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 LT |
14 | #include <linux/smp_lock.h> |
15 | #include <linux/highmem.h> | |
16 | #include <linux/blkdev.h> | |
17 | #include <linux/module.h> | |
18 | #include <linux/blkpg.h> | |
19 | #include <linux/buffer_head.h> | |
20 | #include <linux/mpage.h> | |
21 | #include <linux/mount.h> | |
22 | #include <linux/uio.h> | |
23 | #include <linux/namei.h> | |
24 | #include <asm/uaccess.h> | |
25 | ||
26 | struct bdev_inode { | |
27 | struct block_device bdev; | |
28 | struct inode vfs_inode; | |
29 | }; | |
30 | ||
31 | static inline struct bdev_inode *BDEV_I(struct inode *inode) | |
32 | { | |
33 | return container_of(inode, struct bdev_inode, vfs_inode); | |
34 | } | |
35 | ||
36 | inline struct block_device *I_BDEV(struct inode *inode) | |
37 | { | |
38 | return &BDEV_I(inode)->bdev; | |
39 | } | |
40 | ||
41 | EXPORT_SYMBOL(I_BDEV); | |
42 | ||
43 | static sector_t max_block(struct block_device *bdev) | |
44 | { | |
45 | sector_t retval = ~((sector_t)0); | |
46 | loff_t sz = i_size_read(bdev->bd_inode); | |
47 | ||
48 | if (sz) { | |
49 | unsigned int size = block_size(bdev); | |
50 | unsigned int sizebits = blksize_bits(size); | |
51 | retval = (sz >> sizebits); | |
52 | } | |
53 | return retval; | |
54 | } | |
55 | ||
56 | /* Kill _all_ buffers, dirty or not.. */ | |
57 | static void kill_bdev(struct block_device *bdev) | |
58 | { | |
59 | invalidate_bdev(bdev, 1); | |
60 | truncate_inode_pages(bdev->bd_inode->i_mapping, 0); | |
61 | } | |
62 | ||
63 | int set_blocksize(struct block_device *bdev, int size) | |
64 | { | |
65 | /* Size must be a power of two, and between 512 and PAGE_SIZE */ | |
66 | if (size > PAGE_SIZE || size < 512 || (size & (size-1))) | |
67 | return -EINVAL; | |
68 | ||
69 | /* Size cannot be smaller than the size supported by the device */ | |
70 | if (size < bdev_hardsect_size(bdev)) | |
71 | return -EINVAL; | |
72 | ||
73 | /* Don't change the size if it is same as current */ | |
74 | if (bdev->bd_block_size != size) { | |
75 | sync_blockdev(bdev); | |
76 | bdev->bd_block_size = size; | |
77 | bdev->bd_inode->i_blkbits = blksize_bits(size); | |
78 | kill_bdev(bdev); | |
79 | } | |
80 | return 0; | |
81 | } | |
82 | ||
83 | EXPORT_SYMBOL(set_blocksize); | |
84 | ||
85 | int sb_set_blocksize(struct super_block *sb, int size) | |
86 | { | |
1da177e4 LT |
87 | if (set_blocksize(sb->s_bdev, size)) |
88 | return 0; | |
89 | /* If we get here, we know size is power of two | |
90 | * and it's value is between 512 and PAGE_SIZE */ | |
91 | sb->s_blocksize = size; | |
38885bd4 | 92 | sb->s_blocksize_bits = blksize_bits(size); |
1da177e4 LT |
93 | return sb->s_blocksize; |
94 | } | |
95 | ||
96 | EXPORT_SYMBOL(sb_set_blocksize); | |
97 | ||
98 | int sb_min_blocksize(struct super_block *sb, int size) | |
99 | { | |
100 | int minsize = bdev_hardsect_size(sb->s_bdev); | |
101 | if (size < minsize) | |
102 | size = minsize; | |
103 | return sb_set_blocksize(sb, size); | |
104 | } | |
105 | ||
106 | EXPORT_SYMBOL(sb_min_blocksize); | |
107 | ||
108 | static int | |
109 | blkdev_get_block(struct inode *inode, sector_t iblock, | |
110 | struct buffer_head *bh, int create) | |
111 | { | |
112 | if (iblock >= max_block(I_BDEV(inode))) { | |
113 | if (create) | |
114 | return -EIO; | |
115 | ||
116 | /* | |
117 | * for reads, we're just trying to fill a partial page. | |
118 | * return a hole, they will have to call get_block again | |
119 | * before they can fill it, and they will get -EIO at that | |
120 | * time | |
121 | */ | |
122 | return 0; | |
123 | } | |
124 | bh->b_bdev = I_BDEV(inode); | |
125 | bh->b_blocknr = iblock; | |
126 | set_buffer_mapped(bh); | |
127 | return 0; | |
128 | } | |
129 | ||
130 | static int | |
131 | blkdev_get_blocks(struct inode *inode, sector_t iblock, | |
1d8fa7a2 | 132 | struct buffer_head *bh, int create) |
1da177e4 LT |
133 | { |
134 | sector_t end_block = max_block(I_BDEV(inode)); | |
1d8fa7a2 | 135 | unsigned long max_blocks = bh->b_size >> inode->i_blkbits; |
1da177e4 LT |
136 | |
137 | if ((iblock + max_blocks) > end_block) { | |
138 | max_blocks = end_block - iblock; | |
139 | if ((long)max_blocks <= 0) { | |
140 | if (create) | |
141 | return -EIO; /* write fully beyond EOF */ | |
142 | /* | |
143 | * It is a read which is fully beyond EOF. We return | |
144 | * a !buffer_mapped buffer | |
145 | */ | |
146 | max_blocks = 0; | |
147 | } | |
148 | } | |
149 | ||
150 | bh->b_bdev = I_BDEV(inode); | |
151 | bh->b_blocknr = iblock; | |
152 | bh->b_size = max_blocks << inode->i_blkbits; | |
153 | if (max_blocks) | |
154 | set_buffer_mapped(bh); | |
155 | return 0; | |
156 | } | |
157 | ||
158 | static ssize_t | |
159 | blkdev_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, | |
160 | loff_t offset, unsigned long nr_segs) | |
161 | { | |
162 | struct file *file = iocb->ki_filp; | |
163 | struct inode *inode = file->f_mapping->host; | |
164 | ||
165 | return blockdev_direct_IO_no_locking(rw, iocb, inode, I_BDEV(inode), | |
166 | iov, offset, nr_segs, blkdev_get_blocks, NULL); | |
167 | } | |
168 | ||
169 | static int blkdev_writepage(struct page *page, struct writeback_control *wbc) | |
