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> | |
7db9cfd3 | 14 | #include <linux/device_cgroup.h> |
1da177e4 LT |
15 | #include <linux/highmem.h> |
16 | #include <linux/blkdev.h> | |
17 | #include <linux/module.h> | |
18 | #include <linux/blkpg.h> | |
b502bd11 | 19 | #include <linux/magic.h> |
1da177e4 | 20 | #include <linux/buffer_head.h> |
ff01bb48 | 21 | #include <linux/swap.h> |
585d3bc0 | 22 | #include <linux/pagevec.h> |
811d736f | 23 | #include <linux/writeback.h> |
1da177e4 LT |
24 | #include <linux/mpage.h> |
25 | #include <linux/mount.h> | |
26 | #include <linux/uio.h> | |
27 | #include <linux/namei.h> | |
1368c4f2 | 28 | #include <linux/log2.h> |
ff01bb48 | 29 | #include <linux/cleancache.h> |
a27bb332 | 30 | #include <linux/aio.h> |
1da177e4 | 31 | #include <asm/uaccess.h> |
07f3f05c | 32 | #include "internal.h" |
1da177e4 LT |
33 | |
34 | struct bdev_inode { | |
35 | struct block_device bdev; | |
36 | struct inode vfs_inode; | |
37 | }; | |
38 | ||
4c54ac62 AB |
39 | static const struct address_space_operations def_blk_aops; |
40 | ||
1da177e4 LT |
41 | static inline struct bdev_inode *BDEV_I(struct inode *inode) |
42 | { | |
43 | return container_of(inode, struct bdev_inode, vfs_inode); | |
44 | } | |
45 | ||
46 | inline struct block_device *I_BDEV(struct inode *inode) | |
47 | { | |
48 | return &BDEV_I(inode)->bdev; | |
49 | } | |
1da177e4 LT |
50 | EXPORT_SYMBOL(I_BDEV); |
51 | ||
a5491e0c | 52 | /* |
f758eeab CH |
53 | * Move the inode from its current bdi to a new bdi. If the inode is dirty we |
54 | * need to move it onto the dirty list of @dst so that the inode is always on | |
55 | * the right list. | |
a5491e0c DC |
56 | */ |
57 | static void bdev_inode_switch_bdi(struct inode *inode, | |
58 | struct backing_dev_info *dst) | |
59 | { | |
f758eeab | 60 | struct backing_dev_info *old = inode->i_data.backing_dev_info; |
a5faeaf9 | 61 | bool wakeup_bdi = false; |
f758eeab CH |
62 | |
63 | if (unlikely(dst == old)) /* deadlock avoidance */ | |
64 | return; | |
65 | bdi_lock_two(&old->wb, &dst->wb); | |
250df6ed | 66 | spin_lock(&inode->i_lock); |
a5491e0c | 67 | inode->i_data.backing_dev_info = dst; |
a5faeaf9 JK |
68 | if (inode->i_state & I_DIRTY) { |
69 | if (bdi_cap_writeback_dirty(dst) && !wb_has_dirty_io(&dst->wb)) | |
70 | wakeup_bdi = true; | |
7ccf19a8 | 71 | list_move(&inode->i_wb_list, &dst->wb.b_dirty); |
a5faeaf9 | 72 | } |
250df6ed | 73 | spin_unlock(&inode->i_lock); |
f758eeab CH |
74 | spin_unlock(&old->wb.list_lock); |
75 | spin_unlock(&dst->wb.list_lock); | |
a5faeaf9 JK |
76 | |
77 | if (wakeup_bdi) | |
78 | bdi_wakeup_thread_delayed(dst); | |
a5491e0c DC |
79 | } |
80 | ||
f9a14399 | 81 | /* Kill _all_ buffers and pagecache , dirty or not.. */ |
ff01bb48 | 82 | void kill_bdev(struct block_device *bdev) |
1da177e4 | 83 | { |
ff01bb48 AV |
84 | struct address_space *mapping = bdev->bd_inode->i_mapping; |
85 | ||
91b0abe3 | 86 | if (mapping->nrpages == 0 && mapping->nrshadows == 0) |
f9a14399 | 87 | return; |
ff01bb48 | 88 | |
f9a14399 | 89 | invalidate_bh_lrus(); |
ff01bb48 | 90 | truncate_inode_pages(mapping, 0); |
1da177e4 | 91 | } |
ff01bb48 AV |
92 | EXPORT_SYMBOL(kill_bdev); |
93 | ||
94 | /* Invalidate clean unused buffers and pagecache. */ | |
95 | void invalidate_bdev(struct block_device *bdev) | |
96 | { | |
97 | struct address_space *mapping = bdev->bd_inode->i_mapping; | |
98 | ||
99 | if (mapping->nrpages == 0) | |
100 | return; | |
101 | ||
102 | invalidate_bh_lrus(); | |
103 | lru_add_drain_all(); /* make sure all lru add caches are flushed */ | |
104 | invalidate_mapping_pages(mapping, 0, -1); | |
105 | /* 99% of the time, we don't need to flush the cleancache on the bdev. | |
106 | * But, for the strange corners, lets be cautious | |
107 | */ | |
3167760f | 108 | cleancache_invalidate_inode(mapping); |
ff01bb48 AV |
109 | } |
110 | EXPORT_SYMBOL(invalidate_bdev); | |
1da177e4 LT |
111 | |
112 | int set_blocksize(struct block_device *bdev, int size) | |
113 | { | |
114 | /* Size must be a power of two, and between 512 and PAGE_SIZE */ | |
1368c4f2 | 115 | if (size > PAGE_SIZE || size < 512 || !is_power_of_2(size)) |
1da177e4 LT |
116 | return -EINVAL; |
117 | ||
118 | /* Size cannot be smaller than the size supported by the device */ | |
e1defc4f | 119 | if (size < bdev_logical_block_size(bdev)) |
1da177e4 LT |
120 | return -EINVAL; |
121 | ||
122 | /* Don't change the size if it is same as current */ | |
123 | if (bdev->bd_block_size != size) { | |
124 | sync_blockdev(bdev); | |
125 | bdev->bd_block_size = size; | |
126 | bdev->bd_inode->i_blkbits = blksize_bits(size); | |
127 | kill_bdev(bdev); | |
128 | } | |
129 | return 0; | |
130 | } | |
131 | ||
132 | EXPORT_SYMBOL(set_blocksize); | |
133 | ||
134 | int sb_set_blocksize(struct super_block *sb, int size) | |
135 | { | |
1da177e4 LT |
136 | if (set_blocksize(sb->s_bdev, size)) |
137 | return 0; | |
138 | /* If we get here, we know size is power of two | |
139 | * and it's value is between 512 and PAGE_SIZE */ | |
140 | sb->s_blocksize = size; | |
38885bd4 | 141 | sb->s_blocksize_bits = blksize_bits(size); |
1da177e4 LT |
142 | return sb->s_blocksize; |
143 | } | |
144 | ||
145 | EXPORT_SYMBOL(sb_set_blocksize); | |
146 | ||
147 | int sb_min_blocksize(struct super_block *sb, int size) | |
148 | { | |
e1defc4f | 149 | int minsize = bdev_logical_block_size(sb->s_bdev); |
1da177e4 LT |
150 | if (size < minsize) |
151 | size = minsize; | |
152 | return sb_set_blocksize(sb, size); | |
153 | } | |
154 | ||
155 | EXPORT_SYMBOL(sb_min_blocksize); | |
156 | ||
157 | static int | |
158 | blkdev_get_block(struct inode *inode, sector_t iblock, | |
159 | struct buffer_head *bh, int create) | |
160 | { | |
1da177e4 LT |
161 | bh->b_bdev = I_BDEV(inode); |
162 | bh->b_blocknr = iblock; | |
163 | set_buffer_mapped(bh); | |
164 | return 0; | |
165 | } | |
166 | ||
b2e895db | 167 | static ssize_t |
d8d3d94b AV |
168 | blkdev_direct_IO(int rw, struct kiocb *iocb, struct iov_iter *iter, |
169 | loff_t offset) | |
b2e895db AM |
170 | { |
171 | struct file *file = iocb->ki_filp; | |
172 | struct inode *inode = file->f_mapping->host; | |
173 | ||
31b14039 AV |
174 | return __blockdev_direct_IO(rw, iocb, inode, I_BDEV(inode), iter, |
175 | offset, blkdev_get_block, | |
d8d3d94b | 176 | NULL, NULL, 0); |
b2e895db AM |
177 | } |
178 | ||
5cee5815 JK |
179 | int __sync_blockdev(struct block_device *bdev, int wait) |
180 | { | |
181 | if (!bdev) | |
182 | return 0; | |
183 | if (!wait) | |
184 | return filemap_flush(bdev->bd_inode->i_mapping); | |
185 | return filemap_write_and_wait(bdev->bd_inode->i_mapping); | |
186 | } | |
187 | ||
585d3bc0 NP |
188 | /* |
189 | * Write out and wait upon all the dirty data associated with a block | |
190 | * device via its mapping. Does not take the superblock lock. | |
191 | */ | |
192 | int sync_blockdev(struct block_device *bdev) | |
193 | { | |
5cee5815 | 194 | return __sync_blockdev(bdev, 1); |
585d3bc0 NP |
195 | } |
196 | EXPORT_SYMBOL(sync_blockdev); | |
197 | ||
198 | /* | |
199 | * Write out and wait upon all dirty data associated with this | |
200 | * device. Filesystem data as well as the underlying block | |
201 | * device. Takes the superblock lock. | |
202 | */ | |
203 | int fsync_bdev(struct block_device *bdev) | |
204 | { | |
205 | struct super_block *sb = get_super(bdev); | |
206 | if (sb) { | |
60b0680f | 207 | int res = sync_filesystem(sb); |
585d3bc0 NP |
208 | drop_super(sb); |
209 | return res; | |
210 | } | |
211 | return sync_blockdev(bdev); | |
212 | } | |
47e4491b | 213 | EXPORT_SYMBOL(fsync_bdev); |
585d3bc0 NP |
214 | |
215 | /** | |
216 | * freeze_bdev -- lock a filesystem and force it into a consistent state | |
217 | * @bdev: blockdevice to lock | |
218 | * | |
585d3bc0 NP |
219 | * If a superblock is found on this device, we take the s_umount semaphore |
220 | * on it to make sure nobody unmounts until the snapshot creation is done. | |
221 | * The reference counter (bd_fsfreeze_count) guarantees that only the last | |
222 | * unfreeze process can unfreeze the frozen filesystem actually when multiple | |
223 | * freeze requests arrive simultaneously. It counts up in freeze_bdev() and | |
224 | * count down in thaw_bdev(). When it becomes 0, thaw_bdev() will unfreeze | |
225 | * actually. | |
226 | */ | |
227 | struct super_block *freeze_bdev(struct block_device *bdev) | |
