]>
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> | |
66114cad | 17 | #include <linux/backing-dev.h> |
1da177e4 LT |
18 | #include <linux/module.h> |
19 | #include <linux/blkpg.h> | |
b502bd11 | 20 | #include <linux/magic.h> |
b0686260 | 21 | #include <linux/dax.h> |
1da177e4 | 22 | #include <linux/buffer_head.h> |
ff01bb48 | 23 | #include <linux/swap.h> |
585d3bc0 | 24 | #include <linux/pagevec.h> |
811d736f | 25 | #include <linux/writeback.h> |
1da177e4 LT |
26 | #include <linux/mpage.h> |
27 | #include <linux/mount.h> | |
28 | #include <linux/uio.h> | |
29 | #include <linux/namei.h> | |
1368c4f2 | 30 | #include <linux/log2.h> |
ff01bb48 | 31 | #include <linux/cleancache.h> |
c94c2acf | 32 | #include <linux/dax.h> |
acc93d30 | 33 | #include <linux/badblocks.h> |
189ce2b9 | 34 | #include <linux/task_io_accounting_ops.h> |
25f4c414 | 35 | #include <linux/falloc.h> |
7c0f6ba6 | 36 | #include <linux/uaccess.h> |
07f3f05c | 37 | #include "internal.h" |
1da177e4 LT |
38 | |
39 | struct bdev_inode { | |
40 | struct block_device bdev; | |
41 | struct inode vfs_inode; | |
42 | }; | |
43 | ||
4c54ac62 AB |
44 | static const struct address_space_operations def_blk_aops; |
45 | ||
1da177e4 LT |
46 | static inline struct bdev_inode *BDEV_I(struct inode *inode) |
47 | { | |
48 | return container_of(inode, struct bdev_inode, vfs_inode); | |
49 | } | |
50 | ||
ff5053f6 | 51 | struct block_device *I_BDEV(struct inode *inode) |
1da177e4 LT |
52 | { |
53 | return &BDEV_I(inode)->bdev; | |
54 | } | |
1da177e4 LT |
55 | EXPORT_SYMBOL(I_BDEV); |
56 | ||
dbd3ca50 | 57 | static void bdev_write_inode(struct block_device *bdev) |
564f00f6 | 58 | { |
dbd3ca50 VG |
59 | struct inode *inode = bdev->bd_inode; |
60 | int ret; | |
61 | ||
564f00f6 CH |
62 | spin_lock(&inode->i_lock); |
63 | while (inode->i_state & I_DIRTY) { | |
64 | spin_unlock(&inode->i_lock); | |
dbd3ca50 VG |
65 | ret = write_inode_now(inode, true); |
66 | if (ret) { | |
67 | char name[BDEVNAME_SIZE]; | |
68 | pr_warn_ratelimited("VFS: Dirty inode writeback failed " | |
69 | "for block device %s (err=%d).\n", | |
70 | bdevname(bdev, name), ret); | |
71 | } | |
564f00f6 CH |
72 | spin_lock(&inode->i_lock); |
73 | } | |
74 | spin_unlock(&inode->i_lock); | |
75 | } | |
76 | ||
f9a14399 | 77 | /* Kill _all_ buffers and pagecache , dirty or not.. */ |
ff01bb48 | 78 | void kill_bdev(struct block_device *bdev) |
1da177e4 | 79 | { |
ff01bb48 AV |
80 | struct address_space *mapping = bdev->bd_inode->i_mapping; |
81 | ||
f9fe48be | 82 | if (mapping->nrpages == 0 && mapping->nrexceptional == 0) |
f9a14399 | 83 | return; |
ff01bb48 | 84 | |
f9a14399 | 85 | invalidate_bh_lrus(); |
ff01bb48 | 86 | truncate_inode_pages(mapping, 0); |
1da177e4 | 87 | } |
ff01bb48 AV |
88 | EXPORT_SYMBOL(kill_bdev); |
89 | ||
90 | /* Invalidate clean unused buffers and pagecache. */ | |
91 | void invalidate_bdev(struct block_device *bdev) | |
92 | { | |
93 | struct address_space *mapping = bdev->bd_inode->i_mapping; | |
94 | ||
a5f6a6a9 AR |
95 | if (mapping->nrpages) { |
96 | invalidate_bh_lrus(); | |
97 | lru_add_drain_all(); /* make sure all lru add caches are flushed */ | |
98 | invalidate_mapping_pages(mapping, 0, -1); | |
99 | } | |
ff01bb48 AV |
100 | /* 99% of the time, we don't need to flush the cleancache on the bdev. |
101 | * But, for the strange corners, lets be cautious | |
102 | */ | |
3167760f | 103 | cleancache_invalidate_inode(mapping); |
ff01bb48 AV |
104 | } |
105 | EXPORT_SYMBOL(invalidate_bdev); | |
1da177e4 | 106 | |
c30974a1 JK |
107 | static void set_init_blocksize(struct block_device *bdev) |
108 | { | |
109 | unsigned bsize = bdev_logical_block_size(bdev); | |
110 | loff_t size = i_size_read(bdev->bd_inode); | |
111 | ||
112 | while (bsize < PAGE_SIZE) { | |
113 | if (size & bsize) | |
114 | break; | |
115 | bsize <<= 1; | |
116 | } | |
117 | bdev->bd_block_size = bsize; | |
118 | bdev->bd_inode->i_blkbits = blksize_bits(bsize); | |
119 | } | |
120 | ||
1da177e4 LT |
121 | int set_blocksize(struct block_device *bdev, int size) |
122 | { | |
123 | /* Size must be a power of two, and between 512 and PAGE_SIZE */ | |
1368c4f2 | 124 | if (size > PAGE_SIZE || size < 512 || !is_power_of_2(size)) |
1da177e4 LT |
125 | return -EINVAL; |
126 | ||
127 | /* Size cannot be smaller than the size supported by the device */ | |
e1defc4f | 128 | if (size < bdev_logical_block_size(bdev)) |
1da177e4 LT |
129 | return -EINVAL; |
130 | ||
131 | /* Don't change the size if it is same as current */ | |
132 | if (bdev->bd_block_size != size) { | |
133 | sync_blockdev(bdev); | |
134 | bdev->bd_block_size = size; | |
135 | bdev->bd_inode->i_blkbits = blksize_bits(size); | |
136 | kill_bdev(bdev); | |
137 | } | |
138 | return 0; | |
139 | } | |
140 | ||
141 | EXPORT_SYMBOL(set_blocksize); | |
142 | ||
143 | int sb_set_blocksize(struct super_block *sb, int size) | |
144 | { | |
1da177e4 LT |
145 | if (set_blocksize(sb->s_bdev, size)) |
146 | return 0; | |
147 | /* If we get here, we know size is power of two | |
148 | * and it's value is between 512 and PAGE_SIZE */ | |
149 | sb->s_blocksize = size; | |
38885bd4 | 150 | sb->s_blocksize_bits = blksize_bits(size); |
1da177e4 LT |
151 | return sb->s_blocksize; |
152 | } | |
153 | ||
154 | EXPORT_SYMBOL(sb_set_blocksize); | |
155 | ||
156 | int sb_min_blocksize(struct super_block *sb, int size) | |
157 | { | |
e1defc4f | 158 | int minsize = bdev_logical_block_size(sb->s_bdev); |
1da177e4 LT |
159 | if (size < minsize) |
160 | size = minsize; | |
161 | return sb_set_blocksize(sb, size); | |
162 | } | |
163 | ||
164 | EXPORT_SYMBOL(sb_min_blocksize); | |
165 | ||
166 | static int | |
167 | blkdev_get_block(struct inode *inode, sector_t iblock, | |
168 | struct buffer_head *bh, int create) | |
169 | { | |
1da177e4 LT |
170 | bh->b_bdev = I_BDEV(inode); |
171 | bh->b_blocknr = iblock; | |
172 | set_buffer_mapped(bh); | |
173 | return 0; | |
174 | } | |
175 | ||
4ebb16ca DW |
176 | static struct inode *bdev_file_inode(struct file *file) |
177 | { | |
178 | return file->f_mapping->host; | |
179 | } | |
180 | ||
78250c02 JA |
181 | static unsigned int dio_bio_write_op(struct kiocb *iocb) |
182 | { | |
183 | unsigned int op = REQ_OP_WRITE | REQ_SYNC | REQ_IDLE; | |
184 | ||
185 | /* avoid the need for a I/O completion work item */ | |
186 | if (iocb->ki_flags & IOCB_DSYNC) | |
187 | op |= REQ_FUA; | |
188 | return op; | |
189 | } | |
190 | ||
189ce2b9 CH |
191 | #define DIO_INLINE_BIO_VECS 4 |
192 | ||
193 | static void blkdev_bio_end_io_simple(struct bio *bio) | |
194 | { | |
195 | struct task_struct *waiter = bio->bi_private; | |
196 | ||
197 | WRITE_ONCE(bio->bi_private, NULL); | |
198 | wake_up_process(waiter); | |
199 | } | |
200 | ||
201 | static ssize_t | |
202 | __blkdev_direct_IO_simple(struct kiocb *iocb, struct iov_iter *iter, | |
203 | int nr_pages) | |
204 | { | |
205 | struct file *file = iocb->ki_filp; | |
206 | struct block_device *bdev = I_BDEV(bdev_file_inode(file)); | |
72ecad22 | 207 | struct bio_vec inline_vecs[DIO_INLINE_BIO_VECS], *vecs, *bvec; |
189ce2b9 CH |
208 | loff_t pos = iocb->ki_pos; |
209 | bool should_dirty = false; | |
210 | struct bio bio; | |
211 | ssize_t ret; | |
212 | blk_qc_t qc; | |
213 | int i; | |
214 | ||
9a794fb9 JA |
215 | if ((pos | iov_iter_alignment(iter)) & |
216 | (bdev_logical_block_size(bdev) - 1)) | |
189ce2b9 CH |
217 | return -EINVAL; |
218 | ||
72ecad22 JA |
219 | if (nr_pages <= DIO_INLINE_BIO_VECS) |
220 | vecs = inline_vecs; | |
221 | else { | |
222 | vecs = kmalloc(nr_pages * sizeof(struct bio_vec), GFP_KERNEL); | |
223 | if (!vecs) | |
224 | return -ENOMEM; | |
225 | } | |
226 | ||
3a83f467 | 227 | bio_init(&bio, vecs, nr_pages); |
74d46992 | 228 | bio_set_dev(&bio, bdev); |
4d1a4765 | 229 | bio.bi_iter.bi_sector = pos >> 9; |
45d06cf7 | 230 | bio.bi_write_hint = iocb->ki_hint; |
189ce2b9 CH |
231 | bio.bi_private = current; |
232 | bio.bi_end_io = blkdev_bio_end_io_simple; | |
233 | ||
234 | ret = bio_iov_iter_get_pages(&bio, iter); | |
235 | if (unlikely(ret)) | |
ba0a2578 | 236 | goto out; |
189ce2b9 CH |
237 | ret = bio.bi_iter.bi_size; |
238 | ||
239 | if (iov_iter_rw(iter) == READ) { | |
78250c02 | 240 | bio.bi_opf = REQ_OP_READ; |
189ce2b9 CH |
241 | if (iter_is_iovec(iter)) |
242 | should_dirty = true; | |
243 | } else { | |
78250c02 | 244 | bio.bi_opf = dio_bio_write_op(iocb); |
189ce2b9 CH |
245 | task_io_account_write(ret); |
246 | } | |
247 | ||
248 | qc = submit_bio(&bio); | |
249 | for (;;) { | |
250 | set_current_state(TASK_UNINTERRUPTIBLE); | |
251 | if (!READ_ONCE(bio.bi_private)) | |
252 | break; | |
253 | if (!(iocb->ki_flags & IOCB_HIPRI) || | |
ea435e1b | 254 | !blk_poll(bdev_get_queue(bdev), qc)) |
189ce2b9 CH |
255 | io_schedule(); |
256 | } | |
257 | __set_current_state(TASK_RUNNING); | |
258 | ||
259 | bio_for_each_segment_all(bvec, &bio, i) { | |
260 | if (should_dirty && !PageCompound(bvec->bv_page)) | |
261 | set_page_dirty_lock(bvec->bv_page); | |
262 | put_page(bvec->bv_page); | |
263 | } | |
264 | ||
4e4cbee9 | 265 | if (unlikely(bio.bi_status)) |
c6b1e36c | 266 | ret = blk_status_to_errno(bio.bi_status); |
9ae3b3f5 | 267 | |
ba0a2578 MW |
268 | out: |
269 | if (vecs != inline_vecs) | |
270 | kfree(vecs); | |
271 | ||
9ae3b3f5 JA |
272 | bio_uninit(&bio); |
273 | ||
189ce2b9 CH |
274 | return ret; |
275 | } | |
276 | ||
542ff7bf CH |
277 | struct blkdev_dio { |
278 | union { | |
279 | struct kiocb *iocb; | |
280 | struct task_struct *waiter; | |
281 | }; | |
282 | size_t size; | |
283 | atomic_t ref; | |
284 | bool multi_bio : 1; | |
285 | bool should_dirty : 1; | |
286 | bool is_sync : 1; | |
287 | struct bio bio; | |
288 | }; | |
289 | ||
290 | static struct bio_set *blkdev_dio_pool __read_mostly; | |
291 | ||
292 | static void blkdev_bio_end_io(struct bio *bio) | |
293 | { | |
294 | struct blkdev_dio *dio = bio->bi_private; | |
295 | bool should_dirty = dio->should_dirty; | |
296 | ||
297 | if (dio->multi_bio && !atomic_dec_and_test(&dio->ref)) { | |
4e4cbee9 CH |
298 | if (bio->bi_status && !dio->bio.bi_status) |
299 | dio->bio.bi_status = bio->bi_status; | |
542ff7bf CH |
300 | } else { |
301 | if (!dio->is_sync) { | |
302 | struct kiocb *iocb = dio->iocb; | |
4e4cbee9 | 303 | ssize_t ret; |
542ff7bf | 304 | |
4e4cbee9 | 305 | if (likely(!dio->bio.bi_status)) { |
542ff7bf CH |
306 | ret = dio->size; |
307 | iocb->ki_pos += ret; | |
