Commit | Line | Data |
---|---|---|
457c8996 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
1da177e4 | 2 | /* |
1da177e4 LT |
3 | * Copyright (C) 1991, 1992 Linus Torvalds |
4 | * Copyright (C) 2001 Andrea Arcangeli <andrea@suse.de> SuSE | |
7b51e703 | 5 | * Copyright (C) 2016 - 2020 Christoph Hellwig |
1da177e4 LT |
6 | */ |
7 | ||
1da177e4 LT |
8 | #include <linux/init.h> |
9 | #include <linux/mm.h> | |
1da177e4 LT |
10 | #include <linux/slab.h> |
11 | #include <linux/kmod.h> | |
12 | #include <linux/major.h> | |
7db9cfd3 | 13 | #include <linux/device_cgroup.h> |
1da177e4 | 14 | #include <linux/blkdev.h> |
fe45e630 | 15 | #include <linux/blk-integrity.h> |
66114cad | 16 | #include <linux/backing-dev.h> |
1da177e4 LT |
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> |
811d736f | 22 | #include <linux/writeback.h> |
1da177e4 | 23 | #include <linux/mount.h> |
9030d16e | 24 | #include <linux/pseudo_fs.h> |
1da177e4 LT |
25 | #include <linux/uio.h> |
26 | #include <linux/namei.h> | |
15e3d2c5 | 27 | #include <linux/part_stat.h> |
7c0f6ba6 | 28 | #include <linux/uaccess.h> |
0dca4462 CH |
29 | #include "../fs/internal.h" |
30 | #include "blk.h" | |
1da177e4 LT |
31 | |
32 | struct bdev_inode { | |
33 | struct block_device bdev; | |
34 | struct inode vfs_inode; | |
35 | }; | |
36 | ||
37 | static inline struct bdev_inode *BDEV_I(struct inode *inode) | |
38 | { | |
39 | return container_of(inode, struct bdev_inode, vfs_inode); | |
40 | } | |
41 | ||
ff5053f6 | 42 | struct block_device *I_BDEV(struct inode *inode) |
1da177e4 LT |
43 | { |
44 | return &BDEV_I(inode)->bdev; | |
45 | } | |
1da177e4 LT |
46 | EXPORT_SYMBOL(I_BDEV); |
47 | ||
dbd3ca50 | 48 | static void bdev_write_inode(struct block_device *bdev) |
564f00f6 | 49 | { |
dbd3ca50 VG |
50 | struct inode *inode = bdev->bd_inode; |
51 | int ret; | |
52 | ||
564f00f6 CH |
53 | spin_lock(&inode->i_lock); |
54 | while (inode->i_state & I_DIRTY) { | |
55 | spin_unlock(&inode->i_lock); | |
dbd3ca50 VG |
56 | ret = write_inode_now(inode, true); |
57 | if (ret) { | |
58 | char name[BDEVNAME_SIZE]; | |
59 | pr_warn_ratelimited("VFS: Dirty inode writeback failed " | |
60 | "for block device %s (err=%d).\n", | |
61 | bdevname(bdev, name), ret); | |
62 | } | |
564f00f6 CH |
63 | spin_lock(&inode->i_lock); |
64 | } | |
65 | spin_unlock(&inode->i_lock); | |
66 | } | |
67 | ||
f9a14399 | 68 | /* Kill _all_ buffers and pagecache , dirty or not.. */ |
3373a346 | 69 | static void kill_bdev(struct block_device *bdev) |
1da177e4 | 70 | { |
ff01bb48 AV |
71 | struct address_space *mapping = bdev->bd_inode->i_mapping; |
72 | ||
7716506a | 73 | if (mapping_empty(mapping)) |
f9a14399 | 74 | return; |
ff01bb48 | 75 | |
f9a14399 | 76 | invalidate_bh_lrus(); |
ff01bb48 | 77 | truncate_inode_pages(mapping, 0); |
3373a346 | 78 | } |
ff01bb48 AV |
79 | |
80 | /* Invalidate clean unused buffers and pagecache. */ | |
81 | void invalidate_bdev(struct block_device *bdev) | |
82 | { | |
83 | struct address_space *mapping = bdev->bd_inode->i_mapping; | |
84 | ||
a5f6a6a9 AR |
85 | if (mapping->nrpages) { |
86 | invalidate_bh_lrus(); | |
87 | lru_add_drain_all(); /* make sure all lru add caches are flushed */ | |
88 | invalidate_mapping_pages(mapping, 0, -1); | |
89 | } | |
ff01bb48 AV |
90 | } |
91 | EXPORT_SYMBOL(invalidate_bdev); | |
1da177e4 | 92 | |
384d87ef JK |
93 | /* |
94 | * Drop all buffers & page cache for given bdev range. This function bails | |
95 | * with error if bdev has other exclusive owner (such as filesystem). | |
96 | */ | |
97 | int truncate_bdev_range(struct block_device *bdev, fmode_t mode, | |
98 | loff_t lstart, loff_t lend) | |
99 | { | |
384d87ef JK |
100 | /* |
101 | * If we don't hold exclusive handle for the device, upgrade to it | |
102 | * while we discard the buffer cache to avoid discarding buffers | |
103 | * under live filesystem. | |
104 | */ | |
105 | if (!(mode & FMODE_EXCL)) { | |
37c3fc9a | 106 | int err = bd_prepare_to_claim(bdev, truncate_bdev_range); |
384d87ef | 107 | if (err) |
56887cff | 108 | goto invalidate; |
384d87ef | 109 | } |
37c3fc9a | 110 | |
384d87ef | 111 | truncate_inode_pages_range(bdev->bd_inode->i_mapping, lstart, lend); |
37c3fc9a CH |
112 | if (!(mode & FMODE_EXCL)) |
113 | bd_abort_claiming(bdev, truncate_bdev_range); | |
384d87ef | 114 | return 0; |
56887cff JK |
115 | |
116 | invalidate: | |
117 | /* | |
118 | * Someone else has handle exclusively open. Try invalidating instead. | |
119 | * The 'end' argument is inclusive so the rounding is safe. | |
120 | */ | |
121 | return invalidate_inode_pages2_range(bdev->bd_inode->i_mapping, | |
122 | lstart >> PAGE_SHIFT, | |
123 | lend >> PAGE_SHIFT); | |
384d87ef | 124 | } |
384d87ef | 125 | |
04906b2f JK |
126 | static void set_init_blocksize(struct block_device *bdev) |
127 | { | |
8dc932d3 MM |
128 | unsigned int bsize = bdev_logical_block_size(bdev); |
129 | loff_t size = i_size_read(bdev->bd_inode); | |
130 | ||
131 | while (bsize < PAGE_SIZE) { | |
132 | if (size & bsize) | |
133 | break; | |
134 | bsize <<= 1; | |
135 | } | |
136 | bdev->bd_inode->i_blkbits = blksize_bits(bsize); | |
04906b2f JK |
137 | } |
138 | ||
1da177e4 LT |
139 | int set_blocksize(struct block_device *bdev, int size) |
140 | { | |
141 | /* Size must be a power of two, and between 512 and PAGE_SIZE */ | |
