Commit | Line | Data |
---|---|---|
3dcf60bc | 1 | // SPDX-License-Identifier: GPL-2.0 |
86db1e29 JA |
2 | /* |
3 | * Functions related to setting various queue properties from drivers | |
4 | */ | |
5 | #include <linux/kernel.h> | |
6 | #include <linux/module.h> | |
7 | #include <linux/init.h> | |
8 | #include <linux/bio.h> | |
9 | #include <linux/blkdev.h> | |
4ee60ec1 | 10 | #include <linux/pagemap.h> |
edb0872f | 11 | #include <linux/backing-dev-defs.h> |
70dd5bf3 | 12 | #include <linux/gcd.h> |
2cda2728 | 13 | #include <linux/lcm.h> |
ad5ebd2f | 14 | #include <linux/jiffies.h> |
5a0e3ad6 | 15 | #include <linux/gfp.h> |
45147fb5 | 16 | #include <linux/dma-mapping.h> |
86db1e29 JA |
17 | |
18 | #include "blk.h" | |
0bc65bd4 | 19 | #include "blk-rq-qos.h" |
87760e5e | 20 | #include "blk-wbt.h" |
86db1e29 | 21 | |
242f9dcb JA |
22 | void blk_queue_rq_timeout(struct request_queue *q, unsigned int timeout) |
23 | { | |
24 | q->rq_timeout = timeout; | |
25 | } | |
26 | EXPORT_SYMBOL_GPL(blk_queue_rq_timeout); | |
27 | ||
e475bba2 MP |
28 | /** |
29 | * blk_set_default_limits - reset limits to default values | |
f740f5ca | 30 | * @lim: the queue_limits structure to reset |
e475bba2 MP |
31 | * |
32 | * Description: | |
b1bd055d | 33 | * Returns a queue_limit struct to its default state. |
e475bba2 MP |
34 | */ |
35 | void blk_set_default_limits(struct queue_limits *lim) | |
36 | { | |
8a78362c | 37 | lim->max_segments = BLK_MAX_SEGMENTS; |
1e739730 | 38 | lim->max_discard_segments = 1; |
13f05c8d | 39 | lim->max_integrity_segments = 0; |
e475bba2 | 40 | lim->seg_boundary_mask = BLK_SEG_BOUNDARY_MASK; |
03100aad | 41 | lim->virt_boundary_mask = 0; |
eb28d31b | 42 | lim->max_segment_size = BLK_MAX_SEGMENT_SIZE; |
5f009d3f | 43 | lim->max_sectors = lim->max_hw_sectors = BLK_SAFE_MAX_SECTORS; |
c9c77418 | 44 | lim->max_user_sectors = lim->max_dev_sectors = 0; |
762380ad | 45 | lim->chunk_sectors = 0; |
a6f0788e | 46 | lim->max_write_zeroes_sectors = 0; |
0512a75b | 47 | lim->max_zone_append_sectors = 0; |
86b37281 | 48 | lim->max_discard_sectors = 0; |
0034af03 | 49 | lim->max_hw_discard_sectors = 0; |
44abff2c | 50 | lim->max_secure_erase_sectors = 0; |
3c407dc7 | 51 | lim->discard_granularity = 512; |
86b37281 MP |
52 | lim->discard_alignment = 0; |
53 | lim->discard_misaligned = 0; | |
e475bba2 | 54 | lim->logical_block_size = lim->physical_block_size = lim->io_min = 512; |
9bb33f24 | 55 | lim->bounce = BLK_BOUNCE_NONE; |
e475bba2 MP |
56 | lim->alignment_offset = 0; |
57 | lim->io_opt = 0; | |
58 | lim->misaligned = 0; | |
7437bb73 | 59 | lim->zoned = false; |
a805a4fa | 60 | lim->zone_write_granularity = 0; |
c964d62f | 61 | lim->dma_alignment = 511; |
e475bba2 | 62 | } |
e475bba2 | 63 | |
b1bd055d MP |
64 | /** |
65 | * blk_set_stacking_limits - set default limits for stacking devices | |
66 | * @lim: the queue_limits structure to reset | |
67 | * | |
68 | * Description: | |
69 | * Returns a queue_limit struct to its default state. Should be used | |
70 | * by stacking drivers like DM that have no internal limits. | |
71 | */ | |
72 | void blk_set_stacking_limits(struct queue_limits *lim) | |
73 | { | |
74 | blk_set_default_limits(lim); | |
75 | ||
76 | /* Inherit limits from component devices */ | |
b1bd055d | 77 | lim->max_segments = USHRT_MAX; |
42c9cdfe | 78 | lim->max_discard_segments = USHRT_MAX; |
b1bd055d | 79 | lim->max_hw_sectors = UINT_MAX; |
d82ae52e | 80 | lim->max_segment_size = UINT_MAX; |
fe86cdce | 81 | lim->max_sectors = UINT_MAX; |
ca369d51 | 82 | lim->max_dev_sectors = UINT_MAX; |
