]>
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; | |
18e9e510 JB |
248 | error = freeze_super(sb); |
249 | if (error) { | |
250 | deactivate_super(sb); | |
251 | bdev->bd_fsfreeze_count--; | |
585d3bc0 | 252 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
18e9e510 | 253 | return ERR_PTR(error); |
585d3bc0 | 254 | } |
18e9e510 | 255 | deactivate_super(sb); |
4504230a | 256 | out: |
585d3bc0 NP |
257 | sync_blockdev(bdev); |
258 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
4fadd7bb | 259 | return sb; /* thaw_bdev releases s->s_umount */ |
585d3bc0 NP |
260 | } |
261 | EXPORT_SYMBOL(freeze_bdev); | |
262 | ||
263 | /** | |
264 | * thaw_bdev -- unlock filesystem | |
265 | * @bdev: blockdevice to unlock | |
266 | * @sb: associated superblock | |
267 | * | |
268 | * Unlocks the filesystem and marks it writeable again after freeze_bdev(). | |
269 | */ | |
270 | int thaw_bdev(struct block_device *bdev, struct super_block *sb) | |
271 | { | |
4504230a | 272 | int error = -EINVAL; |
585d3bc0 NP |
273 | |
274 | mutex_lock(&bdev->bd_fsfreeze_mutex); | |
4504230a | 275 | if (!bdev->bd_fsfreeze_count) |
18e9e510 | 276 | goto out; |
4504230a CH |
277 | |
278 | error = 0; | |
279 | if (--bdev->bd_fsfreeze_count > 0) | |
18e9e510 | 280 | goto out; |
4504230a CH |
281 | |
282 | if (!sb) | |
18e9e510 | 283 | goto out; |
4504230a | 284 | |
18e9e510 JB |
285 | error = thaw_super(sb); |
286 | if (error) { | |
287 | bdev->bd_fsfreeze_count++; | |
288 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
289 | return error; | |
290 | } | |
291 | out: | |
585d3bc0 NP |
292 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
293 | return 0; | |
294 | } | |
295 | EXPORT_SYMBOL(thaw_bdev); | |
296 | ||
1da177e4 LT |
297 | static int blkdev_writepage(struct page *page, struct writeback_control *wbc) |
298 | { | |
299 | return block_write_full_page(page, blkdev_get_block, wbc); | |
300 | } | |
301 | ||
302 | static int blkdev_readpage(struct file * file, struct page * page) | |
303 | { | |
304 | return block_read_full_page(page, blkdev_get_block); | |
305 | } | |
306 | ||
6272b5a5 NP |
307 | static int blkdev_write_begin(struct file *file, struct address_space *mapping, |
308 | loff_t pos, unsigned len, unsigned flags, | |
309 | struct page **pagep, void **fsdata) | |
1da177e4 | 310 | { |
6272b5a5 NP |
311 | *pagep = NULL; |
312 | return block_write_begin(file, mapping, pos, len, flags, pagep, fsdata, | |
313 | blkdev_get_block); | |
1da177e4 LT |
314 | } |
315 | ||
6272b5a5 NP |
316 | static int blkdev_write_end(struct file *file, struct address_space *mapping, |
317 | loff_t pos, unsigned len, unsigned copied, | |
318 | struct page *page, void *fsdata) | |
1da177e4 | 319 | { |
6272b5a5 NP |
320 | int ret; |
321 | ret = block_write_end(file, mapping, pos, len, copied, page, fsdata); | |
322 | ||
323 | unlock_page(page); | |
324 | page_cache_release(page); | |
325 | ||
326 | return ret; | |
1da177e4 LT |
327 | } |
328 | ||
329 | /* | |
330 | * private llseek: | |
0f7fc9e4 | 331 | * for a block special file file->f_path.dentry->d_inode->i_size is zero |
1da177e4 LT |
332 | * so we compute the size by hand (just as in block_read/write above) |
333 | */ | |
334 | static loff_t block_llseek(struct file *file, loff_t offset, int origin) | |
335 | { | |
336 | struct inode *bd_inode = file->f_mapping->host; | |
337 | loff_t size; | |
338 | loff_t retval; | |
339 | ||
1b1dcc1b | 340 | mutex_lock(&bd_inode->i_mutex); |
1da177e4 LT |
341 | size = i_size_read(bd_inode); |
342 | ||
343 | switch (origin) { | |
344 | case 2: | |
345 | offset += size; | |
346 | break; | |
347 | case 1: | |
348 | offset += file->f_pos; | |
349 | } | |
350 | retval = -EINVAL; | |
351 | if (offset >= 0 && offset <= size) { | |
352 | if (offset != file->f_pos) { | |
353 | file->f_pos = offset; | |
354 | } | |
355 | retval = offset; | |
356 | } | |
1b1dcc1b | 357 | mutex_unlock(&bd_inode->i_mutex); |
1da177e4 LT |
358 | return retval; |
359 | } | |
360 | ||
7ea80859 | 361 | int blkdev_fsync(struct file *filp, int datasync) |
1da177e4 | 362 | { |
b8af67e2 AB |
363 | struct inode *bd_inode = filp->f_mapping->host; |
364 | struct block_device *bdev = I_BDEV(bd_inode); | |
ab0a9735 CH |
365 | int error; |
366 | ||
b8af67e2 AB |
367 | /* |
368 | * There is no need to serialise calls to blkdev_issue_flush with | |
369 | * i_mutex and doing so causes performance issues with concurrent | |
370 | * O_SYNC writers to a block device. | |
371 | */ | |
372 | mutex_unlock(&bd_inode->i_mutex); | |
373 | ||
7407cf35 | 374 | error = blkdev_issue_flush(bdev, GFP_KERNEL, NULL, BLKDEV_IFL_WAIT); |
ab0a9735 CH |
375 | if (error == -EOPNOTSUPP) |
376 | error = 0; | |
b8af67e2 AB |
377 | |
378 | mutex_lock(&bd_inode->i_mutex); | |
379 | ||
ab0a9735 | 380 | return error; |
1da177e4 | 381 | } |
b1dd3b28 | 382 | EXPORT_SYMBOL(blkdev_fsync); |
1da177e4 LT |
383 | |
384 | /* | |
385 | * pseudo-fs | |
386 | */ | |
387 | ||
388 | static __cacheline_aligned_in_smp DEFINE_SPINLOCK(bdev_lock); | |
e18b890b | 389 | static struct kmem_cache * bdev_cachep __read_mostly; |
1da177e4 LT |
390 | |
391 | static struct inode *bdev_alloc_inode(struct super_block *sb) | |
392 | { | |
e94b1766 | 393 | struct bdev_inode *ei = kmem_cache_alloc(bdev_cachep, GFP_KERNEL); |
1da177e4 LT |
394 | if (!ei) |
395 | return NULL; | |
396 | return &ei->vfs_inode; | |
397 | } | |
398 | ||
399 | static void bdev_destroy_inode(struct inode *inode) | |
400 | { | |
401 | struct bdev_inode *bdi = BDEV_I(inode); | |
402 | ||
1da177e4 LT |
403 | kmem_cache_free(bdev_cachep, bdi); |
404 | } | |
405 | ||
51cc5068 | 406 | static void init_once(void *foo) |
1da177e4 LT |
407 | { |
408 | struct bdev_inode *ei = (struct bdev_inode *) foo; | |
409 | struct block_device *bdev = &ei->bdev; | |
410 | ||
a35afb83 CL |
411 | memset(bdev, 0, sizeof(*bdev)); |
412 | mutex_init(&bdev->bd_mutex); | |
a35afb83 CL |
413 | INIT_LIST_HEAD(&bdev->bd_inodes); |
414 | INIT_LIST_HEAD(&bdev->bd_list); | |
641dc636 | 415 | #ifdef CONFIG_SYSFS |
a35afb83 | 416 | INIT_LIST_HEAD(&bdev->bd_holder_list); |
641dc636 | 417 | #endif |
a35afb83 | 418 | inode_init_once(&ei->vfs_inode); |
fcccf502 TS |
419 | /* Initialize mutex for freeze. */ |
420 | mutex_init(&bdev->bd_fsfreeze_mutex); | |
1da177e4 LT |
421 | } |
422 | ||
423 | static inline void __bd_forget(struct inode *inode) | |
424 | { | |
425 | list_del_init(&inode->i_devices); | |
426 | inode->i_bdev = NULL; | |
427 | inode->i_mapping = &inode->i_data; | |
428 | } | |
429 | ||
430 | static void bdev_clear_inode(struct inode *inode) | |
431 | { | |
432 | struct block_device *bdev = &BDEV_I(inode)->bdev; | |
433 | struct list_head *p; | |
434 | spin_lock(&bdev_lock); | |
435 | while ( (p = bdev->bd_inodes.next) != &bdev->bd_inodes ) { | |
436 | __bd_forget(list_entry(p, struct inode, i_devices)); | |
437 | } | |