1368c4f2 | 142 | if (size > PAGE_SIZE || size < 512 || !is_power_of_2(size)) |
1da177e4 LT |
143 | return -EINVAL; |
144 | ||
145 | /* Size cannot be smaller than the size supported by the device */ | |
e1defc4f | 146 | if (size < bdev_logical_block_size(bdev)) |
1da177e4 LT |
147 | return -EINVAL; |
148 | ||
149 | /* Don't change the size if it is same as current */ | |
6b7b181b | 150 | if (bdev->bd_inode->i_blkbits != blksize_bits(size)) { |
1da177e4 | 151 | sync_blockdev(bdev); |
1da177e4 LT |
152 | bdev->bd_inode->i_blkbits = blksize_bits(size); |
153 | kill_bdev(bdev); | |
154 | } | |
155 | return 0; | |
156 | } | |
157 | ||
158 | EXPORT_SYMBOL(set_blocksize); | |
159 | ||
160 | int sb_set_blocksize(struct super_block *sb, int size) | |
161 | { | |
1da177e4 LT |
162 | if (set_blocksize(sb->s_bdev, size)) |
163 | return 0; | |
164 | /* If we get here, we know size is power of two | |
165 | * and it's value is between 512 and PAGE_SIZE */ | |
166 | sb->s_blocksize = size; | |
38885bd4 | 167 | sb->s_blocksize_bits = blksize_bits(size); |
1da177e4 LT |
168 | return sb->s_blocksize; |
169 | } | |
170 | ||
171 | EXPORT_SYMBOL(sb_set_blocksize); | |
172 | ||
173 | int sb_min_blocksize(struct super_block *sb, int size) | |
174 | { | |
e1defc4f | 175 | int minsize = bdev_logical_block_size(sb->s_bdev); |
1da177e4 LT |
176 | if (size < minsize) |
177 | size = minsize; | |
178 | return sb_set_blocksize(sb, size); | |
179 | } | |
180 | ||
181 | EXPORT_SYMBOL(sb_min_blocksize); | |
182 | ||
70164eb6 | 183 | int sync_blockdev_nowait(struct block_device *bdev) |
5cee5815 JK |
184 | { |
185 | if (!bdev) | |
186 | return 0; | |
70164eb6 | 187 | return filemap_flush(bdev->bd_inode->i_mapping); |
5cee5815 | 188 | } |
70164eb6 | 189 | EXPORT_SYMBOL_GPL(sync_blockdev_nowait); |
5cee5815 | 190 | |
585d3bc0 NP |
191 | /* |
192 | * Write out and wait upon all the dirty data associated with a block | |
193 | * device via its mapping. Does not take the superblock lock. | |
194 | */ | |
195 | int sync_blockdev(struct block_device *bdev) | |
196 | { | |
70164eb6 CH |
197 | if (!bdev) |
198 | return 0; | |
199 | return filemap_write_and_wait(bdev->bd_inode->i_mapping); | |
585d3bc0 NP |
200 | } |
201 | EXPORT_SYMBOL(sync_blockdev); | |
202 | ||
97d6fb1b YM |
203 | int sync_blockdev_range(struct block_device *bdev, loff_t lstart, loff_t lend) |
204 | { | |
205 | return filemap_write_and_wait_range(bdev->bd_inode->i_mapping, | |
206 | lstart, lend); | |
207 | } | |
208 | EXPORT_SYMBOL(sync_blockdev_range); | |
209 | ||
585d3bc0 NP |
210 | /* |
211 | * Write out and wait upon all dirty data associated with this | |
212 | * device. Filesystem data as well as the underlying block | |
213 | * device. Takes the superblock lock. | |
214 | */ | |
215 | int fsync_bdev(struct block_device *bdev) | |
216 | { | |
217 | struct super_block *sb = get_super(bdev); | |
218 | if (sb) { | |
60b0680f | 219 | int res = sync_filesystem(sb); |
585d3bc0 NP |
220 | drop_super(sb); |
221 | return res; | |
222 | } | |
223 | return sync_blockdev(bdev); | |
224 | } | |
47e4491b | 225 | EXPORT_SYMBOL(fsync_bdev); |
585d3bc0 NP |
226 | |
227 | /** | |
228 | * freeze_bdev -- lock a filesystem and force it into a consistent state | |
229 | * @bdev: blockdevice to lock | |
230 | * | |
585d3bc0 NP |
231 | * If a superblock is found on this device, we take the s_umount semaphore |
232 | * on it to make sure nobody unmounts until the snapshot creation is done. | |
233 | * The reference counter (bd_fsfreeze_count) guarantees that only the last | |
234 | * unfreeze process can unfreeze the frozen filesystem actually when multiple | |
235 | * freeze requests arrive simultaneously. It counts up in freeze_bdev() and | |
236 | * count down in thaw_bdev(). When it becomes 0, thaw_bdev() will unfreeze | |
237 | * actually. | |
238 | */ | |
040f04bd | 239 | int freeze_bdev(struct block_device *bdev) |
585d3bc0 NP |
240 | { |
241 | struct super_block *sb; | |
242 | int error = 0; | |
243 | ||
244 | mutex_lock(&bdev->bd_fsfreeze_mutex); | |
040f04bd CH |
245 | if (++bdev->bd_fsfreeze_count > 1) |
246 | goto done; | |
4504230a CH |
247 | |
248 | sb = get_active_super(bdev); | |
249 | if (!sb) | |
040f04bd | 250 | goto sync; |
48b6bca6 BM |
251 | if (sb->s_op->freeze_super) |
252 | error = sb->s_op->freeze_super(sb); | |
253 | else | |
254 | error = freeze_super(sb); | |
040f04bd CH |
255 | deactivate_super(sb); |
256 | ||
18e9e510 | 257 | if (error) { |
18e9e510 | 258 | bdev->bd_fsfreeze_count--; |
040f04bd | 259 | goto done; |
585d3bc0 | 260 | } |
040f04bd CH |
261 | bdev->bd_fsfreeze_sb = sb; |
262 | ||
263 | sync: | |
585d3bc0 | 264 | sync_blockdev(bdev); |
040f04bd | 265 | done: |
585d3bc0 | 266 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
040f04bd | 267 | return error; |
585d3bc0 NP |
268 | } |
269 | EXPORT_SYMBOL(freeze_bdev); | |
270 | ||
271 | /** | |
272 | * thaw_bdev -- unlock filesystem | |
273 | * @bdev: blockdevice to unlock | |
585d3bc0 NP |
274 | * |
275 | * Unlocks the filesystem and marks it writeable again after freeze_bdev(). | |
276 | */ | |
040f04bd | 277 | int thaw_bdev(struct block_device *bdev) |
585d3bc0 | 278 | { |
040f04bd | 279 | struct super_block *sb; |
4504230a | 280 | int error = -EINVAL; |
585d3bc0 NP |
281 | |
282 | mutex_lock(&bdev->bd_fsfreeze_mutex); | |