4e4cbee9 CH |
308 | } else { |
309 | ret = blk_status_to_errno(dio->bio.bi_status); | |
542ff7bf CH |
310 | } |
311 | ||
312 | dio->iocb->ki_complete(iocb, ret, 0); | |
313 | bio_put(&dio->bio); | |
314 | } else { | |
315 | struct task_struct *waiter = dio->waiter; | |
316 | ||
317 | WRITE_ONCE(dio->waiter, NULL); | |
318 | wake_up_process(waiter); | |
319 | } | |
320 | } | |
321 | ||
322 | if (should_dirty) { | |
323 | bio_check_pages_dirty(bio); | |
324 | } else { | |
325 | struct bio_vec *bvec; | |
326 | int i; | |
327 | ||
328 | bio_for_each_segment_all(bvec, bio, i) | |
329 | put_page(bvec->bv_page); | |
330 | bio_put(bio); | |
331 | } | |
332 | } | |
333 | ||
b2e895db | 334 | static ssize_t |
542ff7bf | 335 | __blkdev_direct_IO(struct kiocb *iocb, struct iov_iter *iter, int nr_pages) |
b2e895db AM |
336 | { |
337 | struct file *file = iocb->ki_filp; | |
4ebb16ca | 338 | struct inode *inode = bdev_file_inode(file); |
542ff7bf | 339 | struct block_device *bdev = I_BDEV(inode); |
64d656a1 | 340 | struct blk_plug plug; |
542ff7bf CH |
341 | struct blkdev_dio *dio; |
342 | struct bio *bio; | |
690e5325 | 343 | bool is_read = (iov_iter_rw(iter) == READ), is_sync; |
542ff7bf CH |
344 | loff_t pos = iocb->ki_pos; |
345 | blk_qc_t qc = BLK_QC_T_NONE; | |
36ffc6c1 | 346 | int ret = 0; |
542ff7bf | 347 | |
9a794fb9 JA |
348 | if ((pos | iov_iter_alignment(iter)) & |
349 | (bdev_logical_block_size(bdev) - 1)) | |
542ff7bf CH |
350 | return -EINVAL; |
351 | ||
352 | bio = bio_alloc_bioset(GFP_KERNEL, nr_pages, blkdev_dio_pool); | |
353 | bio_get(bio); /* extra ref for the completion handler */ | |
354 | ||
355 | dio = container_of(bio, struct blkdev_dio, bio); | |
690e5325 | 356 | dio->is_sync = is_sync = is_sync_kiocb(iocb); |
542ff7bf CH |
357 | if (dio->is_sync) |
358 | dio->waiter = current; | |
359 | else | |
360 | dio->iocb = iocb; | |
361 | ||
362 | dio->size = 0; | |
363 | dio->multi_bio = false; | |
364 | dio->should_dirty = is_read && (iter->type == ITER_IOVEC); | |
365 | ||
64d656a1 | 366 | blk_start_plug(&plug); |
542ff7bf | 367 | for (;;) { |
74d46992 | 368 | bio_set_dev(bio, bdev); |
4d1a4765 | 369 | bio->bi_iter.bi_sector = pos >> 9; |
45d06cf7 | 370 | bio->bi_write_hint = iocb->ki_hint; |
542ff7bf CH |
371 | bio->bi_private = dio; |
372 | bio->bi_end_io = blkdev_bio_end_io; | |
373 | ||
374 | ret = bio_iov_iter_get_pages(bio, iter); | |
375 | if (unlikely(ret)) { | |
4e4cbee9 | 376 | bio->bi_status = BLK_STS_IOERR; |
542ff7bf CH |
377 | bio_endio(bio); |
378 | break; | |
379 | } | |
380 | ||
381 | if (is_read) { | |
382 | bio->bi_opf = REQ_OP_READ; | |
383 | if (dio->should_dirty) | |
384 | bio_set_pages_dirty(bio); | |
385 | } else { | |
386 | bio->bi_opf = dio_bio_write_op(iocb); | |
387 | task_io_account_write(bio->bi_iter.bi_size); | |
388 | } | |
389 | ||
390 | dio->size += bio->bi_iter.bi_size; | |
391 | pos += bio->bi_iter.bi_size; | |
392 | ||
393 | nr_pages = iov_iter_npages(iter, BIO_MAX_PAGES); | |
394 | if (!nr_pages) { | |
395 | qc = submit_bio(bio); | |
396 | break; | |
397 | } | |
398 | ||
399 | if (!dio->multi_bio) { | |
400 | dio->multi_bio = true; | |
401 | atomic_set(&dio->ref, 2); | |
402 | } else { | |
403 | atomic_inc(&dio->ref); | |
404 | } | |
405 | ||
406 | submit_bio(bio); | |
407 | bio = bio_alloc(GFP_KERNEL, nr_pages); | |
408 | } | |
64d656a1 | 409 | blk_finish_plug(&plug); |
542ff7bf | 410 | |
690e5325 | 411 | if (!is_sync) |
542ff7bf CH |
412 | return -EIOCBQUEUED; |
413 | ||
414 | for (;;) { | |
415 | set_current_state(TASK_UNINTERRUPTIBLE); | |
416 | if (!READ_ONCE(dio->waiter)) | |
417 | break; | |
418 | ||
419 | if (!(iocb->ki_flags & IOCB_HIPRI) || | |
ea435e1b | 420 | !blk_poll(bdev_get_queue(bdev), qc)) |
542ff7bf CH |
421 | io_schedule(); |
422 | } | |
423 | __set_current_state(TASK_RUNNING); | |
424 | ||
36ffc6c1 | 425 | if (!ret) |
4e4cbee9 | 426 | ret = blk_status_to_errno(dio->bio.bi_status); |
7a62a523 | 427 | if (likely(!ret)) |
542ff7bf | 428 | ret = dio->size; |
542ff7bf CH |
429 | |
430 | bio_put(&dio->bio); | |
431 | return ret; | |
432 | } | |
433 | ||
434 | static ssize_t | |
435 | blkdev_direct_IO(struct kiocb *iocb, struct iov_iter *iter) | |
436 | { | |
189ce2b9 | 437 | int nr_pages; |
b2e895db | 438 | |
72ecad22 | 439 | nr_pages = iov_iter_npages(iter, BIO_MAX_PAGES + 1); |
189ce2b9 CH |
440 | if (!nr_pages) |
441 | return 0; | |
72ecad22 | 442 | if (is_sync_kiocb(iocb) && nr_pages <= BIO_MAX_PAGES) |
189ce2b9 | 443 | return __blkdev_direct_IO_simple(iocb, iter, nr_pages); |
542ff7bf CH |
444 | |
445 | return __blkdev_direct_IO(iocb, iter, min(nr_pages, BIO_MAX_PAGES)); | |
446 | } | |
447 | ||
448 | static __init int blkdev_init(void) | |
449 | { | |
011067b0 | 450 | blkdev_dio_pool = bioset_create(4, offsetof(struct blkdev_dio, bio), BIOSET_NEED_BVECS); |
542ff7bf CH |
451 | if (!blkdev_dio_pool) |
452 | return -ENOMEM; | |
453 | return 0; | |
b2e895db | 454 | } |
542ff7bf | 455 | module_init(blkdev_init); |
b2e895db | 456 | |
5cee5815 JK |
457 | int __sync_blockdev(struct block_device *bdev, int wait) |
458 | { | |
459 | if (!bdev) | |
460 | return 0; | |
461 | if (!wait) | |
462 | return filemap_flush(bdev->bd_inode->i_mapping); | |
463 | return filemap_write_and_wait(bdev->bd_inode->i_mapping); | |
464 | } | |
465 | ||
585d3bc0 NP |
466 | /* |
467 | * Write out and wait upon all the dirty data associated with a block | |
468 | * device via its mapping. Does not take the superblock lock. | |
469 | */ | |
470 | int sync_blockdev(struct block_device *bdev) | |
471 | { | |
5cee5815 | 472 | return __sync_blockdev(bdev, 1); |
585d3bc0 NP |
473 | } |
474 | EXPORT_SYMBOL(sync_blockdev); | |
475 | ||
476 | /* | |
477 | * Write out and wait upon all dirty data associated with this | |
478 | * device. Filesystem data as well as the underlying block | |
479 | * device. Takes the superblock lock. | |
480 | */ | |
481 | int fsync_bdev(struct block_device *bdev) | |
482 | { | |
483 | struct super_block *sb = get_super(bdev); | |
484 | if (sb) { | |
60b0680f | 485 | int res = sync_filesystem(sb); |
585d3bc0 NP |
486 | drop_super(sb); |
487 | return res; | |
488 | } | |
489 | return sync_blockdev(bdev); | |
490 | } | |
47e4491b | 491 | EXPORT_SYMBOL(fsync_bdev); |
585d3bc0 NP |
492 | |
493 | /** | |
494 | * freeze_bdev -- lock a filesystem and force it into a consistent state | |
495 | * @bdev: blockdevice to lock | |
496 | * | |
585d3bc0 NP |
497 | * If a superblock is found on this device, we take the s_umount semaphore |
498 | * on it to make sure nobody unmounts until the snapshot creation is done. | |
499 | * The reference counter (bd_fsfreeze_count) guarantees that only the last | |
500 | * unfreeze process can unfreeze the frozen filesystem actually when multiple | |
501 | * freeze requests arrive simultaneously. It counts up in freeze_bdev() and | |
502 | * count down in thaw_bdev(). When it becomes 0, thaw_bdev() will unfreeze | |
503 | * actually. | |
504 | */ | |
505 | struct super_block *freeze_bdev(struct block_device *bdev) | |
506 | { | |
507 | struct super_block *sb; | |
508 | int error = 0; | |
509 | ||
510 | mutex_lock(&bdev->bd_fsfreeze_mutex); | |
4504230a CH |
511 | if (++bdev->bd_fsfreeze_count > 1) { |
512 | /* | |
513 | * We don't even need to grab a reference - the first call | |
514 | * to freeze_bdev grab an active reference and only the last | |
515 | * thaw_bdev drops it. | |
516 | */ | |
585d3bc0 | 517 | sb = get_super(bdev); |
5bb53c0f AR |
518 | if (sb) |
519 | drop_super(sb); | |
4504230a CH |
520 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
521 | return sb; | |
522 | } | |
523 | ||
524 | sb = get_active_super(bdev); | |
525 | if (!sb) | |
526 | goto out; | |
48b6bca6 BM |
527 | if (sb->s_op->freeze_super) |
528 | error = sb->s_op->freeze_super(sb); | |
529 | else | |
530 | error = freeze_super(sb); | |
18e9e510 JB |
531 | if (error) { |
532 | deactivate_super(sb); | |
533 | bdev->bd_fsfreeze_count--; | |
585d3bc0 | 534 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
18e9e510 | 535 | return ERR_PTR(error); |
585d3bc0 | 536 | } |
18e9e510 | 537 | deactivate_super(sb); |
4504230a | 538 | out: |
585d3bc0 NP |
539 | sync_blockdev(bdev); |
540 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
4fadd7bb | 541 | return sb; /* thaw_bdev releases s->s_umount */ |
585d3bc0 NP |
542 | } |
543 | EXPORT_SYMBOL(freeze_bdev); | |
544 | ||
545 | /** | |
546 | * thaw_bdev -- unlock filesystem | |
547 | * @bdev: blockdevice to unlock | |
548 | * @sb: associated superblock | |
549 | * | |
550 | * Unlocks the filesystem and marks it writeable again after freeze_bdev(). | |
551 | */ | |
552 | int thaw_bdev(struct block_device *bdev, struct super_block *sb) | |
553 | { | |
4504230a | 554 | int error = -EINVAL; |
585d3bc0 NP |
555 | |
556 | mutex_lock(&bdev->bd_fsfreeze_mutex); | |
4504230a | 557 | if (!bdev->bd_fsfreeze_count) |
18e9e510 | 558 | goto out; |
4504230a CH |
559 | |
560 | error = 0; | |
561 | if (--bdev->bd_fsfreeze_count > 0) | |
18e9e510 | 562 | goto out; |
4504230a CH |
563 | |
564 | if (!sb) | |
18e9e510 | 565 | goto out; |
4504230a | 566 | |
48b6bca6 BM |
567 | if (sb->s_op->thaw_super) |
568 | error = sb->s_op->thaw_super(sb); | |
569 | else | |
570 | error = thaw_super(sb); | |
997198ba | 571 | if (error) |
18e9e510 | 572 | bdev->bd_fsfreeze_count++; |
18e9e510 | 573 | out: |
585d3bc0 | 574 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
997198ba | 575 | return error; |
585d3bc0 NP |
576 | } |
577 | EXPORT_SYMBOL(thaw_bdev); | |
578 | ||
1da177e4 LT |
579 | static int blkdev_writepage(struct page *page, struct writeback_control *wbc) |
580 | { | |
581 | return block_write_full_page(page, blkdev_get_block, wbc); | |
582 | } | |
583 | ||
584 | static int blkdev_readpage(struct file * file, struct page * page) | |
585 | { | |
586 | return block_read_full_page(page, blkdev_get_block); | |
587 | } | |
588 | ||
447f05bb AM |
589 | static int blkdev_readpages(struct file *file, struct address_space *mapping, |
590 | struct list_head *pages, unsigned nr_pages) | |
591 | { | |
592 | return mpage_readpages(mapping, pages, nr_pages, blkdev_get_block); | |
593 | } | |
594 | ||
6272b5a5 NP |
595 | static int blkdev_write_begin(struct file *file, struct address_space *mapping, |
596 | loff_t pos, unsigned len, unsigned flags, | |
597 | struct page **pagep, void **fsdata) | |
1da177e4 | 598 | { |
