |
1257 | } | |
1da177e4 | 1258 | if (ret) |
0762b8bd | 1259 | goto out_clear; |
1da177e4 LT |
1260 | } |
1261 | if (!bdev->bd_openers) { | |
1262 | bd_set_size(bdev,(loff_t)get_capacity(disk)<<9); | |
1263 | bdi = blk_get_backing_dev_info(bdev); | |
1264 | if (bdi == NULL) | |
1265 | bdi = &default_backing_dev_info; | |
1266 | bdev->bd_inode->i_data.backing_dev_info = bdi; | |
1267 | } | |
1268 | if (bdev->bd_invalidated) | |
1269 | rescan_partitions(disk, bdev); | |
1270 | } else { | |
1da177e4 LT |
1271 | struct block_device *whole; |
1272 | whole = bdget_disk(disk, 0); | |
1273 | ret = -ENOMEM; | |
1274 | if (!whole) | |
0762b8bd | 1275 | goto out_clear; |
37be4124 | 1276 | BUG_ON(for_part); |
572c4892 | 1277 | ret = __blkdev_get(whole, mode, 1); |
1da177e4 | 1278 | if (ret) |
0762b8bd | 1279 | goto out_clear; |
1da177e4 | 1280 | bdev->bd_contains = whole; |
1da177e4 LT |
1281 | bdev->bd_inode->i_data.backing_dev_info = |
1282 | whole->bd_inode->i_data.backing_dev_info; | |
89f97496 | 1283 | bdev->bd_part = disk_get_part(disk, partno); |
e71bf0d0 | 1284 | if (!(disk->flags & GENHD_FL_UP) || |
89f97496 | 1285 | !bdev->bd_part || !bdev->bd_part->nr_sects) { |
1da177e4 | 1286 | ret = -ENXIO; |
0762b8bd | 1287 | goto out_clear; |
1da177e4 | 1288 | } |
89f97496 | 1289 | bd_set_size(bdev, (loff_t)bdev->bd_part->nr_sects << 9); |
1da177e4 LT |
1290 | } |
1291 | } else { | |
0762b8bd | 1292 | module_put(disk->fops->owner); |
960cc0f4 | 1293 | put_disk(disk); |
0762b8bd | 1294 | disk = NULL; |
1da177e4 LT |
1295 | if (bdev->bd_contains == bdev) { |
1296 | if (bdev->bd_disk->fops->open) { | |
572c4892 | 1297 | ret = bdev->bd_disk->fops->open(bdev, mode); |
1da177e4 | 1298 | if (ret) |
0762b8bd | 1299 | goto out_unlock_bdev; |
1da177e4 LT |
1300 | } |
1301 | if (bdev->bd_invalidated) | |
1302 | rescan_partitions(bdev->bd_disk, bdev); | |
1da177e4 LT |
1303 | } |
1304 | } | |
1305 | bdev->bd_openers++; | |
37be4124 N |
1306 | if (for_part) |
1307 | bdev->bd_part_count++; | |
c039e313 | 1308 | mutex_unlock(&bdev->bd_mutex); |
1da177e4 LT |
1309 | unlock_kernel(); |
1310 | return 0; | |
1311 | ||
0762b8bd | 1312 | out_clear: |
89f97496 | 1313 | disk_put_part(bdev->bd_part); |
1da177e4 | 1314 | bdev->bd_disk = NULL; |
0762b8bd | 1315 | bdev->bd_part = NULL; |
1da177e4 LT |
1316 | bdev->bd_inode->i_data.backing_dev_info = &default_backing_dev_info; |
1317 | if (bdev != bdev->bd_contains) | |
572c4892 | 1318 | __blkdev_put(bdev->bd_contains, mode, 1); |
1da177e4 | 1319 | bdev->bd_contains = NULL; |
0762b8bd | 1320 | out_unlock_bdev: |
c039e313 | 1321 | mutex_unlock(&bdev->bd_mutex); |
0762b8bd | 1322 | out_unlock_kernel: |
1da177e4 | 1323 | unlock_kernel(); |
0762b8bd | 1324 | |
0762b8bd TH |
1325 | if (disk) |
1326 | module_put(disk->fops->owner); | |
1327 | put_disk(disk); | |
1328 | bdput(bdev); | |
1329 | ||
1da177e4 LT |
1330 | return ret; |
1331 | } | |
1332 | ||
572c4892 | 1333 | int blkdev_get(struct block_device *bdev, fmode_t mode) |