4504230a | 283 | if (!bdev->bd_fsfreeze_count) |
18e9e510 | 284 | goto out; |
4504230a CH |
285 | |
286 | error = 0; | |
287 | if (--bdev->bd_fsfreeze_count > 0) | |
18e9e510 | 288 | goto out; |
4504230a | 289 | |
040f04bd | 290 | sb = bdev->bd_fsfreeze_sb; |
4504230a | 291 | if (!sb) |
18e9e510 | 292 | goto out; |
4504230a | 293 | |
48b6bca6 BM |
294 | if (sb->s_op->thaw_super) |
295 | error = sb->s_op->thaw_super(sb); | |
296 | else | |
297 | error = thaw_super(sb); | |
997198ba | 298 | if (error) |
18e9e510 | 299 | bdev->bd_fsfreeze_count++; |
04a6a536 ST |
300 | else |
301 | bdev->bd_fsfreeze_sb = NULL; | |
18e9e510 | 302 | out: |
585d3bc0 | 303 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
997198ba | 304 | return error; |
585d3bc0 NP |
305 | } |
306 | EXPORT_SYMBOL(thaw_bdev); | |
307 | ||
47a191fd MW |
308 | /** |
309 | * bdev_read_page() - Start reading a page from a block device | |
310 | * @bdev: The device to read the page from | |
311 | * @sector: The offset on the device to read the page to (need not be aligned) | |
312 | * @page: The page to read | |
313 | * | |
314 | * On entry, the page should be locked. It will be unlocked when the page | |
315 | * has been read. If the block driver implements rw_page synchronously, | |
316 | * that will be true on exit from this function, but it need not be. | |
317 | * | |
318 | * Errors returned by this function are usually "soft", eg out of memory, or | |
319 | * queue full; callers should try a different route to read this page rather | |
320 | * than propagate an error back up the stack. | |
321 | * | |
322 | * Return: negative errno if an error occurs, 0 if submission was successful. | |
323 | */ | |
324 | int bdev_read_page(struct block_device *bdev, sector_t sector, | |
325 | struct page *page) | |
326 | { | |
327 | const struct block_device_operations *ops = bdev->bd_disk->fops; | |
2e6edc95 DW |
328 | int result = -EOPNOTSUPP; |
329 | ||
f68eb1e7 | 330 | if (!ops->rw_page || bdev_get_integrity(bdev)) |
2e6edc95 DW |
331 | return result; |
332 | ||
025a3865 | 333 | result = blk_queue_enter(bdev_get_queue(bdev), 0); |
2e6edc95 DW |
334 | if (result) |
335 | return result; | |
3f289dcb TH |
336 | result = ops->rw_page(bdev, sector + get_start_sect(bdev), page, |
337 | REQ_OP_READ); | |
025a3865 | 338 | blk_queue_exit(bdev_get_queue(bdev)); |
2e6edc95 | 339 | return result; |
47a191fd | 340 | } |
47a191fd MW |
341 | |
342 | /** | |
343 | * bdev_write_page() - Start writing a page to a block device | |
344 | * @bdev: The device to write the page to | |
345 | * @sector: The offset on the device to write the page to (need not be aligned) | |
346 | * @page: The page to write | |
347 | * @wbc: The writeback_control for the write | |
348 | * | |
349 | * On entry, the page should be locked and not currently under writeback. | |
350 | * On exit, if the write started successfully, the page will be unlocked and | |
351 | * under writeback. If the write failed already (eg the driver failed to | |
352 | * queue the page to the device), the page will still be locked. If the | |
353 | * caller is a ->writepage implementation, it will need to unlock the page. | |
354 | * | |
355 | * Errors returned by this function are usually "soft", eg out of memory, or | |
356 | * queue full; callers should try a different route to write this page rather | |
357 | * than propagate an error back up the stack. | |
358 | * | |
359 | * Return: negative errno if an error occurs, 0 if submission was successful. | |
360 | */ | |
361 | int bdev_write_page(struct block_device *bdev, sector_t sector, | |
362 | struct page *page, struct writeback_control *wbc) | |
363 | { | |
364 | int result; | |
47a191fd | 365 | const struct block_device_operations *ops = bdev->bd_disk->fops; |
2e6edc95 | 366 | |
f68eb1e7 | 367 | if (!ops->rw_page || bdev_get_integrity(bdev)) |
47a191fd | 368 | return -EOPNOTSUPP; |
025a3865 | 369 | result = blk_queue_enter(bdev_get_queue(bdev), 0); |
2e6edc95 DW |
370 | if (result) |
371 | return result; | |
372 | ||
47a191fd | 373 | set_page_writeback(page); |
3f289dcb TH |
374 | result = ops->rw_page(bdev, sector + get_start_sect(bdev), page, |
375 | REQ_OP_WRITE); | |
f892760a | 376 | if (result) { |
47a191fd | 377 | end_page_writeback(page); |
f892760a MW |
378 | } else { |
379 | clean_page_buffers(page); | |
47a191fd | 380 | unlock_page(page); |
f892760a | 381 | } |
025a3865 | 382 | blk_queue_exit(bdev_get_queue(bdev)); |
47a191fd MW |
383 | return result; |
384 | } | |
47a191fd | 385 | |
1da177e4 LT |
386 | /* |
387 | * pseudo-fs | |
388 | */ | |
389 | ||
390 | static __cacheline_aligned_in_smp DEFINE_SPINLOCK(bdev_lock); | |
e18b890b | 391 | static struct kmem_cache * bdev_cachep __read_mostly; |
1da177e4 LT |
392 | |
393 | static struct inode *bdev_alloc_inode(struct super_block *sb) | |
394 | { | |
fd60b288 | 395 | struct bdev_inode *ei = alloc_inode_sb(sb, bdev_cachep, GFP_KERNEL); |
2d2f6f1b | 396 | |
1da177e4 LT |
397 | if (!ei) |
398 | return NULL; | |
2d2f6f1b | 399 | memset(&ei->bdev, 0, sizeof(ei->bdev)); |
1da177e4 LT |
400 | return &ei->vfs_inode; |
401 | } | |
402 | ||
41149cb0 | 403 | static void bdev_free_inode(struct inode *inode) |
1da177e4 | 404 | { |
15e3d2c5 CH |
405 | struct block_device *bdev = I_BDEV(inode); |
406 | ||
407 | free_percpu(bdev->bd_stats); | |
231926db | 408 | kfree(bdev->bd_meta_info); |
