]>
git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blob - block/blk-lib.c
2 * Functions related to generic helpers functions
4 #include <linux/kernel.h>
5 #include <linux/module.h>
7 #include <linux/blkdev.h>
8 #include <linux/scatterlist.h>
15 struct completion
*wait
;
18 static void bio_batch_end_io(struct bio
*bio
)
20 struct bio_batch
*bb
= bio
->bi_private
;
22 if (bio
->bi_error
&& bio
->bi_error
!= -EOPNOTSUPP
)
23 bb
->error
= bio
->bi_error
;
24 if (atomic_dec_and_test(&bb
->done
))
30 * blkdev_issue_discard - queue a discard
31 * @bdev: blockdev to issue discard for
32 * @sector: start sector
33 * @nr_sects: number of sectors to discard
34 * @gfp_mask: memory allocation flags (for bio_alloc)
35 * @flags: BLKDEV_IFL_* flags to control behaviour
38 * Issue a discard request for the sectors in question.
40 int blkdev_issue_discard(struct block_device
*bdev
, sector_t sector
,
41 sector_t nr_sects
, gfp_t gfp_mask
, unsigned long flags
)
43 DECLARE_COMPLETION_ONSTACK(wait
);
44 struct request_queue
*q
= bdev_get_queue(bdev
);
45 int type
= REQ_WRITE
| REQ_DISCARD
;
46 unsigned int max_discard_sectors
, granularity
;
56 if (!blk_queue_discard(q
))
59 /* Zero-sector (unknown) and one-sector granularities are the same. */
60 granularity
= max(q
->limits
.discard_granularity
>> 9, 1U);
61 alignment
= (bdev_discard_alignment(bdev
) >> 9) % granularity
;
64 * Ensure that max_discard_sectors is of the proper
65 * granularity, so that requests stay aligned after a split.
67 max_discard_sectors
= min(q
->limits
.max_discard_sectors
, UINT_MAX
>> 9);
68 max_discard_sectors
-= max_discard_sectors
% granularity
;
69 if (unlikely(!max_discard_sectors
)) {
70 /* Avoid infinite loop below. Being cautious never hurts. */
74 if (flags
& BLKDEV_DISCARD_SECURE
) {
75 if (!blk_queue_secdiscard(q
))
80 atomic_set(&bb
.done
, 1);
84 blk_start_plug(&plug
);
86 unsigned int req_sects
;
87 sector_t end_sect
, tmp
;
89 bio
= bio_alloc(gfp_mask
, 1);
95 req_sects
= min_t(sector_t
, nr_sects
, max_discard_sectors
);
98 * If splitting a request, and the next starting sector would be
99 * misaligned, stop the discard at the previous aligned sector.
101 end_sect
= sector
+ req_sects
;
103 if (req_sects
< nr_sects
&&
104 sector_div(tmp
, granularity
) != alignment
) {
105 end_sect
= end_sect
- alignment
;
106 sector_div(end_sect
, granularity
);
107 end_sect
= end_sect
* granularity
+ alignment
;
108 req_sects
= end_sect
- sector
;
111 bio
->bi_iter
.bi_sector
= sector
;
112 bio
->bi_end_io
= bio_batch_end_io
;
114 bio
->bi_private
= &bb
;
116 bio
->bi_iter
.bi_size
= req_sects
<< 9;
117 nr_sects
-= req_sects
;
120 atomic_inc(&bb
.done
);
121 submit_bio(type
, bio
);
124 * We can loop for a long time in here, if someone does
125 * full device discards (like mkfs). Be nice and allow
126 * us to schedule out to avoid softlocking if preempt
131 blk_finish_plug(&plug
);
133 /* Wait for bios in-flight */
134 if (!atomic_dec_and_test(&bb
.done
))
135 wait_for_completion_io(&wait
);
141 EXPORT_SYMBOL(blkdev_issue_discard
);
144 * blkdev_issue_write_same - queue a write same operation
145 * @bdev: target blockdev
146 * @sector: start sector
147 * @nr_sects: number of sectors to write
148 * @gfp_mask: memory allocation flags (for bio_alloc)
149 * @page: page containing data to write
152 * Issue a write same request for the sectors in question.