228 | { | |
229 | struct super_block *sb; | |
230 | int error = 0; | |
231 | ||
232 | mutex_lock(&bdev->bd_fsfreeze_mutex); | |
4504230a CH |
233 | if (++bdev->bd_fsfreeze_count > 1) { |
234 | /* | |
235 | * We don't even need to grab a reference - the first call | |
236 | * to freeze_bdev grab an active reference and only the last | |
237 | * thaw_bdev drops it. | |
238 | */ | |
585d3bc0 | 239 | sb = get_super(bdev); |
4504230a CH |
240 | drop_super(sb); |
241 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
242 | return sb; | |
243 | } | |
244 | ||
245 | sb = get_active_super(bdev); | |
246 | if (!sb) | |
247 | goto out; | |
18e9e510 JB |
248 | error = freeze_super(sb); |
249 | if (error) { | |
250 | deactivate_super(sb); | |
251 | bdev->bd_fsfreeze_count--; | |
585d3bc0 | 252 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
18e9e510 | 253 | return ERR_PTR(error); |
585d3bc0 | 254 | } |
18e9e510 | 255 | deactivate_super(sb); |
4504230a | 256 | out: |
585d3bc0 NP |
257 | sync_blockdev(bdev); |
258 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
4fadd7bb | 259 | return sb; /* thaw_bdev releases s->s_umount */ |
585d3bc0 NP |
260 | } |
261 | EXPORT_SYMBOL(freeze_bdev); | |
262 | ||
263 | /** | |
264 | * thaw_bdev -- unlock filesystem | |
265 | * @bdev: blockdevice to unlock | |
266 | * @sb: associated superblock | |
267 | * | |
268 | * Unlocks the filesystem and marks it writeable again after freeze_bdev(). | |
269 | */ | |
270 | int thaw_bdev(struct block_device *bdev, struct super_block *sb) | |
271 | { | |
4504230a | 272 | int error = -EINVAL; |
585d3bc0 NP |
273 | |
274 | mutex_lock(&bdev->bd_fsfreeze_mutex); | |
4504230a | 275 | if (!bdev->bd_fsfreeze_count) |
18e9e510 | 276 | goto out; |
4504230a CH |
277 | |
278 | error = 0; | |
279 | if (--bdev->bd_fsfreeze_count > 0) | |
18e9e510 | 280 | goto out; |
4504230a CH |
281 | |
282 | if (!sb) | |
18e9e510 | 283 | goto out; |
4504230a | 284 | |
18e9e510 JB |
285 | error = thaw_super(sb); |
286 | if (error) { | |
287 | bdev->bd_fsfreeze_count++; | |
288 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
289 | return error; | |
290 | } | |
291 | out: | |
585d3bc0 NP |
292 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
293 | return 0; | |
294 | } | |
295 | EXPORT_SYMBOL(thaw_bdev); | |
296 | ||
1da177e4 LT |
297 | static int blkdev_writepage(struct page *page, struct writeback_control *wbc) |
298 | { | |
299 | return block_write_full_page(page, blkdev_get_block, wbc); | |
300 | } | |
301 | ||
302 | static int blkdev_readpage(struct file * file, struct page * page) | |
303 | { | |
304 | return block_read_full_page(page, blkdev_get_block); | |
305 | } | |
306 | ||
447f05bb AM |
307 | static int blkdev_readpages(struct file *file, struct address_space *mapping, |
308 | struct list_head *pages, unsigned nr_pages) | |
309 | { | |
310 | return mpage_readpages(mapping, pages, nr_pages, blkdev_get_block); | |
311 | } | |
312 | ||
6272b5a5 NP |
313 | static int blkdev_write_begin(struct file *file, struct address_space *mapping, |
314 | loff_t pos, unsigned len, unsigned flags, | |
315 | struct page **pagep, void **fsdata) | |
1da177e4 | 316 | { |
155130a4 CH |
317 | return block_write_begin(mapping, pos, len, flags, pagep, |
318 | blkdev_get_block); | |
1da177e4 LT |
319 | } |
320 | ||
6272b5a5 NP |
321 | static int blkdev_write_end(struct file *file, struct address_space *mapping, |
322 | loff_t pos, unsigned len, unsigned copied, | |
323 | struct page *page, void *fsdata) | |
1da177e4 | 324 | { |
6272b5a5 NP |
325 | int ret; |
326 | ret = block_write_end(file, mapping, pos, len, copied, page, fsdata); | |
327 | ||
328 | unlock_page(page); | |
329 | page_cache_release(page); | |
330 | ||
331 | return ret; | |
1da177e4 LT |
332 | } |
333 | ||
334 | /* | |
335 | * private llseek: | |
496ad9aa | 336 | * for a block special file file_inode(file)->i_size is zero |
1da177e4 LT |
337 | * so we compute the size by hand (just as in block_read/write above) |
338 | */ | |
965c8e59 | 339 | static loff_t block_llseek(struct file *file, loff_t offset, int whence) |
1da177e4 LT |
340 | { |
341 | struct inode *bd_inode = file->f_mapping->host; | |
1da177e4 LT |
342 | loff_t retval; |
343 | ||
1b1dcc1b | 344 | mutex_lock(&bd_inode->i_mutex); |
5d48f3a2 | 345 | retval = fixed_size_llseek(file, offset, whence, i_size_read(bd_inode)); |
1b1dcc1b | 346 | mutex_unlock(&bd_inode->i_mutex); |
1da177e4 LT |
347 | return retval; |
348 | } | |
349 | ||
02c24a82 | 350 | int blkdev_fsync(struct file *filp, loff_t start, loff_t end, int datasync) |
1da177e4 | 351 | { |
b8af67e2 AB |
352 | struct inode *bd_inode = filp->f_mapping->host; |
353 | struct block_device *bdev = I_BDEV(bd_inode); | |
ab0a9735 | 354 | int error; |
da5aa861 RW |
355 | |
356 | error = filemap_write_and_wait_range(filp->f_mapping, start, end); | |
357 | if (error) | |
358 | return error; | |
ab0a9735 | 359 | |
b8af67e2 AB |
360 | /* |
361 | * There is no need to serialise calls to blkdev_issue_flush with | |
362 | * i_mutex and doing so causes performance issues with concurrent | |
363 | * O_SYNC writers to a block device. | |
364 | */ | |
dd3932ed | 365 | error = blkdev_issue_flush(bdev, GFP_KERNEL, NULL); |
ab0a9735 CH |
366 | if (error == -EOPNOTSUPP) |
367 | error = 0; | |
b8af67e2 | 368 | |
ab0a9735 | 369 | return error; |
1da177e4 | 370 | } |
b1dd3b28 | 371 | EXPORT_SYMBOL(blkdev_fsync); |
1da177e4 | 372 | |
47a191fd MW |
373 | /** |
374 | * bdev_read_page() - Start reading a page from a block device | |
375 | * @bdev: The device to read the page from | |
376 | * @sector: The offset on the device to read the page to (need not be aligned) | |
377 | * @page: The page to read | |
378 | * | |
379 | * On entry, the page should be locked. It will be unlocked when the page | |
380 | * has been read. If the block driver implements rw_page synchronously, | |
381 | * that will be true on exit from this function, but it need not be. | |
382 | * | |
383 | * Errors returned by this function are usually "soft", eg out of memory, or | |
384 | * queue full; callers should try a different route to read this page rather | |
385 | * than propagate an error back up the stack. | |
386 | * | |
387 | * Return: negative errno if an error occurs, 0 if submission was successful. | |
388 | */ | |
389 | int bdev_read_page(struct block_device *bdev, sector_t sector, | |
390 | struct page *page) | |
391 | { | |
392 | const struct block_device_operations *ops = bdev->bd_disk->fops; | |
393 | if (!ops->rw_page) | |
394 | return -EOPNOTSUPP; | |
395 | return ops->rw_page(bdev, sector + get_start_sect(bdev), page, READ); | |
396 | } | |
397 | EXPORT_SYMBOL_GPL(bdev_read_page); | |
398 | ||
399 | /** | |
400 | * bdev_write_page() - Start writing a page to a block device | |
401 | * @bdev: The device to write the page to | |
402 | * @sector: The offset on the device to write the page to (need not be aligned) | |
403 | * @page: The page to write | |
404 | * @wbc: The writeback_control for the write | |
405 | * | |
406 | * On entry, the page should be locked and not currently under writeback. | |
407 | * On exit, if the write started successfully, the page will be unlocked and | |
408 | * under writeback. If the write failed already (eg the driver failed to | |
409 | * queue the page to the device), the page will still be locked. If the | |
410 | * caller is a ->writepage implementation, it will need to unlock the page. | |
411 | * | |
412 | * Errors returned by this function are usually "soft", eg out of memory, or | |
413 | * queue full; callers should try a different route to write this page rather | |
414 | * than propagate an error back up the stack. | |
415 | * | |
416 | * Return: negative errno if an error occurs, 0 if submission was successful. | |
417 | */ | |
418 | int bdev_write_page(struct block_device *bdev, sector_t sector, | |
419 | struct page *page, struct writeback_control *wbc) | |
420 | { | |
421 | int result; | |
422 | int rw = (wbc->sync_mode == WB_SYNC_ALL) ? WRITE_SYNC : WRITE; | |
423 | const struct block_device_operations *ops = bdev->bd_disk->fops; | |
424 | if (!ops->rw_page) | |
425 | return -EOPNOTSUPP; | |
426 | set_page_writeback(page); | |
427 | result = ops->rw_page(bdev, sector + get_start_sect(bdev), page, rw); | |
428 | if (result) | |
429 | end_page_writeback(page); | |
430 | else | |
431 | unlock_page(page); | |
