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 | * | |
219 | * This takes the block device bd_mount_sem to make sure no new mounts | |
220 | * happen on bdev until thaw_bdev() is called. | |
221 | * If a superblock is found on this device, we take the s_umount semaphore | |
222 | * on it to make sure nobody unmounts until the snapshot creation is done. | |
223 | * The reference counter (bd_fsfreeze_count) guarantees that only the last | |
224 | * unfreeze process can unfreeze the frozen filesystem actually when multiple | |
225 | * freeze requests arrive simultaneously. It counts up in freeze_bdev() and | |
226 | * count down in thaw_bdev(). When it becomes 0, thaw_bdev() will unfreeze | |
227 | * actually. | |
228 | */ | |
229 | struct super_block *freeze_bdev(struct block_device *bdev) | |
230 | { | |
231 | struct super_block *sb; | |
232 | int error = 0; | |
233 | ||
234 | mutex_lock(&bdev->bd_fsfreeze_mutex); | |
235 | if (bdev->bd_fsfreeze_count > 0) { | |
236 | bdev->bd_fsfreeze_count++; | |
237 | sb = get_super(bdev); | |
238 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
239 | return sb; | |
240 | } | |
241 | bdev->bd_fsfreeze_count++; | |
242 | ||
243 | down(&bdev->bd_mount_sem); | |
244 | sb = get_super(bdev); | |
245 | if (sb && !(sb->s_flags & MS_RDONLY)) { | |
246 | sb->s_frozen = SB_FREEZE_WRITE; | |
247 | smp_wmb(); | |
248 | ||
60b0680f | 249 | sync_filesystem(sb); |
585d3bc0 NP |
250 | |
251 | sb->s_frozen = SB_FREEZE_TRANS; | |
252 | smp_wmb(); | |
253 | ||
254 | sync_blockdev(sb->s_bdev); | |
255 | ||
256 | if (sb->s_op->freeze_fs) { | |
257 | error = sb->s_op->freeze_fs(sb); | |
258 | if (error) { | |
259 | printk(KERN_ERR | |
260 | "VFS:Filesystem freeze failed\n"); | |
261 | sb->s_frozen = SB_UNFROZEN; | |
262 | drop_super(sb); | |
263 | up(&bdev->bd_mount_sem); | |
264 | bdev->bd_fsfreeze_count--; | |
265 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
266 | return ERR_PTR(error); | |
267 | } | |
268 | } | |
269 | } | |
270 | ||
271 | sync_blockdev(bdev); | |
272 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
273 | ||
274 | return sb; /* thaw_bdev releases s->s_umount and bd_mount_sem */ | |
275 | } | |
276 | EXPORT_SYMBOL(freeze_bdev); | |
277 | ||
278 | /** | |
279 | * thaw_bdev -- unlock filesystem | |
280 | * @bdev: blockdevice to unlock | |
281 | * @sb: associated superblock | |
282 | * | |
283 | * Unlocks the filesystem and marks it writeable again after freeze_bdev(). | |
284 | */ | |
285 | int thaw_bdev(struct block_device *bdev, struct super_block *sb) | |
286 | { | |
287 | int error = 0; | |
288 | ||
289 | mutex_lock(&bdev->bd_fsfreeze_mutex); | |
290 | if (!bdev->bd_fsfreeze_count) { | |
291 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
292 | return -EINVAL; | |
293 | } | |
294 | ||
295 | bdev->bd_fsfreeze_count--; | |
296 | if (bdev->bd_fsfreeze_count > 0) { | |
297 | if (sb) | |
298 | drop_super(sb); | |
299 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
300 | return 0; | |
301 | } | |
302 | ||
303 | if (sb) { | |
304 | BUG_ON(sb->s_bdev != bdev); | |
305 | if (!(sb->s_flags & MS_RDONLY)) { | |
306 | if (sb->s_op->unfreeze_fs) { | |
307 | error = sb->s_op->unfreeze_fs(sb); | |
308 | if (error) { | |
309 | printk(KERN_ERR | |
310 | "VFS:Filesystem thaw failed\n"); | |
311 | sb->s_frozen = SB_FREEZE_TRANS; | |
312 | bdev->bd_fsfreeze_count++; | |
313 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
314 | return error; | |
315 | } | |
316 | } | |
317 | sb->s_frozen = SB_UNFROZEN; | |
318 | smp_wmb(); | |
319 | wake_up(&sb->s_wait_unfrozen); | |
320 | } | |
321 | drop_super(sb); | |
322 | } | |
323 | ||
324 | up(&bdev->bd_mount_sem); | |
325 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
326 | return 0; | |
327 | } | |
328 | EXPORT_SYMBOL(thaw_bdev); | |
329 | ||
1da177e4 LT |
330 | static int blkdev_writepage(struct page *page, struct writeback_control *wbc) |
331 | { | |
332 | return block_write_full_page(page, blkdev_get_block, wbc); | |
333 | } | |
334 | ||
335 | static int blkdev_readpage(struct file * file, struct page * page) | |
336 | { | |
337 | return block_read_full_page(page, blkdev_get_block); | |
338 | } | |
339 | ||
6272b5a5 NP |
340 | static int blkdev_write_begin(struct file *file, struct address_space *mapping, |
341 | loff_t pos, unsigned len, unsigned flags, | |
342 | struct page **pagep, void **fsdata) | |
1da177e4 | 343 | { |
6272b5a5 NP |
344 | *pagep = NULL; |
345 | return block_write_begin(file, mapping, pos, len, flags, pagep, fsdata, | |
346 | blkdev_get_block); | |
1da177e4 LT |
347 | } |
348 | ||
6272b5a5 NP |
349 | static int blkdev_write_end(struct file *file, struct address_space *mapping, |
350 | loff_t pos, unsigned len, unsigned copied, | |
351 | struct page *page, void *fsdata) | |
1da177e4 | 352 | { |
6272b5a5 NP |
353 | int ret; |
354 | ret = block_write_end(file, mapping, pos, len, copied, page, fsdata); | |
355 | ||
356 | unlock_page(page); | |
357 | page_cache_release(page); | |
358 | ||
359 | return ret; | |
1da177e4 LT |
360 | } |
361 | ||
362 | /* | |
363 | * private llseek: | |
0f7fc9e4 | 364 | * for a block special file file->f_path.dentry->d_inode->i_size is zero |
1da177e4 LT |
365 | * so we compute the size by hand (just as in block_read/write above) |
366 | */ | |
367 | static loff_t block_llseek(struct file *file, loff_t offset, int origin) | |
368 | { | |
369 | struct inode *bd_inode = file->f_mapping->host; | |
370 | loff_t size; | |
371 | loff_t retval; | |
372 | ||
1b1dcc1b | 373 | mutex_lock(&bd_inode->i_mutex); |
1da177e4 LT |
374 | size = i_size_read(bd_inode); |
375 | ||
376 | switch (origin) { | |
377 | case 2: | |
378 | offset += size; | |
379 | break; | |
380 | case 1: | |
381 | offset += file->f_pos; | |
382 | } | |
383 | retval = -EINVAL; | |
384 | if (offset >= 0 && offset <= size) { | |
385 | if (offset != file->f_pos) { | |
386 | file->f_pos = offset; | |
387 | } | |
388 | retval = offset; | |
389 | } | |
1b1dcc1b | 390 | mutex_unlock(&bd_inode->i_mutex); |
