]>
Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
f31e7e40 DM |
2 | /* |
3 | * Functions related to generic helpers functions | |
4 | */ | |
5 | #include <linux/kernel.h> | |
6 | #include <linux/module.h> | |
7 | #include <linux/bio.h> | |
8 | #include <linux/blkdev.h> | |
9 | #include <linux/scatterlist.h> | |
10 | ||
11 | #include "blk.h" | |
12 | ||
e3cc28ea CH |
13 | static sector_t bio_discard_limit(struct block_device *bdev, sector_t sector) |
14 | { | |
7b47ef52 | 15 | unsigned int discard_granularity = bdev_discard_granularity(bdev); |
e3cc28ea CH |
16 | sector_t granularity_aligned_sector; |
17 | ||
18 | if (bdev_is_partition(bdev)) | |
19 | sector += bdev->bd_start_sect; | |
20 | ||
21 | granularity_aligned_sector = | |
22 | round_up(sector, discard_granularity >> SECTOR_SHIFT); | |
23 | ||
24 | /* | |
25 | * Make sure subsequent bios start aligned to the discard granularity if | |
26 | * it needs to be split. | |
27 | */ | |
28 | if (granularity_aligned_sector != sector) | |
29 | return granularity_aligned_sector - sector; | |
30 | ||
31 | /* | |
32 | * Align the bio size to the discard granularity to make splitting the bio | |
33 | * at discard granularity boundaries easier in the driver if needed. | |
34 | */ | |
35 | return round_down(UINT_MAX, discard_granularity) >> SECTOR_SHIFT; | |
36 | } | |
37 | ||
e8b4869b CH |
38 | struct bio *blk_alloc_discard_bio(struct block_device *bdev, |
39 | sector_t *sector, sector_t *nr_sects, gfp_t gfp_mask) | |
f31e7e40 | 40 | { |
e8b4869b CH |
41 | sector_t bio_sects = min(*nr_sects, bio_discard_limit(bdev, *sector)); |
42 | struct bio *bio; | |
a22c4d7e | 43 | |
e8b4869b CH |
44 | if (!bio_sects) |
45 | return NULL; | |
4800bf7b | 46 | |
e8b4869b CH |
47 | bio = bio_alloc(bdev, 0, REQ_OP_DISCARD, gfp_mask); |
48 | if (!bio) | |
49 | return NULL; | |
50 | bio->bi_iter.bi_sector = *sector; | |
51 | bio->bi_iter.bi_size = bio_sects << SECTOR_SHIFT; | |
52 | *sector += bio_sects; | |
53 | *nr_sects -= bio_sects; | |
54 | /* | |
55 | * We can loop for a long time in here if someone does full device | |
56 | * discards (like mkfs). Be nice and allow us to schedule out to avoid | |
57 | * softlocking if preempt is disabled. | |
58 | */ | |
59 | cond_resched(); | |
60 | return bio; | |
61 | } | |
f31e7e40 | 62 | |
e8b4869b CH |
63 | int __blkdev_issue_discard(struct block_device *bdev, sector_t sector, |
64 | sector_t nr_sects, gfp_t gfp_mask, struct bio **biop) | |
65 | { | |
66 | struct bio *bio; | |
38f25255 | 67 | |
e8b4869b CH |
68 | while ((bio = blk_alloc_discard_bio(bdev, §or, &nr_sects, |
69 | gfp_mask))) | |
70 | *biop = bio_chain_and_submit(*biop, bio); | |
38f25255 CH |
71 | return 0; |
72 | } | |
73 | EXPORT_SYMBOL(__blkdev_issue_discard); | |
74 | ||
75 | /** | |
76 | * blkdev_issue_discard - queue a discard | |
77 | * @bdev: blockdev to issue discard for | |
78 | * @sector: start sector | |