438 | list_del_init(&bdev->bd_list); | |
439 | spin_unlock(&bdev_lock); | |
440 | } | |
441 | ||
ee9b6d61 | 442 | static const struct super_operations bdev_sops = { |
1da177e4 LT |
443 | .statfs = simple_statfs, |
444 | .alloc_inode = bdev_alloc_inode, | |
445 | .destroy_inode = bdev_destroy_inode, | |
446 | .drop_inode = generic_delete_inode, | |
447 | .clear_inode = bdev_clear_inode, | |
448 | }; | |
449 | ||
454e2398 DH |
450 | static int bd_get_sb(struct file_system_type *fs_type, |
451 | int flags, const char *dev_name, void *data, struct vfsmount *mnt) | |
1da177e4 | 452 | { |
454e2398 | 453 | return get_sb_pseudo(fs_type, "bdev:", &bdev_sops, 0x62646576, mnt); |
1da177e4 LT |
454 | } |
455 | ||
456 | static struct file_system_type bd_type = { | |
457 | .name = "bdev", | |
458 | .get_sb = bd_get_sb, | |
459 | .kill_sb = kill_anon_super, | |
460 | }; | |
461 | ||
c2acf7b9 | 462 | struct super_block *blockdev_superblock __read_mostly; |
1da177e4 LT |
463 | |
464 | void __init bdev_cache_init(void) | |
465 | { | |
466 | int err; | |
c2acf7b9 DC |
467 | struct vfsmount *bd_mnt; |
468 | ||
1da177e4 | 469 | bdev_cachep = kmem_cache_create("bdev_cache", sizeof(struct bdev_inode), |
fffb60f9 PJ |
470 | 0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| |
471 | SLAB_MEM_SPREAD|SLAB_PANIC), | |
20c2df83 | 472 | init_once); |
1da177e4 LT |
473 | err = register_filesystem(&bd_type); |
474 | if (err) | |
475 | panic("Cannot register bdev pseudo-fs"); | |
476 | bd_mnt = kern_mount(&bd_type); | |
1da177e4 LT |
477 | if (IS_ERR(bd_mnt)) |
478 | panic("Cannot create bdev pseudo-fs"); | |
2e1483c9 CM |
479 | /* |
480 | * This vfsmount structure is only used to obtain the | |
481 | * blockdev_superblock, so tell kmemleak not to report it. | |
482 | */ | |
483 | kmemleak_not_leak(bd_mnt); | |
1da177e4 LT |
484 | blockdev_superblock = bd_mnt->mnt_sb; /* For writeback */ |
485 | } | |
486 | ||
487 | /* | |
488 | * Most likely _very_ bad one - but then it's hardly critical for small | |
489 | * /dev and can be fixed when somebody will need really large one. | |
490 | * Keep in mind that it will be fed through icache hash function too. | |
491 | */ | |
492 | static inline unsigned long hash(dev_t dev) | |
493 | { | |
494 | return MAJOR(dev)+MINOR(dev); | |
495 | } | |
496 | ||
497 | static int bdev_test(struct inode *inode, void *data) | |
498 | { | |
499 | return BDEV_I(inode)->bdev.bd_dev == *(dev_t *)data; | |
500 | } | |
501 | ||
502 | static int bdev_set(struct inode *inode, void *data) | |
503 | { | |
504 | BDEV_I(inode)->bdev.bd_dev = *(dev_t *)data; | |
505 | return 0; | |
506 | } | |
507 | ||
508 | static LIST_HEAD(all_bdevs); | |
509 | ||
510 | struct block_device *bdget(dev_t dev) | |
511 | { | |
512 | struct block_device *bdev; | |
513 | struct inode *inode; | |
514 | ||
c2acf7b9 | 515 | inode = iget5_locked(blockdev_superblock, hash(dev), |
1da177e4 LT |
516 | bdev_test, bdev_set, &dev); |
517 | ||
518 | if (!inode) | |
519 | return NULL; | |
520 | ||
521 | bdev = &BDEV_I(inode)->bdev; | |
522 | ||
523 | if (inode->i_state & I_NEW) { | |
524 | bdev->bd_contains = NULL; | |
525 | bdev->bd_inode = inode; | |
526 | bdev->bd_block_size = (1 << inode->i_blkbits); | |
527 | bdev->bd_part_count = 0; | |
528 | bdev->bd_invalidated = 0; | |
529 | inode->i_mode = S_IFBLK; | |
530 | inode->i_rdev = dev; | |
531 | inode->i_bdev = bdev; | |
532 | inode->i_data.a_ops = &def_blk_aops; | |
533 | mapping_set_gfp_mask(&inode->i_data, GFP_USER); | |
534 | inode->i_data.backing_dev_info = &default_backing_dev_info; | |
535 | spin_lock(&bdev_lock); | |
536 | list_add(&bdev->bd_list, &all_bdevs); | |
537 | spin_unlock(&bdev_lock); | |
538 | unlock_new_inode(inode); | |
539 | } | |
540 | return bdev; | |
541 | } | |
542 | ||
543 | EXPORT_SYMBOL(bdget); | |
544 | ||
dddac6a7 AJ |
545 | /** |
546 | * bdgrab -- Grab a reference to an already referenced block device | |
547 | * @bdev: Block device to grab a reference to. | |
548 | */ | |
549 | struct block_device *bdgrab(struct block_device *bdev) | |
550 | { | |
551 | atomic_inc(&bdev->bd_inode->i_count); | |
552 | return bdev; | |
553 | } | |
554 | ||
1da177e4 LT |
555 | long nr_blockdev_pages(void) |
556 | { | |
203a2935 | 557 | struct block_device *bdev; |
1da177e4 LT |
558 | long ret = 0; |
559 | spin_lock(&bdev_lock); | |
203a2935 | 560 | list_for_each_entry(bdev, &all_bdevs, bd_list) { |
1da177e4 LT |
561 | ret += bdev->bd_inode->i_mapping->nrpages; |
562 | } | |
563 | spin_unlock(&bdev_lock); | |
564 | return ret; | |
565 | } | |
566 | ||
567 | void bdput(struct block_device *bdev) | |
568 | { | |
569 | iput(bdev->bd_inode); | |
570 | } | |
571 | ||
572 | EXPORT_SYMBOL(bdput); | |
573 | ||
574 | static struct block_device *bd_acquire(struct inode *inode) | |
575 | { | |
576 | struct block_device *bdev; | |
09d967c6 | 577 | |
1da177e4 LT |
578 | spin_lock(&bdev_lock); |
579 | bdev = inode->i_bdev; | |
09d967c6 OH |
580 | if (bdev) { |
581 | atomic_inc(&bdev->bd_inode->i_count); | |
1da177e4 LT |
582 | spin_unlock(&bdev_lock); |
583 | return bdev; | |
584 | } | |
585 | spin_unlock(&bdev_lock); | |
09d967c6 | 586 | |
1da177e4 LT |
587 | bdev = bdget(inode->i_rdev); |
588 | if (bdev) { | |
589 | spin_lock(&bdev_lock); | |
09d967c6 OH |
590 | if (!inode->i_bdev) { |
591 | /* | |
592 | * We take an additional bd_inode->i_count for inode, | |
593 | * and it's released in clear_inode() of inode. | |
594 | * So, we can access it via ->i_mapping always | |
595 | * without igrab(). | |
596 | */ | |
597 | atomic_inc(&bdev->bd_inode->i_count); | |
598 | inode->i_bdev = bdev; | |
599 | inode->i_mapping = bdev->bd_inode->i_mapping; | |
600 | list_add(&inode->i_devices, &bdev->bd_inodes); | |
601 | } | |
1da177e4 LT |
602 | spin_unlock(&bdev_lock); |
603 | } | |
604 | return bdev; | |
605 | } | |
606 | ||
607 | /* Call when you free inode */ | |
608 | ||
609 | void bd_forget(struct inode *inode) | |
610 | { | |
09d967c6 OH |
611 | struct block_device *bdev = NULL; |
612 | ||
1da177e4 | 613 | spin_lock(&bdev_lock); |
09d967c6 | 614 | if (inode->i_bdev) { |
c2acf7b9 | 615 | if (!sb_is_blkdev_sb(inode->i_sb)) |
09d967c6 | 616 | bdev = inode->i_bdev; |
1da177e4 | 617 | __bd_forget(inode); |
09d967c6 | 618 | } |
1da177e4 | 619 | spin_unlock(&bdev_lock); |
09d967c6 OH |
620 | |
621 | if (bdev) | |
622 | iput(bdev->bd_inode); | |
1da177e4 LT |
623 | } |
624 | ||
1a3cbbc5 TH |
625 | /** |
626 | * bd_may_claim - test whether a block device can be claimed | |
627 | * @bdev: block device of interest | |
628 | * @whole: whole block device containing @bdev, may equal @bdev | |
629 | * @holder: holder trying to claim @bdev | |
630 | * | |
631 | * Test whther @bdev can be claimed by @holder. | |
632 | * | |
633 | * CONTEXT: | |
634 | * spin_lock(&bdev_lock). | |
635 | * | |
636 | * RETURNS: | |