1da177e4 LT |
391 | return retval; |
392 | } | |
393 | ||
394 | /* | |
395 | * Filp is never NULL; the only case when ->fsync() is called with | |
396 | * NULL first argument is nfsd_sync_dir() and that's not a directory. | |
397 | */ | |
398 | ||
399 | static int block_fsync(struct file *filp, struct dentry *dentry, int datasync) | |
400 | { | |
401 | return sync_blockdev(I_BDEV(filp->f_mapping->host)); | |
402 | } | |
403 | ||
404 | /* | |
405 | * pseudo-fs | |
406 | */ | |
407 | ||
408 | static __cacheline_aligned_in_smp DEFINE_SPINLOCK(bdev_lock); | |
e18b890b | 409 | static struct kmem_cache * bdev_cachep __read_mostly; |
1da177e4 LT |
410 | |
411 | static struct inode *bdev_alloc_inode(struct super_block *sb) | |
412 | { | |
e94b1766 | 413 | struct bdev_inode *ei = kmem_cache_alloc(bdev_cachep, GFP_KERNEL); |
1da177e4 LT |
414 | if (!ei) |
415 | return NULL; | |
416 | return &ei->vfs_inode; | |
417 | } | |
418 | ||
419 | static void bdev_destroy_inode(struct inode *inode) | |
420 | { | |
421 | struct bdev_inode *bdi = BDEV_I(inode); | |
422 | ||
423 | bdi->bdev.bd_inode_backing_dev_info = NULL; | |
424 | kmem_cache_free(bdev_cachep, bdi); | |
425 | } | |
426 | ||
51cc5068 | 427 | static void init_once(void *foo) |
1da177e4 LT |
428 | { |
429 | struct bdev_inode *ei = (struct bdev_inode *) foo; | |
430 | struct block_device *bdev = &ei->bdev; | |
431 | ||
a35afb83 CL |
432 | memset(bdev, 0, sizeof(*bdev)); |
433 | mutex_init(&bdev->bd_mutex); | |
434 | sema_init(&bdev->bd_mount_sem, 1); | |
435 | INIT_LIST_HEAD(&bdev->bd_inodes); | |
436 | INIT_LIST_HEAD(&bdev->bd_list); | |
641dc636 | 437 | #ifdef CONFIG_SYSFS |
a35afb83 | 438 | INIT_LIST_HEAD(&bdev->bd_holder_list); |
641dc636 | 439 | #endif |
a35afb83 | 440 | inode_init_once(&ei->vfs_inode); |
fcccf502 TS |
441 | /* Initialize mutex for freeze. */ |
442 | mutex_init(&bdev->bd_fsfreeze_mutex); | |
1da177e4 LT |
443 | } |
444 | ||
445 | static inline void __bd_forget(struct inode *inode) | |
446 | { | |
447 | list_del_init(&inode->i_devices); | |
448 | inode->i_bdev = NULL; | |
449 | inode->i_mapping = &inode->i_data; | |
450 | } | |
451 | ||
452 | static void bdev_clear_inode(struct inode *inode) | |
453 | { | |
454 | struct block_device *bdev = &BDEV_I(inode)->bdev; | |
455 | struct list_head *p; | |
456 | spin_lock(&bdev_lock); | |
457 | while ( (p = bdev->bd_inodes.next) != &bdev->bd_inodes ) { | |
458 | __bd_forget(list_entry(p, struct inode, i_devices)); | |
459 | } | |
460 | list_del_init(&bdev->bd_list); | |
461 | spin_unlock(&bdev_lock); | |
462 | } | |
463 | ||
ee9b6d61 | 464 | static const struct super_operations bdev_sops = { |
1da177e4 LT |
465 | .statfs = simple_statfs, |
466 | .alloc_inode = bdev_alloc_inode, | |
467 | .destroy_inode = bdev_destroy_inode, | |
468 | .drop_inode = generic_delete_inode, | |
469 | .clear_inode = bdev_clear_inode, | |
470 | }; | |
471 | ||
454e2398 DH |
472 | static int bd_get_sb(struct file_system_type *fs_type, |
473 | int flags, const char *dev_name, void *data, struct vfsmount *mnt) | |
1da177e4 | 474 | { |
454e2398 | 475 | return get_sb_pseudo(fs_type, "bdev:", &bdev_sops, 0x62646576, mnt); |
1da177e4 LT |
476 | } |
477 | ||
478 | static struct file_system_type bd_type = { | |
479 | .name = "bdev", | |
480 | .get_sb = bd_get_sb, | |
481 | .kill_sb = kill_anon_super, | |
482 | }; | |
483 | ||
c2acf7b9 | 484 | struct super_block *blockdev_superblock __read_mostly; |
1da177e4 LT |
485 | |
486 | void __init bdev_cache_init(void) | |
487 | { | |
488 | int err; | |
c2acf7b9 DC |
489 | struct vfsmount *bd_mnt; |
490 | ||
1da177e4 | 491 | bdev_cachep = kmem_cache_create("bdev_cache", sizeof(struct bdev_inode), |
fffb60f9 PJ |
492 | 0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| |
493 | SLAB_MEM_SPREAD|SLAB_PANIC), | |
20c2df83 | 494 | init_once); |
1da177e4 LT |
495 | err = register_filesystem(&bd_type); |
496 | if (err) | |
497 | panic("Cannot register bdev pseudo-fs"); | |
498 | bd_mnt = kern_mount(&bd_type); | |
1da177e4 LT |
499 | if (IS_ERR(bd_mnt)) |
500 | panic("Cannot create bdev pseudo-fs"); | |
2e1483c9 CM |
501 | /* |
502 | * This vfsmount structure is only used to obtain the | |
503 | * blockdev_superblock, so tell kmemleak not to report it. | |
504 | */ | |
505 | kmemleak_not_leak(bd_mnt); | |
1da177e4 LT |
506 | blockdev_superblock = bd_mnt->mnt_sb; /* For writeback */ |
507 | } | |
508 | ||
509 | /* | |
510 | * Most likely _very_ bad one - but then it's hardly critical for small | |
511 | * /dev and can be fixed when somebody will need really large one. | |
512 | * Keep in mind that it will be fed through icache hash function too. | |
513 | */ | |
514 | static inline unsigned long hash(dev_t dev) | |
515 | { | |
516 | return MAJOR(dev)+MINOR(dev); | |
517 | } | |
518 | ||
519 | static int bdev_test(struct inode *inode, void *data) | |
520 | { | |
521 | return BDEV_I(inode)->bdev.bd_dev == *(dev_t *)data; | |
522 | } | |
523 | ||
524 | static int bdev_set(struct inode *inode, void *data) | |
525 | { | |
526 | BDEV_I(inode)->bdev.bd_dev = *(dev_t *)data; | |
527 | return 0; | |
528 | } | |
529 | ||
530 | static LIST_HEAD(all_bdevs); | |
531 | ||
532 | struct block_device *bdget(dev_t dev) | |
533 | { | |
534 | struct block_device *bdev; | |
535 | struct inode *inode; | |
536 | ||
c2acf7b9 | 537 | inode = iget5_locked(blockdev_superblock, hash(dev), |
1da177e4 LT |
538 | bdev_test, bdev_set, &dev); |
539 | ||
540 | if (!inode) | |
541 | return NULL; | |
542 | ||
543 | bdev = &BDEV_I(inode)->bdev; | |
544 | ||
545 | if (inode->i_state & I_NEW) { | |
546 | bdev->bd_contains = NULL; | |
547 | bdev->bd_inode = inode; | |
548 | bdev->bd_block_size = (1 << inode->i_blkbits); | |
549 | bdev->bd_part_count = 0; | |
550 | bdev->bd_invalidated = 0; | |
551 | inode->i_mode = S_IFBLK; | |
552 | inode->i_rdev = dev; | |
553 | inode->i_bdev = bdev; | |
554 | inode->i_data.a_ops = &def_blk_aops; | |
555 | mapping_set_gfp_mask(&inode->i_data, GFP_USER); | |
556 | inode->i_data.backing_dev_info = &default_backing_dev_info; | |
