Commit | Line | Data |
---|---|---|
f31e7e40 DM |
1 | /* |
2 | * Functions related to generic helpers functions | |
3 | */ | |
4 | #include <linux/kernel.h> | |
5 | #include <linux/module.h> | |
6 | #include <linux/bio.h> | |
7 | #include <linux/blkdev.h> | |
8 | #include <linux/scatterlist.h> | |
9 | ||
10 | #include "blk.h" | |
11 | ||
5dba3089 LC |
12 | struct bio_batch { |
13 | atomic_t done; | |
4246a0b6 | 14 | int error; |
5dba3089 LC |
15 | struct completion *wait; |
16 | }; | |
17 | ||
4246a0b6 | 18 | static void bio_batch_end_io(struct bio *bio) |
f31e7e40 | 19 | { |
5dba3089 LC |
20 | struct bio_batch *bb = bio->bi_private; |
21 | ||
4246a0b6 CH |
22 | if (bio->bi_error && bio->bi_error != -EOPNOTSUPP) |
23 | bb->error = bio->bi_error; | |
5dba3089 LC |
24 | if (atomic_dec_and_test(&bb->done)) |
25 | complete(bb->wait); | |
f31e7e40 DM |
26 | bio_put(bio); |
27 | } | |
28 | ||
b49a0871 ML |
29 | /* |
30 | * Ensure that max discard sectors doesn't overflow bi_size and hopefully | |
31 | * it is of the proper granularity as long as the granularity is a power | |
32 | * of two. | |
33 | */ | |
34 | #define MAX_BIO_SECTORS ((1U << 31) >> 9) | |
35 | ||
f31e7e40 DM |
36 | /** |
37 | * blkdev_issue_discard - queue a discard | |
38 | * @bdev: blockdev to issue discard for | |
39 | * @sector: start sector | |
40 | * @nr_sects: number of sectors to discard | |
41 | * @gfp_mask: memory allocation flags (for bio_alloc) | |
42 | * @flags: BLKDEV_IFL_* flags to control behaviour | |
43 | * | |
44 | * Description: | |
45 | * Issue a discard request for the sectors in question. | |
46 | */ | |
47 | int blkdev_issue_discard(struct block_device *bdev, sector_t sector, | |
48 | sector_t nr_sects, gfp_t gfp_mask, unsigned long flags) | |
49 | { | |
50 | DECLARE_COMPLETION_ONSTACK(wait); | |
51 | struct request_queue *q = bdev_get_queue(bdev); | |
8c555367 | 52 | int type = REQ_WRITE | REQ_DISCARD; |
5dba3089 | 53 | struct bio_batch bb; |
f31e7e40 | 54 | struct bio *bio; |
f31e7e40 | 55 | int ret = 0; |
0cfbcafc | 56 | struct blk_plug plug; |
f31e7e40 DM |
57 | |
58 | if (!q) | |
59 | return -ENXIO; | |
60 | ||
61 | if (!blk_queue_discard(q)) | |
62 | return -EOPNOTSUPP; | |
63 | ||
dd3932ed | 64 | if (flags & BLKDEV_DISCARD_SECURE) { |
8d57a98c AH |
65 | if (!blk_queue_secdiscard(q)) |
66 | return -EOPNOTSUPP; | |
8c555367 | 67 | type |= REQ_SECURE; |
8d57a98c AH |
68 | } |
69 | ||
5dba3089 | 70 | atomic_set(&bb.done, 1); |
4246a0b6 | 71 | bb.error = 0; |
5dba3089 LC |
72 | bb.wait = &wait; |
73 | ||
0cfbcafc | 74 | blk_start_plug(&plug); |
5dba3089 | 75 | while (nr_sects) { |
c6e66634 | 76 | unsigned int req_sects; |
b49a0871 | 77 | sector_t end_sect; |
c6e66634 | 78 | |
f31e7e40 | 79 | bio = bio_alloc(gfp_mask, 1); |
66ac0280 CH |
80 | if (!bio) { |
81 | ret = -ENOMEM; | |
82 | break; | |
83 | } | |
84 | ||
b49a0871 | 85 | req_sects = min_t(sector_t, nr_sects, MAX_BIO_SECTORS); |
c6e66634 | 86 | end_sect = sector + req_sects; |
c6e66634 | 87 | |
4f024f37 | 88 | bio->bi_iter.bi_sector = sector; |
5dba3089 | 89 | bio->bi_end_io = bio_batch_end_io; |
f31e7e40 | 90 | bio->bi_bdev = bdev; |
5dba3089 | 91 | bio->bi_private = &bb; |
f31e7e40 | 92 | |
4f024f37 | 93 | bio->bi_iter.bi_size = req_sects << 9; |
c6e66634 PB |
94 | nr_sects -= req_sects; |
95 | sector = end_sect; | |
f31e7e40 | 96 | |
5dba3089 | 97 | atomic_inc(&bb.done); |
f31e7e40 | 98 | submit_bio(type, bio); |
c8123f8c JA |
99 | |
100 | /* | |
101 | * We can loop for a long time in here, if someone does | |
102 | * full device discards (like mkfs). Be nice and allow | |