a6f0788e | 83 | lim->max_write_zeroes_sectors = UINT_MAX; |
0512a75b | 84 | lim->max_zone_append_sectors = UINT_MAX; |
b1bd055d MP |
85 | } |
86 | EXPORT_SYMBOL(blk_set_stacking_limits); | |
87 | ||
86db1e29 JA |
88 | /** |
89 | * blk_queue_bounce_limit - set bounce buffer limit for queue | |
cd0aca2d | 90 | * @q: the request queue for the device |
9bb33f24 | 91 | * @bounce: bounce limit to enforce |
86db1e29 JA |
92 | * |
93 | * Description: | |
9bb33f24 CH |
94 | * Force bouncing for ISA DMA ranges or highmem. |
95 | * | |
96 | * DEPRECATED, don't use in new code. | |
86db1e29 | 97 | **/ |
9bb33f24 | 98 | void blk_queue_bounce_limit(struct request_queue *q, enum blk_bounce bounce) |
86db1e29 | 99 | { |
9bb33f24 | 100 | q->limits.bounce = bounce; |
86db1e29 | 101 | } |
86db1e29 JA |
102 | EXPORT_SYMBOL(blk_queue_bounce_limit); |
103 | ||
104 | /** | |
ca369d51 MP |
105 | * blk_queue_max_hw_sectors - set max sectors for a request for this queue |
106 | * @q: the request queue for the device | |
2800aac1 | 107 | * @max_hw_sectors: max hardware sectors in the usual 512b unit |
86db1e29 JA |
108 | * |
109 | * Description: | |
2800aac1 MP |
110 | * Enables a low level driver to set a hard upper limit, |
111 | * max_hw_sectors, on the size of requests. max_hw_sectors is set by | |
4f258a46 MP |
112 | * the device driver based upon the capabilities of the I/O |
113 | * controller. | |
2800aac1 | 114 | * |
ca369d51 MP |
115 | * max_dev_sectors is a hard limit imposed by the storage device for |
116 | * READ/WRITE requests. It is set by the disk driver. | |
117 | * | |
2800aac1 MP |
118 | * max_sectors is a soft limit imposed by the block layer for |
119 | * filesystem type requests. This value can be overridden on a | |
120 | * per-device basis in /sys/block/<device>/queue/max_sectors_kb. | |
121 | * The soft limit can not exceed max_hw_sectors. | |
86db1e29 | 122 | **/ |
ca369d51 | 123 | void blk_queue_max_hw_sectors(struct request_queue *q, unsigned int max_hw_sectors) |
86db1e29 | 124 | { |
ca369d51 MP |
125 | struct queue_limits *limits = &q->limits; |
126 | unsigned int max_sectors; | |
127 | ||
09cbfeaf KS |
128 | if ((max_hw_sectors << 9) < PAGE_SIZE) { |
129 | max_hw_sectors = 1 << (PAGE_SHIFT - 9); | |
f19d1e3b | 130 | pr_info("%s: set to minimum %u\n", __func__, max_hw_sectors); |
86db1e29 JA |
131 | } |
132 | ||
817046ec DLM |
133 | max_hw_sectors = round_down(max_hw_sectors, |
134 | limits->logical_block_size >> SECTOR_SHIFT); | |
30e2bc08 | 135 | limits->max_hw_sectors = max_hw_sectors; |
817046ec | 136 | |
ca369d51 | 137 | max_sectors = min_not_zero(max_hw_sectors, limits->max_dev_sectors); |
c9c77418 KB |
138 | |
139 | if (limits->max_user_sectors) | |
140 | max_sectors = min(max_sectors, limits->max_user_sectors); | |
141 | else | |
d6b9f4e6 | 142 | max_sectors = min(max_sectors, BLK_DEF_MAX_SECTORS_CAP); |
c9c77418 | 143 | |
817046ec DLM |
144 | max_sectors = round_down(max_sectors, |
145 | limits->logical_block_size >> SECTOR_SHIFT); | |
ca369d51 | 146 | limits->max_sectors = max_sectors; |
817046ec | 147 | |
d152c682 | 148 | if (!q->disk) |
edb0872f | 149 | return; |
d152c682 | 150 | q->disk->bdi->io_pages = max_sectors >> (PAGE_SHIFT - 9); |
86db1e29 | 151 | } |
086fa5ff | 152 | EXPORT_SYMBOL(blk_queue_max_hw_sectors); |
86db1e29 | 153 | |
762380ad JA |
154 | /** |
155 | * blk_queue_chunk_sectors - set size of the chunk for this queue | |
156 | * @q: the request queue for the device | |
157 | * @chunk_sectors: chunk sectors in the usual 512b unit | |
158 | * | |
159 | * Description: | |