637 | * %true if @bdev can be claimed, %false otherwise. | |
638 | */ | |
639 | static bool bd_may_claim(struct block_device *bdev, struct block_device *whole, | |
640 | void *holder) | |
1da177e4 | 641 | { |
1da177e4 | 642 | if (bdev->bd_holder == holder) |
1a3cbbc5 | 643 | return true; /* already a holder */ |
1da177e4 | 644 | else if (bdev->bd_holder != NULL) |
1a3cbbc5 | 645 | return false; /* held by someone else */ |
1da177e4 | 646 | else if (bdev->bd_contains == bdev) |
1a3cbbc5 | 647 | return true; /* is a whole device which isn't held */ |
1da177e4 | 648 | |
1a3cbbc5 TH |
649 | else if (whole->bd_holder == bd_claim) |
650 | return true; /* is a partition of a device that is being partitioned */ | |
651 | else if (whole->bd_holder != NULL) | |
652 | return false; /* is a partition of a held device */ | |
1da177e4 | 653 | else |
1a3cbbc5 TH |
654 | return true; /* is a partition of an un-held device */ |
655 | } | |
656 | ||
6b4517a7 TH |
657 | /** |
658 | * bd_prepare_to_claim - prepare to claim a block device | |
659 | * @bdev: block device of interest | |
660 | * @whole: the whole device containing @bdev, may equal @bdev | |
661 | * @holder: holder trying to claim @bdev | |
662 | * | |
663 | * Prepare to claim @bdev. This function fails if @bdev is already | |
664 | * claimed by another holder and waits if another claiming is in | |
665 | * progress. This function doesn't actually claim. On successful | |
666 | * return, the caller has ownership of bd_claiming and bd_holder[s]. | |
667 | * | |
668 | * CONTEXT: | |
669 | * spin_lock(&bdev_lock). Might release bdev_lock, sleep and regrab | |
670 | * it multiple times. | |
671 | * | |
672 | * RETURNS: | |
673 | * 0 if @bdev can be claimed, -EBUSY otherwise. | |
674 | */ | |
675 | static int bd_prepare_to_claim(struct block_device *bdev, | |
676 | struct block_device *whole, void *holder) | |
677 | { | |
678 | retry: | |
679 | /* if someone else claimed, fail */ | |
680 | if (!bd_may_claim(bdev, whole, holder)) | |
681 | return -EBUSY; | |
682 | ||
683 | /* if someone else is claiming, wait for it to finish */ | |
684 | if (whole->bd_claiming && whole->bd_claiming != holder) { | |
685 | wait_queue_head_t *wq = bit_waitqueue(&whole->bd_claiming, 0); | |
686 | DEFINE_WAIT(wait); | |
687 | ||
688 | prepare_to_wait(wq, &wait, TASK_UNINTERRUPTIBLE); | |
689 | spin_unlock(&bdev_lock); | |
690 | schedule(); | |
691 | finish_wait(wq, &wait); | |
692 | spin_lock(&bdev_lock); | |
693 | goto retry; | |
694 | } | |
695 | ||
696 | /* yay, all mine */ | |
697 | return 0; | |
698 | } | |
699 | ||
700 | /** | |
701 | * bd_start_claiming - start claiming a block device | |
702 | * @bdev: block device of interest | |
703 | * @holder: holder trying to claim @bdev | |
704 | * | |
705 | * @bdev is about to be opened exclusively. Check @bdev can be opened | |
706 | * exclusively and mark that an exclusive open is in progress. Each | |
707 | * successful call to this function must be matched with a call to | |
708 | * either bd_claim() or bd_abort_claiming(). If this function | |
709 | * succeeds, the matching bd_claim() is guaranteed to succeed. | |
710 | * | |
711 | * CONTEXT: | |
712 | * Might sleep. | |
713 | * | |
714 | * RETURNS: | |
715 | * Pointer to the block device containing @bdev on success, ERR_PTR() | |
716 | * value on failure. | |
717 | */ | |
718 | static struct block_device *bd_start_claiming(struct block_device *bdev, | |
719 | void *holder) | |
720 | { | |
721 | struct gendisk *disk; | |
722 | struct block_device *whole; | |
723 | int partno, err; | |
724 | ||
725 | might_sleep(); | |
726 | ||
727 | /* | |
728 | * @bdev might not have been initialized properly yet, look up | |
729 | * and grab the outer block device the hard way. | |
730 | */ | |
731 | disk = get_gendisk(bdev->bd_dev, &partno); | |
732 | if (!disk) | |
733 | return ERR_PTR(-ENXIO); | |
734 | ||
735 | whole = bdget_disk(disk, 0); | |
736 | put_disk(disk); | |
737 | if (!whole) | |
738 | return ERR_PTR(-ENOMEM); | |
739 | ||
740 | /* prepare to claim, if successful, mark claiming in progress */ | |
741 | spin_lock(&bdev_lock); | |
742 | ||
743 | err = bd_prepare_to_claim(bdev, whole, holder); | |
744 | if (err == 0) { | |
745 | whole->bd_claiming = holder; | |
746 | spin_unlock(&bdev_lock); | |
747 | return whole; | |
748 | } else { | |
749 | spin_unlock(&bdev_lock); | |
750 | bdput(whole); | |
751 | return ERR_PTR(err); | |
752 | } | |
753 | } | |
754 | ||
755 | /* releases bdev_lock */ | |
756 | static void __bd_abort_claiming(struct block_device *whole, void *holder) | |
757 | { | |
758 | BUG_ON(whole->bd_claiming != holder); | |
759 | whole->bd_claiming = NULL; | |
760 | wake_up_bit(&whole->bd_claiming, 0); | |
761 | ||
762 | spin_unlock(&bdev_lock); | |
763 | bdput(whole); | |
764 | } | |
765 | ||
766 | /** | |
767 | * bd_abort_claiming - abort claiming a block device | |
768 | * @whole: whole block device returned by bd_start_claiming() | |
769 | * @holder: holder trying to claim @bdev | |
770 | * | |
771 | * Abort a claiming block started by bd_start_claiming(). Note that | |
772 | * @whole is not the block device to be claimed but the whole device | |
773 | * returned by bd_start_claiming(). | |
774 | * | |
775 | * CONTEXT: | |
776 | * Grabs and releases bdev_lock. | |
777 | */ | |
778 | static void bd_abort_claiming(struct block_device *whole, void *holder) | |
779 | { | |
780 | spin_lock(&bdev_lock); | |
781 | __bd_abort_claiming(whole, holder); /* releases bdev_lock */ | |
782 | } | |
783 | ||
1a3cbbc5 TH |
784 | /** |
785 | * bd_claim - claim a block device | |
786 | * @bdev: block device to claim | |
787 | * @holder: holder trying to claim @bdev | |
788 | * | |
6b4517a7 TH |
789 | * Try to claim @bdev which must have been opened successfully. This |
790 | * function may be called with or without preceding | |
791 | * blk_start_claiming(). In the former case, this function is always | |
792 | * successful and terminates the claiming block. | |
793 | * | |
794 | * CONTEXT: | |
795 | * Might sleep. | |
1a3cbbc5 TH |
796 | * |
797 | * RETURNS: | |
798 | * 0 if successful, -EBUSY if @bdev is already claimed. | |
799 | */ | |
800 | int bd_claim(struct block_device *bdev, void *holder) | |
801 | { | |
802 | struct block_device *whole = bdev->bd_contains; | |
6b4517a7 TH |
803 | int res; |
804 | ||
805 | might_sleep(); | |
1a3cbbc5 TH |
806 | |
807 | spin_lock(&bdev_lock); | |
1da177e4 | 808 | |
6b4517a7 TH |
809 | res = bd_prepare_to_claim(bdev, whole, holder); |
810 | if (res == 0) { | |
1da177e4 LT |
811 | /* note that for a whole device bd_holders |
812 | * will be incremented twice, and bd_holder will | |
813 | * be set to bd_claim before being set to holder | |
814 | */ | |
1a3cbbc5 TH |
815 | whole->bd_holders++; |
816 | whole->bd_holder = bd_claim; | |
1da177e4 LT |
817 | bdev->bd_holders++; |
818 | bdev->bd_holder = holder; | |