155130a4 CH |
599 | return block_write_begin(mapping, pos, len, flags, pagep, |
600 | blkdev_get_block); | |
1da177e4 LT |
601 | } |
602 | ||
6272b5a5 NP |
603 | static int blkdev_write_end(struct file *file, struct address_space *mapping, |
604 | loff_t pos, unsigned len, unsigned copied, | |
605 | struct page *page, void *fsdata) | |
1da177e4 | 606 | { |
6272b5a5 NP |
607 | int ret; |
608 | ret = block_write_end(file, mapping, pos, len, copied, page, fsdata); | |
609 | ||
610 | unlock_page(page); | |
09cbfeaf | 611 | put_page(page); |
6272b5a5 NP |
612 | |
613 | return ret; | |
1da177e4 LT |
614 | } |
615 | ||
616 | /* | |
617 | * private llseek: | |
496ad9aa | 618 | * for a block special file file_inode(file)->i_size is zero |
1da177e4 LT |
619 | * so we compute the size by hand (just as in block_read/write above) |
620 | */ | |
965c8e59 | 621 | static loff_t block_llseek(struct file *file, loff_t offset, int whence) |
1da177e4 | 622 | { |
4ebb16ca | 623 | struct inode *bd_inode = bdev_file_inode(file); |
1da177e4 LT |
624 | loff_t retval; |
625 | ||
5955102c | 626 | inode_lock(bd_inode); |
5d48f3a2 | 627 | retval = fixed_size_llseek(file, offset, whence, i_size_read(bd_inode)); |
5955102c | 628 | inode_unlock(bd_inode); |
1da177e4 LT |
629 | return retval; |
630 | } | |
631 | ||
02c24a82 | 632 | int blkdev_fsync(struct file *filp, loff_t start, loff_t end, int datasync) |
1da177e4 | 633 | { |
4ebb16ca | 634 | struct inode *bd_inode = bdev_file_inode(filp); |
b8af67e2 | 635 | struct block_device *bdev = I_BDEV(bd_inode); |
ab0a9735 | 636 | int error; |
da5aa861 | 637 | |
372cf243 | 638 | error = file_write_and_wait_range(filp, start, end); |
da5aa861 RW |
639 | if (error) |
640 | return error; | |
ab0a9735 | 641 | |
b8af67e2 AB |
642 | /* |
643 | * There is no need to serialise calls to blkdev_issue_flush with | |
644 | * i_mutex and doing so causes performance issues with concurrent | |
645 | * O_SYNC writers to a block device. | |
646 | */ | |
dd3932ed | 647 | error = blkdev_issue_flush(bdev, GFP_KERNEL, NULL); |
ab0a9735 CH |
648 | if (error == -EOPNOTSUPP) |
649 | error = 0; | |
b8af67e2 | 650 | |
ab0a9735 | 651 | return error; |
1da177e4 | 652 | } |
b1dd3b28 | 653 | EXPORT_SYMBOL(blkdev_fsync); |
1da177e4 | 654 | |
47a191fd MW |
655 | /** |
656 | * bdev_read_page() - Start reading a page from a block device | |
657 | * @bdev: The device to read the page from | |
658 | * @sector: The offset on the device to read the page to (need not be aligned) | |
659 | * @page: The page to read | |
660 | * | |
661 | * On entry, the page should be locked. It will be unlocked when the page | |
662 | * has been read. If the block driver implements rw_page synchronously, | |
663 | * that will be true on exit from this function, but it need not be. | |
664 | * | |
665 | * Errors returned by this function are usually "soft", eg out of memory, or | |
666 | * queue full; callers should try a different route to read this page rather | |
667 | * than propagate an error back up the stack. | |
668 | * | |
669 | * Return: negative errno if an error occurs, 0 if submission was successful. | |
670 | */ | |
671 | int bdev_read_page(struct block_device *bdev, sector_t sector, | |
672 | struct page *page) | |
673 | { | |
674 | const struct block_device_operations *ops = bdev->bd_disk->fops; | |
2e6edc95 DW |
675 | int result = -EOPNOTSUPP; |
676 | ||
f68eb1e7 | 677 | if (!ops->rw_page || bdev_get_integrity(bdev)) |
2e6edc95 DW |
678 | return result; |
679 | ||
3a0a5299 | 680 | result = blk_queue_enter(bdev->bd_queue, 0); |
2e6edc95 DW |
681 | if (result) |
682 | return result; | |
c11f0c0b | 683 | result = ops->rw_page(bdev, sector + get_start_sect(bdev), page, false); |
2e6edc95 DW |
684 | blk_queue_exit(bdev->bd_queue); |
685 | return result; | |
47a191fd MW |
686 | } |
687 | EXPORT_SYMBOL_GPL(bdev_read_page); | |
688 | ||
689 | /** | |
690 | * bdev_write_page() - Start writing a page to a block device | |
691 | * @bdev: The device to write the page to | |
692 | * @sector: The offset on the device to write the page to (need not be aligned) | |
693 | * @page: The page to write | |
694 | * @wbc: The writeback_control for the write | |
695 | * | |
696 | * On entry, the page should be locked and not currently under writeback. | |
697 | * On exit, if the write started successfully, the page will be unlocked and | |
698 | * under writeback. If the write failed already (eg the driver failed to | |
699 | * queue the page to the device), the page will still be locked. If the | |
700 | * caller is a ->writepage implementation, it will need to unlock the page. | |
701 | * | |
702 | * Errors returned by this function are usually "soft", eg out of memory, or | |
703 | * queue full; callers should try a different route to write this page rather | |
704 | * than propagate an error back up the stack. | |
705 | * | |
706 | * Return: negative errno if an error occurs, 0 if submission was successful. | |
707 | */ | |
708 | int bdev_write_page(struct block_device *bdev, sector_t sector, | |
709 | struct page *page, struct writeback_control *wbc) | |
710 | { | |
711 | int result; | |
47a191fd | 712 | const struct block_device_operations *ops = bdev->bd_disk->fops; |
2e6edc95 | 713 | |
f68eb1e7 | 714 | if (!ops->rw_page || bdev_get_integrity(bdev)) |
47a191fd | 715 | return -EOPNOTSUPP; |
3a0a5299 | 716 | result = blk_queue_enter(bdev->bd_queue, 0); |
2e6edc95 DW |
717 | if (result) |
718 | return result; | |
719 | ||
47a191fd | 720 | set_page_writeback(page); |
c11f0c0b | 721 | result = ops->rw_page(bdev, sector + get_start_sect(bdev), page, true); |
f892760a | 722 | if (result) { |
47a191fd | 723 | end_page_writeback(page); |
f892760a MW |
724 | } else { |
725 | clean_page_buffers(page); | |
47a191fd | 726 | unlock_page(page); |
f892760a | 727 | } |
2e6edc95 | 728 | blk_queue_exit(bdev->bd_queue); |
47a191fd MW |
729 | return result; |
730 | } | |
731 | EXPORT_SYMBOL_GPL(bdev_write_page); | |
732 | ||
1da177e4 LT |
733 | /* |
734 | * pseudo-fs | |
735 | */ | |
736 | ||
737 | static __cacheline_aligned_in_smp DEFINE_SPINLOCK(bdev_lock); | |
e18b890b | 738 | static struct kmem_cache * bdev_cachep __read_mostly; |
1da177e4 LT |
739 | |
740 | static struct inode *bdev_alloc_inode(struct super_block *sb) | |
741 | { | |
e94b1766 | 742 | struct bdev_inode *ei = kmem_cache_alloc(bdev_cachep, GFP_KERNEL); |
1da177e4 LT |
743 | if (!ei) |
744 | return NULL; | |
745 | return &ei->vfs_inode; | |
746 | } | |
747 | ||
fa0d7e3d | 748 | static void bdev_i_callback(struct rcu_head *head) |
1da177e4 | 749 | { |
fa0d7e3d | 750 | struct inode *inode = container_of(head, struct inode, i_rcu); |
1da177e4 LT |
751 | struct bdev_inode *bdi = BDEV_I(inode); |
752 | ||
1da177e4 LT |
753 | kmem_cache_free(bdev_cachep, bdi); |
754 | } | |
755 | ||
fa0d7e3d NP |
756 | static void bdev_destroy_inode(struct inode *inode) |
757 | { | |
758 | call_rcu(&inode->i_rcu, bdev_i_callback); | |
759 | } | |
760 | ||
51cc5068 | 761 | static void init_once(void *foo) |
1da177e4 LT |
762 | { |
763 | struct bdev_inode *ei = (struct bdev_inode *) foo; | |
764 | struct block_device *bdev = &ei->bdev; | |
765 | ||
a35afb83 CL |
766 | memset(bdev, 0, sizeof(*bdev)); |
767 | mutex_init(&bdev->bd_mutex); | |
a35afb83 | 768 | INIT_LIST_HEAD(&bdev->bd_list); |
49731baa TH |
769 | #ifdef CONFIG_SYSFS |
770 | INIT_LIST_HEAD(&bdev->bd_holder_disks); | |
771 | #endif | |
a5a79d00 | 772 | bdev->bd_bdi = &noop_backing_dev_info; |
a35afb83 | 773 | inode_init_once(&ei->vfs_inode); |
fcccf502 TS |
774 | /* Initialize mutex for freeze. */ |
775 | mutex_init(&bdev->bd_fsfreeze_mutex); | |
1da177e4 LT |
776 | } |
777 | ||
b57922d9 | 778 | static void bdev_evict_inode(struct inode *inode) |
1da177e4 LT |
779 | { |
780 | struct block_device *bdev = &BDEV_I(inode)->bdev; | |
91b0abe3 | 781 | truncate_inode_pages_final(&inode->i_data); |
b57922d9 | 782 | invalidate_inode_buffers(inode); /* is it needed here? */ |
dbd5768f | 783 | clear_inode(inode); |
1da177e4 | 784 | spin_lock(&bdev_lock); |
1da177e4 LT |
785 | list_del_init(&bdev->bd_list); |
786 | spin_unlock(&bdev_lock); | |
f759741d JK |
787 | /* Detach inode from wb early as bdi_put() may free bdi->wb */ |
788 | inode_detach_wb(inode); | |
a5a79d00 | 789 | if (bdev->bd_bdi != &noop_backing_dev_info) { |
b1d2dc56 | 790 | bdi_put(bdev->bd_bdi); |
a5a79d00 JK |
791 | bdev->bd_bdi = &noop_backing_dev_info; |
792 | } | |
1da177e4 LT |
793 | } |
794 | ||
ee9b6d61 | 795 | static const struct super_operations bdev_sops = { |
1da177e4 LT |
796 | .statfs = simple_statfs, |
797 | .alloc_inode = bdev_alloc_inode, | |
798 | .destroy_inode = bdev_destroy_inode, | |
799 | .drop_inode = generic_delete_inode, | |
b57922d9 | 800 | .evict_inode = bdev_evict_inode, |
1da177e4 LT |
801 | }; |
802 | ||
51139ada AV |
803 | static struct dentry *bd_mount(struct file_system_type *fs_type, |
804 | int flags, const char *dev_name, void *data) | |
1da177e4 | 805 | { |
3684aa70 SL |
806 | struct dentry *dent; |
807 | dent = mount_pseudo(fs_type, "bdev:", &bdev_sops, NULL, BDEVFS_MAGIC); | |
e9e5e3fa | 808 | if (!IS_ERR(dent)) |
3684aa70 SL |
809 | dent->d_sb->s_iflags |= SB_I_CGROUPWB; |
810 | return dent; | |
1da177e4 LT |
811 | } |
812 | ||
813 | static struct file_system_type bd_type = { | |
814 | .name = "bdev", | |
51139ada | 815 | .mount = bd_mount, |
1da177e4 LT |
816 | .kill_sb = kill_anon_super, |
817 | }; | |
818 | ||
a212b105 TH |
819 | struct super_block *blockdev_superblock __read_mostly; |
820 | EXPORT_SYMBOL_GPL(blockdev_superblock); | |
1da177e4 LT |
821 | |
822 | void __init bdev_cache_init(void) | |
823 | { | |
824 | int err; | |
ace8577a | 825 | static struct vfsmount *bd_mnt; |
c2acf7b9 | 826 | |
1da177e4 | 827 | bdev_cachep = kmem_cache_create("bdev_cache", sizeof(struct bdev_inode), |
fffb60f9 | 828 | 0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| |
5d097056 | 829 | SLAB_MEM_SPREAD|SLAB_ACCOUNT|SLAB_PANIC), |
20c2df83 | 830 | init_once); |
1da177e4 LT |
831 | err = register_filesystem(&bd_type); |
832 | if (err) | |
833 | panic("Cannot register bdev pseudo-fs"); | |
834 | bd_mnt = kern_mount(&bd_type); | |
1da177e4 LT |
835 | if (IS_ERR(bd_mnt)) |
836 | panic("Cannot create bdev pseudo-fs"); | |
ace8577a | 837 | blockdev_superblock = bd_mnt->mnt_sb; /* For writeback */ |
1da177e4 LT |
838 | } |
839 | ||
840 | /* | |