103 | * us to schedule out to avoid softlocking if preempt | |
104 | * is disabled. | |
105 | */ | |
106 | cond_resched(); | |
5dba3089 | 107 | } |
0cfbcafc | 108 | blk_finish_plug(&plug); |
f31e7e40 | 109 | |
5dba3089 LC |
110 | /* Wait for bios in-flight */ |
111 | if (!atomic_dec_and_test(&bb.done)) | |
5577022f | 112 | wait_for_completion_io(&wait); |
f31e7e40 | 113 | |
4246a0b6 CH |
114 | if (bb.error) |
115 | return bb.error; | |
f31e7e40 | 116 | return ret; |
f31e7e40 DM |
117 | } |
118 | EXPORT_SYMBOL(blkdev_issue_discard); | |
3f14d792 | 119 | |
4363ac7c MP |
120 | /** |
121 | * blkdev_issue_write_same - queue a write same operation | |
122 | * @bdev: target blockdev | |
123 | * @sector: start sector | |
124 | * @nr_sects: number of sectors to write | |
125 | * @gfp_mask: memory allocation flags (for bio_alloc) | |
126 | * @page: page containing data to write | |
127 | * | |
128 | * Description: | |
129 | * Issue a write same request for the sectors in question. | |
130 | */ | |
131 | int blkdev_issue_write_same(struct block_device *bdev, sector_t sector, | |
132 | sector_t nr_sects, gfp_t gfp_mask, | |
133 | struct page *page) | |
134 | { | |
135 | DECLARE_COMPLETION_ONSTACK(wait); | |
136 | struct request_queue *q = bdev_get_queue(bdev); | |
137 | unsigned int max_write_same_sectors; | |
138 | struct bio_batch bb; | |
139 | struct bio *bio; | |
140 | int ret = 0; | |
141 | ||
142 | if (!q) | |
143 | return -ENXIO; | |
144 | ||
b49a0871 ML |
145 | /* Ensure that max_write_same_sectors doesn't overflow bi_size */ |
146 | max_write_same_sectors = UINT_MAX >> 9; | |
4363ac7c MP |
147 | |
148 | atomic_set(&bb.done, 1); | |
4246a0b6 | 149 | bb.error = 0; |
4363ac7c MP |
150 | bb.wait = &wait; |
151 | ||
152 | while (nr_sects) { | |
153 | bio = bio_alloc(gfp_mask, 1); | |
154 | if (!bio) { | |
155 | ret = -ENOMEM; | |
156 | break; | |
157 | } | |
158 | ||
4f024f37 | 159 | bio->bi_iter.bi_sector = sector; |
4363ac7c MP |
160 | bio->bi_end_io = bio_batch_end_io; |
161 | bio->bi_bdev = bdev; | |
162 | bio->bi_private = &bb; | |
163 | bio->bi_vcnt = 1; | |
164 | bio->bi_io_vec->bv_page = page; | |
165 | bio->bi_io_vec->bv_offset = 0; | |
166 | bio->bi_io_vec->bv_len = bdev_logical_block_size(bdev); | |
167 | ||
168 | if (nr_sects > max_write_same_sectors) { | |
4f024f37 | 169 | bio->bi_iter.bi_size = max_write_same_sectors << 9; |
4363ac7c MP |
170 | nr_sects -= max_write_same_sectors; |
171 | sector += max_write_same_sectors; | |
172 | } else { | |
4f024f37 | 173 | bio->bi_iter.bi_size = nr_sects << 9; |
4363ac7c MP |
174 | nr_sects = 0; |
175 | } | |
176 | ||
177 | atomic_inc(&bb.done); | |
178 | submit_bio(REQ_WRITE | REQ_WRITE_SAME, bio); | |
179 | } | |
180 | ||
181 | /* Wait for bios in-flight */ | |
182 | if (!atomic_dec_and_test(&bb.done)) | |
5577022f | 183 | wait_for_completion_io(&wait); |
4363ac7c | 184 | |
4246a0b6 CH |
185 | if (bb.error) |
186 | return bb.error; | |
4363ac7c MP |
187 | return ret; |
188 | } | |
189 | EXPORT_SYMBOL(blkdev_issue_write_same); | |
190 | ||
3f14d792 | 191 | /** |
291d24f6 | 192 | * blkdev_issue_zeroout - generate number of zero filed write bios |
3f14d792 DM |
193 | * @bdev: blockdev to issue |
194 | * @sector: start sector | |
195 | * @nr_sects: number of sectors to write | |
196 | * @gfp_mask: memory allocation flags (for bio_alloc) | |
3f14d792 DM |
197 | * |
198 | * Description: | |
199 | * Generate and issue number of bios with zerofiled pages. | |
3f14d792 DM |
200 | */ |
201 | ||
35086784 FF |
202 | static int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, |