819 | } | |
1a3cbbc5 | 820 | |
6b4517a7 TH |
821 | if (whole->bd_claiming) |
822 | __bd_abort_claiming(whole, holder); /* releases bdev_lock */ | |
823 | else | |
824 | spin_unlock(&bdev_lock); | |
825 | ||
1da177e4 LT |
826 | return res; |
827 | } | |
1da177e4 LT |
828 | EXPORT_SYMBOL(bd_claim); |
829 | ||
830 | void bd_release(struct block_device *bdev) | |
831 | { | |
832 | spin_lock(&bdev_lock); | |
833 | if (!--bdev->bd_contains->bd_holders) | |
834 | bdev->bd_contains->bd_holder = NULL; | |
835 | if (!--bdev->bd_holders) | |
836 | bdev->bd_holder = NULL; | |
837 | spin_unlock(&bdev_lock); | |
838 | } | |
839 | ||
840 | EXPORT_SYMBOL(bd_release); | |
841 | ||
641dc636 JN |
842 | #ifdef CONFIG_SYSFS |
843 | /* | |
844 | * Functions for bd_claim_by_kobject / bd_release_from_kobject | |
845 | * | |
846 | * If a kobject is passed to bd_claim_by_kobject() | |
847 | * and the kobject has a parent directory, | |
848 | * following symlinks are created: | |
849 | * o from the kobject to the claimed bdev | |
850 | * o from "holders" directory of the bdev to the parent of the kobject | |
851 | * bd_release_from_kobject() removes these symlinks. | |
852 | * | |
853 | * Example: | |
854 | * If /dev/dm-0 maps to /dev/sda, kobject corresponding to | |
855 | * /sys/block/dm-0/slaves is passed to bd_claim_by_kobject(), then: | |
856 | * /sys/block/dm-0/slaves/sda --> /sys/block/sda | |
857 | * /sys/block/sda/holders/dm-0 --> /sys/block/dm-0 | |
858 | */ | |
859 | ||
4d7dd8fd | 860 | static int add_symlink(struct kobject *from, struct kobject *to) |
641dc636 JN |
861 | { |
862 | if (!from || !to) | |
4d7dd8fd AM |
863 | return 0; |
864 | return sysfs_create_link(from, to, kobject_name(to)); | |
641dc636 JN |
865 | } |
866 | ||
867 | static void del_symlink(struct kobject *from, struct kobject *to) | |
868 | { | |
869 | if (!from || !to) | |
870 | return; | |
871 | sysfs_remove_link(from, kobject_name(to)); | |
872 | } | |
873 | ||
874 | /* | |
875 | * 'struct bd_holder' contains pointers to kobjects symlinked by | |
876 | * bd_claim_by_kobject. | |
877 | * It's connected to bd_holder_list which is protected by bdev->bd_sem. | |
878 | */ | |
879 | struct bd_holder { | |
880 | struct list_head list; /* chain of holders of the bdev */ | |
881 | int count; /* references from the holder */ | |
882 | struct kobject *sdir; /* holder object, e.g. "/block/dm-0/slaves" */ | |
883 | struct kobject *hdev; /* e.g. "/block/dm-0" */ | |
884 | struct kobject *hdir; /* e.g. "/block/sda/holders" */ | |
885 | struct kobject *sdev; /* e.g. "/block/sda" */ | |
886 | }; | |
887 | ||
888 | /* | |
889 | * Get references of related kobjects at once. | |
890 | * Returns 1 on success. 0 on failure. | |
891 | * | |
892 | * Should call bd_holder_release_dirs() after successful use. | |
893 | */ | |
894 | static int bd_holder_grab_dirs(struct block_device *bdev, | |
895 | struct bd_holder *bo) | |
896 | { | |
897 | if (!bdev || !bo) | |
898 | return 0; | |
899 | ||
900 | bo->sdir = kobject_get(bo->sdir); | |
901 | if (!bo->sdir) | |
902 | return 0; | |
903 | ||
904 | bo->hdev = kobject_get(bo->sdir->parent); | |
905 | if (!bo->hdev) | |
906 | goto fail_put_sdir; | |
907 | ||
0762b8bd | 908 | bo->sdev = kobject_get(&part_to_dev(bdev->bd_part)->kobj); |
641dc636 JN |
909 | if (!bo->sdev) |
910 | goto fail_put_hdev; | |
911 | ||
4c46501d | 912 | bo->hdir = kobject_get(bdev->bd_part->holder_dir); |
641dc636 JN |
913 | if (!bo->hdir) |
914 | goto fail_put_sdev; | |
915 | ||
916 | return 1; | |
917 | ||
918 | fail_put_sdev: | |
919 | kobject_put(bo->sdev); | |
920 | fail_put_hdev: | |
921 | kobject_put(bo->hdev); | |
922 | fail_put_sdir: | |
923 | kobject_put(bo->sdir); | |
924 | ||
925 | return 0; | |
926 | } | |
927 | ||
928 | /* Put references of related kobjects at once. */ | |
929 | static void bd_holder_release_dirs(struct bd_holder *bo) | |
930 | { | |
931 | kobject_put(bo->hdir); | |
932 | kobject_put(bo->sdev); | |
933 | kobject_put(bo->hdev); | |
934 | kobject_put(bo->sdir); | |
935 | } | |
936 | ||
937 | static struct bd_holder *alloc_bd_holder(struct kobject *kobj) | |
938 | { | |
939 | struct bd_holder *bo; | |
940 | ||
941 | bo = kzalloc(sizeof(*bo), GFP_KERNEL); | |
942 | if (!bo) | |
943 | return NULL; | |
944 | ||
945 | bo->count = 1; | |
946 | bo->sdir = kobj; | |
947 | ||
948 | return bo; | |
949 | } | |
950 | ||
951 | static void free_bd_holder(struct bd_holder *bo) | |
952 | { | |
953 | kfree(bo); | |
954 | } | |
955 | ||
df6c0cd9 JN |
956 | /** |
957 | * find_bd_holder - find matching struct bd_holder from the block device | |
958 | * | |
959 | * @bdev: struct block device to be searched | |
960 | * @bo: target struct bd_holder | |
961 | * | |
962 | * Returns matching entry with @bo in @bdev->bd_holder_list. | |
963 | * If found, increment the reference count and return the pointer. | |
964 | * If not found, returns NULL. | |
965 | */ | |
36a561d6 AM |
966 | static struct bd_holder *find_bd_holder(struct block_device *bdev, |
967 | struct bd_holder *bo) | |
df6c0cd9 JN |
968 | { |
969 | struct bd_holder *tmp; | |
970 | ||
971 | list_for_each_entry(tmp, &bdev->bd_holder_list, list) | |
972 | if (tmp->sdir == bo->sdir) { | |
973 | tmp->count++; | |
974 | return tmp; | |
975 | } | |
976 | ||
977 | return NULL; | |
978 | } | |
979 | ||
641dc636 JN |
980 | /** |
981 | * add_bd_holder - create sysfs symlinks for bd_claim() relationship | |
982 | * | |
983 | * @bdev: block device to be bd_claimed | |
984 | * @bo: preallocated and initialized by alloc_bd_holder() | |
985 | * | |
df6c0cd9 | 986 | * Add @bo to @bdev->bd_holder_list, create symlinks. |
641dc636 | 987 | * |
df6c0cd9 JN |
988 | * Returns 0 if symlinks are created. |
989 | * Returns -ve if something fails. | |
641dc636 JN |
990 | */ |
991 | static int add_bd_holder(struct block_device *bdev, struct bd_holder *bo) | |
992 | { | |
4e91672c | 993 | int err; |
641dc636 JN |
994 | |
995 | if (!bo) | |
4d7dd8fd | 996 | return -EINVAL; |
641dc636 | 997 | |
641dc636 | 998 | if (!bd_holder_grab_dirs(bdev, bo)) |
4d7dd8fd | 999 | return -EBUSY; |
641dc636 | 1000 | |
4e91672c JW |
1001 | err = add_symlink(bo->sdir, bo->sdev); |
1002 | if (err) | |
1003 | return err; | |
1004 | ||
1005 | err = add_symlink(bo->hdir, bo->hdev); | |
1006 | if (err) { | |
1007 | del_symlink(bo->sdir, bo->sdev); | |
1008 | return err; | |
4d7dd8fd | 1009 | } |
4e91672c JW |
1010 | |
1011 | list_add_tail(&bo->list, &bdev->bd_holder_list); | |
1012 | return 0; | |
641dc636 JN |
1013 | } |
1014 | ||
1015 | /** | |
1016 | * del_bd_holder - delete sysfs symlinks for bd_claim() relationship | |
1017 | * | |
1018 | * @bdev: block device to be bd_claimed | |
1019 | * @kobj: holder's kobject | |
1020 | * | |
1021 | * If there is matching entry with @kobj in @bdev->bd_holder_list | |
1022 | * and no other bd_claim() from the same kobject, | |