557 | spin_lock(&bdev_lock); | |
558 | list_add(&bdev->bd_list, &all_bdevs); | |
559 | spin_unlock(&bdev_lock); | |
560 | unlock_new_inode(inode); | |
561 | } | |
562 | return bdev; | |
563 | } | |
564 | ||
565 | EXPORT_SYMBOL(bdget); | |
566 | ||
567 | long nr_blockdev_pages(void) | |
568 | { | |
203a2935 | 569 | struct block_device *bdev; |
1da177e4 LT |
570 | long ret = 0; |
571 | spin_lock(&bdev_lock); | |
203a2935 | 572 | list_for_each_entry(bdev, &all_bdevs, bd_list) { |
1da177e4 LT |
573 | ret += bdev->bd_inode->i_mapping->nrpages; |
574 | } | |
575 | spin_unlock(&bdev_lock); | |
576 | return ret; | |
577 | } | |
578 | ||
579 | void bdput(struct block_device *bdev) | |
580 | { | |
581 | iput(bdev->bd_inode); | |
582 | } | |
583 | ||
584 | EXPORT_SYMBOL(bdput); | |
585 | ||
586 | static struct block_device *bd_acquire(struct inode *inode) | |
587 | { | |
588 | struct block_device *bdev; | |
09d967c6 | 589 | |
1da177e4 LT |
590 | spin_lock(&bdev_lock); |
591 | bdev = inode->i_bdev; | |
09d967c6 OH |
592 | if (bdev) { |
593 | atomic_inc(&bdev->bd_inode->i_count); | |
1da177e4 LT |
594 | spin_unlock(&bdev_lock); |
595 | return bdev; | |
596 | } | |
597 | spin_unlock(&bdev_lock); | |
09d967c6 | 598 | |
1da177e4 LT |
599 | bdev = bdget(inode->i_rdev); |
600 | if (bdev) { | |
601 | spin_lock(&bdev_lock); | |
09d967c6 OH |
602 | if (!inode->i_bdev) { |
603 | /* | |
604 | * We take an additional bd_inode->i_count for inode, | |
605 | * and it's released in clear_inode() of inode. | |
606 | * So, we can access it via ->i_mapping always | |
607 | * without igrab(). | |
608 | */ | |
609 | atomic_inc(&bdev->bd_inode->i_count); | |
610 | inode->i_bdev = bdev; | |
611 | inode->i_mapping = bdev->bd_inode->i_mapping; | |
612 | list_add(&inode->i_devices, &bdev->bd_inodes); | |
613 | } | |
1da177e4 LT |
614 | spin_unlock(&bdev_lock); |
615 | } | |
616 | return bdev; | |
617 | } | |
618 | ||
619 | /* Call when you free inode */ | |
620 | ||
621 | void bd_forget(struct inode *inode) | |
622 | { | |
09d967c6 OH |
623 | struct block_device *bdev = NULL; |
624 | ||
1da177e4 | 625 | spin_lock(&bdev_lock); |
09d967c6 | 626 | if (inode->i_bdev) { |
c2acf7b9 | 627 | if (!sb_is_blkdev_sb(inode->i_sb)) |
09d967c6 | 628 | bdev = inode->i_bdev; |
1da177e4 | 629 | __bd_forget(inode); |
09d967c6 | 630 | } |
1da177e4 | 631 | spin_unlock(&bdev_lock); |
09d967c6 OH |
632 | |
633 | if (bdev) | |
634 | iput(bdev->bd_inode); | |
1da177e4 LT |
635 | } |
636 | ||
637 | int bd_claim(struct block_device *bdev, void *holder) | |
638 | { | |
639 | int res; | |
640 | spin_lock(&bdev_lock); | |
641 | ||
642 | /* first decide result */ | |
643 | if (bdev->bd_holder == holder) | |
644 | res = 0; /* already a holder */ | |
645 | else if (bdev->bd_holder != NULL) | |
646 | res = -EBUSY; /* held by someone else */ | |
647 | else if (bdev->bd_contains == bdev) | |
648 | res = 0; /* is a whole device which isn't held */ | |
649 | ||
650 | else if (bdev->bd_contains->bd_holder == bd_claim) | |
651 | res = 0; /* is a partition of a device that is being partitioned */ | |
652 | else if (bdev->bd_contains->bd_holder != NULL) | |
653 | res = -EBUSY; /* is a partition of a held device */ | |
654 | else | |
655 | res = 0; /* is a partition of an un-held device */ | |
656 | ||
657 | /* now impose change */ | |
658 | if (res==0) { | |
659 | /* note that for a whole device bd_holders | |
660 | * will be incremented twice, and bd_holder will | |
661 | * be set to bd_claim before being set to holder | |
662 | */ | |
663 | bdev->bd_contains->bd_holders ++; | |
664 | bdev->bd_contains->bd_holder = bd_claim; | |
665 | bdev->bd_holders++; | |
666 | bdev->bd_holder = holder; | |
667 | } | |
668 | spin_unlock(&bdev_lock); | |
669 | return res; | |
670 | } | |
671 | ||
672 | EXPORT_SYMBOL(bd_claim); | |
673 | ||
674 | void bd_release(struct block_device *bdev) | |
675 | { | |
676 | spin_lock(&bdev_lock); | |
677 | if (!--bdev->bd_contains->bd_holders) | |
678 | bdev->bd_contains->bd_holder = NULL; | |
679 | if (!--bdev->bd_holders) | |
680 | bdev->bd_holder = NULL; | |
681 | spin_unlock(&bdev_lock); | |
682 | } | |
683 | ||
684 | EXPORT_SYMBOL(bd_release); | |
685 | ||
641dc636 JN |
686 | #ifdef CONFIG_SYSFS |
687 | /* | |
688 | * Functions for bd_claim_by_kobject / bd_release_from_kobject | |
689 | * | |
690 | * If a kobject is passed to bd_claim_by_kobject() | |
691 | * and the kobject has a parent directory, | |
692 | * following symlinks are created: | |
693 | * o from the kobject to the claimed bdev | |
694 | * o from "holders" directory of the bdev to the parent of the kobject | |
695 | * bd_release_from_kobject() removes these symlinks. | |
696 | * | |
697 | * Example: | |
698 | * If /dev/dm-0 maps to /dev/sda, kobject corresponding to | |
699 | * /sys/block/dm-0/slaves is passed to bd_claim_by_kobject(), then: | |
700 | * /sys/block/dm-0/slaves/sda --> /sys/block/sda | |
701 | * /sys/block/sda/holders/dm-0 --> /sys/block/dm-0 | |
702 | */ | |
703 | ||
4d7dd8fd | 704 | static int add_symlink(struct kobject *from, struct kobject *to) |
641dc636 JN |
705 | { |
706 | if (!from || !to) | |
4d7dd8fd AM |
707 | return 0; |
708 | return sysfs_create_link(from, to, kobject_name(to)); | |
641dc636 JN |
709 | } |
710 | ||
711 | static void del_symlink(struct kobject *from, struct kobject *to) | |
712 | { | |
713 | if (!from || !to) | |
714 | return; | |
715 | sysfs_remove_link(from, kobject_name(to)); | |
716 | } | |
717 | ||
718 | /* | |
719 | * 'struct bd_holder' contains pointers to kobjects symlinked by | |
720 | * bd_claim_by_kobject. | |
721 | * It's connected to bd_holder_list which is protected by bdev->bd_sem. | |
722 | */ | |
723 | struct bd_holder { | |
724 | struct list_head list; /* chain of holders of the bdev */ | |
725 | int count; /* references from the holder */ | |
726 | struct kobject *sdir; /* holder object, e.g. "/block/dm-0/slaves" */ | |
727 | struct kobject *hdev; /* e.g. "/block/dm-0" */ | |
728 | struct kobject *hdir; /* e.g. "/block/sda/holders" */ | |