841 | * Most likely _very_ bad one - but then it's hardly critical for small | |
842 | * /dev and can be fixed when somebody will need really large one. | |
843 | * Keep in mind that it will be fed through icache hash function too. | |
844 | */ | |
845 | static inline unsigned long hash(dev_t dev) | |
846 | { | |
847 | return MAJOR(dev)+MINOR(dev); | |
848 | } | |
849 | ||
850 | static int bdev_test(struct inode *inode, void *data) | |
851 | { | |
852 | return BDEV_I(inode)->bdev.bd_dev == *(dev_t *)data; | |
853 | } | |
854 | ||
855 | static int bdev_set(struct inode *inode, void *data) | |
856 | { | |
857 | BDEV_I(inode)->bdev.bd_dev = *(dev_t *)data; | |
858 | return 0; | |
859 | } | |
860 | ||
861 | static LIST_HEAD(all_bdevs); | |
862 | ||
f44f1ab5 JK |
863 | /* |
864 | * If there is a bdev inode for this device, unhash it so that it gets evicted | |
865 | * as soon as last inode reference is dropped. | |
866 | */ | |
867 | void bdev_unhash_inode(dev_t dev) | |
868 | { | |
869 | struct inode *inode; | |
870 | ||
871 | inode = ilookup5(blockdev_superblock, hash(dev), bdev_test, &dev); | |
872 | if (inode) { | |
873 | remove_inode_hash(inode); | |
874 | iput(inode); | |
875 | } | |
876 | } | |
877 | ||
1da177e4 LT |
878 | struct block_device *bdget(dev_t dev) |
879 | { | |
880 | struct block_device *bdev; | |
881 | struct inode *inode; | |
882 | ||
c2acf7b9 | 883 | inode = iget5_locked(blockdev_superblock, hash(dev), |
1da177e4 LT |
884 | bdev_test, bdev_set, &dev); |
885 | ||
886 | if (!inode) | |
887 | return NULL; | |
888 | ||
889 | bdev = &BDEV_I(inode)->bdev; | |
890 | ||
891 | if (inode->i_state & I_NEW) { | |
892 | bdev->bd_contains = NULL; | |
782b94cd | 893 | bdev->bd_super = NULL; |
1da177e4 | 894 | bdev->bd_inode = inode; |
93407472 | 895 | bdev->bd_block_size = i_blocksize(inode); |
1da177e4 LT |
896 | bdev->bd_part_count = 0; |
897 | bdev->bd_invalidated = 0; | |
898 | inode->i_mode = S_IFBLK; | |
899 | inode->i_rdev = dev; | |
900 | inode->i_bdev = bdev; | |
901 | inode->i_data.a_ops = &def_blk_aops; | |
902 | mapping_set_gfp_mask(&inode->i_data, GFP_USER); | |
1da177e4 LT |
903 | spin_lock(&bdev_lock); |
904 | list_add(&bdev->bd_list, &all_bdevs); | |
905 | spin_unlock(&bdev_lock); | |
906 | unlock_new_inode(inode); | |
907 | } | |
908 | return bdev; | |
909 | } | |
910 | ||
911 | EXPORT_SYMBOL(bdget); | |
912 | ||
dddac6a7 AJ |
913 | /** |
914 | * bdgrab -- Grab a reference to an already referenced block device | |
915 | * @bdev: Block device to grab a reference to. | |
916 | */ | |
917 | struct block_device *bdgrab(struct block_device *bdev) | |
918 | { | |
7de9c6ee | 919 | ihold(bdev->bd_inode); |
dddac6a7 AJ |
920 | return bdev; |
921 | } | |
c1681bf8 | 922 | EXPORT_SYMBOL(bdgrab); |
dddac6a7 | 923 | |
1da177e4 LT |
924 | long nr_blockdev_pages(void) |
925 | { | |
203a2935 | 926 | struct block_device *bdev; |
1da177e4 LT |
927 | long ret = 0; |
928 | spin_lock(&bdev_lock); | |
203a2935 | 929 | list_for_each_entry(bdev, &all_bdevs, bd_list) { |
1da177e4 LT |
930 | ret += bdev->bd_inode->i_mapping->nrpages; |
931 | } | |
932 | spin_unlock(&bdev_lock); | |
933 | return ret; | |
934 | } | |
935 | ||
936 | void bdput(struct block_device *bdev) | |
937 | { | |
938 | iput(bdev->bd_inode); | |
939 | } | |
940 | ||
941 | EXPORT_SYMBOL(bdput); | |
942 | ||
943 | static struct block_device *bd_acquire(struct inode *inode) | |
944 | { | |
945 | struct block_device *bdev; | |
09d967c6 | 946 | |
1da177e4 LT |
947 | spin_lock(&bdev_lock); |
948 | bdev = inode->i_bdev; | |
cccd9fb9 | 949 | if (bdev && !inode_unhashed(bdev->bd_inode)) { |
ed8a9d2c | 950 | bdgrab(bdev); |
1da177e4 LT |
951 | spin_unlock(&bdev_lock); |
952 | return bdev; | |
953 | } | |
954 | spin_unlock(&bdev_lock); | |
09d967c6 | 955 | |
cccd9fb9 JK |
956 | /* |
957 | * i_bdev references block device inode that was already shut down | |
958 | * (corresponding device got removed). Remove the reference and look | |
959 | * up block device inode again just in case new device got | |
960 | * reestablished under the same device number. | |
961 | */ | |
962 | if (bdev) | |
963 | bd_forget(inode); | |
964 | ||
1da177e4 LT |
965 | bdev = bdget(inode->i_rdev); |
966 | if (bdev) { | |
967 | spin_lock(&bdev_lock); | |
09d967c6 OH |
968 | if (!inode->i_bdev) { |
969 | /* | |
7de9c6ee | 970 | * We take an additional reference to bd_inode, |
09d967c6 OH |
971 | * and it's released in clear_inode() of inode. |
972 | * So, we can access it via ->i_mapping always | |
973 | * without igrab(). | |
974 | */ | |
ed8a9d2c | 975 | bdgrab(bdev); |
09d967c6 OH |
976 | inode->i_bdev = bdev; |
977 | inode->i_mapping = bdev->bd_inode->i_mapping; | |
09d967c6 | 978 | } |
1da177e4 LT |
979 | spin_unlock(&bdev_lock); |
980 | } | |
981 | return bdev; | |
982 | } | |
983 | ||
984 | /* Call when you free inode */ | |
985 | ||
986 | void bd_forget(struct inode *inode) | |
987 | { | |
09d967c6 OH |
988 | struct block_device *bdev = NULL; |
989 | ||
1da177e4 | 990 | spin_lock(&bdev_lock); |
b4ea2eaa YH |
991 | if (!sb_is_blkdev_sb(inode->i_sb)) |
992 | bdev = inode->i_bdev; | |
a4a4f943 AV |
993 | inode->i_bdev = NULL; |
994 | inode->i_mapping = &inode->i_data; | |
1da177e4 | 995 | spin_unlock(&bdev_lock); |
09d967c6 OH |
996 | |
997 | if (bdev) | |
ed8a9d2c | 998 | bdput(bdev); |
1da177e4 LT |
999 | } |
1000 | ||
1a3cbbc5 TH |
1001 | /** |
1002 | * bd_may_claim - test whether a block device can be claimed | |
1003 | * @bdev: block device of interest | |
1004 | * @whole: whole block device containing @bdev, may equal @bdev | |
1005 | * @holder: holder trying to claim @bdev | |
1006 | * | |
25985edc | 1007 | * Test whether @bdev can be claimed by @holder. |
1a3cbbc5 TH |
1008 | * |
1009 | * CONTEXT: | |
1010 | * spin_lock(&bdev_lock). | |
1011 | * | |
1012 | * RETURNS: | |
1013 | * %true if @bdev can be claimed, %false otherwise. | |
1014 | */ | |
1015 | static bool bd_may_claim(struct block_device *bdev, struct block_device *whole, | |
1016 | void *holder) | |
1da177e4 | 1017 | { |
1da177e4 | 1018 | if (bdev->bd_holder == holder) |
1a3cbbc5 | 1019 | return true; /* already a holder */ |
1da177e4 | 1020 | else if (bdev->bd_holder != NULL) |
1a3cbbc5 | 1021 | return false; /* held by someone else */ |
bcc7f5b4 | 1022 | else if (whole == bdev) |
1a3cbbc5 | 1023 | return true; /* is a whole device which isn't held */ |
1da177e4 | 1024 | |
e525fd89 | 1025 | else if (whole->bd_holder == bd_may_claim) |
1a3cbbc5 TH |
1026 | return true; /* is a partition of a device that is being partitioned */ |
1027 | else if (whole->bd_holder != NULL) | |
1028 | return false; /* is a partition of a held device */ | |
1da177e4 | 1029 | else |
1a3cbbc5 TH |
1030 | return true; /* is a partition of an un-held device */ |
1031 | } | |
1032 | ||
6b4517a7 TH |
1033 | /** |
1034 | * bd_prepare_to_claim - prepare to claim a block device | |
1035 | * @bdev: block device of interest | |
1036 | * @whole: the whole device containing @bdev, may equal @bdev | |
1037 | * @holder: holder trying to claim @bdev | |
1038 | * | |
1039 | * Prepare to claim @bdev. This function fails if @bdev is already | |
1040 | * claimed by another holder and waits if another claiming is in | |
1041 | * progress. This function doesn't actually claim. On successful | |
1042 | * return, the caller has ownership of bd_claiming and bd_holder[s]. | |
1043 | * | |
1044 | * CONTEXT: | |
1045 | * spin_lock(&bdev_lock). Might release bdev_lock, sleep and regrab | |
1046 | * it multiple times. | |
1047 | * | |
1048 | * RETURNS: | |
1049 | * 0 if @bdev can be claimed, -EBUSY otherwise. | |
1050 | */ | |
1051 | static int bd_prepare_to_claim(struct block_device *bdev, | |
1052 | struct block_device *whole, void *holder) | |
1053 | { | |
1054 | retry: | |
1055 | /* if someone else claimed, fail */ | |
1056 | if (!bd_may_claim(bdev, whole, holder)) | |
1057 | return -EBUSY; | |
1058 | ||
e75aa858 TH |
1059 | /* if claiming is already in progress, wait for it to finish */ |
1060 | if (whole->bd_claiming) { | |
6b4517a7 TH |
1061 | wait_queue_head_t *wq = bit_waitqueue(&whole->bd_claiming, 0); |
1062 | DEFINE_WAIT(wait); | |
1063 | ||
1064 | prepare_to_wait(wq, &wait, TASK_UNINTERRUPTIBLE); | |
1065 | spin_unlock(&bdev_lock); | |
1066 | schedule(); | |
1067 | finish_wait(wq, &wait); | |
1068 | spin_lock(&bdev_lock); | |
1069 | goto retry; | |
1070 | } | |
1071 | ||
1072 | /* yay, all mine */ | |
1073 | return 0; | |
1074 | } | |
1075 | ||
1076 | /** | |
1077 | * bd_start_claiming - start claiming a block device | |
1078 | * @bdev: block device of interest | |
1079 | * @holder: holder trying to claim @bdev | |
1080 | * | |
1081 | * @bdev is about to be opened exclusively. Check @bdev can be opened | |
1082 | * exclusively and mark that an exclusive open is in progress. Each | |
1083 | * successful call to this function must be matched with a call to | |
b0018361 NP |
1084 | * either bd_finish_claiming() or bd_abort_claiming() (which do not |
1085 | * fail). | |
1086 | * | |
1087 | * This function is used to gain exclusive access to the block device | |
1088 | * without actually causing other exclusive open attempts to fail. It | |
1089 | * should be used when the open sequence itself requires exclusive | |
1090 | * access but may subsequently fail. | |
6b4517a7 TH |
1091 | * |
1092 | * CONTEXT: | |
1093 | * Might sleep. | |
1094 | * | |
1095 | * RETURNS: | |
1096 | * Pointer to the block device containing @bdev on success, ERR_PTR() | |
1097 | * value on failure. | |
1098 | */ | |
1099 | static struct block_device *bd_start_claiming(struct block_device *bdev, | |
1100 | void *holder) | |
1101 | { | |
1102 | struct gendisk *disk; | |
1103 | struct block_device *whole; | |
1104 | int partno, err; | |
1105 | ||
1106 | might_sleep(); | |
1107 | ||
1108 | /* | |
1109 | * @bdev might not have been initialized properly yet, look up | |
1110 | * and grab the outer block device the hard way. | |
1111 | */ | |
1112 | disk = get_gendisk(bdev->bd_dev, &partno); | |
1113 | if (!disk) | |
1114 | return ERR_PTR(-ENXIO); | |
1115 | ||
d4c208b8 TH |
1116 | /* |
1117 | * Normally, @bdev should equal what's returned from bdget_disk() | |
1118 | * if partno is 0; however, some drivers (floppy) use multiple | |
1119 | * bdev's for the same physical device and @bdev may be one of the | |
1120 | * aliases. Keep @bdev if partno is 0. This means claimer | |