1023 | * remove the struct bd_holder from the list, delete symlinks for it. | |
1024 | * | |
1025 | * Returns a pointer to the struct bd_holder when it's removed from the list | |
1026 | * and ready to be freed. | |
1027 | * Returns NULL if matching claim isn't found or there is other bd_claim() | |
1028 | * by the same kobject. | |
1029 | */ | |
1030 | static struct bd_holder *del_bd_holder(struct block_device *bdev, | |
1031 | struct kobject *kobj) | |
1032 | { | |
1033 | struct bd_holder *bo; | |
1034 | ||
1035 | list_for_each_entry(bo, &bdev->bd_holder_list, list) { | |
1036 | if (bo->sdir == kobj) { | |
1037 | bo->count--; | |
1038 | BUG_ON(bo->count < 0); | |
1039 | if (!bo->count) { | |
1040 | list_del(&bo->list); | |
1041 | del_symlink(bo->sdir, bo->sdev); | |
1042 | del_symlink(bo->hdir, bo->hdev); | |
1043 | bd_holder_release_dirs(bo); | |
1044 | return bo; | |
1045 | } | |
1046 | break; | |
1047 | } | |
1048 | } | |
1049 | ||
1050 | return NULL; | |
1051 | } | |
1052 | ||
1053 | /** | |
1054 | * bd_claim_by_kobject - bd_claim() with additional kobject signature | |
1055 | * | |
1056 | * @bdev: block device to be claimed | |
1057 | * @holder: holder's signature | |
1058 | * @kobj: holder's kobject | |
1059 | * | |
1060 | * Do bd_claim() and if it succeeds, create sysfs symlinks between | |
1061 | * the bdev and the holder's kobject. | |
1062 | * Use bd_release_from_kobject() when relesing the claimed bdev. | |
1063 | * | |
1064 | * Returns 0 on success. (same as bd_claim()) | |
1065 | * Returns errno on failure. | |
1066 | */ | |
1067 | static int bd_claim_by_kobject(struct block_device *bdev, void *holder, | |
1068 | struct kobject *kobj) | |
1069 | { | |
4e91672c | 1070 | int err; |
df6c0cd9 | 1071 | struct bd_holder *bo, *found; |
641dc636 JN |
1072 | |
1073 | if (!kobj) | |
1074 | return -EINVAL; | |
1075 | ||
1076 | bo = alloc_bd_holder(kobj); | |
1077 | if (!bo) | |
1078 | return -ENOMEM; | |
1079 | ||
2e7b651d | 1080 | mutex_lock(&bdev->bd_mutex); |
df6c0cd9 | 1081 | |
4e91672c JW |
1082 | err = bd_claim(bdev, holder); |
1083 | if (err) | |
4210df28 | 1084 | goto fail; |
4e91672c JW |
1085 | |
1086 | found = find_bd_holder(bdev, bo); | |
1087 | if (found) | |
4210df28 | 1088 | goto fail; |
4e91672c JW |
1089 | |
1090 | err = add_bd_holder(bdev, bo); | |
1091 | if (err) | |
1092 | bd_release(bdev); | |
4210df28 AM |
1093 | else |
1094 | bo = NULL; | |
1095 | fail: | |
b4cf1b72 | 1096 | mutex_unlock(&bdev->bd_mutex); |
4210df28 | 1097 | free_bd_holder(bo); |
4e91672c | 1098 | return err; |
641dc636 JN |
1099 | } |
1100 | ||
1101 | /** | |
1102 | * bd_release_from_kobject - bd_release() with additional kobject signature | |
1103 | * | |
1104 | * @bdev: block device to be released | |
1105 | * @kobj: holder's kobject | |
1106 | * | |
1107 | * Do bd_release() and remove sysfs symlinks created by bd_claim_by_kobject(). | |
1108 | */ | |
1109 | static void bd_release_from_kobject(struct block_device *bdev, | |
1110 | struct kobject *kobj) | |
1111 | { | |
641dc636 JN |
1112 | if (!kobj) |
1113 | return; | |
1114 | ||
2e7b651d | 1115 | mutex_lock(&bdev->bd_mutex); |
641dc636 | 1116 | bd_release(bdev); |
4210df28 | 1117 | free_bd_holder(del_bd_holder(bdev, kobj)); |
b4cf1b72 | 1118 | mutex_unlock(&bdev->bd_mutex); |
641dc636 JN |
1119 | } |
1120 | ||
1121 | /** | |
1122 | * bd_claim_by_disk - wrapper function for bd_claim_by_kobject() | |
1123 | * | |
1124 | * @bdev: block device to be claimed | |
1125 | * @holder: holder's signature | |
1126 | * @disk: holder's gendisk | |
1127 | * | |
1128 | * Call bd_claim_by_kobject() with getting @disk->slave_dir. | |
1129 | */ | |
1130 | int bd_claim_by_disk(struct block_device *bdev, void *holder, | |
1131 | struct gendisk *disk) | |
1132 | { | |
1133 | return bd_claim_by_kobject(bdev, holder, kobject_get(disk->slave_dir)); | |
1134 | } | |
1135 | EXPORT_SYMBOL_GPL(bd_claim_by_disk); | |
1136 | ||
1137 | /** | |
1138 | * bd_release_from_disk - wrapper function for bd_release_from_kobject() | |
1139 | * | |
1140 | * @bdev: block device to be claimed | |
1141 | * @disk: holder's gendisk | |
1142 | * | |
1143 | * Call bd_release_from_kobject() and put @disk->slave_dir. | |
1144 | */ | |
1145 | void bd_release_from_disk(struct block_device *bdev, struct gendisk *disk) | |
1146 | { | |
1147 | bd_release_from_kobject(bdev, disk->slave_dir); | |
1148 | kobject_put(disk->slave_dir); | |
1149 | } | |
1150 | EXPORT_SYMBOL_GPL(bd_release_from_disk); | |
1151 | #endif | |
1152 | ||
1da177e4 LT |
1153 | /* |
1154 | * Tries to open block device by device number. Use it ONLY if you | |
1155 | * really do not have anything better - i.e. when you are behind a | |
1156 | * truly sucky interface and all you are given is a device number. _Never_ | |
1157 | * to be used for internal purposes. If you ever need it - reconsider | |
1158 | * your API. | |
1159 | */ | |
aeb5d727 | 1160 | struct block_device *open_by_devnum(dev_t dev, fmode_t mode) |
1da177e4 LT |
1161 | { |
1162 | struct block_device *bdev = bdget(dev); | |
1163 | int err = -ENOMEM; | |
1da177e4 | 1164 | if (bdev) |
572c4892 | 1165 | err = blkdev_get(bdev, mode); |
1da177e4 LT |
1166 | return err ? ERR_PTR(err) : bdev; |
1167 | } | |
1168 | ||
1169 | EXPORT_SYMBOL(open_by_devnum); | |
1170 | ||
56ade44b AP |
1171 | /** |
1172 | * flush_disk - invalidates all buffer-cache entries on a disk | |
1173 | * | |
1174 | * @bdev: struct block device to be flushed | |
1175 | * | |
1176 | * Invalidates all buffer-cache entries on a disk. It should be called | |
1177 | * when a disk has been changed -- either by a media change or online | |
1178 | * resize. | |
1179 | */ | |
1180 | static void flush_disk(struct block_device *bdev) | |
1181 | { | |
1182 | if (__invalidate_device(bdev)) { | |
1183 | char name[BDEVNAME_SIZE] = ""; | |
1184 | ||
1185 | if (bdev->bd_disk) | |
1186 | disk_name(bdev->bd_disk, 0, name); | |
1187 | printk(KERN_WARNING "VFS: busy inodes on changed media or " | |
1188 | "resized disk %s\n", name); | |
1189 | } | |
1190 | ||
1191 | if (!bdev->bd_disk) | |
1192 | return; | |
1193 | if (disk_partitionable(bdev->bd_disk)) | |
1194 | bdev->bd_invalidated = 1; | |
1195 | } | |
1196 | ||
c3279d14 | 1197 | /** |
57d1b536 | 1198 | * check_disk_size_change - checks for disk size change and adjusts bdev size. |
c3279d14 AP |
1199 | * @disk: struct gendisk to check |
1200 | * @bdev: struct bdev to adjust. | |
1201 | * | |
1202 | * This routine checks to see if the bdev size does not match the disk size | |
1203 | * and adjusts it if it differs. | |
1204 | */ | |
1205 | void check_disk_size_change(struct gendisk *disk, struct block_device *bdev) | |
1206 | { | |
1207 | loff_t disk_size, bdev_size; | |
1208 | ||
1209 | disk_size = (loff_t)get_capacity(disk) << 9; | |
1210 | bdev_size = i_size_read(bdev->bd_inode); | |