729 | struct kobject *sdev; /* e.g. "/block/sda" */ | |
730 | }; | |
731 | ||
732 | /* | |
733 | * Get references of related kobjects at once. | |
734 | * Returns 1 on success. 0 on failure. | |
735 | * | |
736 | * Should call bd_holder_release_dirs() after successful use. | |
737 | */ | |
738 | static int bd_holder_grab_dirs(struct block_device *bdev, | |
739 | struct bd_holder *bo) | |
740 | { | |
741 | if (!bdev || !bo) | |
742 | return 0; | |
743 | ||
744 | bo->sdir = kobject_get(bo->sdir); | |
745 | if (!bo->sdir) | |
746 | return 0; | |
747 | ||
748 | bo->hdev = kobject_get(bo->sdir->parent); | |
749 | if (!bo->hdev) | |
750 | goto fail_put_sdir; | |
751 | ||
0762b8bd | 752 | bo->sdev = kobject_get(&part_to_dev(bdev->bd_part)->kobj); |
641dc636 JN |
753 | if (!bo->sdev) |
754 | goto fail_put_hdev; | |
755 | ||
4c46501d | 756 | bo->hdir = kobject_get(bdev->bd_part->holder_dir); |
641dc636 JN |
757 | if (!bo->hdir) |
758 | goto fail_put_sdev; | |
759 | ||
760 | return 1; | |
761 | ||
762 | fail_put_sdev: | |
763 | kobject_put(bo->sdev); | |
764 | fail_put_hdev: | |
765 | kobject_put(bo->hdev); | |
766 | fail_put_sdir: | |
767 | kobject_put(bo->sdir); | |
768 | ||
769 | return 0; | |
770 | } | |
771 | ||
772 | /* Put references of related kobjects at once. */ | |
773 | static void bd_holder_release_dirs(struct bd_holder *bo) | |
774 | { | |
775 | kobject_put(bo->hdir); | |
776 | kobject_put(bo->sdev); | |
777 | kobject_put(bo->hdev); | |
778 | kobject_put(bo->sdir); | |
779 | } | |
780 | ||
781 | static struct bd_holder *alloc_bd_holder(struct kobject *kobj) | |
782 | { | |
783 | struct bd_holder *bo; | |
784 | ||
785 | bo = kzalloc(sizeof(*bo), GFP_KERNEL); | |
786 | if (!bo) | |
787 | return NULL; | |
788 | ||
789 | bo->count = 1; | |
790 | bo->sdir = kobj; | |
791 | ||
792 | return bo; | |
793 | } | |
794 | ||
795 | static void free_bd_holder(struct bd_holder *bo) | |
796 | { | |
797 | kfree(bo); | |
798 | } | |
799 | ||
df6c0cd9 JN |
800 | /** |
801 | * find_bd_holder - find matching struct bd_holder from the block device | |
802 | * | |
803 | * @bdev: struct block device to be searched | |
804 | * @bo: target struct bd_holder | |
805 | * | |
806 | * Returns matching entry with @bo in @bdev->bd_holder_list. | |
807 | * If found, increment the reference count and return the pointer. | |
808 | * If not found, returns NULL. | |
809 | */ | |
36a561d6 AM |
810 | static struct bd_holder *find_bd_holder(struct block_device *bdev, |
811 | struct bd_holder *bo) | |
df6c0cd9 JN |
812 | { |
813 | struct bd_holder *tmp; | |
814 | ||
815 | list_for_each_entry(tmp, &bdev->bd_holder_list, list) | |
816 | if (tmp->sdir == bo->sdir) { | |
817 | tmp->count++; | |
818 | return tmp; | |
819 | } | |
820 | ||
821 | return NULL; | |
822 | } | |
823 | ||
641dc636 JN |
824 | /** |
825 | * add_bd_holder - create sysfs symlinks for bd_claim() relationship | |
826 | * | |
827 | * @bdev: block device to be bd_claimed | |
828 | * @bo: preallocated and initialized by alloc_bd_holder() | |
829 | * | |
df6c0cd9 | 830 | * Add @bo to @bdev->bd_holder_list, create symlinks. |
641dc636 | 831 | * |
df6c0cd9 JN |
832 | * Returns 0 if symlinks are created. |
833 | * Returns -ve if something fails. | |
641dc636 JN |
834 | */ |
835 | static int add_bd_holder(struct block_device *bdev, struct bd_holder *bo) | |
836 | { | |
4e91672c | 837 | int err; |
641dc636 JN |
838 | |
839 | if (!bo) | |
4d7dd8fd | 840 | return -EINVAL; |
641dc636 | 841 | |
641dc636 | 842 | if (!bd_holder_grab_dirs(bdev, bo)) |
4d7dd8fd | 843 | return -EBUSY; |
641dc636 | 844 | |
4e91672c JW |
845 | err = add_symlink(bo->sdir, bo->sdev); |
846 | if (err) | |
847 | return err; | |
848 | ||
849 | err = add_symlink(bo->hdir, bo->hdev); | |
850 | if (err) { | |
851 | del_symlink(bo->sdir, bo->sdev); | |
852 | return err; | |
4d7dd8fd | 853 | } |
4e91672c JW |
854 | |
855 | list_add_tail(&bo->list, &bdev->bd_holder_list); | |
856 | return 0; | |
641dc636 JN |
857 | } |
858 | ||
859 | /** | |
860 | * del_bd_holder - delete sysfs symlinks for bd_claim() relationship | |
861 | * | |
862 | * @bdev: block device to be bd_claimed | |
863 | * @kobj: holder's kobject | |
864 | * | |
865 | * If there is matching entry with @kobj in @bdev->bd_holder_list | |
866 | * and no other bd_claim() from the same kobject, | |
867 | * remove the struct bd_holder from the list, delete symlinks for it. | |
868 | * | |
869 | * Returns a pointer to the struct bd_holder when it's removed from the list | |
870 | * and ready to be freed. | |
871 | * Returns NULL if matching claim isn't found or there is other bd_claim() | |
872 | * by the same kobject. | |
873 | */ | |
874 | static struct bd_holder *del_bd_holder(struct block_device *bdev, | |
875 | struct kobject *kobj) | |
876 | { | |
877 | struct bd_holder *bo; | |
878 | ||
879 | list_for_each_entry(bo, &bdev->bd_holder_list, list) { | |
880 | if (bo->sdir == kobj) { | |
881 | bo->count--; | |
882 | BUG_ON(bo->count < 0); | |
883 | if (!bo->count) { | |
884 | list_del(&bo->list); | |
885 | del_symlink(bo->sdir, bo->sdev); | |
886 | del_symlink(bo->hdir, bo->hdev); | |
887 | bd_holder_release_dirs(bo); | |
888 | return bo; | |
889 | } | |
890 | break; | |
891 | } | |
892 | } | |
893 | ||
894 | return NULL; | |
895 | } | |
896 | ||
897 | /** | |
898 | * bd_claim_by_kobject - bd_claim() with additional kobject signature | |
899 | * | |
900 | * @bdev: block device to be claimed | |
901 | * @holder: holder's signature | |
902 | * @kobj: holder's kobject | |
903 | * | |
904 | * Do bd_claim() and if it succeeds, create sysfs symlinks between | |
905 | * the bdev and the holder's kobject. | |
906 | * Use bd_release_from_kobject() when relesing the claimed bdev. | |
907 | * | |
908 | * Returns 0 on success. (same as bd_claim()) | |
909 | * Returns errno on failure. | |
910 | */ | |
911 | static int bd_claim_by_kobject(struct block_device *bdev, void *holder, | |
912 | struct kobject *kobj) | |
913 | { | |
4e91672c | 914 | int err; |
df6c0cd9 | 915 | struct bd_holder *bo, *found; |