170 | { | |
171 | return block_write_full_page(page, blkdev_get_block, wbc); | |
172 | } | |
173 | ||
174 | static int blkdev_readpage(struct file * file, struct page * page) | |
175 | { | |
176 | return block_read_full_page(page, blkdev_get_block); | |
177 | } | |
178 | ||
179 | static int blkdev_prepare_write(struct file *file, struct page *page, unsigned from, unsigned to) | |
180 | { | |
181 | return block_prepare_write(page, from, to, blkdev_get_block); | |
182 | } | |
183 | ||
184 | static int blkdev_commit_write(struct file *file, struct page *page, unsigned from, unsigned to) | |
185 | { | |
186 | return block_commit_write(page, from, to); | |
187 | } | |
188 | ||
189 | /* | |
190 | * private llseek: | |
191 | * for a block special file file->f_dentry->d_inode->i_size is zero | |
192 | * so we compute the size by hand (just as in block_read/write above) | |
193 | */ | |
194 | static loff_t block_llseek(struct file *file, loff_t offset, int origin) | |
195 | { | |
196 | struct inode *bd_inode = file->f_mapping->host; | |
197 | loff_t size; | |
198 | loff_t retval; | |
199 | ||
1b1dcc1b | 200 | mutex_lock(&bd_inode->i_mutex); |
1da177e4 LT |
201 | size = i_size_read(bd_inode); |
202 | ||
203 | switch (origin) { | |
204 | case 2: | |
205 | offset += size; | |
206 | break; | |
207 | case 1: | |
208 | offset += file->f_pos; | |
209 | } | |
210 | retval = -EINVAL; | |
211 | if (offset >= 0 && offset <= size) { | |
212 | if (offset != file->f_pos) { | |
213 | file->f_pos = offset; | |
214 | } | |
215 | retval = offset; | |
216 | } | |
1b1dcc1b | 217 | mutex_unlock(&bd_inode->i_mutex); |
1da177e4 LT |
218 | return retval; |
219 | } | |
220 | ||
221 | /* | |
222 | * Filp is never NULL; the only case when ->fsync() is called with | |
223 | * NULL first argument is nfsd_sync_dir() and that's not a directory. | |
224 | */ | |
225 | ||
226 | static int block_fsync(struct file *filp, struct dentry *dentry, int datasync) | |
227 | { | |
228 | return sync_blockdev(I_BDEV(filp->f_mapping->host)); | |
229 | } | |
230 | ||
231 | /* | |
232 | * pseudo-fs | |
233 | */ | |
234 | ||
235 | static __cacheline_aligned_in_smp DEFINE_SPINLOCK(bdev_lock); | |
fa3536cc | 236 | static kmem_cache_t * bdev_cachep __read_mostly; |
1da177e4 LT |
237 | |
238 | static struct inode *bdev_alloc_inode(struct super_block *sb) | |
239 | { | |
240 | struct bdev_inode *ei = kmem_cache_alloc(bdev_cachep, SLAB_KERNEL); | |
241 | if (!ei) | |
242 | return NULL; | |
243 | return &ei->vfs_inode; | |
244 | } | |
245 | ||
246 | static void bdev_destroy_inode(struct inode *inode) | |
247 | { | |
248 | struct bdev_inode *bdi = BDEV_I(inode); | |
249 | ||
250 | bdi->bdev.bd_inode_backing_dev_info = NULL; | |
251 | kmem_cache_free(bdev_cachep, bdi); | |
252 | } | |
253 | ||
254 | static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags) | |
255 | { | |
256 | struct bdev_inode *ei = (struct bdev_inode *) foo; | |
257 | struct block_device *bdev = &ei->bdev; | |
258 | ||
259 | if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) == | |
260 | SLAB_CTOR_CONSTRUCTOR) | |
261 | { | |
262 | memset(bdev, 0, sizeof(*bdev)); | |
c039e313 AV |
263 | mutex_init(&bdev->bd_mutex); |
264 | mutex_init(&bdev->bd_mount_mutex); | |
1da177e4 LT |
265 | INIT_LIST_HEAD(&bdev->bd_inodes); |
266 | INIT_LIST_HEAD(&bdev->bd_list); | |
641dc636 JN |
267 | #ifdef CONFIG_SYSFS |
268 | INIT_LIST_HEAD(&bdev->bd_holder_list); | |
269 | #endif | |
1da177e4 LT |
270 | inode_init_once(&ei->vfs_inode); |
271 | } | |
272 | } | |
273 | ||
274 | static inline void __bd_forget(struct inode *inode) | |
275 | { | |
276 | list_del_init(&inode->i_devices); | |
277 | inode->i_bdev = NULL; | |
278 | inode->i_mapping = &inode->i_data; | |
279 | } | |
280 | ||
281 | static void bdev_clear_inode(struct inode *inode) | |
282 | { | |
283 | struct block_device *bdev = &BDEV_I(inode)->bdev; | |
284 | struct list_head *p; | |
285 | spin_lock(&bdev_lock); | |
286 | while ( (p = bdev->bd_inodes.next) != &bdev->bd_inodes ) { | |
287 | __bd_forget(list_entry(p, struct inode, i_devices)); | |
288 | } | |
289 | list_del_init(&bdev->bd_list); | |
290 | spin_unlock(&bdev_lock); | |
291 | } | |
292 | ||
293 | static struct super_operations bdev_sops = { | |
294 | .statfs = simple_statfs, | |
295 | .alloc_inode = bdev_alloc_inode, | |
296 | .destroy_inode = bdev_destroy_inode, | |
297 | .drop_inode = generic_delete_inode, | |
298 | .clear_inode = bdev_clear_inode, | |
299 | }; | |
300 | ||
454e2398 DH |
301 | static int bd_get_sb(struct file_system_type *fs_type, |
302 | int flags, const char *dev_name, void *data, struct vfsmount *mnt) | |
1da177e4 | 303 | { |
454e2398 | 304 | return get_sb_pseudo(fs_type, "bdev:", &bdev_sops, 0x62646576, mnt); |
1da177e4 LT |
305 | } |
306 | ||
307 | static struct file_system_type bd_type = { | |
308 | .name = "bdev", | |
309 | .get_sb = bd_get_sb, | |
310 | .kill_sb = kill_anon_super, | |
311 | }; | |
312 | ||
fa3536cc | 313 | static struct vfsmount *bd_mnt __read_mostly; |
1da177e4 LT |
314 | struct super_block *blockdev_superblock; |
315 | ||
316 | void __init bdev_cache_init(void) | |
317 | { | |
318 | int err; | |