79 | * @nr_sects: number of sectors to discard | |
80 | * @gfp_mask: memory allocation flags (for bio_alloc) | |
38f25255 CH |
81 | * |
82 | * Description: | |
83 | * Issue a discard request for the sectors in question. | |
84 | */ | |
85 | int blkdev_issue_discard(struct block_device *bdev, sector_t sector, | |
44abff2c | 86 | sector_t nr_sects, gfp_t gfp_mask) |
38f25255 | 87 | { |
38f25255 CH |
88 | struct bio *bio = NULL; |
89 | struct blk_plug plug; | |
90 | int ret; | |
91 | ||
38f25255 | 92 | blk_start_plug(&plug); |
44abff2c | 93 | ret = __blkdev_issue_discard(bdev, sector, nr_sects, gfp_mask, &bio); |
bbd848e0 | 94 | if (!ret && bio) { |
4e49ea4a | 95 | ret = submit_bio_wait(bio); |
48920ff2 | 96 | if (ret == -EOPNOTSUPP) |
bbd848e0 | 97 | ret = 0; |
05bd92dd | 98 | bio_put(bio); |
bbd848e0 | 99 | } |
0cfbcafc | 100 | blk_finish_plug(&plug); |
f31e7e40 | 101 | |
bbd848e0 | 102 | return ret; |
f31e7e40 DM |
103 | } |
104 | EXPORT_SYMBOL(blkdev_issue_discard); | |
3f14d792 | 105 | |
73a768d5 CH |
106 | static sector_t bio_write_zeroes_limit(struct block_device *bdev) |
107 | { | |
108 | sector_t bs_mask = (bdev_logical_block_size(bdev) >> 9) - 1; | |
109 | ||
110 | return min(bdev_write_zeroes_sectors(bdev), | |
111 | (UINT_MAX >> SECTOR_SHIFT) & ~bs_mask); | |
112 | } | |
113 | ||
64b582ca JG |
114 | /* |
115 | * There is no reliable way for the SCSI subsystem to determine whether a | |
116 | * device supports a WRITE SAME operation without actually performing a write | |
117 | * to media. As a result, write_zeroes is enabled by default and will be | |
118 | * disabled if a zeroing operation subsequently fails. This means that this | |
119 | * queue limit is likely to change at runtime. | |
120 | */ | |
f6eacb26 | 121 | static void __blkdev_issue_write_zeroes(struct block_device *bdev, |
a6f0788e | 122 | sector_t sector, sector_t nr_sects, gfp_t gfp_mask, |
64b582ca | 123 | struct bio **biop, unsigned flags, sector_t limit) |
a6f0788e | 124 | { |
64b582ca | 125 | |
a6f0788e | 126 | while (nr_sects) { |
64b582ca | 127 | unsigned int len = min(nr_sects, limit); |
bf86bcdb CH |
128 | struct bio *bio; |
129 | ||
130 | if ((flags & BLKDEV_ZERO_KILLABLE) && | |
131 | fatal_signal_pending(current)) | |
132 | break; | |
76a27e1b | 133 | |
bf86bcdb | 134 | bio = bio_alloc(bdev, 0, REQ_OP_WRITE_ZEROES, gfp_mask); |
a6f0788e | 135 | bio->bi_iter.bi_sector = sector; |
d928be9f CH |
136 | if (flags & BLKDEV_ZERO_NOUNMAP) |
137 | bio->bi_opf |= REQ_NOUNMAP; | |
a6f0788e | 138 | |
76a27e1b | 139 | bio->bi_iter.bi_size = len << SECTOR_SHIFT; |
bf86bcdb CH |
140 | *biop = bio_chain_and_submit(*biop, bio); |
141 | ||
76a27e1b KB |
142 | nr_sects -= len; |
143 | sector += len; | |
a6f0788e CK |
144 | cond_resched(); |
145 | } | |
a6f0788e CK |
146 | } |
147 | ||
99800ced CH |
148 | static int blkdev_issue_write_zeroes(struct block_device *bdev, sector_t sector, |