641dc636 JN |
916 | |
917 | if (!kobj) | |
918 | return -EINVAL; | |
919 | ||
920 | bo = alloc_bd_holder(kobj); | |
921 | if (!bo) | |
922 | return -ENOMEM; | |
923 | ||
2e7b651d | 924 | mutex_lock(&bdev->bd_mutex); |
df6c0cd9 | 925 | |
4e91672c JW |
926 | err = bd_claim(bdev, holder); |
927 | if (err) | |
4210df28 | 928 | goto fail; |
4e91672c JW |
929 | |
930 | found = find_bd_holder(bdev, bo); | |
931 | if (found) | |
4210df28 | 932 | goto fail; |
4e91672c JW |
933 | |
934 | err = add_bd_holder(bdev, bo); | |
935 | if (err) | |
936 | bd_release(bdev); | |
4210df28 AM |
937 | else |
938 | bo = NULL; | |
939 | fail: | |
b4cf1b72 | 940 | mutex_unlock(&bdev->bd_mutex); |
4210df28 | 941 | free_bd_holder(bo); |
4e91672c | 942 | return err; |
641dc636 JN |
943 | } |
944 | ||
945 | /** | |
946 | * bd_release_from_kobject - bd_release() with additional kobject signature | |
947 | * | |
948 | * @bdev: block device to be released | |
949 | * @kobj: holder's kobject | |
950 | * | |
951 | * Do bd_release() and remove sysfs symlinks created by bd_claim_by_kobject(). | |
952 | */ | |
953 | static void bd_release_from_kobject(struct block_device *bdev, | |
954 | struct kobject *kobj) | |
955 | { | |
641dc636 JN |
956 | if (!kobj) |
957 | return; | |
958 | ||
2e7b651d | 959 | mutex_lock(&bdev->bd_mutex); |
641dc636 | 960 | bd_release(bdev); |
4210df28 | 961 | free_bd_holder(del_bd_holder(bdev, kobj)); |
b4cf1b72 | 962 | mutex_unlock(&bdev->bd_mutex); |
641dc636 JN |
963 | } |
964 | ||
965 | /** | |
966 | * bd_claim_by_disk - wrapper function for bd_claim_by_kobject() | |
967 | * | |
968 | * @bdev: block device to be claimed | |
969 | * @holder: holder's signature | |
970 | * @disk: holder's gendisk | |
971 | * | |
972 | * Call bd_claim_by_kobject() with getting @disk->slave_dir. | |
973 | */ | |
974 | int bd_claim_by_disk(struct block_device *bdev, void *holder, | |
975 | struct gendisk *disk) | |
976 | { | |
977 | return bd_claim_by_kobject(bdev, holder, kobject_get(disk->slave_dir)); | |
978 | } | |
979 | EXPORT_SYMBOL_GPL(bd_claim_by_disk); | |
980 | ||
981 | /** | |
982 | * bd_release_from_disk - wrapper function for bd_release_from_kobject() | |
983 | * | |
984 | * @bdev: block device to be claimed | |
985 | * @disk: holder's gendisk | |
986 | * | |
987 | * Call bd_release_from_kobject() and put @disk->slave_dir. | |
988 | */ | |
989 | void bd_release_from_disk(struct block_device *bdev, struct gendisk *disk) | |
990 | { | |
991 | bd_release_from_kobject(bdev, disk->slave_dir); | |
992 | kobject_put(disk->slave_dir); | |
993 | } | |
994 | EXPORT_SYMBOL_GPL(bd_release_from_disk); | |
995 | #endif | |
996 | ||
1da177e4 LT |
997 | /* |
998 | * Tries to open block device by device number. Use it ONLY if you | |
999 | * really do not have anything better - i.e. when you are behind a | |
1000 | * truly sucky interface and all you are given is a device number. _Never_ | |
1001 | * to be used for internal purposes. If you ever need it - reconsider | |
1002 | * your API. | |
1003 | */ | |
aeb5d727 | 1004 | struct block_device *open_by_devnum(dev_t dev, fmode_t mode) |
1da177e4 LT |
1005 | { |
1006 | struct block_device *bdev = bdget(dev); | |
1007 | int err = -ENOMEM; | |
1da177e4 | 1008 | if (bdev) |
572c4892 | 1009 | err = blkdev_get(bdev, mode); |
1da177e4 LT |
1010 | return err ? ERR_PTR(err) : bdev; |
1011 | } | |
1012 | ||
1013 | EXPORT_SYMBOL(open_by_devnum); | |
1014 | ||
56ade44b AP |
1015 | /** |
1016 | * flush_disk - invalidates all buffer-cache entries on a disk | |
1017 | * | |
1018 | * @bdev: struct block device to be flushed | |
1019 | * | |
1020 | * Invalidates all buffer-cache entries on a disk. It should be called | |
1021 | * when a disk has been changed -- either by a media change or online | |
1022 | * resize. | |
1023 | */ | |
1024 | static void flush_disk(struct block_device *bdev) | |
1025 | { | |
1026 | if (__invalidate_device(bdev)) { | |
1027 | char name[BDEVNAME_SIZE] = ""; | |
1028 | ||
1029 | if (bdev->bd_disk) | |
1030 | disk_name(bdev->bd_disk, 0, name); | |
1031 | printk(KERN_WARNING "VFS: busy inodes on changed media or " | |
1032 | "resized disk %s\n", name); | |
1033 | } | |
1034 | ||
1035 | if (!bdev->bd_disk) | |
1036 | return; | |
1037 | if (disk_partitionable(bdev->bd_disk)) | |
1038 | bdev->bd_invalidated = 1; | |
1039 | } | |
1040 | ||
c3279d14 | 1041 | /** |
57d1b536 | 1042 | * check_disk_size_change - checks for disk size change and adjusts bdev size. |
c3279d14 AP |
1043 | * @disk: struct gendisk to check |
1044 | * @bdev: struct bdev to adjust. | |
1045 | * | |
1046 | * This routine checks to see if the bdev size does not match the disk size | |
1047 | * and adjusts it if it differs. | |
1048 | */ | |
1049 | void check_disk_size_change(struct gendisk *disk, struct block_device *bdev) | |
1050 | { | |
1051 | loff_t disk_size, bdev_size; | |
1052 | ||
1053 | disk_size = (loff_t)get_capacity(disk) << 9; | |
1054 | bdev_size = i_size_read(bdev->bd_inode); | |
1055 | if (disk_size != bdev_size) { | |
1056 | char name[BDEVNAME_SIZE]; | |
1057 | ||
1058 | disk_name(disk, 0, name); | |
1059 | printk(KERN_INFO | |
1060 | "%s: detected capacity change from %lld to %lld\n", | |
1061 | name, bdev_size, disk_size); | |
1062 | i_size_write(bdev->bd_inode, disk_size); | |
608aeef1 | 1063 | flush_disk(bdev); |
c3279d14 AP |
1064 | } |
1065 | } | |
1066 | EXPORT_SYMBOL(check_disk_size_change); | |
1067 | ||
0c002c2f | 1068 | /** |
57d1b536 | 1069 | * revalidate_disk - wrapper for lower-level driver's revalidate_disk call-back |
0c002c2f AP |
1070 | * @disk: struct gendisk to be revalidated |
1071 | * | |
1072 | * This routine is a wrapper for lower-level driver's revalidate_disk | |
1073 | * call-backs. It is used to do common pre and post operations needed | |
1074 | * for all revalidate_disk operations. | |
1075 | */ | |
1076 | int revalidate_disk(struct gendisk *disk) | |
1077 | { | |
c3279d14 | 1078 | struct block_device *bdev; |
0c002c2f AP |
1079 | int ret = 0; |
1080 | ||
