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 | ||
1cac41cb MB |
29 | static struct bio *next_bio(struct bio *bio, unsigned int nr_pages, |
30 | int type, gfp_t gfp) | |
31 | { | |
32 | struct bio *new = bio_alloc(gfp, nr_pages); | |
33 | ||
34 | if (bio) { | |
35 | bio_chain(bio, new); | |
36 | submit_bio(type, bio); | |
37 | } | |
38 | ||
39 | return new; | |
40 | } | |
41 | ||
42 | int __blkdev_issue_discard(struct block_device *bdev, sector_t sector, | |
43 | sector_t nr_sects, gfp_t gfp_mask, unsigned long flags, | |
44 | struct bio **biop) | |
f31e7e40 | 45 | { |
f31e7e40 | 46 | struct request_queue *q = bdev_get_queue(bdev); |
1cac41cb MB |
47 | struct bio *bio = *biop; |
48 | int type = REQ_WRITE | REQ_DISCARD | REQ_PRIO; | |
a22c4d7e ML |
49 | unsigned int granularity; |
50 | int alignment; | |
1cac41cb | 51 | sector_t bs_mask; |
f31e7e40 DM |
52 | |
53 | if (!q) | |
54 | return -ENXIO; | |
55 | ||
56 | if (!blk_queue_discard(q)) | |
57 | return -EOPNOTSUPP; | |
58 | ||
1cac41cb MB |
59 | bs_mask = (bdev_logical_block_size(bdev) >> 9) - 1; |
60 | if ((sector | nr_sects) & bs_mask) | |
61 | return -EINVAL; | |
62 | ||
63 | /* Zero-sector (unknown) and one-sector granularities are the same. */ | |
a22c4d7e ML |
64 | granularity = max(q->limits.discard_granularity >> 9, 1U); |
65 | alignment = (bdev_discard_alignment(bdev) >> 9) % granularity; | |
66 | ||
dd3932ed | 67 | if (flags & BLKDEV_DISCARD_SECURE) { |
8d57a98c AH |
68 | if (!blk_queue_secdiscard(q)) |
69 | return -EOPNOTSUPP; | |
8c555367 | 70 | type |= REQ_SECURE; |
8d57a98c AH |
71 | } |
72 | ||
1cac41cb MB |
73 | if (flags & BLKDEV_DISCARD_SYNC) |
74 | type |= REQ_SYNC; | |
5dba3089 LC |
75 | |
76 | while (nr_sects) { | |
c6e66634 | 77 | unsigned int req_sects; |
a22c4d7e | 78 | sector_t end_sect, tmp; |
c6e66634 | 79 | |
a22c4d7e ML |
80 | /* Make sure bi_size doesn't overflow */ |
81 | req_sects = min_t(sector_t, nr_sects, UINT_MAX >> 9); | |
82 | ||
1cac41cb | 83 | /** |
a22c4d7e ML |
84 | * If splitting a request, and the next starting sector would be |
85 | * misaligned, stop the discard at the previous aligned sector. | |
86 | */ | |
c6e66634 | 87 | end_sect = sector + req_sects; |
a22c4d7e ML |
88 | tmp = end_sect; |
89 | if (req_sects < nr_sects && | |
90 | sector_div(tmp, granularity) != alignment) { | |
91 | end_sect = end_sect - alignment; | |
92 | sector_div(end_sect, granularity); | |
93 | end_sect = end_sect * granularity + alignment; | |
94 | req_sects = end_sect - sector; | |
95 | } | |
c6e66634 | 96 | |
1cac41cb | 97 | bio = next_bio(bio, 1, type, gfp_mask); |
4f024f37 | 98 | bio->bi_iter.bi_sector = sector; |
f31e7e40 | 99 | bio->bi_bdev = bdev; |
f31e7e40 | 100 | |
4f024f37 | 101 | bio->bi_iter.bi_size = req_sects << 9; |
c6e66634 PB |
102 | nr_sects -= req_sects; |
103 | sector = end_sect; | |
f31e7e40 | 104 | |
c8123f8c JA |
105 | /* |
106 | * We can loop for a long time in here, if someone does | |
107 | * full device discards (like mkfs). Be nice and allow | |
108 | * us to schedule out to avoid softlocking if preempt | |
109 | * is disabled. | |
110 | */ | |
111 | cond_resched(); | |
5dba3089 | 112 | } |
f31e7e40 | 113 | |
1cac41cb MB |
114 | *biop = bio; |
115 | return 0; | |
116 | } | |
117 | EXPORT_SYMBOL(__blkdev_issue_discard); | |
118 | ||
119 | /** | |
120 | * blkdev_issue_discard - queue a discard | |
121 | * @bdev: blockdev to issue discard for | |
122 | * @sector: start sector | |
123 | * @nr_sects: number of sectors to discard | |
124 | * @gfp_mask: memory allocation flags (for bio_alloc) | |
125 | * @flags: BLKDEV_IFL_* flags to control behaviour | |
126 | * | |
127 | * Description: | |
128 | * Issue a discard request for the sectors in question. | |