160 | * If a driver doesn't want IOs to cross a given chunk size, it can set | |
07d098e6 MS |
161 | * this limit and prevent merging across chunks. Note that the block layer |
162 | * must accept a page worth of data at any offset. So if the crossing of | |
163 | * chunks is a hard limitation in the driver, it must still be prepared | |
164 | * to split single page bios. | |
762380ad JA |
165 | **/ |
166 | void blk_queue_chunk_sectors(struct request_queue *q, unsigned int chunk_sectors) | |
167 | { | |
762380ad JA |
168 | q->limits.chunk_sectors = chunk_sectors; |
169 | } | |
170 | EXPORT_SYMBOL(blk_queue_chunk_sectors); | |
171 | ||
67efc925 CH |
172 | /** |
173 | * blk_queue_max_discard_sectors - set max sectors for a single discard | |
174 | * @q: the request queue for the device | |
c7ebf065 | 175 | * @max_discard_sectors: maximum number of sectors to discard |
67efc925 CH |
176 | **/ |
177 | void blk_queue_max_discard_sectors(struct request_queue *q, | |
178 | unsigned int max_discard_sectors) | |
179 | { | |
0034af03 | 180 | q->limits.max_hw_discard_sectors = max_discard_sectors; |
67efc925 CH |
181 | q->limits.max_discard_sectors = max_discard_sectors; |
182 | } | |
183 | EXPORT_SYMBOL(blk_queue_max_discard_sectors); | |
184 | ||
44abff2c CH |
185 | /** |
186 | * blk_queue_max_secure_erase_sectors - set max sectors for a secure erase | |
187 | * @q: the request queue for the device | |
188 | * @max_sectors: maximum number of sectors to secure_erase | |
189 | **/ | |
190 | void blk_queue_max_secure_erase_sectors(struct request_queue *q, | |
191 | unsigned int max_sectors) | |
192 | { | |
193 | q->limits.max_secure_erase_sectors = max_sectors; | |
194 | } | |
195 | EXPORT_SYMBOL(blk_queue_max_secure_erase_sectors); | |
196 | ||
a6f0788e CK |
197 | /** |
198 | * blk_queue_max_write_zeroes_sectors - set max sectors for a single | |
199 | * write zeroes | |
200 | * @q: the request queue for the device | |
201 | * @max_write_zeroes_sectors: maximum number of sectors to write per command | |
202 | **/ | |
203 | void blk_queue_max_write_zeroes_sectors(struct request_queue *q, | |
204 | unsigned int max_write_zeroes_sectors) | |
205 | { | |
206 | q->limits.max_write_zeroes_sectors = max_write_zeroes_sectors; | |
207 | } | |
208 | EXPORT_SYMBOL(blk_queue_max_write_zeroes_sectors); | |
209 | ||
0512a75b KB |
210 | /** |
211 | * blk_queue_max_zone_append_sectors - set max sectors for a single zone append | |
212 | * @q: the request queue for the device | |
213 | * @max_zone_append_sectors: maximum number of sectors to write per command | |
214 | **/ | |
215 | void blk_queue_max_zone_append_sectors(struct request_queue *q, | |
216 | unsigned int max_zone_append_sectors) | |
217 | { | |
218 | unsigned int max_sectors; | |
219 | ||
220 | if (WARN_ON(!blk_queue_is_zoned(q))) | |
221 | return; | |
222 | ||
223 | max_sectors = min(q->limits.max_hw_sectors, max_zone_append_sectors); | |
224 | max_sectors = min(q->limits.chunk_sectors, max_sectors); | |
225 | ||
226 | /* | |
227 | * Signal eventual driver bugs resulting in the max_zone_append sectors limit | |
228 | * being 0 due to a 0 argument, the chunk_sectors limit (zone size) not set, | |
229 | * or the max_hw_sectors limit not set. | |
230 | */ | |
231 | WARN_ON(!max_sectors); | |
232 | ||
233 | q->limits.max_zone_append_sectors = max_sectors; | |
234 | } | |
235 | EXPORT_SYMBOL_GPL(blk_queue_max_zone_append_sectors); | |
236 | ||
86db1e29 | 237 | /** |
8a78362c | 238 | * blk_queue_max_segments - set max hw segments for a request for this queue |