1211 | if (disk_size != bdev_size) { | |
1212 | char name[BDEVNAME_SIZE]; | |
1213 | ||
1214 | disk_name(disk, 0, name); | |
1215 | printk(KERN_INFO | |
1216 | "%s: detected capacity change from %lld to %lld\n", | |
1217 | name, bdev_size, disk_size); | |
1218 | i_size_write(bdev->bd_inode, disk_size); | |
608aeef1 | 1219 | flush_disk(bdev); |
c3279d14 AP |
1220 | } |
1221 | } | |
1222 | EXPORT_SYMBOL(check_disk_size_change); | |
1223 | ||
0c002c2f | 1224 | /** |
57d1b536 | 1225 | * revalidate_disk - wrapper for lower-level driver's revalidate_disk call-back |
0c002c2f AP |
1226 | * @disk: struct gendisk to be revalidated |
1227 | * | |
1228 | * This routine is a wrapper for lower-level driver's revalidate_disk | |
1229 | * call-backs. It is used to do common pre and post operations needed | |
1230 | * for all revalidate_disk operations. | |
1231 | */ | |
1232 | int revalidate_disk(struct gendisk *disk) | |
1233 | { | |
c3279d14 | 1234 | struct block_device *bdev; |
0c002c2f AP |
1235 | int ret = 0; |
1236 | ||
1237 | if (disk->fops->revalidate_disk) | |
1238 | ret = disk->fops->revalidate_disk(disk); | |
1239 | ||
c3279d14 AP |
1240 | bdev = bdget_disk(disk, 0); |
1241 | if (!bdev) | |
1242 | return ret; | |
1243 | ||
1244 | mutex_lock(&bdev->bd_mutex); | |
1245 | check_disk_size_change(disk, bdev); | |
1246 | mutex_unlock(&bdev->bd_mutex); | |
1247 | bdput(bdev); | |
0c002c2f AP |
1248 | return ret; |
1249 | } | |
1250 | EXPORT_SYMBOL(revalidate_disk); | |
1251 | ||
1da177e4 LT |
1252 | /* |
1253 | * This routine checks whether a removable media has been changed, | |
1254 | * and invalidates all buffer-cache-entries in that case. This | |
1255 | * is a relatively slow routine, so we have to try to minimize using | |
1256 | * it. Thus it is called only upon a 'mount' or 'open'. This | |
1257 | * is the best way of combining speed and utility, I think. | |
1258 | * People changing diskettes in the middle of an operation deserve | |
1259 | * to lose :-) | |
1260 | */ | |
1261 | int check_disk_change(struct block_device *bdev) | |
1262 | { | |
1263 | struct gendisk *disk = bdev->bd_disk; | |
83d5cde4 | 1264 | const struct block_device_operations *bdops = disk->fops; |
1da177e4 LT |
1265 | |
1266 | if (!bdops->media_changed) | |
1267 | return 0; | |
1268 | if (!bdops->media_changed(bdev->bd_disk)) | |
1269 | return 0; | |
1270 | ||
56ade44b | 1271 | flush_disk(bdev); |
1da177e4 LT |
1272 | if (bdops->revalidate_disk) |
1273 | bdops->revalidate_disk(bdev->bd_disk); | |
1da177e4 LT |
1274 | return 1; |
1275 | } | |
1276 | ||
1277 | EXPORT_SYMBOL(check_disk_change); | |
1278 | ||
1279 | void bd_set_size(struct block_device *bdev, loff_t size) | |
1280 | { | |
e1defc4f | 1281 | unsigned bsize = bdev_logical_block_size(bdev); |
1da177e4 LT |
1282 | |
1283 | bdev->bd_inode->i_size = size; | |
1284 | while (bsize < PAGE_CACHE_SIZE) { | |
1285 | if (size & bsize) | |
1286 | break; | |
1287 | bsize <<= 1; | |
1288 | } | |
1289 | bdev->bd_block_size = bsize; | |
1290 | bdev->bd_inode->i_blkbits = blksize_bits(bsize); | |
1291 | } | |
1292 | EXPORT_SYMBOL(bd_set_size); | |
1293 | ||
9a1c3542 | 1294 | static int __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part); |
37be4124 | 1295 | |
6d740cd5 PZ |
1296 | /* |
1297 | * bd_mutex locking: | |
1298 | * | |
1299 | * mutex_lock(part->bd_mutex) | |
1300 | * mutex_lock_nested(whole->bd_mutex, 1) | |
1301 | */ | |
1302 | ||
572c4892 | 1303 | static int __blkdev_get(struct block_device *bdev, fmode_t mode, int for_part) |
1da177e4 | 1304 | { |
1da177e4 | 1305 | struct gendisk *disk; |
7db9cfd3 | 1306 | int ret; |
cf771cb5 | 1307 | int partno; |
fe6e9c1f AV |
1308 | int perm = 0; |
1309 | ||
572c4892 | 1310 | if (mode & FMODE_READ) |
fe6e9c1f | 1311 | perm |= MAY_READ; |
572c4892 | 1312 | if (mode & FMODE_WRITE) |
fe6e9c1f AV |
1313 | perm |= MAY_WRITE; |
1314 | /* | |
1315 | * hooks: /n/, see "layering violations". | |
1316 | */ | |
1317 | ret = devcgroup_inode_permission(bdev->bd_inode, perm); | |
82666020 AV |
1318 | if (ret != 0) { |
1319 | bdput(bdev); | |
7db9cfd3 | 1320 | return ret; |
82666020 | 1321 | } |
7db9cfd3 | 1322 | |
1da177e4 | 1323 | lock_kernel(); |
d3374825 | 1324 | restart: |
0762b8bd | 1325 | |
89f97496 | 1326 | ret = -ENXIO; |
cf771cb5 | 1327 | disk = get_gendisk(bdev->bd_dev, &partno); |
0762b8bd TH |
1328 | if (!disk) |
1329 | goto out_unlock_kernel; | |
1da177e4 | 1330 | |
6796bf54 | 1331 | mutex_lock_nested(&bdev->bd_mutex, for_part); |
1da177e4 LT |
1332 | if (!bdev->bd_openers) { |
1333 | bdev->bd_disk = disk; | |
1334 | bdev->bd_contains = bdev; | |
cf771cb5 | 1335 | if (!partno) { |
1da177e4 | 1336 | struct backing_dev_info *bdi; |
89f97496 TH |
1337 | |
1338 | ret = -ENXIO; | |
1339 | bdev->bd_part = disk_get_part(disk, partno); | |
1340 | if (!bdev->bd_part) | |
1341 | goto out_clear; | |
1342 | ||
1da177e4 | 1343 | if (disk->fops->open) { |
572c4892 | 1344 | ret = disk->fops->open(bdev, mode); |
d3374825 N |
1345 | if (ret == -ERESTARTSYS) { |
1346 | /* Lost a race with 'disk' being | |
1347 | * deleted, try again. | |
1348 | * See md.c | |
1349 | */ | |
1350 | disk_put_part(bdev->bd_part); | |
1351 | bdev->bd_part = NULL; | |
1352 | module_put(disk->fops->owner); | |
1353 | put_disk(disk); | |
1354 | bdev->bd_disk = NULL; | |
1355 | mutex_unlock(&bdev->bd_mutex); | |
1356 | goto restart; | |
1357 | } | |
1da177e4 | 1358 | if (ret) |
0762b8bd | 1359 | goto out_clear; |
1da177e4 LT |
1360 | } |
1361 | if (!bdev->bd_openers) { | |
1362 | bd_set_size(bdev,(loff_t)get_capacity(disk)<<9); | |
1363 | bdi = blk_get_backing_dev_info(bdev); | |
1364 | if (bdi == NULL) | |
1365 | bdi = &default_backing_dev_info; | |
1366 | bdev->bd_inode->i_data.backing_dev_info = bdi; | |
1367 | } | |
1368 | if (bdev->bd_invalidated) | |
1369 | rescan_partitions(disk, bdev); | |
1370 | } else { | |
1da177e4 LT |
1371 | struct block_device *whole; |
1372 | whole = bdget_disk(disk, 0); | |
1373 | ret = -ENOMEM; | |
1374 | if (!whole) | |
0762b8bd | 1375 | goto out_clear; |
37be4124 | 1376 | BUG_ON(for_part); |
572c4892 | 1377 | ret = __blkdev_get(whole, mode, 1); |
1da177e4 | 1378 | if (ret) |
0762b8bd | 1379 | goto out_clear; |
1da177e4 | 1380 | bdev->bd_contains = whole; |
1da177e4 LT |
1381 | bdev->bd_inode->i_data.backing_dev_info = |
1382 | whole->bd_inode->i_data.backing_dev_info; | |
89f97496 | 1383 | bdev->bd_part = disk_get_part(disk, partno); |
e71bf0d0 | 1384 | if (!(disk->flags & GENHD_FL_UP) || |
89f97496 | 1385 | !bdev->bd_part || !bdev->bd_part->nr_sects) { |
1da177e4 | 1386 | ret = -ENXIO; |
0762b8bd | 1387 | goto out_clear; |
1da177e4 | 1388 | } |
89f97496 | 1389 | bd_set_size(bdev, (loff_t)bdev->bd_part->nr_sects << 9); |
1da177e4 LT |
1390 | } |
1391 | } else { | |
0762b8bd | 1392 | module_put(disk->fops->owner); |
960cc0f4 | 1393 | put_disk(disk); |