319 | bdev_cachep = kmem_cache_create("bdev_cache", sizeof(struct bdev_inode), | |
fffb60f9 PJ |
320 | 0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| |
321 | SLAB_MEM_SPREAD|SLAB_PANIC), | |
1da177e4 LT |
322 | init_once, NULL); |
323 | err = register_filesystem(&bd_type); | |
324 | if (err) | |
325 | panic("Cannot register bdev pseudo-fs"); | |
326 | bd_mnt = kern_mount(&bd_type); | |
327 | err = PTR_ERR(bd_mnt); | |
328 | if (IS_ERR(bd_mnt)) | |
329 | panic("Cannot create bdev pseudo-fs"); | |
330 | blockdev_superblock = bd_mnt->mnt_sb; /* For writeback */ | |
331 | } | |
332 | ||
333 | /* | |
334 | * Most likely _very_ bad one - but then it's hardly critical for small | |
335 | * /dev and can be fixed when somebody will need really large one. | |
336 | * Keep in mind that it will be fed through icache hash function too. | |
337 | */ | |
338 | static inline unsigned long hash(dev_t dev) | |
339 | { | |
340 | return MAJOR(dev)+MINOR(dev); | |
341 | } | |
342 | ||
343 | static int bdev_test(struct inode *inode, void *data) | |
344 | { | |
345 | return BDEV_I(inode)->bdev.bd_dev == *(dev_t *)data; | |
346 | } | |
347 | ||
348 | static int bdev_set(struct inode *inode, void *data) | |
349 | { | |
350 | BDEV_I(inode)->bdev.bd_dev = *(dev_t *)data; | |
351 | return 0; | |
352 | } | |
353 | ||
354 | static LIST_HEAD(all_bdevs); | |
355 | ||
356 | struct block_device *bdget(dev_t dev) | |
357 | { | |
358 | struct block_device *bdev; | |
359 | struct inode *inode; | |
360 | ||
361 | inode = iget5_locked(bd_mnt->mnt_sb, hash(dev), | |
362 | bdev_test, bdev_set, &dev); | |
363 | ||
364 | if (!inode) | |
365 | return NULL; | |
366 | ||
367 | bdev = &BDEV_I(inode)->bdev; | |
368 | ||
369 | if (inode->i_state & I_NEW) { | |
370 | bdev->bd_contains = NULL; | |
371 | bdev->bd_inode = inode; | |
372 | bdev->bd_block_size = (1 << inode->i_blkbits); | |
373 | bdev->bd_part_count = 0; | |
374 | bdev->bd_invalidated = 0; | |
375 | inode->i_mode = S_IFBLK; | |
376 | inode->i_rdev = dev; | |
377 | inode->i_bdev = bdev; | |
378 | inode->i_data.a_ops = &def_blk_aops; | |
379 | mapping_set_gfp_mask(&inode->i_data, GFP_USER); | |
380 | inode->i_data.backing_dev_info = &default_backing_dev_info; | |
381 | spin_lock(&bdev_lock); | |
382 | list_add(&bdev->bd_list, &all_bdevs); | |
383 | spin_unlock(&bdev_lock); | |
384 | unlock_new_inode(inode); | |
385 | } | |
386 | return bdev; | |
387 | } | |
388 | ||
389 | EXPORT_SYMBOL(bdget); | |
390 | ||
391 | long nr_blockdev_pages(void) | |
392 | { | |
393 | struct list_head *p; | |
394 | long ret = 0; | |
395 | spin_lock(&bdev_lock); | |
396 | list_for_each(p, &all_bdevs) { | |
397 | struct block_device *bdev; | |
398 | bdev = list_entry(p, struct block_device, bd_list); | |
399 | ret += bdev->bd_inode->i_mapping->nrpages; | |
400 | } | |
401 | spin_unlock(&bdev_lock); | |
402 | return ret; | |
403 | } | |
404 | ||
405 | void bdput(struct block_device *bdev) | |
406 | { | |
407 | iput(bdev->bd_inode); | |
408 | } | |
409 | ||
410 | EXPORT_SYMBOL(bdput); | |
411 | ||
412 | static struct block_device *bd_acquire(struct inode *inode) | |
413 | { | |
414 | struct block_device *bdev; | |
09d967c6 | 415 | |
1da177e4 LT |
416 | spin_lock(&bdev_lock); |
417 | bdev = inode->i_bdev; | |
09d967c6 OH |
418 | if (bdev) { |
419 | atomic_inc(&bdev->bd_inode->i_count); | |
1da177e4 LT |
420 | spin_unlock(&bdev_lock); |
421 | return bdev; | |
422 | } | |
423 | spin_unlock(&bdev_lock); | |
09d967c6 | 424 | |
1da177e4 LT |
425 | bdev = bdget(inode->i_rdev); |
426 | if (bdev) { | |
427 | spin_lock(&bdev_lock); | |
09d967c6 OH |
428 | if (!inode->i_bdev) { |
429 | /* | |
430 | * We take an additional bd_inode->i_count for inode, | |
431 | * and it's released in clear_inode() of inode. | |
432 | * So, we can access it via ->i_mapping always | |
433 | * without igrab(). | |
434 | */ | |
435 | atomic_inc(&bdev->bd_inode->i_count); | |
436 | inode->i_bdev = bdev; | |
437 | inode->i_mapping = bdev->bd_inode->i_mapping; | |
438 | list_add(&inode->i_devices, &bdev->bd_inodes); | |
439 | } | |
1da177e4 LT |
440 | spin_unlock(&bdev_lock); |
441 | } | |
442 | return bdev; | |
443 | } | |
444 | ||
445 | /* Call when you free inode */ | |
446 | ||
447 | void bd_forget(struct inode *inode) | |
448 | { | |
09d967c6 OH |
449 | struct block_device *bdev = NULL; |
450 | ||
1da177e4 | 451 | spin_lock(&bdev_lock); |
09d967c6 OH |
452 | if (inode->i_bdev) { |
453 | if (inode->i_sb != blockdev_superblock) | |
454 | bdev = inode->i_bdev; | |
1da177e4 | 455 | __bd_forget(inode); |
09d967c6 | 456 | } |
1da177e4 | 457 | spin_unlock(&bdev_lock); |
09d967c6 OH |
458 | |
459 | if (bdev) | |
460 | iput(bdev->bd_inode); | |
1da177e4 LT |
461 | } |
462 | ||
463 | int bd_claim(struct block_device *bdev, void *holder) | |
464 | { | |
465 | int res; | |
466 | spin_lock(&bdev_lock); | |
467 | ||
468 | /* first decide result */ | |
469 | if (bdev->bd_holder == holder) | |
470 | res = 0; /* already a holder */ | |
471 | else if (bdev->bd_holder != NULL) | |
472 | res = -EBUSY; /* held by someone else */ | |
473 | else if (bdev->bd_contains == bdev) | |
474 | res = 0; /* is a whole device which isn't held */ | |
475 | ||