15e3d2c5 | 409 | |
889c05cc CH |
410 | if (!bdev_is_partition(bdev)) { |
411 | if (bdev->bd_disk && bdev->bd_disk->bdi) | |
412 | bdi_put(bdev->bd_disk->bdi); | |
340e8457 | 413 | kfree(bdev->bd_disk); |
889c05cc | 414 | } |
9451aa0a CH |
415 | |
416 | if (MAJOR(bdev->bd_dev) == BLOCK_EXT_MAJOR) | |
417 | blk_free_ext_minor(MINOR(bdev->bd_dev)); | |
418 | ||
41149cb0 | 419 | kmem_cache_free(bdev_cachep, BDEV_I(inode)); |
fa0d7e3d NP |
420 | } |
421 | ||
e6cb5382 | 422 | static void init_once(void *data) |
1da177e4 | 423 | { |
e6cb5382 | 424 | struct bdev_inode *ei = data; |
1da177e4 | 425 | |
a35afb83 | 426 | inode_init_once(&ei->vfs_inode); |
1da177e4 LT |
427 | } |
428 | ||
b57922d9 | 429 | static void bdev_evict_inode(struct inode *inode) |
1da177e4 | 430 | { |
91b0abe3 | 431 | truncate_inode_pages_final(&inode->i_data); |
b57922d9 | 432 | invalidate_inode_buffers(inode); /* is it needed here? */ |
dbd5768f | 433 | clear_inode(inode); |
1da177e4 LT |
434 | } |
435 | ||
ee9b6d61 | 436 | static const struct super_operations bdev_sops = { |
1da177e4 LT |
437 | .statfs = simple_statfs, |
438 | .alloc_inode = bdev_alloc_inode, | |
41149cb0 | 439 | .free_inode = bdev_free_inode, |
1da177e4 | 440 | .drop_inode = generic_delete_inode, |
b57922d9 | 441 | .evict_inode = bdev_evict_inode, |
1da177e4 LT |
442 | }; |
443 | ||
9030d16e | 444 | static int bd_init_fs_context(struct fs_context *fc) |
1da177e4 | 445 | { |
9030d16e DH |
446 | struct pseudo_fs_context *ctx = init_pseudo(fc, BDEVFS_MAGIC); |
447 | if (!ctx) | |
448 | return -ENOMEM; | |
449 | fc->s_iflags |= SB_I_CGROUPWB; | |
450 | ctx->ops = &bdev_sops; | |
451 | return 0; | |
1da177e4 LT |
452 | } |
453 | ||
454 | static struct file_system_type bd_type = { | |
455 | .name = "bdev", | |
9030d16e | 456 | .init_fs_context = bd_init_fs_context, |
1da177e4 LT |
457 | .kill_sb = kill_anon_super, |
458 | }; | |
459 | ||
a212b105 TH |
460 | struct super_block *blockdev_superblock __read_mostly; |
461 | EXPORT_SYMBOL_GPL(blockdev_superblock); | |
1da177e4 LT |
462 | |
463 | void __init bdev_cache_init(void) | |
464 | { | |
465 | int err; | |
ace8577a | 466 | static struct vfsmount *bd_mnt; |
c2acf7b9 | 467 | |
1da177e4 | 468 | bdev_cachep = kmem_cache_create("bdev_cache", sizeof(struct bdev_inode), |
fffb60f9 | 469 | 0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| |
5d097056 | 470 | SLAB_MEM_SPREAD|SLAB_ACCOUNT|SLAB_PANIC), |
20c2df83 | 471 | init_once); |
1da177e4 LT |
472 | err = register_filesystem(&bd_type); |
473 | if (err) | |
474 | panic("Cannot register bdev pseudo-fs"); | |
475 | bd_mnt = kern_mount(&bd_type); | |
1da177e4 LT |
476 | if (IS_ERR(bd_mnt)) |
477 | panic("Cannot create bdev pseudo-fs"); | |
ace8577a | 478 | blockdev_superblock = bd_mnt->mnt_sb; /* For writeback */ |
1da177e4 LT |
479 | } |
480 | ||
22ae8ce8 | 481 | struct block_device *bdev_alloc(struct gendisk *disk, u8 partno) |
1da177e4 LT |
482 | { |
483 | struct block_device *bdev; | |
484 | struct inode *inode; | |
485 | ||
22ae8ce8 | 486 | inode = new_inode(blockdev_superblock); |
1da177e4 LT |
487 | if (!inode) |
488 | return NULL; | |
22ae8ce8 CH |
489 | inode->i_mode = S_IFBLK; |
490 | inode->i_rdev = 0; | |
491 | inode->i_data.a_ops = &def_blk_aops; | |
492 | mapping_set_gfp_mask(&inode->i_data, GFP_USER); | |
493 | ||
494 | bdev = I_BDEV(inode); | |
e6cb5382 | 495 | mutex_init(&bdev->bd_fsfreeze_mutex); |
22ae8ce8 | 496 | spin_lock_init(&bdev->bd_size_lock); |
22ae8ce8 | 497 | bdev->bd_partno = partno; |
22ae8ce8 | 498 | bdev->bd_inode = inode; |
17220ca5 | 499 | bdev->bd_queue = disk->queue; |
15e3d2c5 CH |
500 | bdev->bd_stats = alloc_percpu(struct disk_stats); |
501 | if (!bdev->bd_stats) { | |
502 | iput(inode); | |
503 | return NULL; | |
504 | } | |
06cc978d | 505 | bdev->bd_disk = disk; |
22ae8ce8 CH |
506 | return bdev; |
507 | } | |
1da177e4 | 508 | |
22ae8ce8 CH |
509 | void bdev_add(struct block_device *bdev, dev_t dev) |
510 | { | |
511 | bdev->bd_dev = dev; | |
512 | bdev->bd_inode->i_rdev = dev; | |
513 | bdev->bd_inode->i_ino = dev; | |
514 | insert_inode_hash(bdev->bd_inode); | |
515 | } | |
1da177e4 | 516 | |
1da177e4 LT |
517 | long nr_blockdev_pages(void) |
518 | { | |
1008fe6d | 519 | struct inode *inode; |
1da177e4 | 520 | long ret = 0; |
1008fe6d CH |
521 | |
522 | spin_lock(&blockdev_superblock->s_inode_list_lock); | |
523 | list_for_each_entry(inode, &blockdev_superblock->s_inodes, i_sb_list) | |
524 | ret += inode->i_mapping->nrpages; | |
525 | spin_unlock(&blockdev_superblock->s_inode_list_lock); | |
526 | ||
1da177e4 LT |
527 | return ret; |
528 | } | |
529 | ||
1a3cbbc5 TH |
530 | /** |
531 | * bd_may_claim - test whether a block device can be claimed | |
532 | * @bdev: block device of interest | |
533 | * @whole: whole block device containing @bdev, may equal @bdev | |
534 | * @holder: holder trying to claim @bdev | |
535 | * | |
25985edc | 536 | * Test whether @bdev can be claimed by @holder. |
1a3cbbc5 TH |
537 | * |
538 | * CONTEXT: | |
539 | * spin_lock(&bdev_lock). | |
540 | * | |
541 | * RETURNS: | |
542 | * %true if @bdev can be claimed, %false otherwise. | |
543 | */ | |
544 | static bool bd_may_claim(struct block_device *bdev, struct block_device *whole, | |
545 | void *holder) | |
1da177e4 | 546 | { |
1da177e4 | 547 | if (bdev->bd_holder == holder) |