432 | return result; | |
433 | } | |
434 | EXPORT_SYMBOL_GPL(bdev_write_page); | |
435 | ||
1da177e4 LT |
436 | /* |
437 | * pseudo-fs | |
438 | */ | |
439 | ||
440 | static __cacheline_aligned_in_smp DEFINE_SPINLOCK(bdev_lock); | |
e18b890b | 441 | static struct kmem_cache * bdev_cachep __read_mostly; |
1da177e4 LT |
442 | |
443 | static struct inode *bdev_alloc_inode(struct super_block *sb) | |
444 | { | |
e94b1766 | 445 | struct bdev_inode *ei = kmem_cache_alloc(bdev_cachep, GFP_KERNEL); |
1da177e4 LT |
446 | if (!ei) |
447 | return NULL; | |
448 | return &ei->vfs_inode; | |
449 | } | |
450 | ||
fa0d7e3d | 451 | static void bdev_i_callback(struct rcu_head *head) |
1da177e4 | 452 | { |
fa0d7e3d | 453 | struct inode *inode = container_of(head, struct inode, i_rcu); |
1da177e4 LT |
454 | struct bdev_inode *bdi = BDEV_I(inode); |
455 | ||
1da177e4 LT |
456 | kmem_cache_free(bdev_cachep, bdi); |
457 | } | |
458 | ||
fa0d7e3d NP |
459 | static void bdev_destroy_inode(struct inode *inode) |
460 | { | |
461 | call_rcu(&inode->i_rcu, bdev_i_callback); | |
462 | } | |
463 | ||
51cc5068 | 464 | static void init_once(void *foo) |
1da177e4 LT |
465 | { |
466 | struct bdev_inode *ei = (struct bdev_inode *) foo; | |
467 | struct block_device *bdev = &ei->bdev; | |
468 | ||
a35afb83 CL |
469 | memset(bdev, 0, sizeof(*bdev)); |
470 | mutex_init(&bdev->bd_mutex); | |
a35afb83 CL |
471 | INIT_LIST_HEAD(&bdev->bd_inodes); |
472 | INIT_LIST_HEAD(&bdev->bd_list); | |
49731baa TH |
473 | #ifdef CONFIG_SYSFS |
474 | INIT_LIST_HEAD(&bdev->bd_holder_disks); | |
475 | #endif | |
a35afb83 | 476 | inode_init_once(&ei->vfs_inode); |
fcccf502 TS |
477 | /* Initialize mutex for freeze. */ |
478 | mutex_init(&bdev->bd_fsfreeze_mutex); | |
1da177e4 LT |
479 | } |
480 | ||
481 | static inline void __bd_forget(struct inode *inode) | |
482 | { | |
483 | list_del_init(&inode->i_devices); | |
484 | inode->i_bdev = NULL; | |
485 | inode->i_mapping = &inode->i_data; | |
486 | } | |
487 | ||
b57922d9 | 488 | static void bdev_evict_inode(struct inode *inode) |
1da177e4 LT |
489 | { |
490 | struct block_device *bdev = &BDEV_I(inode)->bdev; | |
491 | struct list_head *p; | |
91b0abe3 | 492 | truncate_inode_pages_final(&inode->i_data); |
b57922d9 | 493 | invalidate_inode_buffers(inode); /* is it needed here? */ |
dbd5768f | 494 | clear_inode(inode); |
1da177e4 LT |
495 | spin_lock(&bdev_lock); |
496 | while ( (p = bdev->bd_inodes.next) != &bdev->bd_inodes ) { | |
497 | __bd_forget(list_entry(p, struct inode, i_devices)); | |
498 | } | |
499 | list_del_init(&bdev->bd_list); | |
500 | spin_unlock(&bdev_lock); | |
501 | } | |
502 | ||
ee9b6d61 | 503 | static const struct super_operations bdev_sops = { |
1da177e4 LT |
504 | .statfs = simple_statfs, |
505 | .alloc_inode = bdev_alloc_inode, | |
506 | .destroy_inode = bdev_destroy_inode, | |
507 | .drop_inode = generic_delete_inode, | |
b57922d9 | 508 | .evict_inode = bdev_evict_inode, |
1da177e4 LT |
509 | }; |
510 | ||
51139ada AV |
511 | static struct dentry *bd_mount(struct file_system_type *fs_type, |
512 | int flags, const char *dev_name, void *data) | |
1da177e4 | 513 | { |
b502bd11 | 514 | return mount_pseudo(fs_type, "bdev:", &bdev_sops, NULL, BDEVFS_MAGIC); |
1da177e4 LT |
515 | } |
516 | ||
517 | static struct file_system_type bd_type = { | |
518 | .name = "bdev", | |
51139ada | 519 | .mount = bd_mount, |
1da177e4 LT |
520 | .kill_sb = kill_anon_super, |
521 | }; | |
522 | ||
f47ec3f2 | 523 | static struct super_block *blockdev_superblock __read_mostly; |
1da177e4 LT |
524 | |
525 | void __init bdev_cache_init(void) | |
526 | { | |
527 | int err; | |
ace8577a | 528 | static struct vfsmount *bd_mnt; |
c2acf7b9 | 529 | |
1da177e4 | 530 | bdev_cachep = kmem_cache_create("bdev_cache", sizeof(struct bdev_inode), |
fffb60f9 PJ |
531 | 0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| |
532 | SLAB_MEM_SPREAD|SLAB_PANIC), | |
20c2df83 | 533 | init_once); |
1da177e4 LT |
534 | err = register_filesystem(&bd_type); |
535 | if (err) | |
536 | panic("Cannot register bdev pseudo-fs"); | |
537 | bd_mnt = kern_mount(&bd_type); | |
1da177e4 LT |
538 | if (IS_ERR(bd_mnt)) |
539 | panic("Cannot create bdev pseudo-fs"); | |
ace8577a | 540 | blockdev_superblock = bd_mnt->mnt_sb; /* For writeback */ |
1da177e4 LT |
541 | } |
542 | ||
543 | /* | |
544 | * Most likely _very_ bad one - but then it's hardly critical for small | |
545 | * /dev and can be fixed when somebody will need really large one. | |
546 | * Keep in mind that it will be fed through icache hash function too. | |
547 | */ | |
548 | static inline unsigned long hash(dev_t dev) | |
549 | { | |
550 | return MAJOR(dev)+MINOR(dev); | |
551 | } | |
552 | ||
553 | static int bdev_test(struct inode *inode, void *data) | |
554 | { | |
555 | return BDEV_I(inode)->bdev.bd_dev == *(dev_t *)data; | |
556 | } | |
557 | ||
558 | static int bdev_set(struct inode *inode, void *data) | |
559 | { | |
560 | BDEV_I(inode)->bdev.bd_dev = *(dev_t *)data; | |
561 | return 0; | |
562 | } | |
563 | ||
564 | static LIST_HEAD(all_bdevs); | |
565 | ||
566 | struct block_device *bdget(dev_t dev) | |
567 | { | |
568 | struct block_device *bdev; | |
569 | struct inode *inode; | |
570 | ||
c2acf7b9 | 571 | inode = iget5_locked(blockdev_superblock, hash(dev), |
1da177e4 LT |
572 | bdev_test, bdev_set, &dev); |
573 | ||
574 | if (!inode) | |
575 | return NULL; | |
576 | ||
577 | bdev = &BDEV_I(inode)->bdev; | |
578 | ||
579 | if (inode->i_state & I_NEW) { | |
580 | bdev->bd_contains = NULL; | |
782b94cd | 581 | bdev->bd_super = NULL; |
1da177e4 LT |
582 | bdev->bd_inode = inode; |
583 | bdev->bd_block_size = (1 << inode->i_blkbits); | |
584 | bdev->bd_part_count = 0; | |
585 | bdev->bd_invalidated = 0; | |
586 | inode->i_mode = S_IFBLK; | |
587 | inode->i_rdev = dev; | |
588 | inode->i_bdev = bdev; | |
589 | inode->i_data.a_ops = &def_blk_aops; | |
590 | mapping_set_gfp_mask(&inode->i_data, GFP_USER); | |
591 | inode->i_data.backing_dev_info = &default_backing_dev_info; | |
592 | spin_lock(&bdev_lock); | |
593 | list_add(&bdev->bd_list, &all_bdevs); | |
594 | spin_unlock(&bdev_lock); | |
595 | unlock_new_inode(inode); | |
596 | } | |
597 | return bdev; | |
598 | } | |
599 | ||
600 | EXPORT_SYMBOL(bdget); | |
601 | ||
dddac6a7 AJ |
602 | /** |
603 | * bdgrab -- Grab a reference to an already referenced block device | |
604 | * @bdev: Block device to grab a reference to. | |
605 | */ | |
606 | struct block_device *bdgrab(struct block_device *bdev) | |
607 | { | |
7de9c6ee | 608 | ihold(bdev->bd_inode); |
dddac6a7 AJ |
609 | return bdev; |
610 | } | |
c1681bf8 | 611 | EXPORT_SYMBOL(bdgrab); |
dddac6a7 | 612 | |
1da177e4 LT |
613 | long nr_blockdev_pages(void) |
614 | { | |
203a2935 | 615 | struct block_device *bdev; |
1da177e4 LT |
616 | long ret = 0; |
617 | spin_lock(&bdev_lock); | |
203a2935 | 618 | list_for_each_entry(bdev, &all_bdevs, bd_list) { |
1da177e4 LT |
619 | ret += bdev->bd_inode->i_mapping->nrpages; |
620 | } | |
621 | spin_unlock(&bdev_lock); | |
622 | return ret; | |
623 | } | |
624 | ||
625 | void bdput(struct block_device *bdev) | |
626 | { | |
627 | iput(bdev->bd_inode); | |
628 | } | |
629 | ||
630 | EXPORT_SYMBOL(bdput); | |
631 | ||
632 | static struct block_device *bd_acquire(struct inode *inode) | |
633 | { | |
634 | struct block_device *bdev; | |
09d967c6 | 635 | |
1da177e4 LT |
636 | spin_lock(&bdev_lock); |
637 | bdev = inode->i_bdev; | |
09d967c6 | 638 | if (bdev) { |
7de9c6ee | 639 | ihold(bdev->bd_inode); |
1da177e4 LT |
640 | spin_unlock(&bdev_lock); |
641 | return bdev; | |
642 | } | |
643 | spin_unlock(&bdev_lock); | |
09d967c6 | 644 | |
1da177e4 LT |
645 | bdev = bdget(inode->i_rdev); |
646 | if (bdev) { | |
647 | spin_lock(&bdev_lock); | |
09d967c6 OH |
648 | if (!inode->i_bdev) { |
649 | /* | |
7de9c6ee | 650 | * We take an additional reference to bd_inode, |
09d967c6 OH |
651 | * and it's released in clear_inode() of inode. |
652 | * So, we can access it via ->i_mapping always | |
653 | * without igrab(). | |
654 | */ | |
7de9c6ee | 655 | ihold(bdev->bd_inode); |
09d967c6 OH |
656 | inode->i_bdev = bdev; |
657 | inode->i_mapping = bdev->bd_inode->i_mapping; | |
658 | list_add(&inode->i_devices, &bdev->bd_inodes); | |
659 | } | |
1da177e4 LT |
660 | spin_unlock(&bdev_lock); |
661 | } | |
662 | return bdev; | |
663 | } | |
664 | ||
a8855990 | 665 | int sb_is_blkdev_sb(struct super_block *sb) |