1121 | * tracking is broken for those devices but it has always been that | |
1122 | * way. | |
1123 | */ | |
1124 | if (partno) | |
1125 | whole = bdget_disk(disk, 0); | |
1126 | else | |
1127 | whole = bdgrab(bdev); | |
1128 | ||
cf342570 | 1129 | module_put(disk->fops->owner); |
6b4517a7 TH |
1130 | put_disk(disk); |
1131 | if (!whole) | |
1132 | return ERR_PTR(-ENOMEM); | |
1133 | ||
1134 | /* prepare to claim, if successful, mark claiming in progress */ | |
1135 | spin_lock(&bdev_lock); | |
1136 | ||
1137 | err = bd_prepare_to_claim(bdev, whole, holder); | |
1138 | if (err == 0) { | |
1139 | whole->bd_claiming = holder; | |
1140 | spin_unlock(&bdev_lock); | |
1141 | return whole; | |
1142 | } else { | |
1143 | spin_unlock(&bdev_lock); | |
1144 | bdput(whole); | |
1145 | return ERR_PTR(err); | |
1146 | } | |
1147 | } | |
1148 | ||
641dc636 | 1149 | #ifdef CONFIG_SYSFS |
49731baa TH |
1150 | struct bd_holder_disk { |
1151 | struct list_head list; | |
1152 | struct gendisk *disk; | |
1153 | int refcnt; | |
1154 | }; | |
1155 | ||
1156 | static struct bd_holder_disk *bd_find_holder_disk(struct block_device *bdev, | |
1157 | struct gendisk *disk) | |
1158 | { | |
1159 | struct bd_holder_disk *holder; | |
1160 | ||
1161 | list_for_each_entry(holder, &bdev->bd_holder_disks, list) | |
1162 | if (holder->disk == disk) | |
1163 | return holder; | |
1164 | return NULL; | |
1165 | } | |
1166 | ||
4d7dd8fd | 1167 | static int add_symlink(struct kobject *from, struct kobject *to) |
641dc636 | 1168 | { |
4d7dd8fd | 1169 | return sysfs_create_link(from, to, kobject_name(to)); |
641dc636 JN |
1170 | } |
1171 | ||
1172 | static void del_symlink(struct kobject *from, struct kobject *to) | |
1173 | { | |
641dc636 JN |
1174 | sysfs_remove_link(from, kobject_name(to)); |
1175 | } | |
1176 | ||
df6c0cd9 | 1177 | /** |
e09b457b TH |
1178 | * bd_link_disk_holder - create symlinks between holding disk and slave bdev |
1179 | * @bdev: the claimed slave bdev | |
1180 | * @disk: the holding disk | |
df6c0cd9 | 1181 | * |
49731baa TH |
1182 | * DON'T USE THIS UNLESS YOU'RE ALREADY USING IT. |
1183 | * | |
e09b457b | 1184 | * This functions creates the following sysfs symlinks. |
641dc636 | 1185 | * |
e09b457b TH |
1186 | * - from "slaves" directory of the holder @disk to the claimed @bdev |
1187 | * - from "holders" directory of the @bdev to the holder @disk | |
641dc636 | 1188 | * |
e09b457b TH |
1189 | * For example, if /dev/dm-0 maps to /dev/sda and disk for dm-0 is |
1190 | * passed to bd_link_disk_holder(), then: | |
641dc636 | 1191 | * |
e09b457b TH |
1192 | * /sys/block/dm-0/slaves/sda --> /sys/block/sda |
1193 | * /sys/block/sda/holders/dm-0 --> /sys/block/dm-0 | |
641dc636 | 1194 | * |
e09b457b TH |
1195 | * The caller must have claimed @bdev before calling this function and |
1196 | * ensure that both @bdev and @disk are valid during the creation and | |
1197 | * lifetime of these symlinks. | |
641dc636 | 1198 | * |
e09b457b TH |
1199 | * CONTEXT: |
1200 | * Might sleep. | |
641dc636 | 1201 | * |
e09b457b TH |
1202 | * RETURNS: |
1203 | * 0 on success, -errno on failure. | |
641dc636 | 1204 | */ |
e09b457b | 1205 | int bd_link_disk_holder(struct block_device *bdev, struct gendisk *disk) |
641dc636 | 1206 | { |
49731baa | 1207 | struct bd_holder_disk *holder; |
e09b457b | 1208 | int ret = 0; |
641dc636 | 1209 | |
2e7b651d | 1210 | mutex_lock(&bdev->bd_mutex); |
df6c0cd9 | 1211 | |
49731baa | 1212 | WARN_ON_ONCE(!bdev->bd_holder); |
4e91672c | 1213 | |
e09b457b TH |
1214 | /* FIXME: remove the following once add_disk() handles errors */ |
1215 | if (WARN_ON(!disk->slave_dir || !bdev->bd_part->holder_dir)) | |
1216 | goto out_unlock; | |
4e91672c | 1217 | |
49731baa TH |
1218 | holder = bd_find_holder_disk(bdev, disk); |
1219 | if (holder) { | |
1220 | holder->refcnt++; | |
e09b457b | 1221 | goto out_unlock; |
49731baa | 1222 | } |
641dc636 | 1223 | |
49731baa TH |
1224 | holder = kzalloc(sizeof(*holder), GFP_KERNEL); |
1225 | if (!holder) { | |
1226 | ret = -ENOMEM; | |
e09b457b TH |
1227 | goto out_unlock; |
1228 | } | |
641dc636 | 1229 | |
49731baa TH |
1230 | INIT_LIST_HEAD(&holder->list); |
1231 | holder->disk = disk; | |
1232 | holder->refcnt = 1; | |
1233 | ||
1234 | ret = add_symlink(disk->slave_dir, &part_to_dev(bdev->bd_part)->kobj); | |
1235 | if (ret) | |
1236 | goto out_free; | |
1237 | ||
1238 | ret = add_symlink(bdev->bd_part->holder_dir, &disk_to_dev(disk)->kobj); | |
1239 | if (ret) | |
1240 | goto out_del; | |
e7407d16 TH |
1241 | /* |
1242 | * bdev could be deleted beneath us which would implicitly destroy | |
1243 | * the holder directory. Hold on to it. | |
1244 | */ | |
1245 | kobject_get(bdev->bd_part->holder_dir); | |
49731baa TH |
1246 | |
1247 | list_add(&holder->list, &bdev->bd_holder_disks); | |
1248 | goto out_unlock; | |
1249 | ||
1250 | out_del: | |
1251 | del_symlink(disk->slave_dir, &part_to_dev(bdev->bd_part)->kobj); | |
1252 | out_free: | |
1253 | kfree(holder); | |
e09b457b | 1254 | out_unlock: |
b4cf1b72 | 1255 | mutex_unlock(&bdev->bd_mutex); |
e09b457b | 1256 | return ret; |
641dc636 | 1257 | } |
e09b457b | 1258 | EXPORT_SYMBOL_GPL(bd_link_disk_holder); |
641dc636 | 1259 | |
49731baa TH |
1260 | /** |
1261 | * bd_unlink_disk_holder - destroy symlinks created by bd_link_disk_holder() | |
1262 | * @bdev: the calimed slave bdev | |
1263 | * @disk: the holding disk | |
1264 | * | |
1265 | * DON'T USE THIS UNLESS YOU'RE ALREADY USING IT. | |
1266 | * | |
1267 | * CONTEXT: | |
1268 | * Might sleep. | |
1269 | */ | |
1270 | void bd_unlink_disk_holder(struct block_device *bdev, struct gendisk *disk) | |
641dc636 | 1271 | { |
49731baa | 1272 | struct bd_holder_disk *holder; |
641dc636 | 1273 | |
49731baa | 1274 | mutex_lock(&bdev->bd_mutex); |
641dc636 | 1275 | |
49731baa TH |
1276 | holder = bd_find_holder_disk(bdev, disk); |
1277 | ||
1278 | if (!WARN_ON_ONCE(holder == NULL) && !--holder->refcnt) { | |
1279 | del_symlink(disk->slave_dir, &part_to_dev(bdev->bd_part)->kobj); | |
1280 | del_symlink(bdev->bd_part->holder_dir, | |
1281 | &disk_to_dev(disk)->kobj); | |
e7407d16 | 1282 | kobject_put(bdev->bd_part->holder_dir); |
49731baa TH |
1283 | list_del_init(&holder->list); |
1284 | kfree(holder); | |
1285 | } | |
1286 | ||
1287 | mutex_unlock(&bdev->bd_mutex); | |
1da177e4 | 1288 | } |
49731baa | 1289 | EXPORT_SYMBOL_GPL(bd_unlink_disk_holder); |
641dc636 | 1290 | #endif |
1da177e4 | 1291 | |
56ade44b AP |
1292 | /** |
1293 | * flush_disk - invalidates all buffer-cache entries on a disk | |
1294 | * | |
1295 | * @bdev: struct block device to be flushed | |
e6eb5ce1 | 1296 | * @kill_dirty: flag to guide handling of dirty inodes |
56ade44b AP |
1297 | * |
1298 | * Invalidates all buffer-cache entries on a disk. It should be called | |
1299 | * when a disk has been changed -- either by a media change or online | |
1300 | * resize. | |
1301 | */ | |
93b270f7 | 1302 | static void flush_disk(struct block_device *bdev, bool kill_dirty) |
56ade44b | 1303 | { |
93b270f7 | 1304 | if (__invalidate_device(bdev, kill_dirty)) { |
56ade44b | 1305 | printk(KERN_WARNING "VFS: busy inodes on changed media or " |
424081f3 DM |
1306 | "resized disk %s\n", |
1307 | bdev->bd_disk ? bdev->bd_disk->disk_name : ""); | |
56ade44b AP |
1308 | } |
1309 | ||
1310 | if (!bdev->bd_disk) | |
1311 | return; | |
d27769ec | 1312 | if (disk_part_scan_enabled(bdev->bd_disk)) |
56ade44b AP |
1313 | bdev->bd_invalidated = 1; |
1314 | } | |
1315 | ||
c3279d14 | 1316 | /** |
57d1b536 | 1317 | * check_disk_size_change - checks for disk size change and adjusts bdev size. |
c3279d14 AP |
1318 | * @disk: struct gendisk to check |
1319 | * @bdev: struct bdev to adjust. | |
1320 | * | |
1321 | * This routine checks to see if the bdev size does not match the disk size | |
1322 | * and adjusts it if it differs. | |
1323 | */ | |
1324 | void check_disk_size_change(struct gendisk *disk, struct block_device *bdev) | |
1325 | { | |
1326 | loff_t disk_size, bdev_size; | |
1327 | ||
1328 | disk_size = (loff_t)get_capacity(disk) << 9; | |
1329 | bdev_size = i_size_read(bdev->bd_inode); | |
1330 | if (disk_size != bdev_size) { | |
c3279d14 AP |
1331 | printk(KERN_INFO |
1332 | "%s: detected capacity change from %lld to %lld\n", | |
424081f3 | 1333 | disk->disk_name, bdev_size, disk_size); |
c3279d14 | 1334 | i_size_write(bdev->bd_inode, disk_size); |
93b270f7 | 1335 | flush_disk(bdev, false); |
c3279d14 AP |
1336 | } |
1337 | } | |
1338 | EXPORT_SYMBOL(check_disk_size_change); | |
1339 | ||
0c002c2f | 1340 | /** |
57d1b536 | 1341 | * revalidate_disk - wrapper for lower-level driver's revalidate_disk call-back |
0c002c2f AP |
1342 | * @disk: struct gendisk to be revalidated |
1343 | * | |
1344 | * This routine is a wrapper for lower-level driver's revalidate_disk | |
1345 | * call-backs. It is used to do common pre and post operations needed | |
1346 | * for all revalidate_disk operations. | |
1347 | */ | |
1348 | int revalidate_disk(struct gendisk *disk) | |
1349 | { | |
c3279d14 | 1350 | struct block_device *bdev; |
0c002c2f AP |
1351 | int ret = 0; |
1352 | ||
1353 | if (disk->fops->revalidate_disk) | |
1354 | ret = disk->fops->revalidate_disk(disk); | |
c3279d14 AP |
1355 | bdev = bdget_disk(disk, 0); |
1356 | if (!bdev) | |
1357 | return ret; | |
1358 | ||
1359 | mutex_lock(&bdev->bd_mutex); | |
1360 | check_disk_size_change(disk, bdev); | |
7630b661 | 1361 | bdev->bd_invalidated = 0; |
c3279d14 AP |
1362 | mutex_unlock(&bdev->bd_mutex); |
1363 | bdput(bdev); | |
0c002c2f AP |
1364 | return ret; |
1365 | } | |
1366 | EXPORT_SYMBOL(revalidate_disk); | |
1367 | ||
1da177e4 LT |
1368 | /* |
1369 | * This routine checks whether a removable media has been changed, | |
1370 | * and invalidates all buffer-cache-entries in that case. This | |
1371 | * is a relatively slow routine, so we have to try to minimize using | |
1372 | * it. Thus it is called only upon a 'mount' or 'open'. This | |
1373 | * is the best way of combining speed and utility, I think. | |
1374 | * People changing diskettes in the middle of an operation deserve | |
1375 | * to lose :-) | |
1376 | */ | |
1377 | int check_disk_change(struct block_device *bdev) | |
1378 | { | |
1379 | struct gendisk *disk = bdev->bd_disk; | |
83d5cde4 | 1380 | const struct block_device_operations *bdops = disk->fops; |
77ea887e | 1381 | unsigned int events; |
1da177e4 | 1382 | |