1da177e4 | 1334 | { |
572c4892 | 1335 | return __blkdev_get(bdev, mode, 0); |
37be4124 | 1336 | } |
1da177e4 LT |
1337 | EXPORT_SYMBOL(blkdev_get); |
1338 | ||
1339 | static int blkdev_open(struct inode * inode, struct file * filp) | |
1340 | { | |
1341 | struct block_device *bdev; | |
1342 | int res; | |
1343 | ||
1344 | /* | |
1345 | * Preserve backwards compatibility and allow large file access | |
1346 | * even if userspace doesn't ask for it explicitly. Some mkfs | |
1347 | * binary needs it. We might want to drop this workaround | |
1348 | * during an unstable branch. | |
1349 | */ | |
1350 | filp->f_flags |= O_LARGEFILE; | |
1351 | ||
572c4892 AV |
1352 | if (filp->f_flags & O_NDELAY) |
1353 | filp->f_mode |= FMODE_NDELAY; | |
1354 | if (filp->f_flags & O_EXCL) | |
1355 | filp->f_mode |= FMODE_EXCL; | |
1356 | if ((filp->f_flags & O_ACCMODE) == 3) | |
1357 | filp->f_mode |= FMODE_WRITE_IOCTL; | |
1358 | ||
1da177e4 | 1359 | bdev = bd_acquire(inode); |
6a2aae06 PE |
1360 | if (bdev == NULL) |
1361 | return -ENOMEM; | |
1da177e4 | 1362 | |
572c4892 AV |
1363 | filp->f_mapping = bdev->bd_inode->i_mapping; |
1364 | ||
1365 | res = blkdev_get(bdev, filp->f_mode); | |
1da177e4 LT |
1366 | if (res) |
1367 | return res; | |
1368 | ||
ebbefc01 CH |
1369 | if (filp->f_mode & FMODE_EXCL) { |
1370 | res = bd_claim(bdev, filp); | |
1371 | if (res) | |
1372 | goto out_blkdev_put; | |
1373 | } | |
1da177e4 | 1374 | |
ebbefc01 | 1375 | return 0; |
1da177e4 | 1376 | |
ebbefc01 | 1377 | out_blkdev_put: |
9a1c3542 | 1378 | blkdev_put(bdev, filp->f_mode); |
1da177e4 LT |
1379 | return res; |
1380 | } | |
1381 | ||
9a1c3542 | 1382 | static int __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part) |
2e7b651d PZ |
1383 | { |
1384 | int ret = 0; | |
2e7b651d | 1385 | struct gendisk *disk = bdev->bd_disk; |
37be4124 | 1386 | struct block_device *victim = NULL; |
2e7b651d | 1387 | |
6796bf54 | 1388 | mutex_lock_nested(&bdev->bd_mutex, for_part); |
2e7b651d | 1389 | lock_kernel(); |
37be4124 N |
1390 | if (for_part) |
1391 | bdev->bd_part_count--; | |
1392 | ||
2e7b651d PZ |
1393 | if (!--bdev->bd_openers) { |
1394 | sync_blockdev(bdev); | |
1395 | kill_bdev(bdev); | |
1396 | } | |
1397 | if (bdev->bd_contains == bdev) { | |
1398 | if (disk->fops->release) | |
9a1c3542 | 1399 | ret = disk->fops->release(disk, mode); |
2e7b651d PZ |
1400 | } |
1401 | if (!bdev->bd_openers) { | |
1402 | struct module *owner = disk->fops->owner; | |
1403 | ||
1404 | put_disk(disk); | |
1405 | module_put(owner); | |
0762b8bd TH |
1406 | disk_put_part(bdev->bd_part); |
1407 | bdev->bd_part = NULL; | |
2e7b651d PZ |
1408 | bdev->bd_disk = NULL; |
1409 | bdev->bd_inode->i_data.backing_dev_info = &default_backing_dev_info; | |
37be4124 N |
1410 | if (bdev != bdev->bd_contains) |
1411 | victim = bdev->bd_contains; | |
2e7b651d PZ |
1412 | bdev->bd_contains = NULL; |
1413 | } | |
1414 | unlock_kernel(); | |