0762b8bd | 1394 | disk = NULL; |
1da177e4 LT |
1395 | if (bdev->bd_contains == bdev) { |
1396 | if (bdev->bd_disk->fops->open) { | |
572c4892 | 1397 | ret = bdev->bd_disk->fops->open(bdev, mode); |
1da177e4 | 1398 | if (ret) |
0762b8bd | 1399 | goto out_unlock_bdev; |
1da177e4 LT |
1400 | } |
1401 | if (bdev->bd_invalidated) | |
1402 | rescan_partitions(bdev->bd_disk, bdev); | |
1da177e4 LT |
1403 | } |
1404 | } | |
1405 | bdev->bd_openers++; | |
37be4124 N |
1406 | if (for_part) |
1407 | bdev->bd_part_count++; | |
c039e313 | 1408 | mutex_unlock(&bdev->bd_mutex); |
1da177e4 LT |
1409 | unlock_kernel(); |
1410 | return 0; | |
1411 | ||
0762b8bd | 1412 | out_clear: |
89f97496 | 1413 | disk_put_part(bdev->bd_part); |
1da177e4 | 1414 | bdev->bd_disk = NULL; |
0762b8bd | 1415 | bdev->bd_part = NULL; |
1da177e4 LT |
1416 | bdev->bd_inode->i_data.backing_dev_info = &default_backing_dev_info; |
1417 | if (bdev != bdev->bd_contains) | |
572c4892 | 1418 | __blkdev_put(bdev->bd_contains, mode, 1); |
1da177e4 | 1419 | bdev->bd_contains = NULL; |
0762b8bd | 1420 | out_unlock_bdev: |
c039e313 | 1421 | mutex_unlock(&bdev->bd_mutex); |
0762b8bd | 1422 | out_unlock_kernel: |
1da177e4 | 1423 | unlock_kernel(); |
0762b8bd | 1424 | |
0762b8bd TH |
1425 | if (disk) |
1426 | module_put(disk->fops->owner); | |
1427 | put_disk(disk); | |
1428 | bdput(bdev); | |
1429 | ||
1da177e4 LT |
1430 | return ret; |
1431 | } | |
1432 | ||
572c4892 | 1433 | int blkdev_get(struct block_device *bdev, fmode_t mode) |
1da177e4 | 1434 | { |
572c4892 | 1435 | return __blkdev_get(bdev, mode, 0); |
37be4124 | 1436 | } |
1da177e4 LT |
1437 | EXPORT_SYMBOL(blkdev_get); |
1438 | ||
1439 | static int blkdev_open(struct inode * inode, struct file * filp) | |
1440 | { | |
6b4517a7 | 1441 | struct block_device *whole = NULL; |
1da177e4 LT |
1442 | struct block_device *bdev; |
1443 | int res; | |
1444 | ||
1445 | /* | |
1446 | * Preserve backwards compatibility and allow large file access | |
1447 | * even if userspace doesn't ask for it explicitly. Some mkfs | |
1448 | * binary needs it. We might want to drop this workaround | |
1449 | * during an unstable branch. | |
1450 | */ | |
1451 | filp->f_flags |= O_LARGEFILE; | |
1452 | ||
572c4892 AV |
1453 | if (filp->f_flags & O_NDELAY) |
1454 | filp->f_mode |= FMODE_NDELAY; | |
1455 | if (filp->f_flags & O_EXCL) | |
1456 | filp->f_mode |= FMODE_EXCL; | |
1457 | if ((filp->f_flags & O_ACCMODE) == 3) | |
1458 | filp->f_mode |= FMODE_WRITE_IOCTL; | |
1459 | ||
1da177e4 | 1460 | bdev = bd_acquire(inode); |
6a2aae06 PE |
1461 | if (bdev == NULL) |
1462 | return -ENOMEM; | |
1da177e4 | 1463 | |
6b4517a7 TH |
1464 | if (filp->f_mode & FMODE_EXCL) { |
1465 | whole = bd_start_claiming(bdev, filp); | |
1466 | if (IS_ERR(whole)) { | |
1467 | bdput(bdev); | |
1468 | return PTR_ERR(whole); | |
1469 | } | |
1470 | } | |
1471 | ||
572c4892 AV |
1472 | filp->f_mapping = bdev->bd_inode->i_mapping; |
1473 | ||
1474 | res = blkdev_get(bdev, filp->f_mode); | |
1da177e4 | 1475 | |
6b4517a7 TH |
1476 | if (whole) { |
1477 | if (res == 0) | |
1478 | BUG_ON(bd_claim(bdev, filp) != 0); | |
1479 | else | |
1480 | bd_abort_claiming(whole, filp); | |
ebbefc01 | 1481 | } |
1da177e4 | 1482 | |
1da177e4 LT |
1483 | return res; |
1484 | } | |
1485 | ||
9a1c3542 | 1486 | static int __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part) |
2e7b651d PZ |
1487 | { |
1488 | int ret = 0; | |
2e7b651d | 1489 | struct gendisk *disk = bdev->bd_disk; |
37be4124 | 1490 | struct block_device *victim = NULL; |
2e7b651d | 1491 | |
6796bf54 | 1492 | mutex_lock_nested(&bdev->bd_mutex, for_part); |
2e7b651d | 1493 | lock_kernel(); |
37be4124 N |
1494 | if (for_part) |
1495 | bdev->bd_part_count--; | |
1496 | ||
2e7b651d PZ |
1497 | if (!--bdev->bd_openers) { |
1498 | sync_blockdev(bdev); | |
1499 | kill_bdev(bdev); | |
1500 | } | |
1501 | if (bdev->bd_contains == bdev) { | |
1502 | if (disk->fops->release) | |
9a1c3542 | 1503 | ret = disk->fops->release(disk, mode); |
2e7b651d PZ |
1504 | } |
1505 | if (!bdev->bd_openers) { | |
1506 | struct module *owner = disk->fops->owner; | |
1507 | ||
1508 | put_disk(disk); | |
1509 | module_put(owner); | |
0762b8bd TH |
1510 | disk_put_part(bdev->bd_part); |
1511 | bdev->bd_part = NULL; | |
2e7b651d PZ |
1512 | bdev->bd_disk = NULL; |
1513 | bdev->bd_inode->i_data.backing_dev_info = &default_backing_dev_info; | |
37be4124 N |
1514 | if (bdev != bdev->bd_contains) |
1515 | victim = bdev->bd_contains; | |
2e7b651d PZ |
1516 | bdev->bd_contains = NULL; |
1517 | } | |
1518 | unlock_kernel(); | |
1519 | mutex_unlock(&bdev->bd_mutex); | |
1520 | bdput(bdev); | |
37be4124 | 1521 | if (victim) |
9a1c3542 | 1522 | __blkdev_put(victim, mode, 1); |
2e7b651d PZ |
1523 | return ret; |
1524 | } | |
1525 | ||
9a1c3542 | 1526 | int blkdev_put(struct block_device *bdev, fmode_t mode) |
37be4124 | 1527 | { |
9a1c3542 | 1528 | return __blkdev_put(bdev, mode, 0); |
37be4124 | 1529 | } |
2e7b651d PZ |
1530 | EXPORT_SYMBOL(blkdev_put); |
1531 | ||
1da177e4 LT |
1532 | static int blkdev_close(struct inode * inode, struct file * filp) |
1533 | { | |
1534 | struct block_device *bdev = I_BDEV(filp->f_mapping->host); | |
1535 | if (bdev->bd_holder == filp) | |
1536 | bd_release(bdev); | |
9a1c3542 | 1537 | return blkdev_put(bdev, filp->f_mode); |
1da177e4 LT |
1538 | } |
1539 | ||
bb93e3a5 | 1540 | static long block_ioctl(struct file *file, unsigned cmd, unsigned long arg) |
1da177e4 | 1541 | { |
56b26add AV |
1542 | struct block_device *bdev = I_BDEV(file->f_mapping->host); |
1543 | fmode_t mode = file->f_mode; | |
fd4ce1ac CH |
1544 | |
1545 | /* | |
1546 | * O_NDELAY can be altered using fcntl(.., F_SETFL, ..), so we have | |
1547 | * to updated it before every ioctl. | |
1548 | */ | |
56b26add | 1549 | if (file->f_flags & O_NDELAY) |
fd4ce1ac CH |
1550 | mode |= FMODE_NDELAY; |
1551 | else | |
1552 | mode &= ~FMODE_NDELAY; | |
1553 | ||
56b26add | 1554 | return blkdev_ioctl(bdev, mode, cmd, arg); |
1da177e4 LT |
1555 | } |
1556 | ||
eef99380 CH |
1557 | /* |
1558 | * Write data to the block device. Only intended for the block device itself | |
1559 | * and the raw driver which basically is a fake block device. | |
1560 | * | |
1561 | * Does not take i_mutex for the write and thus is not for general purpose | |
1562 | * use. | |
1563 | */ | |
1564 | ssize_t blkdev_aio_write(struct kiocb *iocb, const struct iovec *iov, | |
1565 | unsigned long nr_segs, loff_t pos) | |
1566 | { | |
1567 | struct file *file = iocb->ki_filp; | |
1568 | ssize_t ret; | |
1569 | ||
1570 | BUG_ON(iocb->ki_pos != pos); | |
1571 | ||
1572 | ret = __generic_file_aio_write(iocb, iov, nr_segs, &iocb->ki_pos); | |
1573 | if (ret > 0 || ret == -EIOCBQUEUED) { | |
1574 | ssize_t err; | |
1575 | ||
1576 | err = generic_write_sync(file, pos, ret); | |
1577 | if (err < 0 && ret > 0) | |
1578 | ret = err; | |