129 | */ | |
130 | int blkdev_issue_discard(struct block_device *bdev, sector_t sector, | |
131 | sector_t nr_sects, gfp_t gfp_mask, unsigned long flags) | |
132 | { | |
133 | struct bio *bio = NULL; | |
134 | struct blk_plug plug; | |
135 | int type = REQ_WRITE | REQ_DISCARD | REQ_PRIO; | |
136 | int ret; | |
137 | ||
138 | blk_start_plug(&plug); | |
139 | ret = __blkdev_issue_discard(bdev, sector, nr_sects, gfp_mask, flags, | |
140 | &bio); | |
141 | if (!ret && bio) { | |
142 | ret = submit_bio_wait(type, bio); | |
143 | if (ret == -EOPNOTSUPP) | |
144 | ret = 0; | |
145 | bio_put(bio); | |
146 | } | |
147 | blk_finish_plug(&plug); | |
f31e7e40 | 148 | |
f31e7e40 | 149 | return ret; |
f31e7e40 DM |
150 | } |
151 | EXPORT_SYMBOL(blkdev_issue_discard); | |
3f14d792 | 152 | |
4363ac7c MP |
153 | /** |
154 | * blkdev_issue_write_same - queue a write same operation | |
155 | * @bdev: target blockdev | |
156 | * @sector: start sector | |
157 | * @nr_sects: number of sectors to write | |
158 | * @gfp_mask: memory allocation flags (for bio_alloc) | |
159 | * @page: page containing data to write | |
160 | * | |
161 | * Description: | |
162 | * Issue a write same request for the sectors in question. | |
163 | */ | |
164 | int blkdev_issue_write_same(struct block_device *bdev, sector_t sector, | |
165 | sector_t nr_sects, gfp_t gfp_mask, | |
166 | struct page *page) | |
167 | { | |
168 | DECLARE_COMPLETION_ONSTACK(wait); | |
169 | struct request_queue *q = bdev_get_queue(bdev); | |
170 | unsigned int max_write_same_sectors; | |
171 | struct bio_batch bb; | |
172 | struct bio *bio; | |
173 | int ret = 0; | |
174 | ||
175 | if (!q) | |
176 | return -ENXIO; | |
177 | ||
b49a0871 ML |
178 | /* Ensure that max_write_same_sectors doesn't overflow bi_size */ |
179 | max_write_same_sectors = UINT_MAX >> 9; | |
4363ac7c MP |
180 | |
181 | atomic_set(&bb.done, 1); | |
4246a0b6 | 182 | bb.error = 0; |
4363ac7c MP |
183 | bb.wait = &wait; |
184 | ||
185 | while (nr_sects) { | |
186 | bio = bio_alloc(gfp_mask, 1); | |
187 | if (!bio) { | |
188 | ret = -ENOMEM; | |
189 | break; | |
190 | } | |
191 | ||
4f024f37 | 192 | bio->bi_iter.bi_sector = sector; |
4363ac7c MP |
193 | bio->bi_end_io = bio_batch_end_io; |
194 | bio->bi_bdev = bdev; | |
195 | bio->bi_private = &bb; | |
196 | bio->bi_vcnt = 1; | |
197 | bio->bi_io_vec->bv_page = page; | |
198 | bio->bi_io_vec->bv_offset = 0; | |
199 | bio->bi_io_vec->bv_len = bdev_logical_block_size(bdev); | |
200 | ||
201 | if (nr_sects > max_write_same_sectors) { | |
4f024f37 | 202 | bio->bi_iter.bi_size = max_write_same_sectors << 9; |
4363ac7c MP |
203 | nr_sects -= max_write_same_sectors; |
204 | sector += max_write_same_sectors; | |
205 | } else { | |
4f024f37 | 206 | bio->bi_iter.bi_size = nr_sects << 9; |
4363ac7c MP |
207 | nr_sects = 0; |
208 | } | |
209 | ||
210 | atomic_inc(&bb.done); | |
211 | submit_bio(REQ_WRITE | REQ_WRITE_SAME, bio); | |
212 | } | |
213 | ||
214 | /* Wait for bios in-flight */ | |
215 | if (!atomic_dec_and_test(&bb.done)) | |
5577022f | 216 | wait_for_completion_io(&wait); |
4363ac7c | 217 | |
4246a0b6 CH |
218 | if (bb.error) |
219 | return bb.error; | |
4363ac7c MP |
220 | return ret; |
221 | } | |
222 | EXPORT_SYMBOL(blkdev_issue_write_same); | |
223 | ||
3f14d792 | 224 | /** |
291d24f6 | 225 | * blkdev_issue_zeroout - generate number of zero filed write bios |
3f14d792 DM |
226 | * @bdev: blockdev to issue |
227 | * @sector: start sector | |
228 | * @nr_sects: number of sectors to write | |
229 | * @gfp_mask: memory allocation flags (for bio_alloc) | |
3f14d792 DM |
230 | * |
231 | * Description: | |
232 | * Generate and issue number of bios with zerofiled pages. | |