1415 | mutex_unlock(&bdev->bd_mutex); | |
1416 | bdput(bdev); | |
37be4124 | 1417 | if (victim) |
9a1c3542 | 1418 | __blkdev_put(victim, mode, 1); |
2e7b651d PZ |
1419 | return ret; |
1420 | } | |
1421 | ||
9a1c3542 | 1422 | int blkdev_put(struct block_device *bdev, fmode_t mode) |
37be4124 | 1423 | { |
9a1c3542 | 1424 | return __blkdev_put(bdev, mode, 0); |
37be4124 | 1425 | } |
2e7b651d PZ |
1426 | EXPORT_SYMBOL(blkdev_put); |
1427 | ||
1da177e4 LT |
1428 | static int blkdev_close(struct inode * inode, struct file * filp) |
1429 | { | |
1430 | struct block_device *bdev = I_BDEV(filp->f_mapping->host); | |
1431 | if (bdev->bd_holder == filp) | |
1432 | bd_release(bdev); | |
9a1c3542 | 1433 | return blkdev_put(bdev, filp->f_mode); |
1da177e4 LT |
1434 | } |
1435 | ||
bb93e3a5 | 1436 | static long block_ioctl(struct file *file, unsigned cmd, unsigned long arg) |
1da177e4 | 1437 | { |
56b26add AV |
1438 | struct block_device *bdev = I_BDEV(file->f_mapping->host); |
1439 | fmode_t mode = file->f_mode; | |
fd4ce1ac CH |
1440 | |
1441 | /* | |
1442 | * O_NDELAY can be altered using fcntl(.., F_SETFL, ..), so we have | |
1443 | * to updated it before every ioctl. | |
1444 | */ | |
56b26add | 1445 | if (file->f_flags & O_NDELAY) |
fd4ce1ac CH |
1446 | mode |= FMODE_NDELAY; |
1447 | else | |
1448 | mode &= ~FMODE_NDELAY; | |
1449 | ||
56b26add | 1450 | return blkdev_ioctl(bdev, mode, cmd, arg); |
1da177e4 LT |
1451 | } |
1452 | ||
eef99380 CH |
1453 | /* |
1454 | * Write data to the block device. Only intended for the block device itself | |
1455 | * and the raw driver which basically is a fake block device. | |
1456 | * | |
1457 | * Does not take i_mutex for the write and thus is not for general purpose | |
1458 | * use. | |
1459 | */ | |
1460 | ssize_t blkdev_aio_write(struct kiocb *iocb, const struct iovec *iov, | |
1461 | unsigned long nr_segs, loff_t pos) | |
1462 | { | |
1463 | struct file *file = iocb->ki_filp; | |
1464 | ssize_t ret; | |
1465 | ||
1466 | BUG_ON(iocb->ki_pos != pos); | |
1467 | ||
1468 | ret = __generic_file_aio_write(iocb, iov, nr_segs, &iocb->ki_pos); | |
1469 | if (ret > 0 || ret == -EIOCBQUEUED) { | |
1470 | ssize_t err; | |
1471 | ||
1472 | err = generic_write_sync(file, pos, ret); | |
1473 | if (err < 0 && ret > 0) | |
1474 | ret = err; | |
1475 | } | |
1476 | return ret; | |
1477 | } | |
1478 | EXPORT_SYMBOL_GPL(blkdev_aio_write); | |
1479 | ||
87d8fe1e TT |
1480 | /* |
1481 | * Try to release a page associated with block device when the system | |
1482 | * is under memory pressure. | |
1483 | */ | |
1484 | static int blkdev_releasepage(struct page *page, gfp_t wait) | |
1485 | { | |
1486 | struct super_block *super = BDEV_I(page->mapping->host)->bdev.bd_super; | |
1487 | ||
1488 | if (super && super->s_op->bdev_try_to_free_page) | |
1489 | return super->s_op->bdev_try_to_free_page(super, page, wait); | |
1490 | ||
1491 | return try_to_free_buffers(page); | |
1492 | } | |