f47ec3f2 AV |
666 | { |
667 | return sb == blockdev_superblock; | |
668 | } | |
669 | ||
1da177e4 LT |
670 | /* Call when you free inode */ |
671 | ||
672 | void bd_forget(struct inode *inode) | |
673 | { | |
09d967c6 OH |
674 | struct block_device *bdev = NULL; |
675 | ||
1da177e4 | 676 | spin_lock(&bdev_lock); |
b4ea2eaa YH |
677 | if (!sb_is_blkdev_sb(inode->i_sb)) |
678 | bdev = inode->i_bdev; | |
679 | __bd_forget(inode); | |
1da177e4 | 680 | spin_unlock(&bdev_lock); |
09d967c6 OH |
681 | |
682 | if (bdev) | |
683 | iput(bdev->bd_inode); | |
1da177e4 LT |
684 | } |
685 | ||
1a3cbbc5 TH |
686 | /** |
687 | * bd_may_claim - test whether a block device can be claimed | |
688 | * @bdev: block device of interest | |
689 | * @whole: whole block device containing @bdev, may equal @bdev | |
690 | * @holder: holder trying to claim @bdev | |
691 | * | |
25985edc | 692 | * Test whether @bdev can be claimed by @holder. |
1a3cbbc5 TH |
693 | * |
694 | * CONTEXT: | |
695 | * spin_lock(&bdev_lock). | |
696 | * | |
697 | * RETURNS: | |
698 | * %true if @bdev can be claimed, %false otherwise. | |
699 | */ | |
700 | static bool bd_may_claim(struct block_device *bdev, struct block_device *whole, | |
701 | void *holder) | |
1da177e4 | 702 | { |
1da177e4 | 703 | if (bdev->bd_holder == holder) |
1a3cbbc5 | 704 | return true; /* already a holder */ |
1da177e4 | 705 | else if (bdev->bd_holder != NULL) |
1a3cbbc5 | 706 | return false; /* held by someone else */ |
1da177e4 | 707 | else if (bdev->bd_contains == bdev) |
1a3cbbc5 | 708 | return true; /* is a whole device which isn't held */ |
1da177e4 | 709 | |
e525fd89 | 710 | else if (whole->bd_holder == bd_may_claim) |
1a3cbbc5 TH |
711 | return true; /* is a partition of a device that is being partitioned */ |
712 | else if (whole->bd_holder != NULL) | |
713 | return false; /* is a partition of a held device */ | |
1da177e4 | 714 | else |
1a3cbbc5 TH |
715 | return true; /* is a partition of an un-held device */ |
716 | } | |
717 | ||
6b4517a7 TH |
718 | /** |
719 | * bd_prepare_to_claim - prepare to claim a block device | |
720 | * @bdev: block device of interest | |
721 | * @whole: the whole device containing @bdev, may equal @bdev | |
722 | * @holder: holder trying to claim @bdev | |
723 | * | |
724 | * Prepare to claim @bdev. This function fails if @bdev is already | |
725 | * claimed by another holder and waits if another claiming is in | |
726 | * progress. This function doesn't actually claim. On successful | |
727 | * return, the caller has ownership of bd_claiming and bd_holder[s]. | |
728 | * | |
729 | * CONTEXT: | |
730 | * spin_lock(&bdev_lock). Might release bdev_lock, sleep and regrab | |
731 | * it multiple times. | |
732 | * | |
733 | * RETURNS: | |
734 | * 0 if @bdev can be claimed, -EBUSY otherwise. | |
735 | */ | |
736 | static int bd_prepare_to_claim(struct block_device *bdev, | |
737 | struct block_device *whole, void *holder) | |
738 | { | |
739 | retry: | |
740 | /* if someone else claimed, fail */ | |
741 | if (!bd_may_claim(bdev, whole, holder)) | |
742 | return -EBUSY; | |
743 | ||
e75aa858 TH |
744 | /* if claiming is already in progress, wait for it to finish */ |
745 | if (whole->bd_claiming) { | |
6b4517a7 TH |
746 | wait_queue_head_t *wq = bit_waitqueue(&whole->bd_claiming, 0); |
747 | DEFINE_WAIT(wait); | |
748 | ||
749 | prepare_to_wait(wq, &wait, TASK_UNINTERRUPTIBLE); | |
750 | spin_unlock(&bdev_lock); | |
751 | schedule(); | |
752 | finish_wait(wq, &wait); | |
753 | spin_lock(&bdev_lock); | |
754 | goto retry; | |
755 | } | |
756 | ||
757 | /* yay, all mine */ | |
758 | return 0; | |
759 | } | |
760 | ||
761 | /** | |
762 | * bd_start_claiming - start claiming a block device | |
763 | * @bdev: block device of interest | |
764 | * @holder: holder trying to claim @bdev | |
765 | * | |
766 | * @bdev is about to be opened exclusively. Check @bdev can be opened | |
767 | * exclusively and mark that an exclusive open is in progress. Each | |
768 | * successful call to this function must be matched with a call to | |
b0018361 NP |
769 | * either bd_finish_claiming() or bd_abort_claiming() (which do not |
770 | * fail). | |
771 | * | |
772 | * This function is used to gain exclusive access to the block device | |
773 | * without actually causing other exclusive open attempts to fail. It | |
774 | * should be used when the open sequence itself requires exclusive | |
775 | * access but may subsequently fail. | |
6b4517a7 TH |
776 | * |
777 | * CONTEXT: | |
778 | * Might sleep. | |
779 | * | |
780 | * RETURNS: | |
781 | * Pointer to the block device containing @bdev on success, ERR_PTR() | |
782 | * value on failure. | |
783 | */ | |
784 | static struct block_device *bd_start_claiming(struct block_device *bdev, | |
785 | void *holder) | |
786 | { | |
787 | struct gendisk *disk; | |
788 | struct block_device *whole; | |
789 | int partno, err; | |
790 | ||
791 | might_sleep(); | |
792 | ||
793 | /* | |
794 | * @bdev might not have been initialized properly yet, look up | |
795 | * and grab the outer block device the hard way. | |
796 | */ | |
797 | disk = get_gendisk(bdev->bd_dev, &partno); | |
798 | if (!disk) | |
799 | return ERR_PTR(-ENXIO); | |
800 | ||
d4c208b8 TH |
801 | /* |
802 | * Normally, @bdev should equal what's returned from bdget_disk() | |
803 | * if partno is 0; however, some drivers (floppy) use multiple | |
804 | * bdev's for the same physical device and @bdev may be one of the | |
805 | * aliases. Keep @bdev if partno is 0. This means claimer | |
806 | * tracking is broken for those devices but it has always been that | |
807 | * way. | |
808 | */ | |
809 | if (partno) | |
810 | whole = bdget_disk(disk, 0); | |
811 | else | |
812 | whole = bdgrab(bdev); | |
813 | ||
cf342570 | 814 | module_put(disk->fops->owner); |
6b4517a7 TH |
815 | put_disk(disk); |
816 | if (!whole) | |
817 | return ERR_PTR(-ENOMEM); | |
818 | ||
819 | /* prepare to claim, if successful, mark claiming in progress */ | |
820 | spin_lock(&bdev_lock); | |
821 | ||
822 | err = bd_prepare_to_claim(bdev, whole, holder); | |
823 | if (err == 0) { | |
824 | whole->bd_claiming = holder; | |
825 | spin_unlock(&bdev_lock); | |
826 | return whole; | |
827 | } else { | |
828 | spin_unlock(&bdev_lock); | |
829 | bdput(whole); | |
830 | return ERR_PTR(err); | |
831 | } | |
832 | } | |
833 | ||
641dc636 | 834 | #ifdef CONFIG_SYSFS |
49731baa TH |
835 | struct bd_holder_disk { |
836 | struct list_head list; | |
837 | struct gendisk *disk; | |
838 | int refcnt; | |
839 | }; | |
840 | ||
841 | static struct bd_holder_disk *bd_find_holder_disk(struct block_device *bdev, | |
842 | struct gendisk *disk) | |
843 | { | |
844 | struct bd_holder_disk *holder; | |
845 | ||
846 | list_for_each_entry(holder, &bdev->bd_holder_disks, list) | |
847 | if (holder->disk == disk) | |
848 | return holder; | |
849 | return NULL; | |
850 | } | |
851 | ||
4d7dd8fd | 852 | static int add_symlink(struct kobject *from, struct kobject *to) |
641dc636 | 853 | { |
4d7dd8fd | 854 | return sysfs_create_link(from, to, kobject_name(to)); |
641dc636 JN |
855 | } |
856 | ||
857 | static void del_symlink(struct kobject *from, struct kobject *to) | |
858 | { | |
641dc636 JN |
859 | sysfs_remove_link(from, kobject_name(to)); |
860 | } | |
861 | ||
df6c0cd9 | 862 | /** |
e09b457b TH |
863 | * bd_link_disk_holder - create symlinks between holding disk and slave bdev |
864 | * @bdev: the claimed slave bdev | |
865 | * @disk: the holding disk | |
df6c0cd9 | 866 | * |
49731baa TH |
867 | * DON'T USE THIS UNLESS YOU'RE ALREADY USING IT. |
868 | * | |
e09b457b | 869 | * This functions creates the following sysfs symlinks. |
641dc636 | 870 | * |
e09b457b TH |
871 | * - from "slaves" directory of the holder @disk to the claimed @bdev |
872 | * - from "holders" directory of the @bdev to the holder @disk | |
641dc636 | 873 | * |
e09b457b TH |
874 | * For example, if /dev/dm-0 maps to /dev/sda and disk for dm-0 is |
875 | * passed to bd_link_disk_holder(), then: | |
641dc636 | 876 | * |
e09b457b TH |
877 | * /sys/block/dm-0/slaves/sda --> /sys/block/sda |
878 | * /sys/block/sda/holders/dm-0 --> /sys/block/dm-0 | |
641dc636 | 879 | * |
e09b457b TH |
880 | * The caller must have claimed @bdev before calling this function and |
881 | * ensure that both @bdev and @disk are valid during the creation and | |
882 | * lifetime of these symlinks. | |