86db1e29 JA |
239 | * @q: the request queue for the device |
240 | * @max_segments: max number of segments | |
241 | * | |
242 | * Description: | |
243 | * Enables a low level driver to set an upper limit on the number of | |
8a78362c | 244 | * hw data segments in a request. |
86db1e29 | 245 | **/ |
8a78362c | 246 | void blk_queue_max_segments(struct request_queue *q, unsigned short max_segments) |
86db1e29 JA |
247 | { |
248 | if (!max_segments) { | |
249 | max_segments = 1; | |
f19d1e3b | 250 | pr_info("%s: set to minimum %u\n", __func__, max_segments); |
86db1e29 JA |
251 | } |
252 | ||
8a78362c | 253 | q->limits.max_segments = max_segments; |
86db1e29 | 254 | } |
8a78362c | 255 | EXPORT_SYMBOL(blk_queue_max_segments); |
86db1e29 | 256 | |
1e739730 CH |
257 | /** |
258 | * blk_queue_max_discard_segments - set max segments for discard requests | |
259 | * @q: the request queue for the device | |
260 | * @max_segments: max number of segments | |
261 | * | |
262 | * Description: | |
263 | * Enables a low level driver to set an upper limit on the number of | |
264 | * segments in a discard request. | |
265 | **/ | |
266 | void blk_queue_max_discard_segments(struct request_queue *q, | |
267 | unsigned short max_segments) | |
268 | { | |
269 | q->limits.max_discard_segments = max_segments; | |
270 | } | |
271 | EXPORT_SYMBOL_GPL(blk_queue_max_discard_segments); | |
272 | ||
86db1e29 JA |
273 | /** |
274 | * blk_queue_max_segment_size - set max segment size for blk_rq_map_sg | |
275 | * @q: the request queue for the device | |
276 | * @max_size: max size of segment in bytes | |
277 | * | |
278 | * Description: | |
279 | * Enables a low level driver to set an upper limit on the size of a | |
280 | * coalesced segment | |
281 | **/ | |
282 | void blk_queue_max_segment_size(struct request_queue *q, unsigned int max_size) | |
283 | { | |
09cbfeaf KS |
284 | if (max_size < PAGE_SIZE) { |
285 | max_size = PAGE_SIZE; | |
f19d1e3b | 286 | pr_info("%s: set to minimum %u\n", __func__, max_size); |
86db1e29 JA |
287 | } |
288 | ||
09324d32 CH |
289 | /* see blk_queue_virt_boundary() for the explanation */ |
290 | WARN_ON_ONCE(q->limits.virt_boundary_mask); | |
291 | ||
025146e1 | 292 | q->limits.max_segment_size = max_size; |
86db1e29 | 293 | } |
86db1e29 JA |
294 | EXPORT_SYMBOL(blk_queue_max_segment_size); |
295 | ||
296 | /** | |
e1defc4f | 297 | * blk_queue_logical_block_size - set logical block size for the queue |
86db1e29 | 298 | * @q: the request queue for the device |
e1defc4f | 299 | * @size: the logical block size, in bytes |
86db1e29 JA |
300 | * |
301 | * Description: | |
e1defc4f MP |
302 | * This should be set to the lowest possible block size that the |
303 | * storage device can address. The default of 512 covers most | |
304 | * hardware. | |
86db1e29 | 305 | **/ |
ad6bf88a | 306 | void blk_queue_logical_block_size(struct request_queue *q, unsigned int size) |
86db1e29 | 307 | { |
817046ec DLM |
308 | struct queue_limits *limits = &q->limits; |
309 | ||
310 | limits->logical_block_size = size; | |
311 | ||
3c407dc7 CH |
312 | if (limits->discard_granularity < limits->logical_block_size) |
313 | limits->discard_granularity = limits->logical_block_size; | |
314 | ||
817046ec DLM |
315 | if (limits->physical_block_size < size) |
316 | limits->physical_block_size = size; | |
c72758f3 | 317 | |
817046ec DLM |
318 | if (limits->io_min < limits->physical_block_size) |
319 | limits->io_min = limits->physical_block_size; | |
c72758f3 | 320 | |
817046ec DLM |
321 | limits->max_hw_sectors = |