1493 | ||
4c54ac62 | 1494 | static const struct address_space_operations def_blk_aops = { |
1da177e4 LT |
1495 | .readpage = blkdev_readpage, |
1496 | .writepage = blkdev_writepage, | |
1497 | .sync_page = block_sync_page, | |
6272b5a5 NP |
1498 | .write_begin = blkdev_write_begin, |
1499 | .write_end = blkdev_write_end, | |
1da177e4 | 1500 | .writepages = generic_writepages, |
87d8fe1e | 1501 | .releasepage = blkdev_releasepage, |
1da177e4 LT |
1502 | .direct_IO = blkdev_direct_IO, |
1503 | }; | |
1504 | ||
4b6f5d20 | 1505 | const struct file_operations def_blk_fops = { |
1da177e4 LT |
1506 | .open = blkdev_open, |
1507 | .release = blkdev_close, | |
1508 | .llseek = block_llseek, | |
543ade1f BP |
1509 | .read = do_sync_read, |
1510 | .write = do_sync_write, | |
1da177e4 | 1511 | .aio_read = generic_file_aio_read, |
eef99380 | 1512 | .aio_write = blkdev_aio_write, |
1da177e4 | 1513 | .mmap = generic_file_mmap, |
b1dd3b28 | 1514 | .fsync = blkdev_fsync, |
bb93e3a5 | 1515 | .unlocked_ioctl = block_ioctl, |
1da177e4 LT |
1516 | #ifdef CONFIG_COMPAT |
1517 | .compat_ioctl = compat_blkdev_ioctl, | |
1518 | #endif | |
7f9c51f0 JA |
1519 | .splice_read = generic_file_splice_read, |
1520 | .splice_write = generic_file_splice_write, | |
1da177e4 LT |
1521 | }; |
1522 | ||
1523 | int ioctl_by_bdev(struct block_device *bdev, unsigned cmd, unsigned long arg) | |
1524 | { | |
1525 | int res; | |
1526 | mm_segment_t old_fs = get_fs(); | |
1527 | set_fs(KERNEL_DS); | |
56b26add | 1528 | res = blkdev_ioctl(bdev, 0, cmd, arg); |
1da177e4 LT |
1529 | set_fs(old_fs); |
1530 | return res; | |
1531 | } | |
1532 | ||
1533 | EXPORT_SYMBOL(ioctl_by_bdev); | |
1534 | ||
1535 | /** | |
1536 | * lookup_bdev - lookup a struct block_device by name | |
94e2959e | 1537 | * @pathname: special file representing the block device |
1da177e4 | 1538 | * |
57d1b536 | 1539 | * Get a reference to the blockdevice at @pathname in the current |
1da177e4 LT |
1540 | * namespace if possible and return it. Return ERR_PTR(error) |
1541 | * otherwise. | |
1542 | */ | |
421748ec | 1543 | struct block_device *lookup_bdev(const char *pathname) |
1da177e4 LT |
1544 | { |
1545 | struct block_device *bdev; | |
1546 | struct inode *inode; | |
421748ec | 1547 | struct path path; |
1da177e4 LT |
1548 | int error; |
1549 | ||
421748ec | 1550 | if (!pathname || !*pathname) |
1da177e4 LT |
1551 | return ERR_PTR(-EINVAL); |
1552 | ||
421748ec | 1553 | error = kern_path(pathname, LOOKUP_FOLLOW, &path); |
1da177e4 LT |
1554 | if (error) |
1555 | return ERR_PTR(error); | |
1556 | ||
421748ec | 1557 | inode = path.dentry->d_inode; |
1da177e4 LT |
1558 | error = -ENOTBLK; |
1559 | if (!S_ISBLK(inode->i_mode)) | |
1560 | goto fail; | |
1561 | error = -EACCES; | |
421748ec | 1562 | if (path.mnt->mnt_flags & MNT_NODEV) |
1da177e4 LT |
1563 | goto fail; |
1564 | error = -ENOMEM; | |
1565 | bdev = bd_acquire(inode); | |
1566 | if (!bdev) | |
1567 | goto fail; | |