1081 | if (disk->fops->revalidate_disk) | |
1082 | ret = disk->fops->revalidate_disk(disk); | |
1083 | ||
c3279d14 AP |
1084 | bdev = bdget_disk(disk, 0); |
1085 | if (!bdev) | |
1086 | return ret; | |
1087 | ||
1088 | mutex_lock(&bdev->bd_mutex); | |
1089 | check_disk_size_change(disk, bdev); | |
1090 | mutex_unlock(&bdev->bd_mutex); | |
1091 | bdput(bdev); | |
0c002c2f AP |
1092 | return ret; |
1093 | } | |
1094 | EXPORT_SYMBOL(revalidate_disk); | |
1095 | ||
1da177e4 LT |
1096 | /* |
1097 | * This routine checks whether a removable media has been changed, | |
1098 | * and invalidates all buffer-cache-entries in that case. This | |
1099 | * is a relatively slow routine, so we have to try to minimize using | |
1100 | * it. Thus it is called only upon a 'mount' or 'open'. This | |
1101 | * is the best way of combining speed and utility, I think. | |
1102 | * People changing diskettes in the middle of an operation deserve | |
1103 | * to lose :-) | |
1104 | */ | |
1105 | int check_disk_change(struct block_device *bdev) | |
1106 | { | |
1107 | struct gendisk *disk = bdev->bd_disk; | |
1108 | struct block_device_operations * bdops = disk->fops; | |
1109 | ||
1110 | if (!bdops->media_changed) | |
1111 | return 0; | |
1112 | if (!bdops->media_changed(bdev->bd_disk)) | |
1113 | return 0; | |
1114 | ||
56ade44b | 1115 | flush_disk(bdev); |
1da177e4 LT |
1116 | if (bdops->revalidate_disk) |
1117 | bdops->revalidate_disk(bdev->bd_disk); | |
1da177e4 LT |
1118 | return 1; |
1119 | } | |
1120 | ||
1121 | EXPORT_SYMBOL(check_disk_change); | |
1122 | ||
1123 | void bd_set_size(struct block_device *bdev, loff_t size) | |
1124 | { | |
e1defc4f | 1125 | unsigned bsize = bdev_logical_block_size(bdev); |
1da177e4 LT |
1126 | |
1127 | bdev->bd_inode->i_size = size; | |
1128 | while (bsize < PAGE_CACHE_SIZE) { | |
1129 | if (size & bsize) | |
1130 | break; | |
1131 | bsize <<= 1; | |
1132 | } | |
1133 | bdev->bd_block_size = bsize; | |
1134 | bdev->bd_inode->i_blkbits = blksize_bits(bsize); | |
1135 | } | |
1136 | EXPORT_SYMBOL(bd_set_size); | |
1137 | ||
9a1c3542 | 1138 | static int __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part); |
37be4124 | 1139 | |
6d740cd5 PZ |
1140 | /* |
1141 | * bd_mutex locking: | |
1142 | * | |
1143 | * mutex_lock(part->bd_mutex) | |
1144 | * mutex_lock_nested(whole->bd_mutex, 1) | |
1145 | */ | |
1146 | ||
572c4892 | 1147 | static int __blkdev_get(struct block_device *bdev, fmode_t mode, int for_part) |
1da177e4 | 1148 | { |
1da177e4 | 1149 | struct gendisk *disk; |
7db9cfd3 | 1150 | int ret; |
cf771cb5 | 1151 | int partno; |
fe6e9c1f AV |
1152 | int perm = 0; |
1153 | ||
572c4892 | 1154 | if (mode & FMODE_READ) |
fe6e9c1f | 1155 | perm |= MAY_READ; |
572c4892 | 1156 | if (mode & FMODE_WRITE) |
fe6e9c1f AV |
1157 | perm |= MAY_WRITE; |
1158 | /* | |
1159 | * hooks: /n/, see "layering violations". | |
1160 | */ | |
1161 | ret = devcgroup_inode_permission(bdev->bd_inode, perm); | |
82666020 AV |
1162 | if (ret != 0) { |
1163 | bdput(bdev); | |
7db9cfd3 | 1164 | return ret; |
82666020 | 1165 | } |
7db9cfd3 | 1166 | |
1da177e4 | 1167 | lock_kernel(); |
d3374825 | 1168 | restart: |
0762b8bd | 1169 | |
89f97496 | 1170 | ret = -ENXIO; |
cf771cb5 | 1171 | disk = get_gendisk(bdev->bd_dev, &partno); |
0762b8bd TH |
1172 | if (!disk) |
1173 | goto out_unlock_kernel; | |
1da177e4 | 1174 | |
6796bf54 | 1175 | mutex_lock_nested(&bdev->bd_mutex, for_part); |
1da177e4 LT |
1176 | if (!bdev->bd_openers) { |
1177 | bdev->bd_disk = disk; | |
1178 | bdev->bd_contains = bdev; | |
cf771cb5 | 1179 | if (!partno) { |
1da177e4 | 1180 | struct backing_dev_info *bdi; |
89f97496 TH |
1181 | |
1182 | ret = -ENXIO; | |
1183 | bdev->bd_part = disk_get_part(disk, partno); | |
1184 | if (!bdev->bd_part) | |
1185 | goto out_clear; | |
1186 | ||
1da177e4 | 1187 | if (disk->fops->open) { |
572c4892 | 1188 | ret = disk->fops->open(bdev, mode); |
d3374825 N |
1189 | if (ret == -ERESTARTSYS) { |
1190 | /* Lost a race with 'disk' being | |
1191 | * deleted, try again. | |
1192 | * See md.c | |
1193 | */ | |
1194 | disk_put_part(bdev->bd_part); | |
1195 | bdev->bd_part = NULL; | |
1196 | module_put(disk->fops->owner); | |
1197 | put_disk(disk); | |
1198 | bdev->bd_disk = NULL; | |
1199 | mutex_unlock(&bdev->bd_mutex); | |
1200 | goto restart; | |
1201 | } | |
1da177e4 | 1202 | if (ret) |
0762b8bd | 1203 | goto out_clear; |
1da177e4 LT |
1204 | } |
1205 | if (!bdev->bd_openers) { | |
1206 | bd_set_size(bdev,(loff_t)get_capacity(disk)<<9); | |
1207 | bdi = blk_get_backing_dev_info(bdev); | |
1208 | if (bdi == NULL) | |
1209 | bdi = &default_backing_dev_info; | |
1210 | bdev->bd_inode->i_data.backing_dev_info = bdi; | |
1211 | } | |
1212 | if (bdev->bd_invalidated) | |
1213 | rescan_partitions(disk, bdev); | |
1214 | } else { | |
1da177e4 LT |
1215 | struct block_device *whole; |
1216 | whole = bdget_disk(disk, 0); | |
1217 | ret = -ENOMEM; | |
1218 | if (!whole) | |
0762b8bd | 1219 | goto out_clear; |
37be4124 | 1220 | BUG_ON(for_part); |
572c4892 | 1221 | ret = __blkdev_get(whole, mode, 1); |
1da177e4 | 1222 | if (ret) |
0762b8bd | 1223 | goto out_clear; |
1da177e4 | 1224 | bdev->bd_contains = whole; |
1da177e4 LT |
1225 | bdev->bd_inode->i_data.backing_dev_info = |
1226 | whole->bd_inode->i_data.backing_dev_info; | |
89f97496 | 1227 | bdev->bd_part = disk_get_part(disk, partno); |
e71bf0d0 | 1228 | if (!(disk->flags & GENHD_FL_UP) || |
89f97496 | 1229 | !bdev->bd_part || !bdev->bd_part->nr_sects) { |
1da177e4 | 1230 | ret = -ENXIO; |
0762b8bd | 1231 | goto out_clear; |
1da177e4 | 1232 | } |
89f97496 | 1233 | bd_set_size(bdev, (loff_t)bdev->bd_part->nr_sects << 9); |
1da177e4 LT |
1234 | } |
1235 | } else { | |
1236 | put_disk(disk); | |
0762b8bd | 1237 | module_put(disk->fops->owner); |
0762b8bd | 1238 | disk = NULL; |
1da177e4 LT |
1239 | if (bdev->bd_contains == bdev) { |
1240 | if (bdev->bd_disk->fops->open) { | |
572c4892 | 1241 | ret = bdev->bd_disk->fops->open(bdev, mode); |
1da177e4 | 1242 | if (ret) |
0762b8bd | 1243 | goto out_unlock_bdev; |