1579 | } | |
1580 | return ret; | |
1581 | } | |
1582 | EXPORT_SYMBOL_GPL(blkdev_aio_write); | |
1583 | ||
87d8fe1e TT |
1584 | /* |
1585 | * Try to release a page associated with block device when the system | |
1586 | * is under memory pressure. | |
1587 | */ | |
1588 | static int blkdev_releasepage(struct page *page, gfp_t wait) | |
1589 | { | |
1590 | struct super_block *super = BDEV_I(page->mapping->host)->bdev.bd_super; | |
1591 | ||
1592 | if (super && super->s_op->bdev_try_to_free_page) | |
1593 | return super->s_op->bdev_try_to_free_page(super, page, wait); | |
1594 | ||
1595 | return try_to_free_buffers(page); | |
1596 | } | |
1597 | ||
4c54ac62 | 1598 | static const struct address_space_operations def_blk_aops = { |
1da177e4 LT |
1599 | .readpage = blkdev_readpage, |
1600 | .writepage = blkdev_writepage, | |
1601 | .sync_page = block_sync_page, | |
6272b5a5 NP |
1602 | .write_begin = blkdev_write_begin, |
1603 | .write_end = blkdev_write_end, | |
1da177e4 | 1604 | .writepages = generic_writepages, |
87d8fe1e | 1605 | .releasepage = blkdev_releasepage, |
1da177e4 LT |
1606 | .direct_IO = blkdev_direct_IO, |
1607 | }; | |
1608 | ||
4b6f5d20 | 1609 | const struct file_operations def_blk_fops = { |
1da177e4 LT |
1610 | .open = blkdev_open, |
1611 | .release = blkdev_close, | |
1612 | .llseek = block_llseek, | |
543ade1f BP |
1613 | .read = do_sync_read, |
1614 | .write = do_sync_write, | |
1da177e4 | 1615 | .aio_read = generic_file_aio_read, |
eef99380 | 1616 | .aio_write = blkdev_aio_write, |
1da177e4 | 1617 | .mmap = generic_file_mmap, |
b1dd3b28 | 1618 | .fsync = blkdev_fsync, |
bb93e3a5 | 1619 | .unlocked_ioctl = block_ioctl, |
1da177e4 LT |
1620 | #ifdef CONFIG_COMPAT |
1621 | .compat_ioctl = compat_blkdev_ioctl, | |
1622 | #endif | |
7f9c51f0 JA |
1623 | .splice_read = generic_file_splice_read, |
1624 | .splice_write = generic_file_splice_write, | |
1da177e4 LT |
1625 | }; |
1626 | ||
1627 | int ioctl_by_bdev(struct block_device *bdev, unsigned cmd, unsigned long arg) | |
1628 | { | |
1629 | int res; | |
1630 | mm_segment_t old_fs = get_fs(); | |
1631 | set_fs(KERNEL_DS); | |
56b26add | 1632 | res = blkdev_ioctl(bdev, 0, cmd, arg); |
1da177e4 LT |
1633 | set_fs(old_fs); |
1634 | return res; | |
1635 | } | |
1636 | ||
1637 | EXPORT_SYMBOL(ioctl_by_bdev); | |
1638 | ||
1639 | /** | |
1640 | * lookup_bdev - lookup a struct block_device by name | |
94e2959e | 1641 | * @pathname: special file representing the block device |
1da177e4 | 1642 | * |
57d1b536 | 1643 | * Get a reference to the blockdevice at @pathname in the current |
1da177e4 LT |
1644 | * namespace if possible and return it. Return ERR_PTR(error) |
1645 | * otherwise. | |
1646 | */ | |
421748ec | 1647 | struct block_device *lookup_bdev(const char *pathname) |
1da177e4 LT |
1648 | { |
1649 | struct block_device *bdev; | |
1650 | struct inode *inode; | |
421748ec | 1651 | struct path path; |
1da177e4 LT |
1652 | int error; |
1653 | ||
421748ec | 1654 | if (!pathname || !*pathname) |
1da177e4 LT |
1655 | return ERR_PTR(-EINVAL); |
1656 | ||
421748ec | 1657 | error = kern_path(pathname, LOOKUP_FOLLOW, &path); |
1da177e4 LT |
1658 | if (error) |
1659 | return ERR_PTR(error); | |
1660 | ||
421748ec | 1661 | inode = path.dentry->d_inode; |
1da177e4 LT |
1662 | error = -ENOTBLK; |
1663 | if (!S_ISBLK(inode->i_mode)) | |
1664 | goto fail; | |
1665 | error = -EACCES; | |
421748ec | 1666 | if (path.mnt->mnt_flags & MNT_NODEV) |
1da177e4 LT |
1667 | goto fail; |
1668 | error = -ENOMEM; | |
1669 | bdev = bd_acquire(inode); | |
1670 | if (!bdev) | |
1671 | goto fail; | |
1672 | out: | |
421748ec | 1673 | path_put(&path); |
1da177e4 LT |
1674 | return bdev; |
1675 | fail: | |
1676 | bdev = ERR_PTR(error); | |
1677 | goto out; | |
1678 | } | |
d5686b44 | 1679 | EXPORT_SYMBOL(lookup_bdev); |
1da177e4 LT |
1680 | |
1681 | /** | |
30c40d2c | 1682 | * open_bdev_exclusive - open a block device by name and set it up for use |
1da177e4 LT |
1683 | * |
1684 | * @path: special file representing the block device | |
30c40d2c | 1685 | * @mode: FMODE_... combination to pass be used |
1da177e4 LT |
1686 | * @holder: owner for exclusion |
1687 | * | |
1688 | * Open the blockdevice described by the special file at @path, claim it | |
1689 | * for the @holder. | |
1690 | */ | |
30c40d2c | 1691 | struct block_device *open_bdev_exclusive(const char *path, fmode_t mode, void *holder) |
1da177e4 | 1692 | { |
6b4517a7 TH |
1693 | struct block_device *bdev, *whole; |
1694 | int error; | |
1da177e4 LT |
1695 | |
1696 | bdev = lookup_bdev(path); | |
1697 | if (IS_ERR(bdev)) | |
1698 | return bdev; | |
1699 | ||
6b4517a7 TH |
1700 | whole = bd_start_claiming(bdev, holder); |
1701 | if (IS_ERR(whole)) { | |
1702 | bdput(bdev); | |
1703 | return whole; | |
1704 | } | |
1705 | ||
572c4892 | 1706 | error = blkdev_get(bdev, mode); |
1da177e4 | 1707 | if (error) |
6b4517a7 TH |
1708 | goto out_abort_claiming; |
1709 | ||
1da177e4 | 1710 | error = -EACCES; |
30c40d2c | 1711 | if ((mode & FMODE_WRITE) && bdev_read_only(bdev)) |
6b4517a7 | 1712 | goto out_blkdev_put; |
1da177e4 | 1713 | |
6b4517a7 | 1714 | BUG_ON(bd_claim(bdev, holder) != 0); |
1da177e4 | 1715 | return bdev; |
6b4517a7 TH |
1716 | |
1717 | out_blkdev_put: | |
9a1c3542 | 1718 | blkdev_put(bdev, mode); |
6b4517a7 TH |
1719 | out_abort_claiming: |
1720 | bd_abort_claiming(whole, holder); | |
1da177e4 LT |
1721 | return ERR_PTR(error); |
1722 | } | |
1723 | ||
30c40d2c | 1724 | EXPORT_SYMBOL(open_bdev_exclusive); |
1da177e4 LT |
1725 | |
1726 | /** | |
30c40d2c | 1727 | * close_bdev_exclusive - close a blockdevice opened by open_bdev_exclusive() |
1da177e4 LT |
1728 | * |
1729 | * @bdev: blockdevice to close | |
30c40d2c | 1730 | * @mode: mode, must match that used to open. |
1da177e4 | 1731 | * |
30c40d2c | 1732 | * This is the counterpart to open_bdev_exclusive(). |
1da177e4 | 1733 | */ |
30c40d2c | 1734 | void close_bdev_exclusive(struct block_device *bdev, fmode_t mode) |
1da177e4 LT |
1735 | { |
1736 | bd_release(bdev); | |
30c40d2c | 1737 | blkdev_put(bdev, mode); |
1da177e4 LT |
1738 | } |
1739 | ||
30c40d2c | 1740 | EXPORT_SYMBOL(close_bdev_exclusive); |
b71e8a4c DH |
1741 | |
1742 | int __invalidate_device(struct block_device *bdev) | |
1743 | { | |
1744 | struct super_block *sb = get_super(bdev); | |
1745 | int res = 0; | |
1746 | ||
1747 | if (sb) { | |
1748 | /* | |
1749 | * no need to lock the super, get_super holds the | |
1750 | * read mutex so the filesystem cannot go away | |
1751 | * under us (->put_super runs with the write lock | |
1752 | * hold). | |
1753 | */ | |
1754 | shrink_dcache_sb(sb); | |
1755 | res = invalidate_inodes(sb); | |
1756 | drop_super(sb); | |
1757 | } | |
f98393a6 | 1758 | invalidate_bdev(bdev); |
b71e8a4c DH |
1759 | return res; |
1760 | } | |
1761 | EXPORT_SYMBOL(__invalidate_device); |