154 int blkdev_issue_write_same(struct block_device
*bdev
, sector_t sector
,
155 sector_t nr_sects
, gfp_t gfp_mask
,
158 DECLARE_COMPLETION_ONSTACK(wait
);
159 struct request_queue
*q
= bdev_get_queue(bdev
);
160 unsigned int max_write_same_sectors
;
168 max_write_same_sectors
= q
->limits
.max_write_same_sectors
;
170 if (max_write_same_sectors
== 0)
173 atomic_set(&bb
.done
, 1);
178 bio
= bio_alloc(gfp_mask
, 1);
184 bio
->bi_iter
.bi_sector
= sector
;
185 bio
->bi_end_io
= bio_batch_end_io
;
187 bio
->bi_private
= &bb
;
189 bio
->bi_io_vec
->bv_page
= page
;
190 bio
->bi_io_vec
->bv_offset
= 0;
191 bio
->bi_io_vec
->bv_len
= bdev_logical_block_size(bdev
);
193 if (nr_sects
> max_write_same_sectors
) {
194 bio
->bi_iter
.bi_size
= max_write_same_sectors
<< 9;
195 nr_sects
-= max_write_same_sectors
;
196 sector
+= max_write_same_sectors
;
198 bio
->bi_iter
.bi_size
= nr_sects
<< 9;
202 atomic_inc(&bb
.done
);
203 submit_bio(REQ_WRITE
| REQ_WRITE_SAME
, bio
);
206 /* Wait for bios in-flight */
207 if (!atomic_dec_and_test(&bb
.done
))
208 wait_for_completion_io(&wait
);
214 EXPORT_SYMBOL(blkdev_issue_write_same
);
217 * blkdev_issue_zeroout - generate number of zero filed write bios
218 * @bdev: blockdev to issue
219 * @sector: start sector
220 * @nr_sects: number of sectors to write
221 * @gfp_mask: memory allocation flags (for bio_alloc)
224 * Generate and issue number of bios with zerofiled pages.
227 static int __blkdev_issue_zeroout(struct block_device
*bdev
, sector_t sector
,
228 sector_t nr_sects
, gfp_t gfp_mask
)
234 DECLARE_COMPLETION_ONSTACK(wait
);
236 atomic_set(&bb
.done
, 1);
241 while (nr_sects
!= 0) {
242 bio
= bio_alloc(gfp_mask
,
243 min(nr_sects
, (sector_t
)BIO_MAX_PAGES
));
249 bio
->bi_iter
.bi_sector
= sector
;
251 bio
->bi_end_io
= bio_batch_end_io
;
252 bio
->bi_private
= &bb
;
254 while (nr_sects
!= 0) {
255 sz
= min((sector_t
) PAGE_SIZE
>> 9 , nr_sects
);
256 ret
= bio_add_page(bio
, ZERO_PAGE(0), sz
<< 9, 0);
257 nr_sects
-= ret
>> 9;
263 atomic_inc(&bb
.done
);
264 submit_bio(WRITE
, bio
);
267 /* Wait for bios in-flight */
268 if (!atomic_dec_and_test(&bb
.done
))
269 wait_for_completion_io(&wait
);
277 * blkdev_issue_zeroout - zero-fill a block range
278 * @bdev: blockdev to write
279 * @sector: start sector
280 * @nr_sects: number of sectors to write
281 * @gfp_mask: memory allocation flags (for bio_alloc)
282 * @discard: whether to discard the block range
285 * Zero-fill a block range. If the discard flag is set and the block
286 * device guarantees that subsequent READ operations to the block range
287 * in question will return zeroes, the blocks will be discarded. Should
288 * the discard request fail, if the discard flag is not set, or if
289 * discard_zeroes_data is not supported, this function will resort to
290 * zeroing the blocks manually, thus provisioning (allocating,
291 * anchoring) them. If the block device supports the WRITE SAME command
292 * blkdev_issue_zeroout() will use it to optimize the process of
293 * clearing the block range. Otherwise the zeroing will be performed
294 * using regular WRITE calls.
297 int blkdev_issue_zeroout(struct block_device
*bdev
, sector_t sector
,
298 sector_t nr_sects
, gfp_t gfp_mask
, bool discard
)
300 struct request_queue
*q
= bdev_get_queue(bdev
);
302 if (discard
&& blk_queue_discard(q
) && q
->limits
.discard_zeroes_data
&&
303 blkdev_issue_discard(bdev
, sector
, nr_sects
, gfp_mask
, 0) == 0)
306 if (bdev_write_same(bdev
) &&
307 blkdev_issue_write_same(bdev
, sector
, nr_sects
, gfp_mask
,
311 return __blkdev_issue_zeroout(bdev
, sector
, nr_sects
, gfp_mask
);
313 EXPORT_SYMBOL(blkdev_issue_zeroout
);