322 | round_down(limits->max_hw_sectors, size >> SECTOR_SHIFT); | |
323 | limits->max_sectors = | |
324 | round_down(limits->max_sectors, size >> SECTOR_SHIFT); | |
86db1e29 | 325 | } |
e1defc4f | 326 | EXPORT_SYMBOL(blk_queue_logical_block_size); |
86db1e29 | 327 | |
c72758f3 MP |
328 | /** |
329 | * blk_queue_physical_block_size - set physical block size for the queue | |
330 | * @q: the request queue for the device | |
331 | * @size: the physical block size, in bytes | |
332 | * | |
333 | * Description: | |
334 | * This should be set to the lowest possible sector size that the | |
335 | * hardware can operate on without reverting to read-modify-write | |
336 | * operations. | |
337 | */ | |
892b6f90 | 338 | void blk_queue_physical_block_size(struct request_queue *q, unsigned int size) |
c72758f3 MP |
339 | { |
340 | q->limits.physical_block_size = size; | |
341 | ||
342 | if (q->limits.physical_block_size < q->limits.logical_block_size) | |
343 | q->limits.physical_block_size = q->limits.logical_block_size; | |
344 | ||
458aa1a0 CH |
345 | if (q->limits.discard_granularity < q->limits.physical_block_size) |
346 | q->limits.discard_granularity = q->limits.physical_block_size; | |
347 | ||
c72758f3 MP |
348 | if (q->limits.io_min < q->limits.physical_block_size) |
349 | q->limits.io_min = q->limits.physical_block_size; | |
350 | } | |
351 | EXPORT_SYMBOL(blk_queue_physical_block_size); | |
352 | ||
a805a4fa DLM |
353 | /** |
354 | * blk_queue_zone_write_granularity - set zone write granularity for the queue | |
355 | * @q: the request queue for the zoned device | |
356 | * @size: the zone write granularity size, in bytes | |
357 | * | |
358 | * Description: | |
359 | * This should be set to the lowest possible size allowing to write in | |
360 | * sequential zones of a zoned block device. | |
361 | */ | |
362 | void blk_queue_zone_write_granularity(struct request_queue *q, | |
363 | unsigned int size) | |
364 | { | |
365 | if (WARN_ON_ONCE(!blk_queue_is_zoned(q))) | |
366 | return; | |
367 | ||
368 | q->limits.zone_write_granularity = size; | |
369 | ||
370 | if (q->limits.zone_write_granularity < q->limits.logical_block_size) | |
371 | q->limits.zone_write_granularity = q->limits.logical_block_size; | |
372 | } | |
373 | EXPORT_SYMBOL_GPL(blk_queue_zone_write_granularity); | |
374 | ||
c72758f3 MP |
375 | /** |
376 | * blk_queue_alignment_offset - set physical block alignment offset | |
377 | * @q: the request queue for the device | |
8ebf9756 | 378 | * @offset: alignment offset in bytes |
c72758f3 MP |
379 | * |
380 | * Description: | |
381 | * Some devices are naturally misaligned to compensate for things like | |
382 | * the legacy DOS partition table 63-sector offset. Low-level drivers | |
383 | * should call this function for devices whose first sector is not | |
384 | * naturally aligned. | |
385 | */ | |
386 | void blk_queue_alignment_offset(struct request_queue *q, unsigned int offset) | |
387 | { | |
388 | q->limits.alignment_offset = | |
389 | offset & (q->limits.physical_block_size - 1); | |
390 | q->limits.misaligned = 0; | |
391 | } | |
392 | EXPORT_SYMBOL(blk_queue_alignment_offset); | |
393 | ||
471aa704 | 394 | void disk_update_readahead(struct gendisk *disk) |
c2e4cd57 | 395 | { |
471aa704 CH |
396 | struct request_queue *q = disk->queue; |
397 | ||
c2e4cd57 CH |
398 | /* |
399 | * For read-ahead of large files to be effective, we need to read ahead | |
400 | * at least twice the optimal I/O size. | |
401 | */ | |
edb0872f | 402 | disk->bdi->ra_pages = |