149 | sector_t nr_sects, gfp_t gfp, unsigned flags) | |
150 | { | |
64b582ca | 151 | sector_t limit = bio_write_zeroes_limit(bdev); |
99800ced CH |
152 | struct bio *bio = NULL; |
153 | struct blk_plug plug; | |
154 | int ret = 0; | |
155 | ||
156 | blk_start_plug(&plug); | |
64b582ca JG |
157 | __blkdev_issue_write_zeroes(bdev, sector, nr_sects, gfp, &bio, |
158 | flags, limit); | |
99800ced | 159 | if (bio) { |
bf86bcdb CH |
160 | if ((flags & BLKDEV_ZERO_KILLABLE) && |
161 | fatal_signal_pending(current)) { | |
162 | bio_await_chain(bio); | |
163 | blk_finish_plug(&plug); | |
164 | return -EINTR; | |
165 | } | |
99800ced CH |
166 | ret = submit_bio_wait(bio); |
167 | bio_put(bio); | |
168 | } | |
169 | blk_finish_plug(&plug); | |
170 | ||
171 | /* | |
172 | * For some devices there is no non-destructive way to verify whether | |
173 | * WRITE ZEROES is actually supported. These will clear the capability | |
174 | * on an I/O error, in which case we'll turn any error into | |
175 | * "not supported" here. | |
176 | */ | |
e33a97a8 | 177 | if (ret && !bdev_write_zeroes_sectors(bdev)) |
99800ced CH |
178 | return -EOPNOTSUPP; |
179 | return ret; | |
180 | } | |
181 | ||
615d22a5 DLM |
182 | /* |
183 | * Convert a number of 512B sectors to a number of pages. | |
184 | * The result is limited to a number of pages that can fit into a BIO. | |
185 | * Also make sure that the result is always at least 1 (page) for the cases | |
186 | * where nr_sects is lower than the number of sectors in a page. | |
187 | */ | |
188 | static unsigned int __blkdev_sectors_to_bio_pages(sector_t nr_sects) | |
189 | { | |
09c2c359 | 190 | sector_t pages = DIV_ROUND_UP_SECTOR_T(nr_sects, PAGE_SIZE / 512); |
615d22a5 | 191 | |
a8affc03 | 192 | return min(pages, (sector_t)BIO_MAX_VECS); |
615d22a5 DLM |
193 | } |
194 | ||
f6eacb26 | 195 | static void __blkdev_issue_zero_pages(struct block_device *bdev, |
425a4dba | 196 | sector_t sector, sector_t nr_sects, gfp_t gfp_mask, |
bf86bcdb | 197 | struct bio **biop, unsigned int flags) |
425a4dba | 198 | { |
bf86bcdb CH |
199 | while (nr_sects) { |
200 | unsigned int nr_vecs = __blkdev_sectors_to_bio_pages(nr_sects); | |
201 | struct bio *bio; | |
425a4dba | 202 | |
bf86bcdb | 203 | bio = bio_alloc(bdev, nr_vecs, REQ_OP_WRITE, gfp_mask); |
425a4dba | 204 | bio->bi_iter.bi_sector = sector; |
425a4dba | 205 | |
bf86bcdb CH |
206 | if ((flags & BLKDEV_ZERO_KILLABLE) && |
207 | fatal_signal_pending(current)) | |
208 | break; | |
209 | ||
210 | do { | |
211 | unsigned int len, added; | |
212 | ||
213 | len = min_t(sector_t, | |
214 | PAGE_SIZE, nr_sects << SECTOR_SHIFT); | |
215 | added = bio_add_page(bio, ZERO_PAGE(0), len, 0); | |
216 | if (added < len) | |
425a4dba | 217 | break; |
bf86bcdb CH |
218 | nr_sects -= added >> SECTOR_SHIFT; |
219 | sector += added >> SECTOR_SHIFT; | |
220 | } while (nr_sects); | |
221 | ||
222 | *biop = bio_chain_and_submit(*biop, bio); | |