476 | else if (bdev->bd_contains->bd_holder == bd_claim) | |
477 | res = 0; /* is a partition of a device that is being partitioned */ | |
478 | else if (bdev->bd_contains->bd_holder != NULL) | |
479 | res = -EBUSY; /* is a partition of a held device */ | |
480 | else | |
481 | res = 0; /* is a partition of an un-held device */ | |
482 | ||
483 | /* now impose change */ | |
484 | if (res==0) { | |
485 | /* note that for a whole device bd_holders | |
486 | * will be incremented twice, and bd_holder will | |
487 | * be set to bd_claim before being set to holder | |
488 | */ | |
489 | bdev->bd_contains->bd_holders ++; | |
490 | bdev->bd_contains->bd_holder = bd_claim; | |
491 | bdev->bd_holders++; | |
492 | bdev->bd_holder = holder; | |
493 | } | |
494 | spin_unlock(&bdev_lock); | |
495 | return res; | |
496 | } | |
497 | ||
498 | EXPORT_SYMBOL(bd_claim); | |
499 | ||
500 | void bd_release(struct block_device *bdev) | |
501 | { | |
502 | spin_lock(&bdev_lock); | |
503 | if (!--bdev->bd_contains->bd_holders) | |
504 | bdev->bd_contains->bd_holder = NULL; | |
505 | if (!--bdev->bd_holders) | |
506 | bdev->bd_holder = NULL; | |
507 | spin_unlock(&bdev_lock); | |
508 | } | |
509 | ||
510 | EXPORT_SYMBOL(bd_release); | |
511 | ||
641dc636 JN |
512 | #ifdef CONFIG_SYSFS |
513 | /* | |
514 | * Functions for bd_claim_by_kobject / bd_release_from_kobject | |
515 | * | |
516 | * If a kobject is passed to bd_claim_by_kobject() | |
517 | * and the kobject has a parent directory, | |
518 | * following symlinks are created: | |
519 | * o from the kobject to the claimed bdev | |
520 | * o from "holders" directory of the bdev to the parent of the kobject | |
521 | * bd_release_from_kobject() removes these symlinks. | |
522 | * | |
523 | * Example: | |
524 | * If /dev/dm-0 maps to /dev/sda, kobject corresponding to | |
525 | * /sys/block/dm-0/slaves is passed to bd_claim_by_kobject(), then: | |
526 | * /sys/block/dm-0/slaves/sda --> /sys/block/sda | |
527 | * /sys/block/sda/holders/dm-0 --> /sys/block/dm-0 | |
528 | */ | |
529 | ||
530 | static struct kobject *bdev_get_kobj(struct block_device *bdev) | |
531 | { | |
532 | if (bdev->bd_contains != bdev) | |
533 | return kobject_get(&bdev->bd_part->kobj); | |
534 | else | |
535 | return kobject_get(&bdev->bd_disk->kobj); | |
536 | } | |
537 | ||
538 | static struct kobject *bdev_get_holder(struct block_device *bdev) | |
539 | { | |
540 | if (bdev->bd_contains != bdev) | |
541 | return kobject_get(bdev->bd_part->holder_dir); | |
542 | else | |
543 | return kobject_get(bdev->bd_disk->holder_dir); | |
544 | } | |
545 | ||
546 | static void add_symlink(struct kobject *from, struct kobject *to) | |
547 | { | |
548 | if (!from || !to) | |
549 | return; | |
550 | sysfs_create_link(from, to, kobject_name(to)); | |
551 | } | |
552 | ||
553 | static void del_symlink(struct kobject *from, struct kobject *to) | |
554 | { | |
555 | if (!from || !to) | |
556 | return; | |
557 | sysfs_remove_link(from, kobject_name(to)); | |
558 | } | |
559 | ||
560 | /* | |
561 | * 'struct bd_holder' contains pointers to kobjects symlinked by | |
562 | * bd_claim_by_kobject. | |
563 | * It's connected to bd_holder_list which is protected by bdev->bd_sem. | |
564 | */ | |
565 | struct bd_holder { | |
566 | struct list_head list; /* chain of holders of the bdev */ | |
567 | int count; /* references from the holder */ | |
568 | struct kobject *sdir; /* holder object, e.g. "/block/dm-0/slaves" */ | |
569 | struct kobject *hdev; /* e.g. "/block/dm-0" */ | |
570 | struct kobject *hdir; /* e.g. "/block/sda/holders" */ | |
571 | struct kobject *sdev; /* e.g. "/block/sda" */ | |
572 | }; | |
573 | ||
574 | /* | |
575 | * Get references of related kobjects at once. | |
576 | * Returns 1 on success. 0 on failure. | |
577 | * | |
578 | * Should call bd_holder_release_dirs() after successful use. | |
579 | */ | |
580 | static int bd_holder_grab_dirs(struct block_device *bdev, | |
581 | struct bd_holder *bo) | |
582 | { | |
583 | if (!bdev || !bo) | |
584 | return 0; | |
585 | ||
586 | bo->sdir = kobject_get(bo->sdir); | |
587 | if (!bo->sdir) | |
588 | return 0; | |
589 | ||
590 | bo->hdev = kobject_get(bo->sdir->parent); | |
591 | if (!bo->hdev) | |
592 | goto fail_put_sdir; | |
593 | ||
594 | bo->sdev = bdev_get_kobj(bdev); | |
595 | if (!bo->sdev) | |
596 | goto fail_put_hdev; | |
597 | ||
598 | bo->hdir = bdev_get_holder(bdev); | |
599 | if (!bo->hdir) | |
600 | goto fail_put_sdev; | |
601 | ||
602 | return 1; | |
603 | ||
604 | fail_put_sdev: | |
605 | kobject_put(bo->sdev); | |
606 | fail_put_hdev: | |
607 | kobject_put(bo->hdev); | |
608 | fail_put_sdir: | |
609 | kobject_put(bo->sdir); | |
610 | ||
611 | return 0; | |
612 | } | |
613 | ||
614 | /* Put references of related kobjects at once. */ | |
615 | static void bd_holder_release_dirs(struct bd_holder *bo) | |
616 | { | |
617 | kobject_put(bo->hdir); | |
618 | kobject_put(bo->sdev); | |
619 | kobject_put(bo->hdev); | |
620 | kobject_put(bo->sdir); | |
621 | } | |
622 | ||
623 | static struct bd_holder *alloc_bd_holder(struct kobject *kobj) | |
624 | { | |
625 | struct bd_holder *bo; | |
626 | ||
627 | bo = kzalloc(sizeof(*bo), GFP_KERNEL); | |