1da177e4 LT |
1244 | } |
1245 | if (bdev->bd_invalidated) | |
1246 | rescan_partitions(bdev->bd_disk, bdev); | |
1da177e4 LT |
1247 | } |
1248 | } | |
1249 | bdev->bd_openers++; | |
37be4124 N |
1250 | if (for_part) |
1251 | bdev->bd_part_count++; | |
c039e313 | 1252 | mutex_unlock(&bdev->bd_mutex); |
1da177e4 LT |
1253 | unlock_kernel(); |
1254 | return 0; | |
1255 | ||
0762b8bd | 1256 | out_clear: |
89f97496 | 1257 | disk_put_part(bdev->bd_part); |
1da177e4 | 1258 | bdev->bd_disk = NULL; |
0762b8bd | 1259 | bdev->bd_part = NULL; |
1da177e4 LT |
1260 | bdev->bd_inode->i_data.backing_dev_info = &default_backing_dev_info; |
1261 | if (bdev != bdev->bd_contains) | |
572c4892 | 1262 | __blkdev_put(bdev->bd_contains, mode, 1); |
1da177e4 | 1263 | bdev->bd_contains = NULL; |
0762b8bd | 1264 | out_unlock_bdev: |
c039e313 | 1265 | mutex_unlock(&bdev->bd_mutex); |
0762b8bd | 1266 | out_unlock_kernel: |
1da177e4 | 1267 | unlock_kernel(); |
0762b8bd | 1268 | |
0762b8bd TH |
1269 | if (disk) |
1270 | module_put(disk->fops->owner); | |
1271 | put_disk(disk); | |
1272 | bdput(bdev); | |
1273 | ||
1da177e4 LT |
1274 | return ret; |
1275 | } | |
1276 | ||
572c4892 | 1277 | int blkdev_get(struct block_device *bdev, fmode_t mode) |
1da177e4 | 1278 | { |
572c4892 | 1279 | return __blkdev_get(bdev, mode, 0); |
37be4124 | 1280 | } |
1da177e4 LT |
1281 | EXPORT_SYMBOL(blkdev_get); |
1282 | ||
1283 | static int blkdev_open(struct inode * inode, struct file * filp) | |
1284 | { | |
1285 | struct block_device *bdev; | |
1286 | int res; | |
1287 | ||
1288 | /* | |
1289 | * Preserve backwards compatibility and allow large file access | |
1290 | * even if userspace doesn't ask for it explicitly. Some mkfs | |
1291 | * binary needs it. We might want to drop this workaround | |
1292 | * during an unstable branch. | |
1293 | */ | |
1294 | filp->f_flags |= O_LARGEFILE; | |
1295 | ||
572c4892 AV |
1296 | if (filp->f_flags & O_NDELAY) |
1297 | filp->f_mode |= FMODE_NDELAY; | |
1298 | if (filp->f_flags & O_EXCL) | |
1299 | filp->f_mode |= FMODE_EXCL; | |
1300 | if ((filp->f_flags & O_ACCMODE) == 3) | |
1301 | filp->f_mode |= FMODE_WRITE_IOCTL; | |
1302 | ||
1da177e4 | 1303 | bdev = bd_acquire(inode); |
6a2aae06 PE |
1304 | if (bdev == NULL) |
1305 | return -ENOMEM; | |
1da177e4 | 1306 | |
572c4892 AV |
1307 | filp->f_mapping = bdev->bd_inode->i_mapping; |
1308 | ||
1309 | res = blkdev_get(bdev, filp->f_mode); | |
1da177e4 LT |
1310 | if (res) |
1311 | return res; | |
1312 | ||
ebbefc01 CH |
1313 | if (filp->f_mode & FMODE_EXCL) { |
1314 | res = bd_claim(bdev, filp); | |
1315 | if (res) | |
1316 | goto out_blkdev_put; | |
1317 | } | |
1da177e4 | 1318 | |
ebbefc01 | 1319 | return 0; |
1da177e4 | 1320 | |
ebbefc01 | 1321 | out_blkdev_put: |
9a1c3542 | 1322 | blkdev_put(bdev, filp->f_mode); |
1da177e4 LT |
1323 | return res; |
1324 | } | |
1325 | ||
9a1c3542 | 1326 | static int __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part) |
2e7b651d PZ |
1327 | { |
1328 | int ret = 0; | |
2e7b651d | 1329 | struct gendisk *disk = bdev->bd_disk; |
37be4124 | 1330 | struct block_device *victim = NULL; |
2e7b651d | 1331 | |
6796bf54 | 1332 | mutex_lock_nested(&bdev->bd_mutex, for_part); |
2e7b651d | 1333 | lock_kernel(); |
37be4124 N |
1334 | if (for_part) |
1335 | bdev->bd_part_count--; | |
1336 | ||
2e7b651d PZ |
1337 | if (!--bdev->bd_openers) { |
1338 | sync_blockdev(bdev); | |
1339 | kill_bdev(bdev); | |
1340 | } | |
1341 | if (bdev->bd_contains == bdev) { | |
1342 | if (disk->fops->release) | |
9a1c3542 | 1343 | ret = disk->fops->release(disk, mode); |
2e7b651d PZ |
1344 | } |
1345 | if (!bdev->bd_openers) { | |
1346 | struct module *owner = disk->fops->owner; | |
1347 | ||
1348 | put_disk(disk); | |
1349 | module_put(owner); | |
0762b8bd TH |
1350 | disk_put_part(bdev->bd_part); |
1351 | bdev->bd_part = NULL; | |
2e7b651d PZ |
1352 | bdev->bd_disk = NULL; |
1353 | bdev->bd_inode->i_data.backing_dev_info = &default_backing_dev_info; | |
37be4124 N |
1354 | if (bdev != bdev->bd_contains) |
1355 | victim = bdev->bd_contains; | |
2e7b651d PZ |
1356 | bdev->bd_contains = NULL; |
1357 | } | |
1358 | unlock_kernel(); | |
1359 | mutex_unlock(&bdev->bd_mutex); | |
1360 | bdput(bdev); | |
37be4124 | 1361 | if (victim) |
9a1c3542 | 1362 | __blkdev_put(victim, mode, 1); |
2e7b651d PZ |
1363 | return ret; |
1364 | } | |
1365 | ||
9a1c3542 | 1366 | int blkdev_put(struct block_device *bdev, fmode_t mode) |
37be4124 | 1367 | { |
9a1c3542 | 1368 | return __blkdev_put(bdev, mode, 0); |
37be4124 | 1369 | } |
2e7b651d PZ |
1370 | EXPORT_SYMBOL(blkdev_put); |
1371 | ||
1da177e4 LT |
1372 | static int blkdev_close(struct inode * inode, struct file * filp) |
1373 | { | |
1374 | struct block_device *bdev = I_BDEV(filp->f_mapping->host); | |
1375 | if (bdev->bd_holder == filp) | |
1376 | bd_release(bdev); | |
9a1c3542 | 1377 | return blkdev_put(bdev, filp->f_mode); |
1da177e4 LT |
1378 | } |
1379 | ||
bb93e3a5 | 1380 | static long block_ioctl(struct file *file, unsigned cmd, unsigned long arg) |
1da177e4 | 1381 | { |
56b26add AV |
1382 | struct block_device *bdev = I_BDEV(file->f_mapping->host); |
1383 | fmode_t mode = file->f_mode; | |
fd4ce1ac CH |
1384 | |
1385 | /* | |
1386 | * O_NDELAY can be altered using fcntl(.., F_SETFL, ..), so we have | |
1387 | * to updated it before every ioctl. | |
1388 | */ | |
56b26add | 1389 | if (file->f_flags & O_NDELAY) |
fd4ce1ac CH |
1390 | mode |= FMODE_NDELAY; |
1391 | else | |
1392 | mode &= ~FMODE_NDELAY; | |
1393 | ||
56b26add | 1394 | return blkdev_ioctl(bdev, mode, cmd, arg); |
1da177e4 LT |
1395 | } |
1396 | ||
87d8fe1e TT |
1397 | /* |
1398 | * Try to release a page associated with block device when the system | |
1399 | * is under memory pressure. | |
1400 | */ | |
1401 | static int blkdev_releasepage(struct page *page, gfp_t wait) | |
1402 | { | |
1403 | struct super_block *super = BDEV_I(page->mapping->host)->bdev.bd_super; | |
1404 | ||
1405 | if (super && super->s_op->bdev_try_to_free_page) | |