77ea887e TH |
1383 | events = disk_clear_events(disk, DISK_EVENT_MEDIA_CHANGE | |
1384 | DISK_EVENT_EJECT_REQUEST); | |
1385 | if (!(events & DISK_EVENT_MEDIA_CHANGE)) | |
1da177e4 LT |
1386 | return 0; |
1387 | ||
93b270f7 | 1388 | flush_disk(bdev, true); |
1da177e4 LT |
1389 | if (bdops->revalidate_disk) |
1390 | bdops->revalidate_disk(bdev->bd_disk); | |
1da177e4 LT |
1391 | return 1; |
1392 | } | |
1393 | ||
1394 | EXPORT_SYMBOL(check_disk_change); | |
1395 | ||
1396 | void bd_set_size(struct block_device *bdev, loff_t size) | |
1397 | { | |
5955102c | 1398 | inode_lock(bdev->bd_inode); |
d646a02a | 1399 | i_size_write(bdev->bd_inode, size); |
5955102c | 1400 | inode_unlock(bdev->bd_inode); |
1da177e4 LT |
1401 | } |
1402 | EXPORT_SYMBOL(bd_set_size); | |
1403 | ||
4385bab1 | 1404 | static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part); |
37be4124 | 1405 | |
6d740cd5 PZ |
1406 | /* |
1407 | * bd_mutex locking: | |
1408 | * | |
1409 | * mutex_lock(part->bd_mutex) | |
1410 | * mutex_lock_nested(whole->bd_mutex, 1) | |
1411 | */ | |
1412 | ||
572c4892 | 1413 | static int __blkdev_get(struct block_device *bdev, fmode_t mode, int for_part) |
1da177e4 | 1414 | { |
1da177e4 | 1415 | struct gendisk *disk; |
523e1d39 | 1416 | struct module *owner; |
7db9cfd3 | 1417 | int ret; |
cf771cb5 | 1418 | int partno; |
fe6e9c1f AV |
1419 | int perm = 0; |
1420 | ||
572c4892 | 1421 | if (mode & FMODE_READ) |
fe6e9c1f | 1422 | perm |= MAY_READ; |
572c4892 | 1423 | if (mode & FMODE_WRITE) |
fe6e9c1f AV |
1424 | perm |= MAY_WRITE; |
1425 | /* | |
1426 | * hooks: /n/, see "layering violations". | |
1427 | */ | |
b7300b78 CW |
1428 | if (!for_part) { |
1429 | ret = devcgroup_inode_permission(bdev->bd_inode, perm); | |
1430 | if (ret != 0) { | |
1431 | bdput(bdev); | |
1432 | return ret; | |
1433 | } | |
82666020 | 1434 | } |
7db9cfd3 | 1435 | |
d3374825 | 1436 | restart: |
0762b8bd | 1437 | |
89f97496 | 1438 | ret = -ENXIO; |
cf771cb5 | 1439 | disk = get_gendisk(bdev->bd_dev, &partno); |
0762b8bd | 1440 | if (!disk) |
6e9624b8 | 1441 | goto out; |
523e1d39 | 1442 | owner = disk->fops->owner; |
1da177e4 | 1443 | |
69e02c59 | 1444 | disk_block_events(disk); |
6796bf54 | 1445 | mutex_lock_nested(&bdev->bd_mutex, for_part); |
1da177e4 LT |
1446 | if (!bdev->bd_openers) { |
1447 | bdev->bd_disk = disk; | |
87192a2a | 1448 | bdev->bd_queue = disk->queue; |
1da177e4 | 1449 | bdev->bd_contains = bdev; |
c2ee070f | 1450 | bdev->bd_partno = partno; |
03cdadb0 | 1451 | |
cf771cb5 | 1452 | if (!partno) { |
89f97496 TH |
1453 | ret = -ENXIO; |
1454 | bdev->bd_part = disk_get_part(disk, partno); | |
1455 | if (!bdev->bd_part) | |
1456 | goto out_clear; | |
1457 | ||
1196f8b8 | 1458 | ret = 0; |
1da177e4 | 1459 | if (disk->fops->open) { |
572c4892 | 1460 | ret = disk->fops->open(bdev, mode); |
d3374825 N |
1461 | if (ret == -ERESTARTSYS) { |
1462 | /* Lost a race with 'disk' being | |
1463 | * deleted, try again. | |
1464 | * See md.c | |
1465 | */ | |
1466 | disk_put_part(bdev->bd_part); | |
1467 | bdev->bd_part = NULL; | |
d3374825 | 1468 | bdev->bd_disk = NULL; |
87192a2a | 1469 | bdev->bd_queue = NULL; |
d3374825 | 1470 | mutex_unlock(&bdev->bd_mutex); |
69e02c59 | 1471 | disk_unblock_events(disk); |
69e02c59 | 1472 | put_disk(disk); |
523e1d39 | 1473 | module_put(owner); |
d3374825 N |
1474 | goto restart; |
1475 | } | |
1da177e4 | 1476 | } |
7e69723f | 1477 | |
c30974a1 | 1478 | if (!ret) { |
7e69723f | 1479 | bd_set_size(bdev,(loff_t)get_capacity(disk)<<9); |
c30974a1 JK |
1480 | set_init_blocksize(bdev); |
1481 | } | |
7e69723f | 1482 | |
1196f8b8 TH |
1483 | /* |
1484 | * If the device is invalidated, rescan partition | |
1485 | * if open succeeded or failed with -ENOMEDIUM. | |
1486 | * The latter is necessary to prevent ghost | |
1487 | * partitions on a removed medium. | |
1488 | */ | |
fe316bf2 JN |
1489 | if (bdev->bd_invalidated) { |
1490 | if (!ret) | |
1491 | rescan_partitions(disk, bdev); | |
1492 | else if (ret == -ENOMEDIUM) | |
1493 | invalidate_partitions(disk, bdev); | |
1494 | } | |
5a023cdb | 1495 | |
1196f8b8 TH |
1496 | if (ret) |
1497 | goto out_clear; | |
1da177e4 | 1498 | } else { |
1da177e4 LT |
1499 | struct block_device *whole; |
1500 | whole = bdget_disk(disk, 0); | |
1501 | ret = -ENOMEM; | |
1502 | if (!whole) | |
0762b8bd | 1503 | goto out_clear; |
37be4124 | 1504 | BUG_ON(for_part); |
572c4892 | 1505 | ret = __blkdev_get(whole, mode, 1); |
1da177e4 | 1506 | if (ret) |
0762b8bd | 1507 | goto out_clear; |
1da177e4 | 1508 | bdev->bd_contains = whole; |
89f97496 | 1509 | bdev->bd_part = disk_get_part(disk, partno); |
e71bf0d0 | 1510 | if (!(disk->flags & GENHD_FL_UP) || |
89f97496 | 1511 | !bdev->bd_part || !bdev->bd_part->nr_sects) { |
1da177e4 | 1512 | ret = -ENXIO; |
0762b8bd | 1513 | goto out_clear; |
1da177e4 | 1514 | } |
89f97496 | 1515 | bd_set_size(bdev, (loff_t)bdev->bd_part->nr_sects << 9); |
c30974a1 | 1516 | set_init_blocksize(bdev); |
1da177e4 | 1517 | } |
03e26279 JK |
1518 | |
1519 | if (bdev->bd_bdi == &noop_backing_dev_info) | |
1520 | bdev->bd_bdi = bdi_get(disk->queue->backing_dev_info); | |
1da177e4 | 1521 | } else { |
1da177e4 | 1522 | if (bdev->bd_contains == bdev) { |
1196f8b8 TH |
1523 | ret = 0; |
1524 | if (bdev->bd_disk->fops->open) | |
572c4892 | 1525 | ret = bdev->bd_disk->fops->open(bdev, mode); |
1196f8b8 | 1526 | /* the same as first opener case, read comment there */ |
fe316bf2 JN |
1527 | if (bdev->bd_invalidated) { |
1528 | if (!ret) | |
1529 | rescan_partitions(bdev->bd_disk, bdev); | |
1530 | else if (ret == -ENOMEDIUM) | |
1531 | invalidate_partitions(bdev->bd_disk, bdev); | |
1532 | } | |
1196f8b8 TH |
1533 | if (ret) |
1534 | goto out_unlock_bdev; | |
1da177e4 | 1535 | } |
69e02c59 | 1536 | /* only one opener holds refs to the module and disk */ |
69e02c59 | 1537 | put_disk(disk); |
523e1d39 | 1538 | module_put(owner); |
1da177e4 LT |
1539 | } |
1540 | bdev->bd_openers++; | |
37be4124 N |
1541 | if (for_part) |
1542 | bdev->bd_part_count++; | |
c039e313 | 1543 | mutex_unlock(&bdev->bd_mutex); |
69e02c59 | 1544 | disk_unblock_events(disk); |
1da177e4 LT |
1545 | return 0; |
1546 | ||
0762b8bd | 1547 | out_clear: |
89f97496 | 1548 | disk_put_part(bdev->bd_part); |
1da177e4 | 1549 | bdev->bd_disk = NULL; |
0762b8bd | 1550 | bdev->bd_part = NULL; |
87192a2a | 1551 | bdev->bd_queue = NULL; |
1da177e4 | 1552 | if (bdev != bdev->bd_contains) |
572c4892 | 1553 | __blkdev_put(bdev->bd_contains, mode, 1); |
1da177e4 | 1554 | bdev->bd_contains = NULL; |
0762b8bd | 1555 | out_unlock_bdev: |
c039e313 | 1556 | mutex_unlock(&bdev->bd_mutex); |
69e02c59 | 1557 | disk_unblock_events(disk); |
0762b8bd | 1558 | put_disk(disk); |
523e1d39 | 1559 | module_put(owner); |
4345caba | 1560 | out: |
0762b8bd TH |
1561 | bdput(bdev); |
1562 | ||
1da177e4 LT |
1563 | return ret; |
1564 | } | |
1565 | ||
d4d77629 TH |
1566 | /** |
1567 | * blkdev_get - open a block device | |
1568 | * @bdev: block_device to open | |
1569 | * @mode: FMODE_* mask | |
1570 | * @holder: exclusive holder identifier | |
1571 | * | |
1572 | * Open @bdev with @mode. If @mode includes %FMODE_EXCL, @bdev is | |
1573 | * open with exclusive access. Specifying %FMODE_EXCL with %NULL | |
1574 | * @holder is invalid. Exclusive opens may nest for the same @holder. | |
1575 | * | |
1576 | * On success, the reference count of @bdev is unchanged. On failure, | |
1577 | * @bdev is put. | |
1578 | * | |
1579 | * CONTEXT: | |
1580 | * Might sleep. | |
1581 | * | |
1582 | * RETURNS: | |
1583 | * 0 on success, -errno on failure. | |
1584 | */ | |
e525fd89 | 1585 | int blkdev_get(struct block_device *bdev, fmode_t mode, void *holder) |
1da177e4 | 1586 | { |
e525fd89 TH |
1587 | struct block_device *whole = NULL; |
1588 | int res; | |
1589 | ||
1590 | WARN_ON_ONCE((mode & FMODE_EXCL) && !holder); | |
1591 | ||
1592 | if ((mode & FMODE_EXCL) && holder) { | |
1593 | whole = bd_start_claiming(bdev, holder); | |
1594 | if (IS_ERR(whole)) { | |
1595 | bdput(bdev); | |
1596 | return PTR_ERR(whole); | |
1597 | } | |
1598 | } | |
1599 | ||
1600 | res = __blkdev_get(bdev, mode, 0); | |
1601 | ||
1602 | if (whole) { | |
d4dc210f TH |
1603 | struct gendisk *disk = whole->bd_disk; |
1604 | ||
6a027eff | 1605 | /* finish claiming */ |
77ea887e | 1606 | mutex_lock(&bdev->bd_mutex); |
6a027eff TH |
1607 | spin_lock(&bdev_lock); |
1608 | ||
77ea887e | 1609 | if (!res) { |
6a027eff TH |
1610 | BUG_ON(!bd_may_claim(bdev, whole, holder)); |
1611 | /* | |
1612 | * Note that for a whole device bd_holders | |
1613 | * will be incremented twice, and bd_holder | |
1614 | * will be set to bd_may_claim before being | |
1615 | * set to holder | |
1616 | */ | |
1617 | whole->bd_holders++; | |
1618 | whole->bd_holder = bd_may_claim; | |
1619 | bdev->bd_holders++; | |
1620 | bdev->bd_holder = holder; | |
1621 | } | |
1622 | ||
1623 | /* tell others that we're done */ | |
1624 | BUG_ON(whole->bd_claiming != holder); | |
1625 | whole->bd_claiming = NULL; | |
1626 | wake_up_bit(&whole->bd_claiming, 0); | |
1627 | ||
1628 | spin_unlock(&bdev_lock); | |
77ea887e TH |
1629 | |
1630 | /* | |
d4dc210f TH |
1631 | * Block event polling for write claims if requested. Any |
1632 | * write holder makes the write_holder state stick until | |
1633 | * all are released. This is good enough and tracking | |
1634 | * individual writeable reference is too fragile given the | |
1635 | * way @mode is used in blkdev_get/put(). | |
77ea887e | 1636 | */ |
4c49ff3f TH |
1637 | if (!res && (mode & FMODE_WRITE) && !bdev->bd_write_holder && |
1638 | (disk->flags & GENHD_FL_BLOCK_EVENTS_ON_EXCL_WRITE)) { | |
77ea887e | 1639 | bdev->bd_write_holder = true; |
d4dc210f | 1640 | disk_block_events(disk); |
77ea887e TH |
1641 | } |
1642 | ||
1643 | mutex_unlock(&bdev->bd_mutex); | |
6a027eff | 1644 | bdput(whole); |
e525fd89 TH |
1645 | } |
1646 | ||
1647 | return res; | |
37be4124 | 1648 | } |
1da177e4 LT |
1649 | EXPORT_SYMBOL(blkdev_get); |
1650 | ||
d4d77629 TH |
1651 | /** |
1652 | * blkdev_get_by_path - open a block device by name | |
1653 | * @path: path to the block device to open | |
1654 | * @mode: FMODE_* mask | |
1655 | * @holder: exclusive holder identifier | |
1656 | * | |
1657 | * Open the blockdevice described by the device file at @path. @mode | |
1658 | * and @holder are identical to blkdev_get(). | |
1659 | * | |