628 | if (!bo) | |
629 | return NULL; | |
630 | ||
631 | bo->count = 1; | |
632 | bo->sdir = kobj; | |
633 | ||
634 | return bo; | |
635 | } | |
636 | ||
637 | static void free_bd_holder(struct bd_holder *bo) | |
638 | { | |
639 | kfree(bo); | |
640 | } | |
641 | ||
642 | /** | |
643 | * add_bd_holder - create sysfs symlinks for bd_claim() relationship | |
644 | * | |
645 | * @bdev: block device to be bd_claimed | |
646 | * @bo: preallocated and initialized by alloc_bd_holder() | |
647 | * | |
648 | * If there is no matching entry with @bo in @bdev->bd_holder_list, | |
649 | * add @bo to the list, create symlinks. | |
650 | * | |
651 | * Returns 1 if @bo was added to the list. | |
652 | * Returns 0 if @bo wasn't used by any reason and should be freed. | |
653 | */ | |
654 | static int add_bd_holder(struct block_device *bdev, struct bd_holder *bo) | |
655 | { | |
656 | struct bd_holder *tmp; | |
657 | ||
658 | if (!bo) | |
659 | return 0; | |
660 | ||
661 | list_for_each_entry(tmp, &bdev->bd_holder_list, list) { | |
662 | if (tmp->sdir == bo->sdir) { | |
663 | tmp->count++; | |
664 | return 0; | |
665 | } | |
666 | } | |
667 | ||
668 | if (!bd_holder_grab_dirs(bdev, bo)) | |
669 | return 0; | |
670 | ||
671 | add_symlink(bo->sdir, bo->sdev); | |
672 | add_symlink(bo->hdir, bo->hdev); | |
673 | list_add_tail(&bo->list, &bdev->bd_holder_list); | |
674 | return 1; | |
675 | } | |
676 | ||
677 | /** | |
678 | * del_bd_holder - delete sysfs symlinks for bd_claim() relationship | |
679 | * | |
680 | * @bdev: block device to be bd_claimed | |
681 | * @kobj: holder's kobject | |
682 | * | |
683 | * If there is matching entry with @kobj in @bdev->bd_holder_list | |
684 | * and no other bd_claim() from the same kobject, | |
685 | * remove the struct bd_holder from the list, delete symlinks for it. | |
686 | * | |
687 | * Returns a pointer to the struct bd_holder when it's removed from the list | |
688 | * and ready to be freed. | |
689 | * Returns NULL if matching claim isn't found or there is other bd_claim() | |
690 | * by the same kobject. | |
691 | */ | |
692 | static struct bd_holder *del_bd_holder(struct block_device *bdev, | |
693 | struct kobject *kobj) | |
694 | { | |
695 | struct bd_holder *bo; | |
696 | ||
697 | list_for_each_entry(bo, &bdev->bd_holder_list, list) { | |
698 | if (bo->sdir == kobj) { | |
699 | bo->count--; | |
700 | BUG_ON(bo->count < 0); | |
701 | if (!bo->count) { | |
702 | list_del(&bo->list); | |
703 | del_symlink(bo->sdir, bo->sdev); | |
704 | del_symlink(bo->hdir, bo->hdev); | |
705 | bd_holder_release_dirs(bo); | |
706 | return bo; | |
707 | } | |
708 | break; | |
709 | } | |
710 | } | |
711 | ||
712 | return NULL; | |
713 | } | |
714 | ||
715 | /** | |
716 | * bd_claim_by_kobject - bd_claim() with additional kobject signature | |
717 | * | |
718 | * @bdev: block device to be claimed | |
719 | * @holder: holder's signature | |
720 | * @kobj: holder's kobject | |
721 | * | |
722 | * Do bd_claim() and if it succeeds, create sysfs symlinks between | |
723 | * the bdev and the holder's kobject. | |
724 | * Use bd_release_from_kobject() when relesing the claimed bdev. | |
725 | * | |
726 | * Returns 0 on success. (same as bd_claim()) | |
727 | * Returns errno on failure. | |
728 | */ | |
729 | static int bd_claim_by_kobject(struct block_device *bdev, void *holder, | |
730 | struct kobject *kobj) | |
731 | { | |
732 | int res; | |
733 | struct bd_holder *bo; | |
734 | ||
735 | if (!kobj) | |
736 | return -EINVAL; | |
737 | ||
738 | bo = alloc_bd_holder(kobj); | |
739 | if (!bo) | |
740 | return -ENOMEM; | |
741 | ||
b4cf1b72 | 742 | mutex_lock(&bdev->bd_mutex); |
641dc636 JN |
743 | res = bd_claim(bdev, holder); |
744 | if (res || !add_bd_holder(bdev, bo)) | |
745 | free_bd_holder(bo); | |
b4cf1b72 | 746 | mutex_unlock(&bdev->bd_mutex); |
641dc636 JN |
747 | |
748 | return res; | |
749 | } | |
750 | ||
751 | /** | |
752 | * bd_release_from_kobject - bd_release() with additional kobject signature | |
753 | * | |
754 | * @bdev: block device to be released | |
755 | * @kobj: holder's kobject | |
756 | * | |
757 | * Do bd_release() and remove sysfs symlinks created by bd_claim_by_kobject(). | |
758 | */ | |
759 | static void bd_release_from_kobject(struct block_device *bdev, | |
760 | struct kobject *kobj) | |
761 | { | |
762 | struct bd_holder *bo; | |
763 | ||
764 | if (!kobj) | |
765 | return; | |
766 | ||
b4cf1b72 | 767 | mutex_lock(&bdev->bd_mutex); |
641dc636 JN |
768 | bd_release(bdev); |
769 | if ((bo = del_bd_holder(bdev, kobj))) | |
770 | free_bd_holder(bo); | |
b4cf1b72 | 771 | mutex_unlock(&bdev->bd_mutex); |
641dc636 JN |
772 | } |
773 | ||
774 | /** | |
775 | * bd_claim_by_disk - wrapper function for bd_claim_by_kobject() | |
776 | * | |
777 | * @bdev: block device to be claimed | |
778 | * @holder: holder's signature | |
779 | * @disk: holder's gendisk | |
780 | * | |
781 | * Call bd_claim_by_kobject() with getting @disk->slave_dir. | |
782 | */ | |
783 | int bd_claim_by_disk(struct block_device *bdev, void *holder, | |
784 | struct gendisk *disk) | |
785 | { | |
786 | return bd_claim_by_kobject(bdev, holder, kobject_get(disk->slave_dir)); | |