1568 | out: | |
421748ec | 1569 | path_put(&path); |
1da177e4 LT |
1570 | return bdev; |
1571 | fail: | |
1572 | bdev = ERR_PTR(error); | |
1573 | goto out; | |
1574 | } | |
d5686b44 | 1575 | EXPORT_SYMBOL(lookup_bdev); |
1da177e4 LT |
1576 | |
1577 | /** | |
30c40d2c | 1578 | * open_bdev_exclusive - open a block device by name and set it up for use |
1da177e4 LT |
1579 | * |
1580 | * @path: special file representing the block device | |
30c40d2c | 1581 | * @mode: FMODE_... combination to pass be used |
1da177e4 LT |
1582 | * @holder: owner for exclusion |
1583 | * | |
1584 | * Open the blockdevice described by the special file at @path, claim it | |
1585 | * for the @holder. | |
1586 | */ | |
30c40d2c | 1587 | struct block_device *open_bdev_exclusive(const char *path, fmode_t mode, void *holder) |
1da177e4 LT |
1588 | { |
1589 | struct block_device *bdev; | |
1da177e4 LT |
1590 | int error = 0; |
1591 | ||
1592 | bdev = lookup_bdev(path); | |
1593 | if (IS_ERR(bdev)) | |
1594 | return bdev; | |
1595 | ||
572c4892 | 1596 | error = blkdev_get(bdev, mode); |
1da177e4 LT |
1597 | if (error) |
1598 | return ERR_PTR(error); | |
1599 | error = -EACCES; | |
30c40d2c | 1600 | if ((mode & FMODE_WRITE) && bdev_read_only(bdev)) |
1da177e4 LT |
1601 | goto blkdev_put; |
1602 | error = bd_claim(bdev, holder); | |
1603 | if (error) | |
1604 | goto blkdev_put; | |
1605 | ||
1606 | return bdev; | |
1607 | ||
1608 | blkdev_put: | |
9a1c3542 | 1609 | blkdev_put(bdev, mode); |
1da177e4 LT |
1610 | return ERR_PTR(error); |
1611 | } | |
1612 | ||
30c40d2c | 1613 | EXPORT_SYMBOL(open_bdev_exclusive); |
1da177e4 LT |
1614 | |
1615 | /** | |
30c40d2c | 1616 | * close_bdev_exclusive - close a blockdevice opened by open_bdev_exclusive() |
1da177e4 LT |
1617 | * |
1618 | * @bdev: blockdevice to close | |
30c40d2c | 1619 | * @mode: mode, must match that used to open. |
1da177e4 | 1620 | * |
30c40d2c | 1621 | * This is the counterpart to open_bdev_exclusive(). |
1da177e4 | 1622 | */ |
30c40d2c | 1623 | void close_bdev_exclusive(struct block_device *bdev, fmode_t mode) |
1da177e4 LT |
1624 | { |
1625 | bd_release(bdev); | |
30c40d2c | 1626 | blkdev_put(bdev, mode); |
1da177e4 LT |
1627 | } |
1628 | ||
30c40d2c | 1629 | EXPORT_SYMBOL(close_bdev_exclusive); |
b71e8a4c DH |
1630 | |
1631 | int __invalidate_device(struct block_device *bdev) | |
1632 | { | |
1633 | struct super_block *sb = get_super(bdev); | |
1634 | int res = 0; | |
1635 | ||
1636 | if (sb) { | |
1637 | /* | |
1638 | * no need to lock the super, get_super holds the | |
1639 | * read mutex so the filesystem cannot go away | |
1640 | * under us (->put_super runs with the write lock | |
1641 | * hold). | |
1642 | */ | |
1643 | shrink_dcache_sb(sb); | |
1644 | res = invalidate_inodes(sb); | |
1645 | drop_super(sb); | |
1646 | } | |
f98393a6 | 1647 | invalidate_bdev(bdev); |
b71e8a4c DH |
1648 | return res; |
1649 | } | |
1650 | EXPORT_SYMBOL(__invalidate_device); |