3f14d792 DM |
233 | */ |
234 | ||
35086784 FF |
235 | static int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, |
236 | sector_t nr_sects, gfp_t gfp_mask) | |
3f14d792 | 237 | { |
18edc8ea | 238 | int ret; |
3f14d792 DM |
239 | struct bio *bio; |
240 | struct bio_batch bb; | |
0aeea189 | 241 | unsigned int sz; |
3f14d792 DM |
242 | DECLARE_COMPLETION_ONSTACK(wait); |
243 | ||
0aeea189 | 244 | atomic_set(&bb.done, 1); |
4246a0b6 | 245 | bb.error = 0; |
3f14d792 | 246 | bb.wait = &wait; |
3f14d792 | 247 | |
18edc8ea | 248 | ret = 0; |
3f14d792 DM |
249 | while (nr_sects != 0) { |
250 | bio = bio_alloc(gfp_mask, | |
251 | min(nr_sects, (sector_t)BIO_MAX_PAGES)); | |
18edc8ea DM |
252 | if (!bio) { |
253 | ret = -ENOMEM; | |
3f14d792 | 254 | break; |
18edc8ea | 255 | } |
3f14d792 | 256 | |
4f024f37 | 257 | bio->bi_iter.bi_sector = sector; |
3f14d792 DM |
258 | bio->bi_bdev = bdev; |
259 | bio->bi_end_io = bio_batch_end_io; | |
dd3932ed | 260 | bio->bi_private = &bb; |
3f14d792 | 261 | |
0341aafb JA |
262 | while (nr_sects != 0) { |
263 | sz = min((sector_t) PAGE_SIZE >> 9 , nr_sects); | |
3f14d792 DM |
264 | ret = bio_add_page(bio, ZERO_PAGE(0), sz << 9, 0); |
265 | nr_sects -= ret >> 9; | |
266 | sector += ret >> 9; | |
267 | if (ret < (sz << 9)) | |
268 | break; | |
269 | } | |
18edc8ea | 270 | ret = 0; |
0aeea189 | 271 | atomic_inc(&bb.done); |
3f14d792 DM |
272 | submit_bio(WRITE, bio); |
273 | } | |
3f14d792 | 274 | |
dd3932ed | 275 | /* Wait for bios in-flight */ |
0aeea189 | 276 | if (!atomic_dec_and_test(&bb.done)) |
5577022f | 277 | wait_for_completion_io(&wait); |
3f14d792 | 278 | |
4246a0b6 CH |
279 | if (bb.error) |
280 | return bb.error; | |
3f14d792 DM |
281 | return ret; |
282 | } | |
579e8f3c MP |
283 | |
284 | /** | |
285 | * blkdev_issue_zeroout - zero-fill a block range | |
286 | * @bdev: blockdev to write | |
287 | * @sector: start sector | |
288 | * @nr_sects: number of sectors to write | |
289 | * @gfp_mask: memory allocation flags (for bio_alloc) | |
d93ba7a5 | 290 | * @discard: whether to discard the block range |
579e8f3c MP |
291 | * |
292 | * Description: | |
d93ba7a5 MP |
293 | * Zero-fill a block range. If the discard flag is set and the block |
294 | * device guarantees that subsequent READ operations to the block range | |
295 | * in question will return zeroes, the blocks will be discarded. Should | |
296 | * the discard request fail, if the discard flag is not set, or if | |
297 | * discard_zeroes_data is not supported, this function will resort to | |
298 | * zeroing the blocks manually, thus provisioning (allocating, | |
299 | * anchoring) them. If the block device supports the WRITE SAME command | |
300 | * blkdev_issue_zeroout() will use it to optimize the process of | |
301 | * clearing the block range. Otherwise the zeroing will be performed | |
302 | * using regular WRITE calls. | |
579e8f3c MP |
303 | */ |
304 | ||
305 | int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, | |
d93ba7a5 | 306 | sector_t nr_sects, gfp_t gfp_mask, bool discard) |
579e8f3c | 307 | { |
d93ba7a5 | 308 | struct request_queue *q = bdev_get_queue(bdev); |
d93ba7a5 | 309 | |
9f9ee1f2 MP |
310 | if (discard && blk_queue_discard(q) && q->limits.discard_zeroes_data && |
311 | blkdev_issue_discard(bdev, sector, nr_sects, gfp_mask, 0) == 0) | |
312 | return 0; | |
d93ba7a5 | 313 | |
9f9ee1f2 MP |
314 | if (bdev_write_same(bdev) && |
315 | blkdev_issue_write_same(bdev, sector, nr_sects, gfp_mask, | |
316 | ZERO_PAGE(0)) == 0) | |
317 | return 0; | |
579e8f3c MP |
318 | |
319 | return __blkdev_issue_zeroout(bdev, sector, nr_sects, gfp_mask); | |
320 | } | |
3f14d792 | 321 | EXPORT_SYMBOL(blkdev_issue_zeroout); |