787 | } | |
788 | EXPORT_SYMBOL_GPL(bd_claim_by_disk); | |
789 | ||
790 | /** | |
791 | * bd_release_from_disk - wrapper function for bd_release_from_kobject() | |
792 | * | |
793 | * @bdev: block device to be claimed | |
794 | * @disk: holder's gendisk | |
795 | * | |
796 | * Call bd_release_from_kobject() and put @disk->slave_dir. | |
797 | */ | |
798 | void bd_release_from_disk(struct block_device *bdev, struct gendisk *disk) | |
799 | { | |
800 | bd_release_from_kobject(bdev, disk->slave_dir); | |
801 | kobject_put(disk->slave_dir); | |
802 | } | |
803 | EXPORT_SYMBOL_GPL(bd_release_from_disk); | |
804 | #endif | |
805 | ||
1da177e4 LT |
806 | /* |
807 | * Tries to open block device by device number. Use it ONLY if you | |
808 | * really do not have anything better - i.e. when you are behind a | |
809 | * truly sucky interface and all you are given is a device number. _Never_ | |
810 | * to be used for internal purposes. If you ever need it - reconsider | |
811 | * your API. | |
812 | */ | |
813 | struct block_device *open_by_devnum(dev_t dev, unsigned mode) | |
814 | { | |
815 | struct block_device *bdev = bdget(dev); | |
816 | int err = -ENOMEM; | |
817 | int flags = mode & FMODE_WRITE ? O_RDWR : O_RDONLY; | |
818 | if (bdev) | |
819 | err = blkdev_get(bdev, mode, flags); | |
820 | return err ? ERR_PTR(err) : bdev; | |
821 | } | |
822 | ||
823 | EXPORT_SYMBOL(open_by_devnum); | |
824 | ||
825 | /* | |
826 | * This routine checks whether a removable media has been changed, | |
827 | * and invalidates all buffer-cache-entries in that case. This | |
828 | * is a relatively slow routine, so we have to try to minimize using | |
829 | * it. Thus it is called only upon a 'mount' or 'open'. This | |
830 | * is the best way of combining speed and utility, I think. | |
831 | * People changing diskettes in the middle of an operation deserve | |
832 | * to lose :-) | |
833 | */ | |
834 | int check_disk_change(struct block_device *bdev) | |
835 | { | |
836 | struct gendisk *disk = bdev->bd_disk; | |
837 | struct block_device_operations * bdops = disk->fops; | |
838 | ||
839 | if (!bdops->media_changed) | |
840 | return 0; | |
841 | if (!bdops->media_changed(bdev->bd_disk)) | |
842 | return 0; | |
843 | ||
2ef41634 | 844 | if (__invalidate_device(bdev)) |
1da177e4 LT |
845 | printk("VFS: busy inodes on changed media.\n"); |
846 | ||
847 | if (bdops->revalidate_disk) | |
848 | bdops->revalidate_disk(bdev->bd_disk); | |
849 | if (bdev->bd_disk->minors > 1) | |
850 | bdev->bd_invalidated = 1; | |
851 | return 1; | |
852 | } | |
853 | ||
854 | EXPORT_SYMBOL(check_disk_change); | |
855 | ||
856 | void bd_set_size(struct block_device *bdev, loff_t size) | |
857 | { | |
858 | unsigned bsize = bdev_hardsect_size(bdev); | |
859 | ||
860 | bdev->bd_inode->i_size = size; | |
861 | while (bsize < PAGE_CACHE_SIZE) { | |
862 | if (size & bsize) | |
863 | break; | |
864 | bsize <<= 1; | |
865 | } | |
866 | bdev->bd_block_size = bsize; | |
867 | bdev->bd_inode->i_blkbits = blksize_bits(bsize); | |
868 | } | |
869 | EXPORT_SYMBOL(bd_set_size); | |
870 | ||
871 | static int do_open(struct block_device *bdev, struct file *file) | |
872 | { | |
873 | struct module *owner = NULL; | |
874 | struct gendisk *disk; | |
875 | int ret = -ENXIO; | |
876 | int part; | |
877 | ||
878 | file->f_mapping = bdev->bd_inode->i_mapping; | |
879 | lock_kernel(); | |
880 | disk = get_gendisk(bdev->bd_dev, &part); | |
881 | if (!disk) { | |
882 | unlock_kernel(); | |
883 | bdput(bdev); | |
884 | return ret; | |
885 | } | |
886 | owner = disk->fops->owner; | |
887 | ||
c039e313 | 888 | mutex_lock(&bdev->bd_mutex); |
1da177e4 LT |
889 | if (!bdev->bd_openers) { |
890 | bdev->bd_disk = disk; | |
891 | bdev->bd_contains = bdev; | |
892 | if (!part) { | |
893 | struct backing_dev_info *bdi; | |
894 | if (disk->fops->open) { | |
895 | ret = disk->fops->open(bdev->bd_inode, file); | |
896 | if (ret) | |
897 | goto out_first; | |
898 | } | |
899 | if (!bdev->bd_openers) { | |
900 | bd_set_size(bdev,(loff_t)get_capacity(disk)<<9); | |
901 | bdi = blk_get_backing_dev_info(bdev); | |
902 | if (bdi == NULL) | |
903 | bdi = &default_backing_dev_info; | |
904 | bdev->bd_inode->i_data.backing_dev_info = bdi; | |
905 | } | |
906 | if (bdev->bd_invalidated) | |
907 | rescan_partitions(disk, bdev); | |
908 | } else { | |
909 | struct hd_struct *p; | |
910 | struct block_device *whole; | |
911 | whole = bdget_disk(disk, 0); | |
912 | ret = -ENOMEM; | |
913 | if (!whole) | |
914 | goto out_first; | |
915 | ret = blkdev_get(whole, file->f_mode, file->f_flags); | |
916 | if (ret) | |
917 | goto out_first; | |
918 | bdev->bd_contains = whole; | |
c039e313 | 919 | mutex_lock(&whole->bd_mutex); |
1da177e4 LT |
920 | whole->bd_part_count++; |
921 | p = disk->part[part - 1]; | |
922 | bdev->bd_inode->i_data.backing_dev_info = | |
923 | whole->bd_inode->i_data.backing_dev_info; | |
924 | if (!(disk->flags & GENHD_FL_UP) || !p || !p->nr_sects) { | |
925 | whole->bd_part_count--; | |
c039e313 | 926 | mutex_unlock(&whole->bd_mutex); |
1da177e4 LT |
927 | ret = -ENXIO; |
928 | goto out_first; | |
929 | } | |
930 | kobject_get(&p->kobj); | |