1a3cbbc5 | 548 | return true; /* already a holder */ |
1da177e4 | 549 | else if (bdev->bd_holder != NULL) |
1a3cbbc5 | 550 | return false; /* held by someone else */ |
bcc7f5b4 | 551 | else if (whole == bdev) |
1a3cbbc5 | 552 | return true; /* is a whole device which isn't held */ |
1da177e4 | 553 | |
e525fd89 | 554 | else if (whole->bd_holder == bd_may_claim) |
1a3cbbc5 TH |
555 | return true; /* is a partition of a device that is being partitioned */ |
556 | else if (whole->bd_holder != NULL) | |
557 | return false; /* is a partition of a held device */ | |
1da177e4 | 558 | else |
1a3cbbc5 TH |
559 | return true; /* is a partition of an un-held device */ |
560 | } | |
561 | ||
6b4517a7 | 562 | /** |
58e46ed9 | 563 | * bd_prepare_to_claim - claim a block device |
6b4517a7 | 564 | * @bdev: block device of interest |
6b4517a7 TH |
565 | * @holder: holder trying to claim @bdev |
566 | * | |
58e46ed9 CH |
567 | * Claim @bdev. This function fails if @bdev is already claimed by another |
568 | * holder and waits if another claiming is in progress. return, the caller | |
569 | * has ownership of bd_claiming and bd_holder[s]. | |
6b4517a7 TH |
570 | * |
571 | * RETURNS: | |
572 | * 0 if @bdev can be claimed, -EBUSY otherwise. | |
573 | */ | |
37c3fc9a | 574 | int bd_prepare_to_claim(struct block_device *bdev, void *holder) |
6b4517a7 | 575 | { |
37c3fc9a CH |
576 | struct block_device *whole = bdev_whole(bdev); |
577 | ||
578 | if (WARN_ON_ONCE(!holder)) | |
579 | return -EINVAL; | |
6b4517a7 | 580 | retry: |
58e46ed9 | 581 | spin_lock(&bdev_lock); |
6b4517a7 | 582 | /* if someone else claimed, fail */ |
58e46ed9 CH |
583 | if (!bd_may_claim(bdev, whole, holder)) { |
584 | spin_unlock(&bdev_lock); | |
6b4517a7 | 585 | return -EBUSY; |
58e46ed9 | 586 | } |
6b4517a7 | 587 | |
e75aa858 TH |
588 | /* if claiming is already in progress, wait for it to finish */ |
589 | if (whole->bd_claiming) { | |
6b4517a7 TH |
590 | wait_queue_head_t *wq = bit_waitqueue(&whole->bd_claiming, 0); |
591 | DEFINE_WAIT(wait); | |
592 | ||
593 | prepare_to_wait(wq, &wait, TASK_UNINTERRUPTIBLE); | |
594 | spin_unlock(&bdev_lock); | |
595 | schedule(); | |
596 | finish_wait(wq, &wait); | |
6b4517a7 TH |
597 | goto retry; |
598 | } | |
599 | ||
600 | /* yay, all mine */ | |
58e46ed9 CH |
601 | whole->bd_claiming = holder; |
602 | spin_unlock(&bdev_lock); | |
6b4517a7 TH |
603 | return 0; |
604 | } | |
ecbe6bc0 | 605 | EXPORT_SYMBOL_GPL(bd_prepare_to_claim); /* only for the loop driver */ |
6b4517a7 | 606 | |
89e524c0 JK |
607 | static void bd_clear_claiming(struct block_device *whole, void *holder) |
608 | { | |
609 | lockdep_assert_held(&bdev_lock); | |
610 | /* tell others that we're done */ | |
611 | BUG_ON(whole->bd_claiming != holder); | |
612 | whole->bd_claiming = NULL; | |
613 | wake_up_bit(&whole->bd_claiming, 0); | |
614 | } | |
615 | ||
616 | /** | |
617 | * bd_finish_claiming - finish claiming of a block device | |
618 | * @bdev: block device of interest | |
89e524c0 JK |
619 | * @holder: holder that has claimed @bdev |
620 | * | |
621 | * Finish exclusive open of a block device. Mark the device as exlusively | |
622 | * open by the holder and wake up all waiters for exclusive open to finish. | |
623 | */ | |
37c3fc9a | 624 | static void bd_finish_claiming(struct block_device *bdev, void *holder) |
89e524c0 | 625 | { |
37c3fc9a CH |
626 | struct block_device *whole = bdev_whole(bdev); |
627 | ||
89e524c0 JK |
628 | spin_lock(&bdev_lock); |
629 | BUG_ON(!bd_may_claim(bdev, whole, holder)); | |
630 | /* | |
631 | * Note that for a whole device bd_holders will be incremented twice, | |
632 | * and bd_holder will be set to bd_may_claim before being set to holder | |
633 | */ | |
634 | whole->bd_holders++; | |
635 | whole->bd_holder = bd_may_claim; | |
636 | bdev->bd_holders++; | |
637 | bdev->bd_holder = holder; | |
638 | bd_clear_claiming(whole, holder); | |
639 | spin_unlock(&bdev_lock); | |
640 | } | |
89e524c0 JK |
641 | |
642 | /** | |
643 | * bd_abort_claiming - abort claiming of a block device | |
644 | * @bdev: block device of interest | |
89e524c0 JK |
645 | * @holder: holder that has claimed @bdev |
646 | * | |
647 | * Abort claiming of a block device when the exclusive open failed. This can be | |
648 | * also used when exclusive open is not actually desired and we just needed | |
649 | * to block other exclusive openers for a while. | |
650 | */ | |
37c3fc9a | 651 | void bd_abort_claiming(struct block_device *bdev, void *holder) |
89e524c0 JK |
652 | { |
653 | spin_lock(&bdev_lock); | |
37c3fc9a | 654 | bd_clear_claiming(bdev_whole(bdev), holder); |
89e524c0 JK |
655 | spin_unlock(&bdev_lock); |
656 | } | |
657 | EXPORT_SYMBOL(bd_abort_claiming); | |
6b4517a7 | 658 | |
c8276b95 CH |
659 | static void blkdev_flush_mapping(struct block_device *bdev) |
660 | { | |
661 | WARN_ON_ONCE(bdev->bd_holders); | |
662 | sync_blockdev(bdev); | |
663 | kill_bdev(bdev); | |
664 | bdev_write_inode(bdev); | |
665 | } | |
37be4124 | 666 | |
362529d9 | 667 | static int blkdev_get_whole(struct block_device *bdev, fmode_t mode) |
a1548b67 | 668 | { |
142fe8f4 | 669 | struct gendisk *disk = bdev->bd_disk; |
af22fef3 | 670 | int ret; |
a1548b67 | 671 | |
362529d9 CH |
672 | if (disk->fops->open) { |
673 | ret = disk->fops->open(bdev, mode); | |
674 | if (ret) { | |