1406 | return super->s_op->bdev_try_to_free_page(super, page, wait); | |
1407 | ||
1408 | return try_to_free_buffers(page); | |
1409 | } | |
1410 | ||
4c54ac62 | 1411 | static const struct address_space_operations def_blk_aops = { |
1da177e4 LT |
1412 | .readpage = blkdev_readpage, |
1413 | .writepage = blkdev_writepage, | |
1414 | .sync_page = block_sync_page, | |
6272b5a5 NP |
1415 | .write_begin = blkdev_write_begin, |
1416 | .write_end = blkdev_write_end, | |
1da177e4 | 1417 | .writepages = generic_writepages, |
87d8fe1e | 1418 | .releasepage = blkdev_releasepage, |
1da177e4 LT |
1419 | .direct_IO = blkdev_direct_IO, |
1420 | }; | |
1421 | ||
4b6f5d20 | 1422 | const struct file_operations def_blk_fops = { |
1da177e4 LT |
1423 | .open = blkdev_open, |
1424 | .release = blkdev_close, | |
1425 | .llseek = block_llseek, | |
543ade1f BP |
1426 | .read = do_sync_read, |
1427 | .write = do_sync_write, | |
1da177e4 | 1428 | .aio_read = generic_file_aio_read, |
027445c3 | 1429 | .aio_write = generic_file_aio_write_nolock, |
1da177e4 LT |
1430 | .mmap = generic_file_mmap, |
1431 | .fsync = block_fsync, | |
bb93e3a5 | 1432 | .unlocked_ioctl = block_ioctl, |
1da177e4 LT |
1433 | #ifdef CONFIG_COMPAT |
1434 | .compat_ioctl = compat_blkdev_ioctl, | |
1435 | #endif | |
7f9c51f0 JA |
1436 | .splice_read = generic_file_splice_read, |
1437 | .splice_write = generic_file_splice_write, | |
1da177e4 LT |
1438 | }; |
1439 | ||
1440 | int ioctl_by_bdev(struct block_device *bdev, unsigned cmd, unsigned long arg) | |
1441 | { | |
1442 | int res; | |
1443 | mm_segment_t old_fs = get_fs(); | |
1444 | set_fs(KERNEL_DS); | |
56b26add | 1445 | res = blkdev_ioctl(bdev, 0, cmd, arg); |
1da177e4 LT |
1446 | set_fs(old_fs); |
1447 | return res; | |
1448 | } | |
1449 | ||
1450 | EXPORT_SYMBOL(ioctl_by_bdev); | |
1451 | ||
1452 | /** | |
1453 | * lookup_bdev - lookup a struct block_device by name | |
94e2959e | 1454 | * @pathname: special file representing the block device |
1da177e4 | 1455 | * |
57d1b536 | 1456 | * Get a reference to the blockdevice at @pathname in the current |
1da177e4 LT |
1457 | * namespace if possible and return it. Return ERR_PTR(error) |
1458 | * otherwise. | |
1459 | */ | |
421748ec | 1460 | struct block_device *lookup_bdev(const char *pathname) |
1da177e4 LT |
1461 | { |
1462 | struct block_device *bdev; | |
1463 | struct inode *inode; | |
421748ec | 1464 | struct path path; |
1da177e4 LT |
1465 | int error; |
1466 | ||
421748ec | 1467 | if (!pathname || !*pathname) |
1da177e4 LT |
1468 | return ERR_PTR(-EINVAL); |
1469 | ||
421748ec | 1470 | error = kern_path(pathname, LOOKUP_FOLLOW, &path); |
1da177e4 LT |
1471 | if (error) |
1472 | return ERR_PTR(error); | |
1473 | ||
421748ec | 1474 | inode = path.dentry->d_inode; |
1da177e4 LT |
1475 | error = -ENOTBLK; |
1476 | if (!S_ISBLK(inode->i_mode)) | |
1477 | goto fail; | |
1478 | error = -EACCES; | |
421748ec | 1479 | if (path.mnt->mnt_flags & MNT_NODEV) |
1da177e4 LT |
1480 | goto fail; |
1481 | error = -ENOMEM; | |
1482 | bdev = bd_acquire(inode); | |
1483 | if (!bdev) | |
1484 | goto fail; | |
1485 | out: | |
421748ec | 1486 | path_put(&path); |
1da177e4 LT |
1487 | return bdev; |
1488 | fail: | |
1489 | bdev = ERR_PTR(error); | |
1490 | goto out; | |
1491 | } | |
d5686b44 | 1492 | EXPORT_SYMBOL(lookup_bdev); |
1da177e4 LT |
1493 | |
1494 | /** | |
30c40d2c | 1495 | * open_bdev_exclusive - open a block device by name and set it up for use |
1da177e4 LT |
1496 | * |
1497 | * @path: special file representing the block device | |
30c40d2c | 1498 | * @mode: FMODE_... combination to pass be used |
1da177e4 LT |
1499 | * @holder: owner for exclusion |
1500 | * | |
1501 | * Open the blockdevice described by the special file at @path, claim it | |
1502 | * for the @holder. | |
1503 | */ | |
30c40d2c | 1504 | struct block_device *open_bdev_exclusive(const char *path, fmode_t mode, void *holder) |
1da177e4 LT |
1505 | { |
1506 | struct block_device *bdev; | |
1da177e4 LT |
1507 | int error = 0; |
1508 | ||
1509 | bdev = lookup_bdev(path); | |
1510 | if (IS_ERR(bdev)) | |
1511 | return bdev; | |
1512 | ||
572c4892 | 1513 | error = blkdev_get(bdev, mode); |
1da177e4 LT |
1514 | if (error) |
1515 | return ERR_PTR(error); | |
1516 | error = -EACCES; | |
30c40d2c | 1517 | if ((mode & FMODE_WRITE) && bdev_read_only(bdev)) |
1da177e4 LT |
1518 | goto blkdev_put; |
1519 | error = bd_claim(bdev, holder); | |
1520 | if (error) | |
1521 | goto blkdev_put; | |
1522 | ||
1523 | return bdev; | |
1524 | ||
1525 | blkdev_put: | |
9a1c3542 | 1526 | blkdev_put(bdev, mode); |
1da177e4 LT |
1527 | return ERR_PTR(error); |
1528 | } | |
1529 | ||
30c40d2c | 1530 | EXPORT_SYMBOL(open_bdev_exclusive); |
1da177e4 LT |
1531 | |
1532 | /** | |
30c40d2c | 1533 | * close_bdev_exclusive - close a blockdevice opened by open_bdev_exclusive() |
1da177e4 LT |
1534 | * |
1535 | * @bdev: blockdevice to close | |
30c40d2c | 1536 | * @mode: mode, must match that used to open. |
1da177e4 | 1537 | * |
30c40d2c | 1538 | * This is the counterpart to open_bdev_exclusive(). |
1da177e4 | 1539 | */ |
30c40d2c | 1540 | void close_bdev_exclusive(struct block_device *bdev, fmode_t mode) |
1da177e4 LT |
1541 | { |
1542 | bd_release(bdev); | |
30c40d2c | 1543 | blkdev_put(bdev, mode); |
1da177e4 LT |
1544 | } |
1545 | ||
30c40d2c | 1546 | EXPORT_SYMBOL(close_bdev_exclusive); |
b71e8a4c DH |
1547 | |
1548 | int __invalidate_device(struct block_device *bdev) | |
1549 | { | |
1550 | struct super_block *sb = get_super(bdev); | |
1551 | int res = 0; | |
1552 | ||
1553 | if (sb) { | |
1554 | /* | |
1555 | * no need to lock the super, get_super holds the | |
1556 | * read mutex so the filesystem cannot go away | |
1557 | * under us (->put_super runs with the write lock | |
1558 | * hold). | |
1559 | */ | |
1560 | shrink_dcache_sb(sb); | |
1561 | res = invalidate_inodes(sb); | |
1562 | drop_super(sb); | |
1563 | } | |
f98393a6 | 1564 | invalidate_bdev(bdev); |
b71e8a4c DH |
1565 | return res; |
1566 | } | |
1567 | EXPORT_SYMBOL(__invalidate_device); |