931 | bdev->bd_part = p; | |
932 | bd_set_size(bdev, (loff_t) p->nr_sects << 9); | |
c039e313 | 933 | mutex_unlock(&whole->bd_mutex); |
1da177e4 LT |
934 | } |
935 | } else { | |
936 | put_disk(disk); | |
937 | module_put(owner); | |
938 | if (bdev->bd_contains == bdev) { | |
939 | if (bdev->bd_disk->fops->open) { | |
940 | ret = bdev->bd_disk->fops->open(bdev->bd_inode, file); | |
941 | if (ret) | |
942 | goto out; | |
943 | } | |
944 | if (bdev->bd_invalidated) | |
945 | rescan_partitions(bdev->bd_disk, bdev); | |
946 | } else { | |
c039e313 | 947 | mutex_lock(&bdev->bd_contains->bd_mutex); |
1da177e4 | 948 | bdev->bd_contains->bd_part_count++; |
c039e313 | 949 | mutex_unlock(&bdev->bd_contains->bd_mutex); |
1da177e4 LT |
950 | } |
951 | } | |
952 | bdev->bd_openers++; | |
c039e313 | 953 | mutex_unlock(&bdev->bd_mutex); |
1da177e4 LT |
954 | unlock_kernel(); |
955 | return 0; | |
956 | ||
957 | out_first: | |
958 | bdev->bd_disk = NULL; | |
959 | bdev->bd_inode->i_data.backing_dev_info = &default_backing_dev_info; | |
960 | if (bdev != bdev->bd_contains) | |
961 | blkdev_put(bdev->bd_contains); | |
962 | bdev->bd_contains = NULL; | |
963 | put_disk(disk); | |
964 | module_put(owner); | |
965 | out: | |
c039e313 | 966 | mutex_unlock(&bdev->bd_mutex); |
1da177e4 LT |
967 | unlock_kernel(); |
968 | if (ret) | |
969 | bdput(bdev); | |
970 | return ret; | |
971 | } | |
972 | ||
973 | int blkdev_get(struct block_device *bdev, mode_t mode, unsigned flags) | |
974 | { | |
975 | /* | |
976 | * This crockload is due to bad choice of ->open() type. | |
977 | * It will go away. | |
978 | * For now, block device ->open() routine must _not_ | |
979 | * examine anything in 'inode' argument except ->i_rdev. | |
980 | */ | |
981 | struct file fake_file = {}; | |
982 | struct dentry fake_dentry = {}; | |
983 | fake_file.f_mode = mode; | |
984 | fake_file.f_flags = flags; | |
985 | fake_file.f_dentry = &fake_dentry; | |
986 | fake_dentry.d_inode = bdev->bd_inode; | |
987 | ||
988 | return do_open(bdev, &fake_file); | |
989 | } | |
990 | ||
991 | EXPORT_SYMBOL(blkdev_get); | |
992 | ||
993 | static int blkdev_open(struct inode * inode, struct file * filp) | |
994 | { | |
995 | struct block_device *bdev; | |
996 | int res; | |
997 | ||
998 | /* | |
999 | * Preserve backwards compatibility and allow large file access | |
1000 | * even if userspace doesn't ask for it explicitly. Some mkfs | |
1001 | * binary needs it. We might want to drop this workaround | |
1002 | * during an unstable branch. | |
1003 | */ | |
1004 | filp->f_flags |= O_LARGEFILE; | |
1005 | ||
1006 | bdev = bd_acquire(inode); | |
1007 | ||
1008 | res = do_open(bdev, filp); | |
1009 | if (res) | |
1010 | return res; | |
1011 | ||
1012 | if (!(filp->f_flags & O_EXCL) ) | |
1013 | return 0; | |
1014 | ||
1015 | if (!(res = bd_claim(bdev, filp))) | |
1016 | return 0; | |
1017 | ||
1018 | blkdev_put(bdev); | |
1019 | return res; | |
1020 | } | |
1021 | ||
1022 | int blkdev_put(struct block_device *bdev) | |
1023 | { | |
1024 | int ret = 0; | |
1025 | struct inode *bd_inode = bdev->bd_inode; | |
1026 | struct gendisk *disk = bdev->bd_disk; | |
1027 | ||
c039e313 | 1028 | mutex_lock(&bdev->bd_mutex); |
1da177e4 LT |
1029 | lock_kernel(); |
1030 | if (!--bdev->bd_openers) { | |
1031 | sync_blockdev(bdev); | |
1032 | kill_bdev(bdev); | |
1033 | } | |
1034 | if (bdev->bd_contains == bdev) { | |
1035 | if (disk->fops->release) | |
1036 | ret = disk->fops->release(bd_inode, NULL); | |
1037 | } else { | |
c039e313 | 1038 | mutex_lock(&bdev->bd_contains->bd_mutex); |
1da177e4 | 1039 | bdev->bd_contains->bd_part_count--; |
c039e313 | 1040 | mutex_unlock(&bdev->bd_contains->bd_mutex); |
1da177e4 LT |
1041 | } |
1042 | if (!bdev->bd_openers) { | |
1043 | struct module *owner = disk->fops->owner; | |
1044 | ||
1045 | put_disk(disk); | |
1046 | module_put(owner); | |
1047 | ||
1048 | if (bdev->bd_contains != bdev) { | |
1049 | kobject_put(&bdev->bd_part->kobj); | |
1050 | bdev->bd_part = NULL; | |
1051 | } | |
1052 | bdev->bd_disk = NULL; | |
1053 | bdev->bd_inode->i_data.backing_dev_info = &default_backing_dev_info; | |
1054 | if (bdev != bdev->bd_contains) { | |
1055 | blkdev_put(bdev->bd_contains); | |
1056 | } | |
1057 | bdev->bd_contains = NULL; | |
1058 | } | |
1059 | unlock_kernel(); | |
c039e313 | 1060 | mutex_unlock(&bdev->bd_mutex); |
1da177e4 LT |
1061 | bdput(bdev); |
1062 | return ret; | |
1063 | } | |
1064 | ||
1065 | EXPORT_SYMBOL(blkdev_put); | |
1066 | ||
1067 | static int blkdev_close(struct inode * inode, struct file * filp) | |
1068 | { | |
1069 | struct block_device *bdev = I_BDEV(filp->f_mapping->host); | |
1070 | if (bdev->bd_holder == filp) | |
1071 | bd_release(bdev); | |
1072 | return blkdev_put(bdev); | |
1073 | } | |
1074 | ||
1075 | static ssize_t blkdev_file_write(struct file *file, const char __user *buf, | |
1076 | size_t count, loff_t *ppos) | |
1077 | { | |
1078 | struct iovec local_iov = { .iov_base = (void __user *)buf, .iov_len = count }; | |
1079 | ||
1080 | return generic_file_write_nolock(file, &local_iov, 1, ppos); | |
1081 | } | |
1082 | ||
1083 | static ssize_t blkdev_file_aio_write(struct kiocb *iocb, const char __user *buf, | |