675 | /* avoid ghost partitions on a removed medium */ | |
676 | if (ret == -ENOMEDIUM && | |
677 | test_bit(GD_NEED_PART_SCAN, &disk->state)) | |
0384264e | 678 | bdev_disk_changed(disk, true); |
362529d9 CH |
679 | return ret; |
680 | } | |
d981cb5b | 681 | } |
a1548b67 | 682 | |
9acf381f | 683 | if (!atomic_read(&bdev->bd_openers)) |
362529d9 | 684 | set_init_blocksize(bdev); |
362529d9 | 685 | if (test_bit(GD_NEED_PART_SCAN, &disk->state)) |
0384264e | 686 | bdev_disk_changed(disk, false); |
9acf381f | 687 | atomic_inc(&bdev->bd_openers); |
483546c1 | 688 | return 0; |
362529d9 | 689 | } |
a1548b67 | 690 | |
c8276b95 CH |
691 | static void blkdev_put_whole(struct block_device *bdev, fmode_t mode) |
692 | { | |
9acf381f | 693 | if (atomic_dec_and_test(&bdev->bd_openers)) |
c8276b95 CH |
694 | blkdev_flush_mapping(bdev); |
695 | if (bdev->bd_disk->fops->release) | |
696 | bdev->bd_disk->fops->release(bdev->bd_disk, mode); | |
a1548b67 CH |
697 | } |
698 | ||
362529d9 | 699 | static int blkdev_get_part(struct block_device *part, fmode_t mode) |
1da177e4 | 700 | { |
362529d9 | 701 | struct gendisk *disk = part->bd_disk; |
362529d9 | 702 | int ret; |
5b642d8b | 703 | |
9acf381f | 704 | if (atomic_read(&part->bd_openers)) |
362529d9 | 705 | goto done; |
6c60ff04 | 706 | |
9d3b8813 | 707 | ret = blkdev_get_whole(bdev_whole(part), mode); |
a8698707 | 708 | if (ret) |
9d3b8813 | 709 | return ret; |
7e69723f | 710 | |
362529d9 CH |
711 | ret = -ENXIO; |
712 | if (!bdev_nr_sectors(part)) | |
713 | goto out_blkdev_put; | |
7e69723f | 714 | |
ab4b5705 | 715 | disk->open_partitions++; |
362529d9 | 716 | set_init_blocksize(part); |
362529d9 | 717 | done: |
9acf381f | 718 | atomic_inc(&part->bd_openers); |
1da177e4 | 719 | return 0; |
5a023cdb | 720 | |
362529d9 | 721 | out_blkdev_put: |
9d3b8813 | 722 | blkdev_put_whole(bdev_whole(part), mode); |
362529d9 | 723 | return ret; |
1da177e4 | 724 | } |
5b56b6ed | 725 | |
c8276b95 CH |
726 | static void blkdev_put_part(struct block_device *part, fmode_t mode) |
727 | { | |
728 | struct block_device *whole = bdev_whole(part); | |
03e26279 | 729 | |
9acf381f | 730 | if (!atomic_dec_and_test(&part->bd_openers)) |
c8276b95 CH |
731 | return; |
732 | blkdev_flush_mapping(part); | |
ab4b5705 | 733 | whole->bd_disk->open_partitions--; |
c8276b95 | 734 | blkdev_put_whole(whole, mode); |
1da177e4 LT |
735 | } |
736 | ||
22ae8ce8 CH |
737 | struct block_device *blkdev_get_no_open(dev_t dev) |
738 | { | |
739 | struct block_device *bdev; | |
9d3b8813 | 740 | struct inode *inode; |
22ae8ce8 | 741 | |
9d3b8813 | 742 | inode = ilookup(blockdev_superblock, dev); |
fbdee71b | 743 | if (!inode && IS_ENABLED(CONFIG_BLOCK_LEGACY_AUTOLOAD)) { |
22ae8ce8 | 744 | blk_request_module(dev); |
9d3b8813 | 745 | inode = ilookup(blockdev_superblock, dev); |
fbdee71b CH |
746 | if (inode) |
747 | pr_warn_ratelimited( | |
451f0b6f | 748 | "block device autoloading is deprecated and will be removed.\n"); |
22ae8ce8 | 749 | } |
fbdee71b CH |
750 | if (!inode) |
751 | return NULL; | |
22ae8ce8 | 752 | |
9d3b8813 CH |
753 | /* switch from the inode reference to a device mode one: */ |
754 | bdev = &BDEV_I(inode)->bdev; | |
755 | if (!kobject_get_unless_zero(&bdev->bd_device.kobj)) | |
756 | bdev = NULL; | |
757 | iput(inode); | |
22ae8ce8 | 758 | return bdev; |
22ae8ce8 CH |
759 | } |
760 | ||
761 | void blkdev_put_no_open(struct block_device *bdev) | |
762 | { | |
9d3b8813 | 763 | put_device(&bdev->bd_device); |
22ae8ce8 CH |
764 | } |
765 | ||
d4d77629 | 766 | /** |
4e7b5671 CH |
767 | * blkdev_get_by_dev - open a block device by device number |
768 | * @dev: device number of block device to open | |
d4d77629 TH |
769 | * @mode: FMODE_* mask |
770 | * @holder: exclusive holder identifier | |
771 | * | |
4e7b5671 CH |
772 | * Open the block device described by device number @dev. If @mode includes |
773 | * %FMODE_EXCL, the block device is opened with exclusive access. Specifying | |
774 | * %FMODE_EXCL with a %NULL @holder is invalid. Exclusive opens may nest for | |
775 | * the same @holder. | |
d4d77629 | 776 | * |
4e7b5671 CH |
777 | * Use this interface ONLY if you really do not have anything better - i.e. when |
778 | * you are behind a truly sucky interface and all you are given is a device | |
779 | * number. Everything else should use blkdev_get_by_path(). | |
d4d77629 TH |
780 | * |
781 | * CONTEXT: | |
782 | * Might sleep. | |
783 | * | |
784 | * RETURNS: | |
4e7b5671 | 785 | * Reference to the block_device on success, ERR_PTR(-errno) on failure. |
d4d77629 | 786 | */ |
4e7b5671 | 787 | struct block_device *blkdev_get_by_dev(dev_t dev, fmode_t mode, void *holder) |
1da177e4 | 788 | { |
5b56b6ed | 789 | bool unblock_events = true; |
4e7b5671 | 790 | struct block_device *bdev; |
5b56b6ed | 791 | struct gendisk *disk; |
5b56b6ed | 792 | int ret; |
e525fd89 | 793 | |
7918f0f6 | 794 | ret = devcgroup_check_permission(DEVCG_DEV_BLOCK, |
4e7b5671 | 795 | MAJOR(dev), MINOR(dev), |
7918f0f6 CH |
796 | ((mode & FMODE_READ) ? DEVCG_ACC_READ : 0) | |
797 | ((mode & FMODE_WRITE) ? DEVCG_ACC_WRITE : 0)); | |
e5c7fb40 | 798 | if (ret) |
4e7b5671 CH |
799 | return ERR_PTR(ret); |
800 | ||
22ae8ce8 CH |
801 | bdev = blkdev_get_no_open(dev); |
802 | if (!bdev) | |
803 | return ERR_PTR(-ENXIO); | |
804 | disk = bdev->bd_disk; | |
e5c7fb40 | 805 | |