c2e4cd57 | 403 | max(queue_io_opt(q) * 2 / PAGE_SIZE, VM_READAHEAD_PAGES); |
edb0872f | 404 | disk->bdi->io_pages = queue_max_sectors(q) >> (PAGE_SHIFT - 9); |
c2e4cd57 | 405 | } |
471aa704 | 406 | EXPORT_SYMBOL_GPL(disk_update_readahead); |
c2e4cd57 | 407 | |
7c958e32 MP |
408 | /** |
409 | * blk_limits_io_min - set minimum request size for a device | |
410 | * @limits: the queue limits | |
411 | * @min: smallest I/O size in bytes | |
412 | * | |
413 | * Description: | |
414 | * Some devices have an internal block size bigger than the reported | |
415 | * hardware sector size. This function can be used to signal the | |
416 | * smallest I/O the device can perform without incurring a performance | |
417 | * penalty. | |
418 | */ | |
419 | void blk_limits_io_min(struct queue_limits *limits, unsigned int min) | |
420 | { | |
421 | limits->io_min = min; | |
422 | ||
423 | if (limits->io_min < limits->logical_block_size) | |
424 | limits->io_min = limits->logical_block_size; | |
425 | ||
426 | if (limits->io_min < limits->physical_block_size) | |
427 | limits->io_min = limits->physical_block_size; | |
428 | } | |
429 | EXPORT_SYMBOL(blk_limits_io_min); | |
430 | ||
c72758f3 MP |
431 | /** |
432 | * blk_queue_io_min - set minimum request size for the queue | |
433 | * @q: the request queue for the device | |
8ebf9756 | 434 | * @min: smallest I/O size in bytes |
c72758f3 MP |
435 | * |
436 | * Description: | |
7e5f5fb0 MP |
437 | * Storage devices may report a granularity or preferred minimum I/O |
438 | * size which is the smallest request the device can perform without | |
439 | * incurring a performance penalty. For disk drives this is often the | |
440 | * physical block size. For RAID arrays it is often the stripe chunk | |
441 | * size. A properly aligned multiple of minimum_io_size is the | |
442 | * preferred request size for workloads where a high number of I/O | |
443 | * operations is desired. | |
c72758f3 MP |
444 | */ |
445 | void blk_queue_io_min(struct request_queue *q, unsigned int min) | |
446 | { | |
7c958e32 | 447 | blk_limits_io_min(&q->limits, min); |
c72758f3 MP |
448 | } |
449 | EXPORT_SYMBOL(blk_queue_io_min); | |
450 | ||
3c5820c7 MP |
451 | /** |
452 | * blk_limits_io_opt - set optimal request size for a device | |
453 | * @limits: the queue limits | |
454 | * @opt: smallest I/O size in bytes | |
455 | * | |
456 | * Description: | |
457 | * Storage devices may report an optimal I/O size, which is the | |
458 | * device's preferred unit for sustained I/O. This is rarely reported | |
459 | * for disk drives. For RAID arrays it is usually the stripe width or | |
460 | * the internal track size. A properly aligned multiple of | |
461 | * optimal_io_size is the preferred request size for workloads where | |
462 | * sustained throughput is desired. | |
463 | */ | |
464 | void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt) | |
465 | { | |
466 | limits->io_opt = opt; | |
467 | } | |
468 | EXPORT_SYMBOL(blk_limits_io_opt); | |
469 | ||
c72758f3 MP |
470 | /** |
471 | * blk_queue_io_opt - set optimal request size for the queue | |
472 | * @q: the request queue for the device | |
8ebf9756 | 473 | * @opt: optimal request size in bytes |
c72758f3 MP |
474 | * |
475 | * Description: | |
7e5f5fb0 MP |
476 | * Storage devices may report an optimal I/O size, which is the |
477 | * device's preferred unit for sustained I/O. This is rarely reported | |
478 | * for disk drives. For RAID arrays it is usually the stripe width or | |
479 | * the internal track size. A properly aligned multiple of | |