1084 | size_t count, loff_t pos) | |
1085 | { | |
1086 | struct iovec local_iov = { .iov_base = (void __user *)buf, .iov_len = count }; | |
1087 | ||
1088 | return generic_file_aio_write_nolock(iocb, &local_iov, 1, &iocb->ki_pos); | |
1089 | } | |
1090 | ||
bb93e3a5 | 1091 | static long block_ioctl(struct file *file, unsigned cmd, unsigned long arg) |
1da177e4 LT |
1092 | { |
1093 | return blkdev_ioctl(file->f_mapping->host, file, cmd, arg); | |
1094 | } | |
1095 | ||
f5e54d6e | 1096 | const struct address_space_operations def_blk_aops = { |
1da177e4 LT |
1097 | .readpage = blkdev_readpage, |
1098 | .writepage = blkdev_writepage, | |
1099 | .sync_page = block_sync_page, | |
1100 | .prepare_write = blkdev_prepare_write, | |
1101 | .commit_write = blkdev_commit_write, | |
1102 | .writepages = generic_writepages, | |
1103 | .direct_IO = blkdev_direct_IO, | |
1104 | }; | |
1105 | ||
4b6f5d20 | 1106 | const struct file_operations def_blk_fops = { |
1da177e4 LT |
1107 | .open = blkdev_open, |
1108 | .release = blkdev_close, | |
1109 | .llseek = block_llseek, | |
1110 | .read = generic_file_read, | |
1111 | .write = blkdev_file_write, | |
1112 | .aio_read = generic_file_aio_read, | |
1113 | .aio_write = blkdev_file_aio_write, | |
1114 | .mmap = generic_file_mmap, | |
1115 | .fsync = block_fsync, | |
bb93e3a5 | 1116 | .unlocked_ioctl = block_ioctl, |
1da177e4 LT |
1117 | #ifdef CONFIG_COMPAT |
1118 | .compat_ioctl = compat_blkdev_ioctl, | |
1119 | #endif | |
1120 | .readv = generic_file_readv, | |
1121 | .writev = generic_file_write_nolock, | |
1122 | .sendfile = generic_file_sendfile, | |
7f9c51f0 JA |
1123 | .splice_read = generic_file_splice_read, |
1124 | .splice_write = generic_file_splice_write, | |
1da177e4 LT |
1125 | }; |
1126 | ||
1127 | int ioctl_by_bdev(struct block_device *bdev, unsigned cmd, unsigned long arg) | |
1128 | { | |
1129 | int res; | |
1130 | mm_segment_t old_fs = get_fs(); | |
1131 | set_fs(KERNEL_DS); | |
1132 | res = blkdev_ioctl(bdev->bd_inode, NULL, cmd, arg); | |
1133 | set_fs(old_fs); | |
1134 | return res; | |
1135 | } | |
1136 | ||
1137 | EXPORT_SYMBOL(ioctl_by_bdev); | |
1138 | ||
1139 | /** | |
1140 | * lookup_bdev - lookup a struct block_device by name | |
1141 | * | |
1142 | * @path: special file representing the block device | |
1143 | * | |
1144 | * Get a reference to the blockdevice at @path in the current | |
1145 | * namespace if possible and return it. Return ERR_PTR(error) | |
1146 | * otherwise. | |
1147 | */ | |
1148 | struct block_device *lookup_bdev(const char *path) | |
1149 | { | |
1150 | struct block_device *bdev; | |
1151 | struct inode *inode; | |
1152 | struct nameidata nd; | |
1153 | int error; | |
1154 | ||
1155 | if (!path || !*path) | |
1156 | return ERR_PTR(-EINVAL); | |
1157 | ||
1158 | error = path_lookup(path, LOOKUP_FOLLOW, &nd); | |
1159 | if (error) | |
1160 | return ERR_PTR(error); | |
1161 | ||
1162 | inode = nd.dentry->d_inode; | |
1163 | error = -ENOTBLK; | |
1164 | if (!S_ISBLK(inode->i_mode)) | |
1165 | goto fail; | |
1166 | error = -EACCES; | |
1167 | if (nd.mnt->mnt_flags & MNT_NODEV) | |
1168 | goto fail; | |
1169 | error = -ENOMEM; | |
1170 | bdev = bd_acquire(inode); | |
1171 | if (!bdev) | |
1172 | goto fail; | |
1173 | out: | |
1174 | path_release(&nd); | |
1175 | return bdev; | |
1176 | fail: | |
1177 | bdev = ERR_PTR(error); | |
1178 | goto out; | |
1179 | } | |
1180 | ||
1181 | /** | |
1182 | * open_bdev_excl - open a block device by name and set it up for use | |
1183 | * | |
1184 | * @path: special file representing the block device | |
1185 | * @flags: %MS_RDONLY for opening read-only | |
1186 | * @holder: owner for exclusion | |
1187 | * | |
1188 | * Open the blockdevice described by the special file at @path, claim it | |
1189 | * for the @holder. | |
1190 | */ | |
1191 | struct block_device *open_bdev_excl(const char *path, int flags, void *holder) | |
1192 | { | |
1193 | struct block_device *bdev; | |
1194 | mode_t mode = FMODE_READ; | |
1195 | int error = 0; | |
1196 | ||
1197 | bdev = lookup_bdev(path); | |
1198 | if (IS_ERR(bdev)) | |
1199 | return bdev; | |
1200 | ||
1201 | if (!(flags & MS_RDONLY)) | |
1202 | mode |= FMODE_WRITE; | |
1203 | error = blkdev_get(bdev, mode, 0); | |
1204 | if (error) | |
1205 | return ERR_PTR(error); | |
1206 | error = -EACCES; | |
1207 | if (!(flags & MS_RDONLY) && bdev_read_only(bdev)) | |
1208 | goto blkdev_put; | |
1209 | error = bd_claim(bdev, holder); | |
1210 | if (error) | |
1211 | goto blkdev_put; | |
1212 | ||
1213 | return bdev; | |
1214 | ||
1215 | blkdev_put: | |
1216 | blkdev_put(bdev); | |
1217 | return ERR_PTR(error); | |
1218 | } | |
1219 | ||
1220 | EXPORT_SYMBOL(open_bdev_excl); | |
1221 | ||
1222 | /** | |
1223 | * close_bdev_excl - release a blockdevice openen by open_bdev_excl() | |
1224 | * | |
1225 | * @bdev: blockdevice to close | |
1226 | * | |
1227 | * This is the counterpart to open_bdev_excl(). | |
1228 | */ | |
1229 | void close_bdev_excl(struct block_device *bdev) | |
1230 | { | |
1231 | bd_release(bdev); | |
1232 | blkdev_put(bdev); | |
1233 | } | |
1234 | ||
1235 | EXPORT_SYMBOL(close_bdev_excl); |