5b56b6ed | 806 | if (mode & FMODE_EXCL) { |
37c3fc9a | 807 | ret = bd_prepare_to_claim(bdev, holder); |
5b56b6ed | 808 | if (ret) |
37c3fc9a | 809 | goto put_blkdev; |
5b56b6ed CH |
810 | } |
811 | ||
812 | disk_block_events(disk); | |
813 | ||
a8698707 | 814 | mutex_lock(&disk->open_mutex); |
362529d9 | 815 | ret = -ENXIO; |
50b4aecf | 816 | if (!disk_live(disk)) |
362529d9 | 817 | goto abort_claiming; |
efcf5932 ML |
818 | if (!try_module_get(disk->fops->owner)) |
819 | goto abort_claiming; | |
362529d9 CH |
820 | if (bdev_is_partition(bdev)) |
821 | ret = blkdev_get_part(bdev, mode); | |
822 | else | |
823 | ret = blkdev_get_whole(bdev, mode); | |
22ae8ce8 | 824 | if (ret) |
efcf5932 | 825 | goto put_module; |
22ae8ce8 | 826 | if (mode & FMODE_EXCL) { |
37c3fc9a | 827 | bd_finish_claiming(bdev, holder); |
5b56b6ed CH |
828 | |
829 | /* | |
830 | * Block event polling for write claims if requested. Any write | |
831 | * holder makes the write_holder state stick until all are | |
832 | * released. This is good enough and tracking individual | |
833 | * writeable reference is too fragile given the way @mode is | |
834 | * used in blkdev_get/put(). | |
835 | */ | |
836 | if ((mode & FMODE_WRITE) && !bdev->bd_write_holder && | |
1545e0b4 | 837 | (disk->event_flags & DISK_EVENT_FLAG_BLOCK_ON_EXCL_WRITE)) { |
5b56b6ed CH |
838 | bdev->bd_write_holder = true; |
839 | unblock_events = false; | |
840 | } | |
841 | } | |
a8698707 | 842 | mutex_unlock(&disk->open_mutex); |
5b56b6ed CH |
843 | |
844 | if (unblock_events) | |
845 | disk_unblock_events(disk); | |
22ae8ce8 | 846 | return bdev; |
efcf5932 ML |
847 | put_module: |
848 | module_put(disk->fops->owner); | |
22ae8ce8 CH |
849 | abort_claiming: |
850 | if (mode & FMODE_EXCL) | |
37c3fc9a | 851 | bd_abort_claiming(bdev, holder); |
a8698707 | 852 | mutex_unlock(&disk->open_mutex); |
22ae8ce8 | 853 | disk_unblock_events(disk); |
22ae8ce8 CH |
854 | put_blkdev: |
855 | blkdev_put_no_open(bdev); | |
22ae8ce8 | 856 | return ERR_PTR(ret); |
37be4124 | 857 | } |
4e7b5671 | 858 | EXPORT_SYMBOL(blkdev_get_by_dev); |
1da177e4 | 859 | |
d4d77629 TH |
860 | /** |
861 | * blkdev_get_by_path - open a block device by name | |
862 | * @path: path to the block device to open | |
863 | * @mode: FMODE_* mask | |
864 | * @holder: exclusive holder identifier | |
865 | * | |
4e7b5671 CH |
866 | * Open the block device described by the device file at @path. If @mode |
867 | * includes %FMODE_EXCL, the block device is opened with exclusive access. | |
868 | * Specifying %FMODE_EXCL with a %NULL @holder is invalid. Exclusive opens may | |
869 | * nest for the same @holder. | |
d4d77629 TH |
870 | * |
871 | * CONTEXT: | |
872 | * Might sleep. | |
873 | * | |
874 | * RETURNS: | |
4e7b5671 | 875 | * Reference to the block_device on success, ERR_PTR(-errno) on failure. |
d4d77629 TH |
876 | */ |
877 | struct block_device *blkdev_get_by_path(const char *path, fmode_t mode, | |
878 | void *holder) | |
879 | { | |
880 | struct block_device *bdev; | |
4e7b5671 CH |
881 | dev_t dev; |
882 | int error; | |
d4d77629 | 883 | |
4e7b5671 CH |
884 | error = lookup_bdev(path, &dev); |
885 | if (error) | |
886 | return ERR_PTR(error); | |
d4d77629 | 887 | |
4e7b5671 CH |
888 | bdev = blkdev_get_by_dev(dev, mode, holder); |
889 | if (!IS_ERR(bdev) && (mode & FMODE_WRITE) && bdev_read_only(bdev)) { | |
e51900f7 CE |
890 | blkdev_put(bdev, mode); |
891 | return ERR_PTR(-EACCES); | |
892 | } | |
893 | ||
d4d77629 TH |
894 | return bdev; |
895 | } | |
896 | EXPORT_SYMBOL(blkdev_get_by_path); | |
897 | ||
4385bab1 | 898 | void blkdev_put(struct block_device *bdev, fmode_t mode) |
2e7b651d | 899 | { |
2e7b651d PZ |
900 | struct gendisk *disk = bdev->bd_disk; |
901 | ||
b849dd84 DA |
902 | /* |
903 | * Sync early if it looks like we're the last one. If someone else | |
904 | * opens the block device between now and the decrement of bd_openers | |
905 | * then we did a sync that we didn't need to, but that's not the end | |
906 | * of the world and we want to avoid long (could be several minute) | |
907 | * syncs while holding the mutex. | |
908 | */ | |
9acf381f | 909 | if (atomic_read(&bdev->bd_openers) == 1) |
b849dd84 DA |
910 | sync_blockdev(bdev); |
911 | ||
a8698707 | 912 | mutex_lock(&disk->open_mutex); |
e525fd89 | 913 | if (mode & FMODE_EXCL) { |
a954ea81 | 914 | struct block_device *whole = bdev_whole(bdev); |
6a027eff TH |
915 | bool bdev_free; |
916 | ||
917 | /* | |
918 | * Release a claim on the device. The holder fields | |
a8698707 | 919 | * are protected with bdev_lock. open_mutex is to |
6a027eff TH |
920 | * synchronize disk_holder unlinking. |
921 | */ | |
6a027eff TH |
922 | spin_lock(&bdev_lock); |
923 | ||
924 | WARN_ON_ONCE(--bdev->bd_holders < 0); | |
a954ea81 | 925 | WARN_ON_ONCE(--whole->bd_holders < 0); |
6a027eff | 926 | |
6a027eff TH |
927 | if ((bdev_free = !bdev->bd_holders)) |
928 | bdev->bd_holder = NULL; | |
a954ea81 CH |
929 | if (!whole->bd_holders) |
930 | whole->bd_holder = NULL; | |
6a027eff TH |
931 | |
932 | spin_unlock(&bdev_lock); | |
933 | ||
77ea887e TH |
934 | /* |
935 | * If this was the last claim, remove holder link and | |
936 | * unblock evpoll if it was a write holder. | |
937 | */ | |
85ef06d1 | 938 | if (bdev_free && bdev->bd_write_holder) { |
5b56b6ed | 939 | disk_unblock_events(disk); |