641dc636 | 883 | * |
e09b457b TH |
884 | * CONTEXT: |
885 | * Might sleep. | |
641dc636 | 886 | * |
e09b457b TH |
887 | * RETURNS: |
888 | * 0 on success, -errno on failure. | |
641dc636 | 889 | */ |
e09b457b | 890 | int bd_link_disk_holder(struct block_device *bdev, struct gendisk *disk) |
641dc636 | 891 | { |
49731baa | 892 | struct bd_holder_disk *holder; |
e09b457b | 893 | int ret = 0; |
641dc636 | 894 | |
2e7b651d | 895 | mutex_lock(&bdev->bd_mutex); |
df6c0cd9 | 896 | |
49731baa | 897 | WARN_ON_ONCE(!bdev->bd_holder); |
4e91672c | 898 | |
e09b457b TH |
899 | /* FIXME: remove the following once add_disk() handles errors */ |
900 | if (WARN_ON(!disk->slave_dir || !bdev->bd_part->holder_dir)) | |
901 | goto out_unlock; | |
4e91672c | 902 | |
49731baa TH |
903 | holder = bd_find_holder_disk(bdev, disk); |
904 | if (holder) { | |
905 | holder->refcnt++; | |
e09b457b | 906 | goto out_unlock; |
49731baa | 907 | } |
641dc636 | 908 | |
49731baa TH |
909 | holder = kzalloc(sizeof(*holder), GFP_KERNEL); |
910 | if (!holder) { | |
911 | ret = -ENOMEM; | |
e09b457b TH |
912 | goto out_unlock; |
913 | } | |
641dc636 | 914 | |
49731baa TH |
915 | INIT_LIST_HEAD(&holder->list); |
916 | holder->disk = disk; | |
917 | holder->refcnt = 1; | |
918 | ||
919 | ret = add_symlink(disk->slave_dir, &part_to_dev(bdev->bd_part)->kobj); | |
920 | if (ret) | |
921 | goto out_free; | |
922 | ||
923 | ret = add_symlink(bdev->bd_part->holder_dir, &disk_to_dev(disk)->kobj); | |
924 | if (ret) | |
925 | goto out_del; | |
e7407d16 TH |
926 | /* |
927 | * bdev could be deleted beneath us which would implicitly destroy | |
928 | * the holder directory. Hold on to it. | |
929 | */ | |
930 | kobject_get(bdev->bd_part->holder_dir); | |
49731baa TH |
931 | |
932 | list_add(&holder->list, &bdev->bd_holder_disks); | |
933 | goto out_unlock; | |
934 | ||
935 | out_del: | |
936 | del_symlink(disk->slave_dir, &part_to_dev(bdev->bd_part)->kobj); | |
937 | out_free: | |
938 | kfree(holder); | |
e09b457b | 939 | out_unlock: |
b4cf1b72 | 940 | mutex_unlock(&bdev->bd_mutex); |
e09b457b | 941 | return ret; |
641dc636 | 942 | } |
e09b457b | 943 | EXPORT_SYMBOL_GPL(bd_link_disk_holder); |
641dc636 | 944 | |
49731baa TH |
945 | /** |
946 | * bd_unlink_disk_holder - destroy symlinks created by bd_link_disk_holder() | |
947 | * @bdev: the calimed slave bdev | |
948 | * @disk: the holding disk | |
949 | * | |
950 | * DON'T USE THIS UNLESS YOU'RE ALREADY USING IT. | |
951 | * | |
952 | * CONTEXT: | |
953 | * Might sleep. | |
954 | */ | |
955 | void bd_unlink_disk_holder(struct block_device *bdev, struct gendisk *disk) | |
641dc636 | 956 | { |
49731baa | 957 | struct bd_holder_disk *holder; |
641dc636 | 958 | |
49731baa | 959 | mutex_lock(&bdev->bd_mutex); |
641dc636 | 960 | |
49731baa TH |
961 | holder = bd_find_holder_disk(bdev, disk); |
962 | ||
963 | if (!WARN_ON_ONCE(holder == NULL) && !--holder->refcnt) { | |
964 | del_symlink(disk->slave_dir, &part_to_dev(bdev->bd_part)->kobj); | |
965 | del_symlink(bdev->bd_part->holder_dir, | |
966 | &disk_to_dev(disk)->kobj); | |
e7407d16 | 967 | kobject_put(bdev->bd_part->holder_dir); |
49731baa TH |
968 | list_del_init(&holder->list); |
969 | kfree(holder); | |
970 | } | |
971 | ||
972 | mutex_unlock(&bdev->bd_mutex); | |
1da177e4 | 973 | } |
49731baa | 974 | EXPORT_SYMBOL_GPL(bd_unlink_disk_holder); |
641dc636 | 975 | #endif |
1da177e4 | 976 | |
56ade44b AP |
977 | /** |
978 | * flush_disk - invalidates all buffer-cache entries on a disk | |
979 | * | |
980 | * @bdev: struct block device to be flushed | |
e6eb5ce1 | 981 | * @kill_dirty: flag to guide handling of dirty inodes |
56ade44b AP |
982 | * |
983 | * Invalidates all buffer-cache entries on a disk. It should be called | |
984 | * when a disk has been changed -- either by a media change or online | |
985 | * resize. | |
986 | */ | |
93b270f7 | 987 | static void flush_disk(struct block_device *bdev, bool kill_dirty) |
56ade44b | 988 | { |
93b270f7 | 989 | if (__invalidate_device(bdev, kill_dirty)) { |
56ade44b AP |
990 | char name[BDEVNAME_SIZE] = ""; |
991 | ||
992 | if (bdev->bd_disk) | |
993 | disk_name(bdev->bd_disk, 0, name); | |
994 | printk(KERN_WARNING "VFS: busy inodes on changed media or " | |
995 | "resized disk %s\n", name); | |
996 | } | |
997 | ||
998 | if (!bdev->bd_disk) | |
999 | return; | |
d27769ec | 1000 | if (disk_part_scan_enabled(bdev->bd_disk)) |
56ade44b AP |
1001 | bdev->bd_invalidated = 1; |
1002 | } | |
1003 | ||
c3279d14 | 1004 | /** |
57d1b536 | 1005 | * check_disk_size_change - checks for disk size change and adjusts bdev size. |
c3279d14 AP |
1006 | * @disk: struct gendisk to check |
1007 | * @bdev: struct bdev to adjust. | |
1008 | * | |
1009 | * This routine checks to see if the bdev size does not match the disk size | |
1010 | * and adjusts it if it differs. | |
1011 | */ | |
1012 | void check_disk_size_change(struct gendisk *disk, struct block_device *bdev) | |
1013 | { | |
1014 | loff_t disk_size, bdev_size; | |
1015 | ||
1016 | disk_size = (loff_t)get_capacity(disk) << 9; | |
1017 | bdev_size = i_size_read(bdev->bd_inode); | |
1018 | if (disk_size != bdev_size) { | |
1019 | char name[BDEVNAME_SIZE]; | |
1020 | ||
1021 | disk_name(disk, 0, name); | |
1022 | printk(KERN_INFO | |
1023 | "%s: detected capacity change from %lld to %lld\n", | |
1024 | name, bdev_size, disk_size); | |
1025 | i_size_write(bdev->bd_inode, disk_size); | |
93b270f7 | 1026 | flush_disk(bdev, false); |
c3279d14 AP |
1027 | } |
1028 | } | |
1029 | EXPORT_SYMBOL(check_disk_size_change); | |
1030 | ||
0c002c2f | 1031 | /** |
57d1b536 | 1032 | * revalidate_disk - wrapper for lower-level driver's revalidate_disk call-back |
0c002c2f AP |
1033 | * @disk: struct gendisk to be revalidated |
1034 | * | |
1035 | * This routine is a wrapper for lower-level driver's revalidate_disk | |
1036 | * call-backs. It is used to do common pre and post operations needed | |
1037 | * for all revalidate_disk operations. | |
1038 | */ | |
1039 | int revalidate_disk(struct gendisk *disk) | |
1040 | { | |
c3279d14 | 1041 | struct block_device *bdev; |
0c002c2f AP |
1042 | int ret = 0; |
1043 | ||
1044 | if (disk->fops->revalidate_disk) | |
1045 | ret = disk->fops->revalidate_disk(disk); | |
1046 | ||
c3279d14 AP |
1047 | bdev = bdget_disk(disk, 0); |
1048 | if (!bdev) | |
1049 | return ret; | |
1050 | ||
1051 | mutex_lock(&bdev->bd_mutex); | |
1052 | check_disk_size_change(disk, bdev); | |
7630b661 | 1053 | bdev->bd_invalidated = 0; |
c3279d14 AP |
1054 | mutex_unlock(&bdev->bd_mutex); |
1055 | bdput(bdev); | |
0c002c2f AP |
1056 | return ret; |
1057 | } | |
1058 | EXPORT_SYMBOL(revalidate_disk); | |
1059 | ||
1da177e4 LT |
1060 | /* |
1061 | * This routine checks whether a removable media has been changed, | |
1062 | * and invalidates all buffer-cache-entries in that case. This | |
1063 | * is a relatively slow routine, so we have to try to minimize using | |
1064 | * it. Thus it is called only upon a 'mount' or 'open'. This | |
1065 | * is the best way of combining speed and utility, I think. | |
1066 | * People changing diskettes in the middle of an operation deserve | |
1067 | * to lose :-) | |
1068 | */ | |
1069 | int check_disk_change(struct block_device *bdev) | |
1070 | { | |
1071 | struct gendisk *disk = bdev->bd_disk; | |
83d5cde4 | 1072 | const struct block_device_operations *bdops = disk->fops; |
77ea887e | 1073 | unsigned int events; |
1da177e4 | 1074 | |
77ea887e TH |
1075 | events = disk_clear_events(disk, DISK_EVENT_MEDIA_CHANGE | |
1076 | DISK_EVENT_EJECT_REQUEST); | |
1077 | if (!(events & DISK_EVENT_MEDIA_CHANGE)) | |
1da177e4 LT |
1078 | return 0; |
1079 | ||
93b270f7 | 1080 | flush_disk(bdev, true); |
1da177e4 LT |
1081 | if (bdops->revalidate_disk) |
1082 | bdops->revalidate_disk(bdev->bd_disk); | |
1da177e4 LT |
1083 | return 1; |
1084 | } | |
1085 | ||
1086 | EXPORT_SYMBOL(check_disk_change); | |
1087 | ||
1088 | void bd_set_size(struct block_device *bdev, loff_t size) | |
1089 | { | |
e1defc4f | 1090 | unsigned bsize = bdev_logical_block_size(bdev); |
1da177e4 | 1091 | |
d646a02a GC |
1092 | mutex_lock(&bdev->bd_inode->i_mutex); |
1093 | i_size_write(bdev->bd_inode, size); | |
1094 | mutex_unlock(&bdev->bd_inode->i_mutex); | |
1da177e4 LT |
1095 | while (bsize < PAGE_CACHE_SIZE) { |
1096 | if (size & bsize) | |
1097 | break; | |
1098 | bsize <<= 1; | |
1099 | } | |