203 | sector_t nr_sects, gfp_t gfp_mask) | |
3f14d792 | 204 | { |
18edc8ea | 205 | int ret; |
3f14d792 DM |
206 | struct bio *bio; |
207 | struct bio_batch bb; | |
0aeea189 | 208 | unsigned int sz; |
3f14d792 DM |
209 | DECLARE_COMPLETION_ONSTACK(wait); |
210 | ||
0aeea189 | 211 | atomic_set(&bb.done, 1); |
4246a0b6 | 212 | bb.error = 0; |
3f14d792 | 213 | bb.wait = &wait; |
3f14d792 | 214 | |
18edc8ea | 215 | ret = 0; |
3f14d792 DM |
216 | while (nr_sects != 0) { |
217 | bio = bio_alloc(gfp_mask, | |
218 | min(nr_sects, (sector_t)BIO_MAX_PAGES)); | |
18edc8ea DM |
219 | if (!bio) { |
220 | ret = -ENOMEM; | |
3f14d792 | 221 | break; |
18edc8ea | 222 | } |
3f14d792 | 223 | |
4f024f37 | 224 | bio->bi_iter.bi_sector = sector; |
3f14d792 DM |
225 | bio->bi_bdev = bdev; |
226 | bio->bi_end_io = bio_batch_end_io; | |
dd3932ed | 227 | bio->bi_private = &bb; |
3f14d792 | 228 | |
0341aafb JA |
229 | while (nr_sects != 0) { |
230 | sz = min((sector_t) PAGE_SIZE >> 9 , nr_sects); | |
3f14d792 DM |
231 | ret = bio_add_page(bio, ZERO_PAGE(0), sz << 9, 0); |
232 | nr_sects -= ret >> 9; | |
233 | sector += ret >> 9; | |
234 | if (ret < (sz << 9)) | |
235 | break; | |
236 | } | |
18edc8ea | 237 | ret = 0; |
0aeea189 | 238 | atomic_inc(&bb.done); |
3f14d792 DM |
239 | submit_bio(WRITE, bio); |
240 | } | |
3f14d792 | 241 | |
dd3932ed | 242 | /* Wait for bios in-flight */ |
0aeea189 | 243 | if (!atomic_dec_and_test(&bb.done)) |
5577022f | 244 | wait_for_completion_io(&wait); |
3f14d792 | 245 | |
4246a0b6 CH |
246 | if (bb.error) |
247 | return bb.error; | |
3f14d792 DM |
248 | return ret; |
249 | } | |
579e8f3c MP |
250 | |
251 | /** | |
252 | * blkdev_issue_zeroout - zero-fill a block range | |
253 | * @bdev: blockdev to write | |
254 | * @sector: start sector | |
255 | * @nr_sects: number of sectors to write | |
256 | * @gfp_mask: memory allocation flags (for bio_alloc) | |
d93ba7a5 | 257 | * @discard: whether to discard the block range |
579e8f3c MP |
258 | * |
259 | * Description: | |
d93ba7a5 MP |
260 | * Zero-fill a block range. If the discard flag is set and the block |
261 | * device guarantees that subsequent READ operations to the block range | |
262 | * in question will return zeroes, the blocks will be discarded. Should | |
263 | * the discard request fail, if the discard flag is not set, or if | |
264 | * discard_zeroes_data is not supported, this function will resort to | |
265 | * zeroing the blocks manually, thus provisioning (allocating, | |
266 | * anchoring) them. If the block device supports the WRITE SAME command | |
267 | * blkdev_issue_zeroout() will use it to optimize the process of | |
268 | * clearing the block range. Otherwise the zeroing will be performed | |
269 | * using regular WRITE calls. | |
579e8f3c MP |
270 | */ |
271 | ||
272 | int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, | |
d93ba7a5 | 273 | sector_t nr_sects, gfp_t gfp_mask, bool discard) |
579e8f3c | 274 | { |
d93ba7a5 | 275 | struct request_queue *q = bdev_get_queue(bdev); |
d93ba7a5 | 276 | |
9f9ee1f2 MP |
277 | if (discard && blk_queue_discard(q) && q->limits.discard_zeroes_data && |
278 | blkdev_issue_discard(bdev, sector, nr_sects, gfp_mask, 0) == 0) | |
279 | return 0; | |
d93ba7a5 | 280 | |
9f9ee1f2 MP |
281 | if (bdev_write_same(bdev) && |
282 | blkdev_issue_write_same(bdev, sector, nr_sects, gfp_mask, | |
283 | ZERO_PAGE(0)) == 0) | |
284 | return 0; | |
579e8f3c MP |
285 | |
286 | return __blkdev_issue_zeroout(bdev, sector, nr_sects, gfp_mask); | |
287 | } | |
3f14d792 | 288 | EXPORT_SYMBOL(blkdev_issue_zeroout); |