1660 | * On success, the returned block_device has reference count of one. | |
1661 | * | |
1662 | * CONTEXT: | |
1663 | * Might sleep. | |
1664 | * | |
1665 | * RETURNS: | |
1666 | * Pointer to block_device on success, ERR_PTR(-errno) on failure. | |
1667 | */ | |
1668 | struct block_device *blkdev_get_by_path(const char *path, fmode_t mode, | |
1669 | void *holder) | |
1670 | { | |
1671 | struct block_device *bdev; | |
b77dabf4 | 1672 | int perm = 0; |
d4d77629 TH |
1673 | int err; |
1674 | ||
b77dabf4 SF |
1675 | if (mode & FMODE_READ) |
1676 | perm |= MAY_READ; | |
1677 | if (mode & FMODE_WRITE) | |
1678 | perm |= MAY_WRITE; | |
1679 | bdev = lookup_bdev(path, perm); | |
d4d77629 TH |
1680 | if (IS_ERR(bdev)) |
1681 | return bdev; | |
1682 | ||
1683 | err = blkdev_get(bdev, mode, holder); | |
1684 | if (err) | |
1685 | return ERR_PTR(err); | |
1686 | ||
e51900f7 CE |
1687 | if ((mode & FMODE_WRITE) && bdev_read_only(bdev)) { |
1688 | blkdev_put(bdev, mode); | |
1689 | return ERR_PTR(-EACCES); | |
1690 | } | |
1691 | ||
d4d77629 TH |
1692 | return bdev; |
1693 | } | |
1694 | EXPORT_SYMBOL(blkdev_get_by_path); | |
1695 | ||
1696 | /** | |
1697 | * blkdev_get_by_dev - open a block device by device number | |
1698 | * @dev: device number of block device to open | |
1699 | * @mode: FMODE_* mask | |
1700 | * @holder: exclusive holder identifier | |
1701 | * | |
1702 | * Open the blockdevice described by device number @dev. @mode and | |
1703 | * @holder are identical to blkdev_get(). | |
1704 | * | |
1705 | * Use it ONLY if you really do not have anything better - i.e. when | |
1706 | * you are behind a truly sucky interface and all you are given is a | |
1707 | * device number. _Never_ to be used for internal purposes. If you | |
1708 | * ever need it - reconsider your API. | |
1709 | * | |
1710 | * On success, the returned block_device has reference count of one. | |
1711 | * | |
1712 | * CONTEXT: | |
1713 | * Might sleep. | |
1714 | * | |
1715 | * RETURNS: | |
1716 | * Pointer to block_device on success, ERR_PTR(-errno) on failure. | |
1717 | */ | |
1718 | struct block_device *blkdev_get_by_dev(dev_t dev, fmode_t mode, void *holder) | |
1719 | { | |
1720 | struct block_device *bdev; | |
1721 | int err; | |
1722 | ||
1723 | bdev = bdget(dev); | |
1724 | if (!bdev) | |
1725 | return ERR_PTR(-ENOMEM); | |
1726 | ||
1727 | err = blkdev_get(bdev, mode, holder); | |
1728 | if (err) | |
1729 | return ERR_PTR(err); | |
1730 | ||
1731 | return bdev; | |
1732 | } | |
1733 | EXPORT_SYMBOL(blkdev_get_by_dev); | |
1734 | ||
1da177e4 LT |
1735 | static int blkdev_open(struct inode * inode, struct file * filp) |
1736 | { | |
1737 | struct block_device *bdev; | |
1da177e4 LT |
1738 | |
1739 | /* | |
1740 | * Preserve backwards compatibility and allow large file access | |
1741 | * even if userspace doesn't ask for it explicitly. Some mkfs | |
1742 | * binary needs it. We might want to drop this workaround | |
1743 | * during an unstable branch. | |
1744 | */ | |
1745 | filp->f_flags |= O_LARGEFILE; | |
1746 | ||
c35fc7a5 CH |
1747 | filp->f_mode |= FMODE_NOWAIT; |
1748 | ||
572c4892 AV |
1749 | if (filp->f_flags & O_NDELAY) |
1750 | filp->f_mode |= FMODE_NDELAY; | |
1751 | if (filp->f_flags & O_EXCL) | |
1752 | filp->f_mode |= FMODE_EXCL; | |
1753 | if ((filp->f_flags & O_ACCMODE) == 3) | |
1754 | filp->f_mode |= FMODE_WRITE_IOCTL; | |
1755 | ||
1da177e4 | 1756 | bdev = bd_acquire(inode); |
6a2aae06 PE |
1757 | if (bdev == NULL) |
1758 | return -ENOMEM; | |
1da177e4 | 1759 | |
dbf2d0a4 SF |
1760 | /* |
1761 | * A negative i_writecount for bdev->bd_inode means that the bdev | |
1762 | * or one of its paritions is mounted in a user namespace. Deny | |
1763 | * writing for non-root in this case, otherwise an unprivileged | |
1764 | * user can attack the kernel by modifying the backing store of a | |
1765 | * mounted filesystem. | |
1766 | */ | |
1767 | if ((filp->f_mode & FMODE_WRITE) && | |
1768 | !file_ns_capable(filp, &init_user_ns, CAP_SYS_ADMIN) && | |
1769 | !atomic_inc_unless_negative(&bdev->bd_inode->i_writecount)) { | |
1770 | bdput(bdev); | |
1771 | return -EBUSY; | |
1772 | } | |
1773 | ||
572c4892 | 1774 | filp->f_mapping = bdev->bd_inode->i_mapping; |
5660e13d | 1775 | filp->f_wb_err = filemap_sample_wb_err(filp->f_mapping); |
572c4892 | 1776 | |
e525fd89 | 1777 | return blkdev_get(bdev, filp->f_mode, filp); |
1da177e4 LT |
1778 | } |
1779 | ||
4385bab1 | 1780 | static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part) |
2e7b651d | 1781 | { |
2e7b651d | 1782 | struct gendisk *disk = bdev->bd_disk; |
37be4124 | 1783 | struct block_device *victim = NULL; |
2e7b651d | 1784 | |
6796bf54 | 1785 | mutex_lock_nested(&bdev->bd_mutex, for_part); |
37be4124 N |
1786 | if (for_part) |
1787 | bdev->bd_part_count--; | |
1788 | ||
2e7b651d | 1789 | if (!--bdev->bd_openers) { |
6a027eff | 1790 | WARN_ON_ONCE(bdev->bd_holders); |
2e7b651d PZ |
1791 | sync_blockdev(bdev); |
1792 | kill_bdev(bdev); | |
43d1c0eb ID |
1793 | |
1794 | bdev_write_inode(bdev); | |
2e7b651d PZ |
1795 | } |
1796 | if (bdev->bd_contains == bdev) { | |
1797 | if (disk->fops->release) | |
db2a144b | 1798 | disk->fops->release(disk, mode); |
2e7b651d PZ |
1799 | } |
1800 | if (!bdev->bd_openers) { | |
1801 | struct module *owner = disk->fops->owner; | |
1802 | ||
0762b8bd TH |
1803 | disk_put_part(bdev->bd_part); |
1804 | bdev->bd_part = NULL; | |
2e7b651d | 1805 | bdev->bd_disk = NULL; |
37be4124 N |
1806 | if (bdev != bdev->bd_contains) |
1807 | victim = bdev->bd_contains; | |
2e7b651d | 1808 | bdev->bd_contains = NULL; |
523e1d39 TH |
1809 | |
1810 | put_disk(disk); | |
1811 | module_put(owner); | |
2e7b651d | 1812 | } |
2e7b651d PZ |
1813 | mutex_unlock(&bdev->bd_mutex); |
1814 | bdput(bdev); | |
37be4124 | 1815 | if (victim) |
9a1c3542 | 1816 | __blkdev_put(victim, mode, 1); |
2e7b651d PZ |
1817 | } |
1818 | ||
4385bab1 | 1819 | void blkdev_put(struct block_device *bdev, fmode_t mode) |
37be4124 | 1820 | { |
85ef06d1 TH |
1821 | mutex_lock(&bdev->bd_mutex); |
1822 | ||
e525fd89 | 1823 | if (mode & FMODE_EXCL) { |
6a027eff TH |
1824 | bool bdev_free; |
1825 | ||
1826 | /* | |
1827 | * Release a claim on the device. The holder fields | |
1828 | * are protected with bdev_lock. bd_mutex is to | |
1829 | * synchronize disk_holder unlinking. | |
1830 | */ | |
6a027eff TH |
1831 | spin_lock(&bdev_lock); |
1832 | ||
1833 | WARN_ON_ONCE(--bdev->bd_holders < 0); | |
1834 | WARN_ON_ONCE(--bdev->bd_contains->bd_holders < 0); | |
1835 | ||
1836 | /* bd_contains might point to self, check in a separate step */ | |
1837 | if ((bdev_free = !bdev->bd_holders)) | |
1838 | bdev->bd_holder = NULL; | |
1839 | if (!bdev->bd_contains->bd_holders) | |
1840 | bdev->bd_contains->bd_holder = NULL; | |
1841 | ||
1842 | spin_unlock(&bdev_lock); | |
1843 | ||
77ea887e TH |
1844 | /* |
1845 | * If this was the last claim, remove holder link and | |
1846 | * unblock evpoll if it was a write holder. | |
1847 | */ | |
85ef06d1 TH |
1848 | if (bdev_free && bdev->bd_write_holder) { |
1849 | disk_unblock_events(bdev->bd_disk); | |
1850 | bdev->bd_write_holder = false; | |
77ea887e | 1851 | } |
6936217c | 1852 | } |
77ea887e | 1853 | |
85ef06d1 TH |
1854 | /* |
1855 | * Trigger event checking and tell drivers to flush MEDIA_CHANGE | |
1856 | * event. This is to ensure detection of media removal commanded | |
1857 | * from userland - e.g. eject(1). | |
1858 | */ | |
1859 | disk_flush_events(bdev->bd_disk, DISK_EVENT_MEDIA_CHANGE); | |
1860 | ||
1861 | mutex_unlock(&bdev->bd_mutex); | |
1862 | ||
4385bab1 | 1863 | __blkdev_put(bdev, mode, 0); |
37be4124 | 1864 | } |
2e7b651d PZ |
1865 | EXPORT_SYMBOL(blkdev_put); |
1866 | ||
1da177e4 LT |
1867 | static int blkdev_close(struct inode * inode, struct file * filp) |
1868 | { | |
4ebb16ca | 1869 | struct block_device *bdev = I_BDEV(bdev_file_inode(filp)); |
dbf2d0a4 SF |
1870 | if (filp->f_mode & FMODE_WRITE && |
1871 | !file_ns_capable(filp, &init_user_ns, CAP_SYS_ADMIN)) | |
1872 | atomic_dec(&bdev->bd_inode->i_writecount); | |
4385bab1 AV |
1873 | blkdev_put(bdev, filp->f_mode); |
1874 | return 0; | |
1da177e4 LT |
1875 | } |
1876 | ||
bb93e3a5 | 1877 | static long block_ioctl(struct file *file, unsigned cmd, unsigned long arg) |
1da177e4 | 1878 | { |
4ebb16ca | 1879 | struct block_device *bdev = I_BDEV(bdev_file_inode(file)); |
56b26add | 1880 | fmode_t mode = file->f_mode; |
fd4ce1ac CH |
1881 | |
1882 | /* | |
1883 | * O_NDELAY can be altered using fcntl(.., F_SETFL, ..), so we have | |
1884 | * to updated it before every ioctl. | |
1885 | */ | |
56b26add | 1886 | if (file->f_flags & O_NDELAY) |
fd4ce1ac CH |
1887 | mode |= FMODE_NDELAY; |
1888 | else | |
1889 | mode &= ~FMODE_NDELAY; | |
1890 | ||
56b26add | 1891 | return blkdev_ioctl(bdev, mode, cmd, arg); |
1da177e4 LT |
1892 | } |
1893 | ||
eef99380 CH |
1894 | /* |
1895 | * Write data to the block device. Only intended for the block device itself | |
1896 | * and the raw driver which basically is a fake block device. | |
1897 | * | |
1898 | * Does not take i_mutex for the write and thus is not for general purpose | |
1899 | * use. | |
1900 | */ | |
1456c0a8 | 1901 | ssize_t blkdev_write_iter(struct kiocb *iocb, struct iov_iter *from) |
eef99380 CH |
1902 | { |
1903 | struct file *file = iocb->ki_filp; | |
4ebb16ca | 1904 | struct inode *bd_inode = bdev_file_inode(file); |
7ec7b94a | 1905 | loff_t size = i_size_read(bd_inode); |
53362a05 | 1906 | struct blk_plug plug; |
eef99380 | 1907 | ssize_t ret; |
5f380c7f | 1908 | |
7ec7b94a AV |
1909 | if (bdev_read_only(I_BDEV(bd_inode))) |
1910 | return -EPERM; | |
5f380c7f | 1911 | |
7ec7b94a | 1912 | if (!iov_iter_count(from)) |
5f380c7f AV |
1913 | return 0; |
1914 | ||
7ec7b94a AV |
1915 | if (iocb->ki_pos >= size) |
1916 | return -ENOSPC; | |
1917 | ||
c35fc7a5 CH |
1918 | if ((iocb->ki_flags & (IOCB_NOWAIT | IOCB_DIRECT)) == IOCB_NOWAIT) |
1919 | return -EOPNOTSUPP; | |
1920 | ||
7ec7b94a | 1921 | iov_iter_truncate(from, size - iocb->ki_pos); |
eef99380 | 1922 | |
53362a05 | 1923 | blk_start_plug(&plug); |
1456c0a8 | 1924 | ret = __generic_file_write_iter(iocb, from); |
e2592217 CH |
1925 | if (ret > 0) |
1926 | ret = generic_write_sync(iocb, ret); | |
53362a05 | 1927 | blk_finish_plug(&plug); |
eef99380 CH |
1928 | return ret; |
1929 | } | |
1456c0a8 | 1930 | EXPORT_SYMBOL_GPL(blkdev_write_iter); |
eef99380 | 1931 | |