1100 | bdev->bd_block_size = bsize; | |
1101 | bdev->bd_inode->i_blkbits = blksize_bits(bsize); | |
1102 | } | |
1103 | EXPORT_SYMBOL(bd_set_size); | |
1104 | ||
4385bab1 | 1105 | static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part); |
37be4124 | 1106 | |
6d740cd5 PZ |
1107 | /* |
1108 | * bd_mutex locking: | |
1109 | * | |
1110 | * mutex_lock(part->bd_mutex) | |
1111 | * mutex_lock_nested(whole->bd_mutex, 1) | |
1112 | */ | |
1113 | ||
572c4892 | 1114 | static int __blkdev_get(struct block_device *bdev, fmode_t mode, int for_part) |
1da177e4 | 1115 | { |
1da177e4 | 1116 | struct gendisk *disk; |
523e1d39 | 1117 | struct module *owner; |
7db9cfd3 | 1118 | int ret; |
cf771cb5 | 1119 | int partno; |
fe6e9c1f AV |
1120 | int perm = 0; |
1121 | ||
572c4892 | 1122 | if (mode & FMODE_READ) |
fe6e9c1f | 1123 | perm |= MAY_READ; |
572c4892 | 1124 | if (mode & FMODE_WRITE) |
fe6e9c1f AV |
1125 | perm |= MAY_WRITE; |
1126 | /* | |
1127 | * hooks: /n/, see "layering violations". | |
1128 | */ | |
b7300b78 CW |
1129 | if (!for_part) { |
1130 | ret = devcgroup_inode_permission(bdev->bd_inode, perm); | |
1131 | if (ret != 0) { | |
1132 | bdput(bdev); | |
1133 | return ret; | |
1134 | } | |
82666020 | 1135 | } |
7db9cfd3 | 1136 | |
d3374825 | 1137 | restart: |
0762b8bd | 1138 | |
89f97496 | 1139 | ret = -ENXIO; |
cf771cb5 | 1140 | disk = get_gendisk(bdev->bd_dev, &partno); |
0762b8bd | 1141 | if (!disk) |
6e9624b8 | 1142 | goto out; |
523e1d39 | 1143 | owner = disk->fops->owner; |
1da177e4 | 1144 | |
69e02c59 | 1145 | disk_block_events(disk); |
6796bf54 | 1146 | mutex_lock_nested(&bdev->bd_mutex, for_part); |
1da177e4 LT |
1147 | if (!bdev->bd_openers) { |
1148 | bdev->bd_disk = disk; | |
87192a2a | 1149 | bdev->bd_queue = disk->queue; |
1da177e4 | 1150 | bdev->bd_contains = bdev; |
cf771cb5 | 1151 | if (!partno) { |
1da177e4 | 1152 | struct backing_dev_info *bdi; |
89f97496 TH |
1153 | |
1154 | ret = -ENXIO; | |
1155 | bdev->bd_part = disk_get_part(disk, partno); | |
1156 | if (!bdev->bd_part) | |
1157 | goto out_clear; | |
1158 | ||
1196f8b8 | 1159 | ret = 0; |
1da177e4 | 1160 | if (disk->fops->open) { |
572c4892 | 1161 | ret = disk->fops->open(bdev, mode); |
d3374825 N |
1162 | if (ret == -ERESTARTSYS) { |
1163 | /* Lost a race with 'disk' being | |
1164 | * deleted, try again. | |
1165 | * See md.c | |
1166 | */ | |
1167 | disk_put_part(bdev->bd_part); | |
1168 | bdev->bd_part = NULL; | |
d3374825 | 1169 | bdev->bd_disk = NULL; |
87192a2a | 1170 | bdev->bd_queue = NULL; |
d3374825 | 1171 | mutex_unlock(&bdev->bd_mutex); |
69e02c59 | 1172 | disk_unblock_events(disk); |
69e02c59 | 1173 | put_disk(disk); |
523e1d39 | 1174 | module_put(owner); |
d3374825 N |
1175 | goto restart; |
1176 | } | |
1da177e4 | 1177 | } |
7e69723f | 1178 | |
de33127d | 1179 | if (!ret) { |
7e69723f TH |
1180 | bd_set_size(bdev,(loff_t)get_capacity(disk)<<9); |
1181 | bdi = blk_get_backing_dev_info(bdev); | |
1182 | if (bdi == NULL) | |
1183 | bdi = &default_backing_dev_info; | |
1184 | bdev_inode_switch_bdi(bdev->bd_inode, bdi); | |
1185 | } | |
1186 | ||
1196f8b8 TH |
1187 | /* |
1188 | * If the device is invalidated, rescan partition | |
1189 | * if open succeeded or failed with -ENOMEDIUM. | |
1190 | * The latter is necessary to prevent ghost | |
1191 | * partitions on a removed medium. | |
1192 | */ | |
fe316bf2 JN |
1193 | if (bdev->bd_invalidated) { |
1194 | if (!ret) | |
1195 | rescan_partitions(disk, bdev); | |
1196 | else if (ret == -ENOMEDIUM) | |
1197 | invalidate_partitions(disk, bdev); | |
1198 | } | |
1196f8b8 TH |
1199 | if (ret) |
1200 | goto out_clear; | |
1da177e4 | 1201 | } else { |
1da177e4 LT |
1202 | struct block_device *whole; |
1203 | whole = bdget_disk(disk, 0); | |
1204 | ret = -ENOMEM; | |
1205 | if (!whole) | |
0762b8bd | 1206 | goto out_clear; |
37be4124 | 1207 | BUG_ON(for_part); |
572c4892 | 1208 | ret = __blkdev_get(whole, mode, 1); |
1da177e4 | 1209 | if (ret) |
0762b8bd | 1210 | goto out_clear; |
1da177e4 | 1211 | bdev->bd_contains = whole; |
a5491e0c DC |
1212 | bdev_inode_switch_bdi(bdev->bd_inode, |
1213 | whole->bd_inode->i_data.backing_dev_info); | |
89f97496 | 1214 | bdev->bd_part = disk_get_part(disk, partno); |
e71bf0d0 | 1215 | if (!(disk->flags & GENHD_FL_UP) || |
89f97496 | 1216 | !bdev->bd_part || !bdev->bd_part->nr_sects) { |
1da177e4 | 1217 | ret = -ENXIO; |
0762b8bd | 1218 | goto out_clear; |
1da177e4 | 1219 | } |
89f97496 | 1220 | bd_set_size(bdev, (loff_t)bdev->bd_part->nr_sects << 9); |
1da177e4 LT |
1221 | } |
1222 | } else { | |
1da177e4 | 1223 | if (bdev->bd_contains == bdev) { |
1196f8b8 TH |
1224 | ret = 0; |
1225 | if (bdev->bd_disk->fops->open) | |
572c4892 | 1226 | ret = bdev->bd_disk->fops->open(bdev, mode); |
1196f8b8 | 1227 | /* the same as first opener case, read comment there */ |
fe316bf2 JN |
1228 | if (bdev->bd_invalidated) { |
1229 | if (!ret) | |
1230 | rescan_partitions(bdev->bd_disk, bdev); | |
1231 | else if (ret == -ENOMEDIUM) | |
1232 | invalidate_partitions(bdev->bd_disk, bdev); | |
1233 | } | |
1196f8b8 TH |
1234 | if (ret) |
1235 | goto out_unlock_bdev; | |
1da177e4 | 1236 | } |
69e02c59 | 1237 | /* only one opener holds refs to the module and disk */ |
69e02c59 | 1238 | put_disk(disk); |
523e1d39 | 1239 | module_put(owner); |
1da177e4 LT |
1240 | } |
1241 | bdev->bd_openers++; | |
37be4124 N |
1242 | if (for_part) |
1243 | bdev->bd_part_count++; | |
c039e313 | 1244 | mutex_unlock(&bdev->bd_mutex); |
69e02c59 | 1245 | disk_unblock_events(disk); |
1da177e4 LT |
1246 | return 0; |
1247 | ||
0762b8bd | 1248 | out_clear: |
89f97496 | 1249 | disk_put_part(bdev->bd_part); |
1da177e4 | 1250 | bdev->bd_disk = NULL; |
0762b8bd | 1251 | bdev->bd_part = NULL; |
87192a2a | 1252 | bdev->bd_queue = NULL; |
a5491e0c | 1253 | bdev_inode_switch_bdi(bdev->bd_inode, &default_backing_dev_info); |
1da177e4 | 1254 | if (bdev != bdev->bd_contains) |
572c4892 | 1255 | __blkdev_put(bdev->bd_contains, mode, 1); |
1da177e4 | 1256 | bdev->bd_contains = NULL; |
0762b8bd | 1257 | out_unlock_bdev: |
c039e313 | 1258 | mutex_unlock(&bdev->bd_mutex); |
69e02c59 | 1259 | disk_unblock_events(disk); |
0762b8bd | 1260 | put_disk(disk); |
523e1d39 | 1261 | module_put(owner); |
4345caba | 1262 | out: |
0762b8bd TH |
1263 | bdput(bdev); |
1264 | ||
1da177e4 LT |
1265 | return ret; |
1266 | } | |
1267 | ||
d4d77629 TH |
1268 | /** |
1269 | * blkdev_get - open a block device | |
1270 | * @bdev: block_device to open | |
1271 | * @mode: FMODE_* mask | |
1272 | * @holder: exclusive holder identifier | |
1273 | * | |
1274 | * Open @bdev with @mode. If @mode includes %FMODE_EXCL, @bdev is | |
1275 | * open with exclusive access. Specifying %FMODE_EXCL with %NULL | |
1276 | * @holder is invalid. Exclusive opens may nest for the same @holder. | |
1277 | * | |
1278 | * On success, the reference count of @bdev is unchanged. On failure, | |
1279 | * @bdev is put. | |
1280 | * | |
1281 | * CONTEXT: | |
1282 | * Might sleep. | |
1283 | * | |
1284 | * RETURNS: | |
1285 | * 0 on success, -errno on failure. | |
1286 | */ | |
e525fd89 | 1287 | int blkdev_get(struct block_device *bdev, fmode_t mode, void *holder) |
1da177e4 | 1288 | { |
e525fd89 TH |
1289 | struct block_device *whole = NULL; |
1290 | int res; | |
1291 | ||
1292 | WARN_ON_ONCE((mode & FMODE_EXCL) && !holder); | |
1293 | ||
1294 | if ((mode & FMODE_EXCL) && holder) { | |
1295 | whole = bd_start_claiming(bdev, holder); | |
1296 | if (IS_ERR(whole)) { | |
1297 | bdput(bdev); | |
1298 | return PTR_ERR(whole); | |
1299 | } | |
1300 | } | |
1301 | ||
1302 | res = __blkdev_get(bdev, mode, 0); | |
1303 | ||
1304 | if (whole) { | |
d4dc210f TH |
1305 | struct gendisk *disk = whole->bd_disk; |
1306 | ||
6a027eff | 1307 | /* finish claiming */ |
77ea887e | 1308 | mutex_lock(&bdev->bd_mutex); |
6a027eff TH |
1309 | spin_lock(&bdev_lock); |
1310 | ||
77ea887e | 1311 | if (!res) { |
6a027eff TH |
1312 | BUG_ON(!bd_may_claim(bdev, whole, holder)); |
1313 | /* | |
1314 | * Note that for a whole device bd_holders | |
1315 | * will be incremented twice, and bd_holder | |
1316 | * will be set to bd_may_claim before being | |
1317 | * set to holder | |
1318 | */ | |
1319 | whole->bd_holders++; | |
1320 | whole->bd_holder = bd_may_claim; | |
1321 | bdev->bd_holders++; | |
1322 | bdev->bd_holder = holder; | |
1323 | } | |
1324 | ||
1325 | /* tell others that we're done */ | |
1326 | BUG_ON(whole->bd_claiming != holder); | |