425a4dba ID |
223 | cond_resched(); |
224 | } | |
425a4dba ID |
225 | } |
226 | ||
99800ced CH |
227 | static int blkdev_issue_zero_pages(struct block_device *bdev, sector_t sector, |
228 | sector_t nr_sects, gfp_t gfp, unsigned flags) | |
229 | { | |
230 | struct bio *bio = NULL; | |
231 | struct blk_plug plug; | |
232 | int ret = 0; | |
233 | ||
234 | if (flags & BLKDEV_ZERO_NOFALLBACK) | |
235 | return -EOPNOTSUPP; | |
236 | ||
237 | blk_start_plug(&plug); | |
bf86bcdb | 238 | __blkdev_issue_zero_pages(bdev, sector, nr_sects, gfp, &bio, flags); |
99800ced | 239 | if (bio) { |
bf86bcdb CH |
240 | if ((flags & BLKDEV_ZERO_KILLABLE) && |
241 | fatal_signal_pending(current)) { | |
242 | bio_await_chain(bio); | |
243 | blk_finish_plug(&plug); | |
244 | return -EINTR; | |
245 | } | |
99800ced CH |
246 | ret = submit_bio_wait(bio); |
247 | bio_put(bio); | |
248 | } | |
249 | blk_finish_plug(&plug); | |
250 | ||
251 | return ret; | |
252 | } | |
253 | ||
3f14d792 | 254 | /** |
e73c23ff | 255 | * __blkdev_issue_zeroout - generate number of zero filed write bios |
3f14d792 DM |
256 | * @bdev: blockdev to issue |
257 | * @sector: start sector | |
258 | * @nr_sects: number of sectors to write | |
259 | * @gfp_mask: memory allocation flags (for bio_alloc) | |
e73c23ff | 260 | * @biop: pointer to anchor bio |
ee472d83 | 261 | * @flags: controls detailed behavior |
3f14d792 DM |
262 | * |
263 | * Description: | |
ee472d83 CH |
264 | * Zero-fill a block range, either using hardware offload or by explicitly |
265 | * writing zeroes to the device. | |
266 | * | |
267 | * If a device is using logical block provisioning, the underlying space will | |
268 | * not be released if %flags contains BLKDEV_ZERO_NOUNMAP. | |
cb365b96 CH |
269 | * |
270 | * If %flags contains BLKDEV_ZERO_NOFALLBACK, the function will return | |
271 | * -EOPNOTSUPP if no explicit hardware offload for zeroing is provided. | |
3f14d792 | 272 | */ |
e73c23ff CK |
273 | int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, |
274 | sector_t nr_sects, gfp_t gfp_mask, struct bio **biop, | |
ee472d83 | 275 | unsigned flags) |
3f14d792 | 276 | { |
64b582ca JG |
277 | sector_t limit = bio_write_zeroes_limit(bdev); |
278 | ||
f6eacb26 CH |
279 | if (bdev_read_only(bdev)) |
280 | return -EPERM; | |
3f14d792 | 281 | |
64b582ca | 282 | if (limit) { |
f6eacb26 | 283 | __blkdev_issue_write_zeroes(bdev, sector, nr_sects, |
64b582ca | 284 | gfp_mask, biop, flags, limit); |
f6eacb26 CH |
285 | } else { |
286 | if (flags & BLKDEV_ZERO_NOFALLBACK) | |
287 | return -EOPNOTSUPP; | |
288 | __blkdev_issue_zero_pages(bdev, sector, nr_sects, gfp_mask, | |
bf86bcdb | 289 | biop, flags); |
f6eacb26 CH |
290 | } |
291 | return 0; | |
3f14d792 | 292 | } |
e73c23ff | 293 | EXPORT_SYMBOL(__blkdev_issue_zeroout); |
579e8f3c MP |
294 | |
295 | /** | |
296 | * blkdev_issue_zeroout - zero-fill a block range | |
297 | * @bdev: blockdev to write | |