b2de525f | 1932 | ssize_t blkdev_read_iter(struct kiocb *iocb, struct iov_iter *to) |
684c9aae LT |
1933 | { |
1934 | struct file *file = iocb->ki_filp; | |
4ebb16ca | 1935 | struct inode *bd_inode = bdev_file_inode(file); |
684c9aae | 1936 | loff_t size = i_size_read(bd_inode); |
a886038b | 1937 | loff_t pos = iocb->ki_pos; |
684c9aae LT |
1938 | |
1939 | if (pos >= size) | |
1940 | return 0; | |
1941 | ||
1942 | size -= pos; | |
a886038b AV |
1943 | iov_iter_truncate(to, size); |
1944 | return generic_file_read_iter(iocb, to); | |
684c9aae | 1945 | } |
b2de525f | 1946 | EXPORT_SYMBOL_GPL(blkdev_read_iter); |
684c9aae | 1947 | |
87d8fe1e TT |
1948 | /* |
1949 | * Try to release a page associated with block device when the system | |
1950 | * is under memory pressure. | |
1951 | */ | |
1952 | static int blkdev_releasepage(struct page *page, gfp_t wait) | |
1953 | { | |
1954 | struct super_block *super = BDEV_I(page->mapping->host)->bdev.bd_super; | |
1955 | ||
1956 | if (super && super->s_op->bdev_try_to_free_page) | |
1957 | return super->s_op->bdev_try_to_free_page(super, page, wait); | |
1958 | ||
1959 | return try_to_free_buffers(page); | |
1960 | } | |
1961 | ||
7f6d5b52 RZ |
1962 | static int blkdev_writepages(struct address_space *mapping, |
1963 | struct writeback_control *wbc) | |
1964 | { | |
1965 | if (dax_mapping(mapping)) { | |
1966 | struct block_device *bdev = I_BDEV(mapping->host); | |
1967 | ||
1968 | return dax_writeback_mapping_range(mapping, bdev, wbc); | |
1969 | } | |
1970 | return generic_writepages(mapping, wbc); | |
1971 | } | |
1972 | ||
4c54ac62 | 1973 | static const struct address_space_operations def_blk_aops = { |
1da177e4 | 1974 | .readpage = blkdev_readpage, |
447f05bb | 1975 | .readpages = blkdev_readpages, |
1da177e4 | 1976 | .writepage = blkdev_writepage, |
6272b5a5 NP |
1977 | .write_begin = blkdev_write_begin, |
1978 | .write_end = blkdev_write_end, | |
7f6d5b52 | 1979 | .writepages = blkdev_writepages, |
87d8fe1e | 1980 | .releasepage = blkdev_releasepage, |
1da177e4 | 1981 | .direct_IO = blkdev_direct_IO, |
b4597226 | 1982 | .is_dirty_writeback = buffer_check_dirty_writeback, |
1da177e4 LT |
1983 | }; |
1984 | ||
25f4c414 DW |
1985 | #define BLKDEV_FALLOC_FL_SUPPORTED \ |
1986 | (FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | \ | |
1987 | FALLOC_FL_ZERO_RANGE | FALLOC_FL_NO_HIDE_STALE) | |
1988 | ||
1989 | static long blkdev_fallocate(struct file *file, int mode, loff_t start, | |
1990 | loff_t len) | |
1991 | { | |
1992 | struct block_device *bdev = I_BDEV(bdev_file_inode(file)); | |
25f4c414 DW |
1993 | struct address_space *mapping; |
1994 | loff_t end = start + len - 1; | |
1995 | loff_t isize; | |
1996 | int error; | |
1997 | ||
1998 | /* Fail if we don't recognize the flags. */ | |
1999 | if (mode & ~BLKDEV_FALLOC_FL_SUPPORTED) | |
2000 | return -EOPNOTSUPP; | |
2001 | ||
2002 | /* Don't go off the end of the device. */ | |
2003 | isize = i_size_read(bdev->bd_inode); | |
2004 | if (start >= isize) | |
2005 | return -EINVAL; | |
2006 | if (end >= isize) { | |
2007 | if (mode & FALLOC_FL_KEEP_SIZE) { | |
2008 | len = isize - start; | |
2009 | end = start + len - 1; | |
2010 | } else | |
2011 | return -EINVAL; | |
2012 | } | |
2013 | ||
2014 | /* | |
2015 | * Don't allow IO that isn't aligned to logical block size. | |
2016 | */ | |
2017 | if ((start | len) & (bdev_logical_block_size(bdev) - 1)) | |
2018 | return -EINVAL; | |
2019 | ||
2020 | /* Invalidate the page cache, including dirty pages. */ | |
2021 | mapping = bdev->bd_inode->i_mapping; | |
2022 | truncate_inode_pages_range(mapping, start, end); | |
2023 | ||
2024 | switch (mode) { | |
2025 | case FALLOC_FL_ZERO_RANGE: | |
2026 | case FALLOC_FL_ZERO_RANGE | FALLOC_FL_KEEP_SIZE: | |
2027 | error = blkdev_issue_zeroout(bdev, start >> 9, len >> 9, | |
ee472d83 | 2028 | GFP_KERNEL, BLKDEV_ZERO_NOUNMAP); |
25f4c414 DW |
2029 | break; |
2030 | case FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE: | |
34045129 CH |
2031 | error = blkdev_issue_zeroout(bdev, start >> 9, len >> 9, |
2032 | GFP_KERNEL, BLKDEV_ZERO_NOFALLBACK); | |
25f4c414 DW |
2033 | break; |
2034 | case FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE | FALLOC_FL_NO_HIDE_STALE: | |
25f4c414 DW |
2035 | error = blkdev_issue_discard(bdev, start >> 9, len >> 9, |
2036 | GFP_KERNEL, 0); | |
2037 | break; | |
2038 | default: | |
2039 | return -EOPNOTSUPP; | |
2040 | } | |
2041 | if (error) | |
2042 | return error; | |
2043 | ||
2044 | /* | |
2045 | * Invalidate again; if someone wandered in and dirtied a page, | |
2046 | * the caller will be given -EBUSY. The third argument is | |
2047 | * inclusive, so the rounding here is safe. | |
2048 | */ | |
2049 | return invalidate_inode_pages2_range(mapping, | |
2050 | start >> PAGE_SHIFT, | |
2051 | end >> PAGE_SHIFT); | |
2052 | } | |
2053 | ||
4b6f5d20 | 2054 | const struct file_operations def_blk_fops = { |
1da177e4 LT |
2055 | .open = blkdev_open, |
2056 | .release = blkdev_close, | |
2057 | .llseek = block_llseek, | |
a886038b | 2058 | .read_iter = blkdev_read_iter, |
1456c0a8 | 2059 | .write_iter = blkdev_write_iter, |
acc93d30 | 2060 | .mmap = generic_file_mmap, |
b1dd3b28 | 2061 | .fsync = blkdev_fsync, |
bb93e3a5 | 2062 | .unlocked_ioctl = block_ioctl, |
1da177e4 LT |
2063 | #ifdef CONFIG_COMPAT |
2064 | .compat_ioctl = compat_blkdev_ioctl, | |
2065 | #endif | |
1e8b3332 | 2066 | .splice_read = generic_file_splice_read, |
8d020765 | 2067 | .splice_write = iter_file_splice_write, |
25f4c414 | 2068 | .fallocate = blkdev_fallocate, |
1da177e4 LT |
2069 | }; |
2070 | ||
2071 | int ioctl_by_bdev(struct block_device *bdev, unsigned cmd, unsigned long arg) | |
2072 | { | |
2073 | int res; | |
2074 | mm_segment_t old_fs = get_fs(); | |
2075 | set_fs(KERNEL_DS); | |
56b26add | 2076 | res = blkdev_ioctl(bdev, 0, cmd, arg); |
1da177e4 LT |
2077 | set_fs(old_fs); |
2078 | return res; | |
2079 | } | |
2080 | ||
2081 | EXPORT_SYMBOL(ioctl_by_bdev); | |
2082 | ||
2083 | /** | |
2084 | * lookup_bdev - lookup a struct block_device by name | |
94e2959e | 2085 | * @pathname: special file representing the block device |
6bdf7d68 | 2086 | * @mask: rights to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) |
1da177e4 | 2087 | * |
57d1b536 | 2088 | * Get a reference to the blockdevice at @pathname in the current |
1da177e4 | 2089 | * namespace if possible and return it. Return ERR_PTR(error) |
6bdf7d68 SF |
2090 | * otherwise. If @mask is non-zero, check for access rights to the |
2091 | * inode at @pathname. | |
1da177e4 | 2092 | */ |
6bdf7d68 | 2093 | struct block_device *lookup_bdev(const char *pathname, int mask) |
1da177e4 LT |
2094 | { |
2095 | struct block_device *bdev; | |
2096 | struct inode *inode; | |
421748ec | 2097 | struct path path; |
1da177e4 LT |
2098 | int error; |
2099 | ||
421748ec | 2100 | if (!pathname || !*pathname) |
1da177e4 LT |
2101 | return ERR_PTR(-EINVAL); |
2102 | ||
421748ec | 2103 | error = kern_path(pathname, LOOKUP_FOLLOW, &path); |
1da177e4 LT |
2104 | if (error) |
2105 | return ERR_PTR(error); | |
2106 | ||
bb668734 | 2107 | inode = d_backing_inode(path.dentry); |
6bdf7d68 SF |
2108 | if (mask != 0 && !capable(CAP_SYS_ADMIN)) { |
2109 | error = __inode_permission(inode, mask); | |
2110 | if (error) | |
2111 | goto fail; | |
2112 | } | |
1da177e4 LT |
2113 | error = -ENOTBLK; |
2114 | if (!S_ISBLK(inode->i_mode)) | |
2115 | goto fail; | |
2116 | error = -EACCES; | |
a2982cc9 | 2117 | if (!may_open_dev(&path)) |
1da177e4 LT |
2118 | goto fail; |
2119 | error = -ENOMEM; | |
2120 | bdev = bd_acquire(inode); | |
2121 | if (!bdev) | |
2122 | goto fail; | |
2123 | out: | |
421748ec | 2124 | path_put(&path); |
1da177e4 LT |
2125 | return bdev; |
2126 | fail: | |
2127 | bdev = ERR_PTR(error); | |
2128 | goto out; | |
2129 | } | |
d5686b44 | 2130 | EXPORT_SYMBOL(lookup_bdev); |
1da177e4 | 2131 | |
93b270f7 | 2132 | int __invalidate_device(struct block_device *bdev, bool kill_dirty) |
b71e8a4c DH |
2133 | { |
2134 | struct super_block *sb = get_super(bdev); | |
2135 | int res = 0; | |
2136 | ||
2137 | if (sb) { | |
2138 | /* | |
2139 | * no need to lock the super, get_super holds the | |
2140 | * read mutex so the filesystem cannot go away | |
2141 | * under us (->put_super runs with the write lock | |
2142 | * hold). | |
2143 | */ | |
2144 | shrink_dcache_sb(sb); | |
93b270f7 | 2145 | res = invalidate_inodes(sb, kill_dirty); |
b71e8a4c DH |
2146 | drop_super(sb); |
2147 | } | |
f98393a6 | 2148 | invalidate_bdev(bdev); |
b71e8a4c DH |
2149 | return res; |
2150 | } | |
2151 | EXPORT_SYMBOL(__invalidate_device); | |
5c0d6b60 JK |
2152 | |
2153 | void iterate_bdevs(void (*func)(struct block_device *, void *), void *arg) | |
2154 | { | |
2155 | struct inode *inode, *old_inode = NULL; | |
2156 | ||
74278da9 | 2157 | spin_lock(&blockdev_superblock->s_inode_list_lock); |
5c0d6b60 JK |
2158 | list_for_each_entry(inode, &blockdev_superblock->s_inodes, i_sb_list) { |
2159 | struct address_space *mapping = inode->i_mapping; | |
af309226 | 2160 | struct block_device *bdev; |
5c0d6b60 JK |
2161 | |
2162 | spin_lock(&inode->i_lock); | |
2163 | if (inode->i_state & (I_FREEING|I_WILL_FREE|I_NEW) || | |
2164 | mapping->nrpages == 0) { | |
2165 | spin_unlock(&inode->i_lock); | |
2166 | continue; | |
2167 | } | |
2168 | __iget(inode); | |
2169 | spin_unlock(&inode->i_lock); | |
74278da9 | 2170 | spin_unlock(&blockdev_superblock->s_inode_list_lock); |
5c0d6b60 JK |
2171 | /* |
2172 | * We hold a reference to 'inode' so it couldn't have been | |
2173 | * removed from s_inodes list while we dropped the | |
74278da9 | 2174 | * s_inode_list_lock We cannot iput the inode now as we can |
5c0d6b60 | 2175 | * be holding the last reference and we cannot iput it under |
74278da9 | 2176 | * s_inode_list_lock. So we keep the reference and iput it |
5c0d6b60 JK |
2177 | * later. |
2178 | */ | |
2179 | iput(old_inode); | |
2180 | old_inode = inode; | |
af309226 | 2181 | bdev = I_BDEV(inode); |
5c0d6b60 | 2182 | |
af309226 RV |
2183 | mutex_lock(&bdev->bd_mutex); |
2184 | if (bdev->bd_openers) | |
2185 | func(bdev, arg); | |
2186 | mutex_unlock(&bdev->bd_mutex); | |
5c0d6b60 | 2187 | |
74278da9 | 2188 | spin_lock(&blockdev_superblock->s_inode_list_lock); |
5c0d6b60 | 2189 | } |
74278da9 | 2190 | spin_unlock(&blockdev_superblock->s_inode_list_lock); |
5c0d6b60 JK |
2191 | iput(old_inode); |
2192 | } |