1327 | whole->bd_claiming = NULL; | |
1328 | wake_up_bit(&whole->bd_claiming, 0); | |
1329 | ||
1330 | spin_unlock(&bdev_lock); | |
77ea887e TH |
1331 | |
1332 | /* | |
d4dc210f TH |
1333 | * Block event polling for write claims if requested. Any |
1334 | * write holder makes the write_holder state stick until | |
1335 | * all are released. This is good enough and tracking | |
1336 | * individual writeable reference is too fragile given the | |
1337 | * way @mode is used in blkdev_get/put(). | |
77ea887e | 1338 | */ |
4c49ff3f TH |
1339 | if (!res && (mode & FMODE_WRITE) && !bdev->bd_write_holder && |
1340 | (disk->flags & GENHD_FL_BLOCK_EVENTS_ON_EXCL_WRITE)) { | |
77ea887e | 1341 | bdev->bd_write_holder = true; |
d4dc210f | 1342 | disk_block_events(disk); |
77ea887e TH |
1343 | } |
1344 | ||
1345 | mutex_unlock(&bdev->bd_mutex); | |
6a027eff | 1346 | bdput(whole); |
e525fd89 TH |
1347 | } |
1348 | ||
1349 | return res; | |
37be4124 | 1350 | } |
1da177e4 LT |
1351 | EXPORT_SYMBOL(blkdev_get); |
1352 | ||
d4d77629 TH |
1353 | /** |
1354 | * blkdev_get_by_path - open a block device by name | |
1355 | * @path: path to the block device to open | |
1356 | * @mode: FMODE_* mask | |
1357 | * @holder: exclusive holder identifier | |
1358 | * | |
1359 | * Open the blockdevice described by the device file at @path. @mode | |
1360 | * and @holder are identical to blkdev_get(). | |
1361 | * | |
1362 | * On success, the returned block_device has reference count of one. | |
1363 | * | |
1364 | * CONTEXT: | |
1365 | * Might sleep. | |
1366 | * | |
1367 | * RETURNS: | |
1368 | * Pointer to block_device on success, ERR_PTR(-errno) on failure. | |
1369 | */ | |
1370 | struct block_device *blkdev_get_by_path(const char *path, fmode_t mode, | |
1371 | void *holder) | |
1372 | { | |
1373 | struct block_device *bdev; | |
1374 | int err; | |
1375 | ||
1376 | bdev = lookup_bdev(path); | |
1377 | if (IS_ERR(bdev)) | |
1378 | return bdev; | |
1379 | ||
1380 | err = blkdev_get(bdev, mode, holder); | |
1381 | if (err) | |
1382 | return ERR_PTR(err); | |
1383 | ||
e51900f7 CE |
1384 | if ((mode & FMODE_WRITE) && bdev_read_only(bdev)) { |
1385 | blkdev_put(bdev, mode); | |
1386 | return ERR_PTR(-EACCES); | |
1387 | } | |
1388 | ||
d4d77629 TH |
1389 | return bdev; |
1390 | } | |
1391 | EXPORT_SYMBOL(blkdev_get_by_path); | |
1392 | ||
1393 | /** | |
1394 | * blkdev_get_by_dev - open a block device by device number | |
1395 | * @dev: device number of block device to open | |
1396 | * @mode: FMODE_* mask | |
1397 | * @holder: exclusive holder identifier | |
1398 | * | |
1399 | * Open the blockdevice described by device number @dev. @mode and | |
1400 | * @holder are identical to blkdev_get(). | |
1401 | * | |
1402 | * Use it ONLY if you really do not have anything better - i.e. when | |
1403 | * you are behind a truly sucky interface and all you are given is a | |
1404 | * device number. _Never_ to be used for internal purposes. If you | |
1405 | * ever need it - reconsider your API. | |
1406 | * | |
1407 | * On success, the returned block_device has reference count of one. | |
1408 | * | |
1409 | * CONTEXT: | |
1410 | * Might sleep. | |
1411 | * | |
1412 | * RETURNS: | |
1413 | * Pointer to block_device on success, ERR_PTR(-errno) on failure. | |
1414 | */ | |
1415 | struct block_device *blkdev_get_by_dev(dev_t dev, fmode_t mode, void *holder) | |
1416 | { | |
1417 | struct block_device *bdev; | |
1418 | int err; | |
1419 | ||
1420 | bdev = bdget(dev); | |
1421 | if (!bdev) | |
1422 | return ERR_PTR(-ENOMEM); | |
1423 | ||
1424 | err = blkdev_get(bdev, mode, holder); | |
1425 | if (err) | |
1426 | return ERR_PTR(err); | |
1427 | ||
1428 | return bdev; | |
1429 | } | |
1430 | EXPORT_SYMBOL(blkdev_get_by_dev); | |
1431 | ||
1da177e4 LT |
1432 | static int blkdev_open(struct inode * inode, struct file * filp) |
1433 | { | |
1434 | struct block_device *bdev; | |
1da177e4 LT |
1435 | |
1436 | /* | |
1437 | * Preserve backwards compatibility and allow large file access | |
1438 | * even if userspace doesn't ask for it explicitly. Some mkfs | |
1439 | * binary needs it. We might want to drop this workaround | |
1440 | * during an unstable branch. | |
1441 | */ | |
1442 | filp->f_flags |= O_LARGEFILE; | |
1443 | ||
572c4892 AV |
1444 | if (filp->f_flags & O_NDELAY) |
1445 | filp->f_mode |= FMODE_NDELAY; | |
1446 | if (filp->f_flags & O_EXCL) | |
1447 | filp->f_mode |= FMODE_EXCL; | |
1448 | if ((filp->f_flags & O_ACCMODE) == 3) | |
1449 | filp->f_mode |= FMODE_WRITE_IOCTL; | |
1450 | ||
1da177e4 | 1451 | bdev = bd_acquire(inode); |
6a2aae06 PE |
1452 | if (bdev == NULL) |
1453 | return -ENOMEM; | |
1da177e4 | 1454 | |
572c4892 AV |
1455 | filp->f_mapping = bdev->bd_inode->i_mapping; |
1456 | ||
e525fd89 | 1457 | return blkdev_get(bdev, filp->f_mode, filp); |
1da177e4 LT |
1458 | } |
1459 | ||
4385bab1 | 1460 | static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part) |
2e7b651d | 1461 | { |
2e7b651d | 1462 | struct gendisk *disk = bdev->bd_disk; |
37be4124 | 1463 | struct block_device *victim = NULL; |
2e7b651d | 1464 | |
6796bf54 | 1465 | mutex_lock_nested(&bdev->bd_mutex, for_part); |
37be4124 N |
1466 | if (for_part) |
1467 | bdev->bd_part_count--; | |
1468 | ||
2e7b651d | 1469 | if (!--bdev->bd_openers) { |
6a027eff | 1470 | WARN_ON_ONCE(bdev->bd_holders); |
2e7b651d PZ |
1471 | sync_blockdev(bdev); |
1472 | kill_bdev(bdev); | |
94007751 N |
1473 | /* ->release can cause the old bdi to disappear, |
1474 | * so must switch it out first | |
1475 | */ | |
1476 | bdev_inode_switch_bdi(bdev->bd_inode, | |
1477 | &default_backing_dev_info); | |
2e7b651d PZ |
1478 | } |
1479 | if (bdev->bd_contains == bdev) { | |
1480 | if (disk->fops->release) | |
db2a144b | 1481 | disk->fops->release(disk, mode); |
2e7b651d PZ |
1482 | } |
1483 | if (!bdev->bd_openers) { | |
1484 | struct module *owner = disk->fops->owner; | |
1485 | ||
0762b8bd TH |
1486 | disk_put_part(bdev->bd_part); |
1487 | bdev->bd_part = NULL; | |
2e7b651d | 1488 | bdev->bd_disk = NULL; |
37be4124 N |
1489 | if (bdev != bdev->bd_contains) |
1490 | victim = bdev->bd_contains; | |
2e7b651d | 1491 | bdev->bd_contains = NULL; |
523e1d39 TH |
1492 | |
1493 | put_disk(disk); | |
1494 | module_put(owner); | |
2e7b651d | 1495 | } |
2e7b651d PZ |
1496 | mutex_unlock(&bdev->bd_mutex); |
1497 | bdput(bdev); | |
37be4124 | 1498 | if (victim) |
9a1c3542 | 1499 | __blkdev_put(victim, mode, 1); |
2e7b651d PZ |
1500 | } |
1501 | ||
4385bab1 | 1502 | void blkdev_put(struct block_device *bdev, fmode_t mode) |
37be4124 | 1503 | { |
85ef06d1 TH |
1504 | mutex_lock(&bdev->bd_mutex); |
1505 | ||
e525fd89 | 1506 | if (mode & FMODE_EXCL) { |
6a027eff TH |
1507 | bool bdev_free; |
1508 | ||
1509 | /* | |
1510 | * Release a claim on the device. The holder fields | |
1511 | * are protected with bdev_lock. bd_mutex is to | |
1512 | * synchronize disk_holder unlinking. | |
1513 | */ | |
6a027eff TH |
1514 | spin_lock(&bdev_lock); |
1515 | ||
1516 | WARN_ON_ONCE(--bdev->bd_holders < 0); | |
1517 | WARN_ON_ONCE(--bdev->bd_contains->bd_holders < 0); | |
1518 | ||
1519 | /* bd_contains might point to self, check in a separate step */ | |
1520 | if ((bdev_free = !bdev->bd_holders)) | |
1521 | bdev->bd_holder = NULL; | |
1522 | if (!bdev->bd_contains->bd_holders) | |
1523 | bdev->bd_contains->bd_holder = NULL; | |
1524 | ||
1525 | spin_unlock(&bdev_lock); | |
1526 | ||
77ea887e TH |
1527 | /* |
1528 | * If this was the last claim, remove holder link and | |
1529 | * unblock evpoll if it was a write holder. | |
1530 | */ | |
85ef06d1 TH |
1531 | if (bdev_free && bdev->bd_write_holder) { |
1532 | disk_unblock_events(bdev->bd_disk); | |
1533 | bdev->bd_write_holder = false; | |
77ea887e | 1534 | } |
6936217c | 1535 | } |
77ea887e | 1536 | |
85ef06d1 TH |
1537 | /* |
1538 | * Trigger event checking and tell drivers to flush MEDIA_CHANGE | |
1539 | * event. This is to ensure detection of media removal commanded | |
1540 | * from userland - e.g. eject(1). | |
1541 | */ | |
1542 | disk_flush_events(bdev->bd_disk, DISK_EVENT_MEDIA_CHANGE); | |
1543 | ||
1544 | mutex_unlock(&bdev->bd_mutex); | |
1545 | ||
4385bab1 | 1546 | __blkdev_put(bdev, mode, 0); |
37be4124 | 1547 | } |
2e7b651d PZ |
1548 | EXPORT_SYMBOL(blkdev_put); |
1549 | ||
1da177e4 LT |
1550 | static int blkdev_close(struct inode * inode, struct file * filp) |
1551 | { | |