85ef06d1 | 940 | bdev->bd_write_holder = false; |
77ea887e | 941 | } |
6936217c | 942 | } |
77ea887e | 943 | |
85ef06d1 TH |
944 | /* |
945 | * Trigger event checking and tell drivers to flush MEDIA_CHANGE | |
946 | * event. This is to ensure detection of media removal commanded | |
947 | * from userland - e.g. eject(1). | |
948 | */ | |
5b56b6ed | 949 | disk_flush_events(disk, DISK_EVENT_MEDIA_CHANGE); |
85ef06d1 | 950 | |
c8276b95 CH |
951 | if (bdev_is_partition(bdev)) |
952 | blkdev_put_part(bdev, mode); | |
953 | else | |
954 | blkdev_put_whole(bdev, mode); | |
a8698707 CH |
955 | mutex_unlock(&disk->open_mutex); |
956 | ||
efcf5932 | 957 | module_put(disk->fops->owner); |
22ae8ce8 | 958 | blkdev_put_no_open(bdev); |
37be4124 | 959 | } |
2e7b651d PZ |
960 | EXPORT_SYMBOL(blkdev_put); |
961 | ||
1da177e4 | 962 | /** |
0ba4566c MWO |
963 | * lookup_bdev() - Look up a struct block_device by name. |
964 | * @pathname: Name of the block device in the filesystem. | |
965 | * @dev: Pointer to the block device's dev_t, if found. | |
1da177e4 | 966 | * |
057178cf | 967 | * Lookup the block device's dev_t at @pathname in the current |
0ba4566c | 968 | * namespace if possible and return it in @dev. |
057178cf | 969 | * |
0ba4566c MWO |
970 | * Context: May sleep. |
971 | * Return: 0 if succeeded, negative errno otherwise. | |
1da177e4 | 972 | */ |
4e7b5671 | 973 | int lookup_bdev(const char *pathname, dev_t *dev) |
1da177e4 | 974 | { |
1da177e4 | 975 | struct inode *inode; |
421748ec | 976 | struct path path; |
1da177e4 LT |
977 | int error; |
978 | ||
421748ec | 979 | if (!pathname || !*pathname) |
4e7b5671 | 980 | return -EINVAL; |
1da177e4 | 981 | |
421748ec | 982 | error = kern_path(pathname, LOOKUP_FOLLOW, &path); |
1da177e4 | 983 | if (error) |
4e7b5671 | 984 | return error; |
1da177e4 | 985 | |
bb668734 | 986 | inode = d_backing_inode(path.dentry); |
1da177e4 LT |
987 | error = -ENOTBLK; |
988 | if (!S_ISBLK(inode->i_mode)) | |
4e7b5671 | 989 | goto out_path_put; |
1da177e4 | 990 | error = -EACCES; |
a2982cc9 | 991 | if (!may_open_dev(&path)) |
4e7b5671 CH |
992 | goto out_path_put; |
993 | ||
994 | *dev = inode->i_rdev; | |
995 | error = 0; | |
996 | out_path_put: | |
421748ec | 997 | path_put(&path); |
4e7b5671 | 998 | return error; |
1da177e4 | 999 | } |
d5686b44 | 1000 | EXPORT_SYMBOL(lookup_bdev); |
1da177e4 | 1001 | |
93b270f7 | 1002 | int __invalidate_device(struct block_device *bdev, bool kill_dirty) |
b71e8a4c DH |
1003 | { |
1004 | struct super_block *sb = get_super(bdev); | |
1005 | int res = 0; | |
1006 | ||
1007 | if (sb) { | |
1008 | /* | |
1009 | * no need to lock the super, get_super holds the | |
1010 | * read mutex so the filesystem cannot go away | |
1011 | * under us (->put_super runs with the write lock | |
1012 | * hold). | |
1013 | */ | |
1014 | shrink_dcache_sb(sb); | |
93b270f7 | 1015 | res = invalidate_inodes(sb, kill_dirty); |
b71e8a4c DH |
1016 | drop_super(sb); |
1017 | } | |
f98393a6 | 1018 | invalidate_bdev(bdev); |
b71e8a4c DH |
1019 | return res; |
1020 | } | |
1021 | EXPORT_SYMBOL(__invalidate_device); | |
5c0d6b60 | 1022 | |
1e03a36b | 1023 | void sync_bdevs(bool wait) |
5c0d6b60 JK |
1024 | { |
1025 | struct inode *inode, *old_inode = NULL; | |
1026 | ||
74278da9 | 1027 | spin_lock(&blockdev_superblock->s_inode_list_lock); |
5c0d6b60 JK |
1028 | list_for_each_entry(inode, &blockdev_superblock->s_inodes, i_sb_list) { |
1029 | struct address_space *mapping = inode->i_mapping; | |
af309226 | 1030 | struct block_device *bdev; |
5c0d6b60 JK |
1031 | |
1032 | spin_lock(&inode->i_lock); | |
1033 | if (inode->i_state & (I_FREEING|I_WILL_FREE|I_NEW) || | |
1034 | mapping->nrpages == 0) { | |
1035 | spin_unlock(&inode->i_lock); | |
1036 | continue; | |
1037 | } | |
1038 | __iget(inode); | |
1039 | spin_unlock(&inode->i_lock); | |
74278da9 | 1040 | spin_unlock(&blockdev_superblock->s_inode_list_lock); |
5c0d6b60 JK |
1041 | /* |
1042 | * We hold a reference to 'inode' so it couldn't have been | |
1043 | * removed from s_inodes list while we dropped the | |
74278da9 | 1044 | * s_inode_list_lock We cannot iput the inode now as we can |
5c0d6b60 | 1045 | * be holding the last reference and we cannot iput it under |
74278da9 | 1046 | * s_inode_list_lock. So we keep the reference and iput it |
5c0d6b60 JK |
1047 | * later. |
1048 | */ | |
1049 | iput(old_inode); | |
1050 | old_inode = inode; | |
af309226 | 1051 | bdev = I_BDEV(inode); |
5c0d6b60 | 1052 | |
a8698707 | 1053 | mutex_lock(&bdev->bd_disk->open_mutex); |
9acf381f | 1054 | if (!atomic_read(&bdev->bd_openers)) { |
1e03a36b CH |
1055 | ; /* skip */ |
1056 | } else if (wait) { | |
1057 | /* | |
1058 | * We keep the error status of individual mapping so | |
1059 | * that applications can catch the writeback error using | |
1060 | * fsync(2). See filemap_fdatawait_keep_errors() for | |
1061 | * details. | |
1062 | */ | |
1063 | filemap_fdatawait_keep_errors(inode->i_mapping); | |
1064 | } else { | |
1065 | filemap_fdatawrite(inode->i_mapping); | |
1066 | } | |
a8698707 | 1067 | mutex_unlock(&bdev->bd_disk->open_mutex); |
5c0d6b60 | 1068 | |
74278da9 | 1069 | spin_lock(&blockdev_superblock->s_inode_list_lock); |
5c0d6b60 | 1070 | } |
74278da9 | 1071 | spin_unlock(&blockdev_superblock->s_inode_list_lock); |
5c0d6b60 JK |
1072 | iput(old_inode); |
1073 | } |