298 | * @sector: start sector | |
299 | * @nr_sects: number of sectors to write | |
300 | * @gfp_mask: memory allocation flags (for bio_alloc) | |
ee472d83 | 301 | * @flags: controls detailed behavior |
579e8f3c MP |
302 | * |
303 | * Description: | |
ee472d83 CH |
304 | * Zero-fill a block range, either using hardware offload or by explicitly |
305 | * writing zeroes to the device. See __blkdev_issue_zeroout() for the | |
306 | * valid values for %flags. | |
579e8f3c | 307 | */ |
579e8f3c | 308 | int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, |
ee472d83 | 309 | sector_t nr_sects, gfp_t gfp_mask, unsigned flags) |
579e8f3c | 310 | { |
99800ced | 311 | int ret; |
d93ba7a5 | 312 | |
99800ced | 313 | if ((sector | nr_sects) & ((bdev_logical_block_size(bdev) >> 9) - 1)) |
d5ce4c31 | 314 | return -EINVAL; |
f6eacb26 CH |
315 | if (bdev_read_only(bdev)) |
316 | return -EPERM; | |
d5ce4c31 | 317 | |
99800ced CH |
318 | if (bdev_write_zeroes_sectors(bdev)) { |
319 | ret = blkdev_issue_write_zeroes(bdev, sector, nr_sects, | |
320 | gfp_mask, flags); | |
39722a2f | 321 | if (ret != -EOPNOTSUPP) |
99800ced | 322 | return ret; |
d5ce4c31 | 323 | } |
579e8f3c | 324 | |
99800ced | 325 | return blkdev_issue_zero_pages(bdev, sector, nr_sects, gfp_mask, flags); |
579e8f3c | 326 | } |
3f14d792 | 327 | EXPORT_SYMBOL(blkdev_issue_zeroout); |
44abff2c CH |
328 | |
329 | int blkdev_issue_secure_erase(struct block_device *bdev, sector_t sector, | |
330 | sector_t nr_sects, gfp_t gfp) | |
331 | { | |
332 | sector_t bs_mask = (bdev_logical_block_size(bdev) >> 9) - 1; | |
333 | unsigned int max_sectors = bdev_max_secure_erase_sectors(bdev); | |
334 | struct bio *bio = NULL; | |
335 | struct blk_plug plug; | |
336 | int ret = 0; | |
337 | ||
c4fa3684 MP |
338 | /* make sure that "len << SECTOR_SHIFT" doesn't overflow */ |
339 | if (max_sectors > UINT_MAX >> SECTOR_SHIFT) | |
340 | max_sectors = UINT_MAX >> SECTOR_SHIFT; | |
341 | max_sectors &= ~bs_mask; | |
342 | ||
44abff2c CH |
343 | if (max_sectors == 0) |
344 | return -EOPNOTSUPP; | |
345 | if ((sector | nr_sects) & bs_mask) | |
346 | return -EINVAL; | |
347 | if (bdev_read_only(bdev)) | |
348 | return -EPERM; | |
349 | ||
350 | blk_start_plug(&plug); | |
5affe497 | 351 | while (nr_sects) { |
44abff2c CH |
352 | unsigned int len = min_t(sector_t, nr_sects, max_sectors); |
353 | ||
354 | bio = blk_next_bio(bio, bdev, 0, REQ_OP_SECURE_ERASE, gfp); | |
355 | bio->bi_iter.bi_sector = sector; | |
c4fa3684 | 356 | bio->bi_iter.bi_size = len << SECTOR_SHIFT; |
44abff2c | 357 | |
c4fa3684 MP |
358 | sector += len; |
359 | nr_sects -= len; | |
44abff2c CH |
360 | cond_resched(); |
361 | } | |
5affe497 KB |
362 | if (bio) { |
363 | ret = submit_bio_wait(bio); | |
364 | bio_put(bio); | |
365 | } | |
44abff2c CH |
366 | blk_finish_plug(&plug); |
367 | ||
368 | return ret; | |
369 | } | |
370 | EXPORT_SYMBOL(blkdev_issue_secure_erase); |