1552 | struct block_device *bdev = I_BDEV(filp->f_mapping->host); | |
4385bab1 AV |
1553 | blkdev_put(bdev, filp->f_mode); |
1554 | return 0; | |
1da177e4 LT |
1555 | } |
1556 | ||
bb93e3a5 | 1557 | static long block_ioctl(struct file *file, unsigned cmd, unsigned long arg) |
1da177e4 | 1558 | { |
56b26add AV |
1559 | struct block_device *bdev = I_BDEV(file->f_mapping->host); |
1560 | fmode_t mode = file->f_mode; | |
fd4ce1ac CH |
1561 | |
1562 | /* | |
1563 | * O_NDELAY can be altered using fcntl(.., F_SETFL, ..), so we have | |
1564 | * to updated it before every ioctl. | |
1565 | */ | |
56b26add | 1566 | if (file->f_flags & O_NDELAY) |
fd4ce1ac CH |
1567 | mode |= FMODE_NDELAY; |
1568 | else | |
1569 | mode &= ~FMODE_NDELAY; | |
1570 | ||
56b26add | 1571 | return blkdev_ioctl(bdev, mode, cmd, arg); |
1da177e4 LT |
1572 | } |
1573 | ||
eef99380 CH |
1574 | /* |
1575 | * Write data to the block device. Only intended for the block device itself | |
1576 | * and the raw driver which basically is a fake block device. | |
1577 | * | |
1578 | * Does not take i_mutex for the write and thus is not for general purpose | |
1579 | * use. | |
1580 | */ | |
1456c0a8 | 1581 | ssize_t blkdev_write_iter(struct kiocb *iocb, struct iov_iter *from) |
eef99380 CH |
1582 | { |
1583 | struct file *file = iocb->ki_filp; | |
53362a05 | 1584 | struct blk_plug plug; |
eef99380 CH |
1585 | ssize_t ret; |
1586 | ||
53362a05 | 1587 | blk_start_plug(&plug); |
1456c0a8 | 1588 | ret = __generic_file_write_iter(iocb, from); |
02afc27f | 1589 | if (ret > 0) { |
eef99380 | 1590 | ssize_t err; |
1456c0a8 | 1591 | err = generic_write_sync(file, iocb->ki_pos - ret, ret); |
45d4f855 | 1592 | if (err < 0) |
eef99380 CH |
1593 | ret = err; |
1594 | } | |
53362a05 | 1595 | blk_finish_plug(&plug); |
eef99380 CH |
1596 | return ret; |
1597 | } | |
1456c0a8 | 1598 | EXPORT_SYMBOL_GPL(blkdev_write_iter); |
eef99380 | 1599 | |
a886038b | 1600 | static ssize_t blkdev_read_iter(struct kiocb *iocb, struct iov_iter *to) |
684c9aae LT |
1601 | { |
1602 | struct file *file = iocb->ki_filp; | |
1603 | struct inode *bd_inode = file->f_mapping->host; | |
1604 | loff_t size = i_size_read(bd_inode); | |
a886038b | 1605 | loff_t pos = iocb->ki_pos; |
684c9aae LT |
1606 | |
1607 | if (pos >= size) | |
1608 | return 0; | |
1609 | ||
1610 | size -= pos; | |
a886038b AV |
1611 | iov_iter_truncate(to, size); |
1612 | return generic_file_read_iter(iocb, to); | |
684c9aae LT |
1613 | } |
1614 | ||
87d8fe1e TT |
1615 | /* |
1616 | * Try to release a page associated with block device when the system | |
1617 | * is under memory pressure. | |
1618 | */ | |
1619 | static int blkdev_releasepage(struct page *page, gfp_t wait) | |
1620 | { | |
1621 | struct super_block *super = BDEV_I(page->mapping->host)->bdev.bd_super; | |
1622 | ||
1623 | if (super && super->s_op->bdev_try_to_free_page) | |
1624 | return super->s_op->bdev_try_to_free_page(super, page, wait); | |
1625 | ||
1626 | return try_to_free_buffers(page); | |
1627 | } | |
1628 | ||
4c54ac62 | 1629 | static const struct address_space_operations def_blk_aops = { |
1da177e4 | 1630 | .readpage = blkdev_readpage, |
447f05bb | 1631 | .readpages = blkdev_readpages, |
1da177e4 | 1632 | .writepage = blkdev_writepage, |
6272b5a5 NP |
1633 | .write_begin = blkdev_write_begin, |
1634 | .write_end = blkdev_write_end, | |
1da177e4 | 1635 | .writepages = generic_writepages, |
87d8fe1e | 1636 | .releasepage = blkdev_releasepage, |
1da177e4 | 1637 | .direct_IO = blkdev_direct_IO, |
b4597226 | 1638 | .is_dirty_writeback = buffer_check_dirty_writeback, |
1da177e4 LT |
1639 | }; |
1640 | ||
4b6f5d20 | 1641 | const struct file_operations def_blk_fops = { |
1da177e4 LT |
1642 | .open = blkdev_open, |
1643 | .release = blkdev_close, | |
1644 | .llseek = block_llseek, | |
a886038b | 1645 | .read = new_sync_read, |
1456c0a8 | 1646 | .write = new_sync_write, |
a886038b | 1647 | .read_iter = blkdev_read_iter, |
1456c0a8 | 1648 | .write_iter = blkdev_write_iter, |
1e8b3332 | 1649 | .mmap = generic_file_mmap, |
b1dd3b28 | 1650 | .fsync = blkdev_fsync, |
bb93e3a5 | 1651 | .unlocked_ioctl = block_ioctl, |
1da177e4 LT |
1652 | #ifdef CONFIG_COMPAT |
1653 | .compat_ioctl = compat_blkdev_ioctl, | |
1654 | #endif | |
1e8b3332 | 1655 | .splice_read = generic_file_splice_read, |
8d020765 | 1656 | .splice_write = iter_file_splice_write, |
1da177e4 LT |
1657 | }; |
1658 | ||
1659 | int ioctl_by_bdev(struct block_device *bdev, unsigned cmd, unsigned long arg) | |
1660 | { | |
1661 | int res; | |
1662 | mm_segment_t old_fs = get_fs(); | |
1663 | set_fs(KERNEL_DS); | |
56b26add | 1664 | res = blkdev_ioctl(bdev, 0, cmd, arg); |
1da177e4 LT |
1665 | set_fs(old_fs); |
1666 | return res; | |
1667 | } | |
1668 | ||
1669 | EXPORT_SYMBOL(ioctl_by_bdev); | |
1670 | ||
1671 | /** | |
1672 | * lookup_bdev - lookup a struct block_device by name | |
94e2959e | 1673 | * @pathname: special file representing the block device |
1da177e4 | 1674 | * |
57d1b536 | 1675 | * Get a reference to the blockdevice at @pathname in the current |
1da177e4 LT |
1676 | * namespace if possible and return it. Return ERR_PTR(error) |
1677 | * otherwise. | |
1678 | */ | |
421748ec | 1679 | struct block_device *lookup_bdev(const char *pathname) |
1da177e4 LT |
1680 | { |
1681 | struct block_device *bdev; | |
1682 | struct inode *inode; | |
421748ec | 1683 | struct path path; |
1da177e4 LT |
1684 | int error; |
1685 | ||
421748ec | 1686 | if (!pathname || !*pathname) |
1da177e4 LT |
1687 | return ERR_PTR(-EINVAL); |
1688 | ||
421748ec | 1689 | error = kern_path(pathname, LOOKUP_FOLLOW, &path); |
1da177e4 LT |
1690 | if (error) |
1691 | return ERR_PTR(error); | |
1692 | ||
421748ec | 1693 | inode = path.dentry->d_inode; |
1da177e4 LT |
1694 | error = -ENOTBLK; |
1695 | if (!S_ISBLK(inode->i_mode)) | |
1696 | goto fail; | |
1697 | error = -EACCES; | |
421748ec | 1698 | if (path.mnt->mnt_flags & MNT_NODEV) |
1da177e4 LT |
1699 | goto fail; |
1700 | error = -ENOMEM; | |
1701 | bdev = bd_acquire(inode); | |
1702 | if (!bdev) | |
1703 | goto fail; | |
1704 | out: | |
421748ec | 1705 | path_put(&path); |
1da177e4 LT |
1706 | return bdev; |
1707 | fail: | |
1708 | bdev = ERR_PTR(error); | |
1709 | goto out; | |
1710 | } | |
d5686b44 | 1711 | EXPORT_SYMBOL(lookup_bdev); |
1da177e4 | 1712 | |
93b270f7 | 1713 | int __invalidate_device(struct block_device *bdev, bool kill_dirty) |
b71e8a4c DH |
1714 | { |
1715 | struct super_block *sb = get_super(bdev); | |
1716 | int res = 0; | |
1717 | ||
1718 | if (sb) { | |
1719 | /* | |
1720 | * no need to lock the super, get_super holds the | |
1721 | * read mutex so the filesystem cannot go away | |
1722 | * under us (->put_super runs with the write lock | |
1723 | * hold). | |
1724 | */ | |
1725 | shrink_dcache_sb(sb); | |
93b270f7 | 1726 | res = invalidate_inodes(sb, kill_dirty); |
b71e8a4c DH |
1727 | drop_super(sb); |
1728 | } | |
f98393a6 | 1729 | invalidate_bdev(bdev); |
b71e8a4c DH |
1730 | return res; |
1731 | } | |
1732 | EXPORT_SYMBOL(__invalidate_device); | |
5c0d6b60 JK |
1733 | |
1734 | void iterate_bdevs(void (*func)(struct block_device *, void *), void *arg) | |
1735 | { | |
1736 | struct inode *inode, *old_inode = NULL; | |
1737 | ||
1738 | spin_lock(&inode_sb_list_lock); | |
1739 | list_for_each_entry(inode, &blockdev_superblock->s_inodes, i_sb_list) { | |
1740 | struct address_space *mapping = inode->i_mapping; | |
1741 | ||
1742 | spin_lock(&inode->i_lock); | |
1743 | if (inode->i_state & (I_FREEING|I_WILL_FREE|I_NEW) || | |
1744 | mapping->nrpages == 0) { | |
1745 | spin_unlock(&inode->i_lock); | |
1746 | continue; | |
1747 | } | |
1748 | __iget(inode); | |
1749 | spin_unlock(&inode->i_lock); | |
1750 | spin_unlock(&inode_sb_list_lock); | |
1751 | /* | |
1752 | * We hold a reference to 'inode' so it couldn't have been | |
1753 | * removed from s_inodes list while we dropped the | |
1754 | * inode_sb_list_lock. We cannot iput the inode now as we can | |
1755 | * be holding the last reference and we cannot iput it under | |
1756 | * inode_sb_list_lock. So we keep the reference and iput it | |
1757 | * later. | |
1758 | */ | |
1759 | iput(old_inode); | |
1760 | old_inode = inode; | |
1761 | ||
1762 | func(I_BDEV(inode), arg); | |
1763 | ||
1764 | spin_lock(&inode_sb_list_lock); | |
1765 | } | |
1766 | spin_unlock(&inode_sb_list_lock); | |
1767 | iput(old_inode); | |
1768 | } |