480 | * optimal_io_size is the preferred request size for workloads where | |
481 | * sustained throughput is desired. | |
c72758f3 MP |
482 | */ |
483 | void blk_queue_io_opt(struct request_queue *q, unsigned int opt) | |
484 | { | |
3c5820c7 | 485 | blk_limits_io_opt(&q->limits, opt); |
d152c682 | 486 | if (!q->disk) |
edb0872f | 487 | return; |
d152c682 | 488 | q->disk->bdi->ra_pages = |
c2e4cd57 | 489 | max(queue_io_opt(q) * 2 / PAGE_SIZE, VM_READAHEAD_PAGES); |
c72758f3 MP |
490 | } |
491 | EXPORT_SYMBOL(blk_queue_io_opt); | |
492 | ||
aa261f20 | 493 | static int queue_limit_alignment_offset(const struct queue_limits *lim, |
89098b07 CH |
494 | sector_t sector) |
495 | { | |
496 | unsigned int granularity = max(lim->physical_block_size, lim->io_min); | |
497 | unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT) | |
498 | << SECTOR_SHIFT; | |
499 | ||
500 | return (granularity + lim->alignment_offset - alignment) % granularity; | |
501 | } | |
502 | ||
aa261f20 BVA |
503 | static unsigned int queue_limit_discard_alignment( |
504 | const struct queue_limits *lim, sector_t sector) | |
5c4b4a5c CH |
505 | { |
506 | unsigned int alignment, granularity, offset; | |
507 | ||
508 | if (!lim->max_discard_sectors) | |
509 | return 0; | |
510 | ||
511 | /* Why are these in bytes, not sectors? */ | |
512 | alignment = lim->discard_alignment >> SECTOR_SHIFT; | |
513 | granularity = lim->discard_granularity >> SECTOR_SHIFT; | |
514 | if (!granularity) | |
515 | return 0; | |
516 | ||
517 | /* Offset of the partition start in 'granularity' sectors */ | |
518 | offset = sector_div(sector, granularity); | |
519 | ||
520 | /* And why do we do this modulus *again* in blkdev_issue_discard()? */ | |
521 | offset = (granularity + alignment - offset) % granularity; | |
522 | ||
523 | /* Turn it back into bytes, gaah */ | |
524 | return offset << SECTOR_SHIFT; | |
525 | } | |
526 | ||
97f433c3 MP |
527 | static unsigned int blk_round_down_sectors(unsigned int sectors, unsigned int lbs) |
528 | { | |
529 | sectors = round_down(sectors, lbs >> SECTOR_SHIFT); | |
530 | if (sectors < PAGE_SIZE >> SECTOR_SHIFT) | |
531 | sectors = PAGE_SIZE >> SECTOR_SHIFT; | |
532 | return sectors; | |
533 | } | |
534 | ||
c72758f3 MP |
535 | /** |
536 | * blk_stack_limits - adjust queue_limits for stacked devices | |
81744ee4 MP |
537 | * @t: the stacking driver limits (top device) |
538 | * @b: the underlying queue limits (bottom, component device) | |
e03a72e1 | 539 | * @start: first data sector within component device |
c72758f3 MP |
540 | * |
541 | * Description: | |
81744ee4 MP |
542 | * This function is used by stacking drivers like MD and DM to ensure |
543 | * that all component devices have compatible block sizes and | |
544 | * alignments. The stacking driver must provide a queue_limits | |
545 | * struct (top) and then iteratively call the stacking function for | |
546 | * all component (bottom) devices. The stacking function will | |
547 | * attempt to combine the values and ensure proper alignment. | |
548 | * | |
549 | * Returns 0 if the top and bottom queue_limits are compatible. The | |
550 | * top device's block sizes and alignment offsets may be adjusted to | |
551 | * ensure alignment with the bottom device. If no compatible sizes | |
552 | * and alignments exist, -1 is returned and the resulting top | |
553 | * queue_limits will have the misaligned flag set to indicate that | |
554 | * the alignment_offset is undefined. | |
c72758f3 MP |
555 | */ |
556 | int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, | |
e03a72e1 | 557 | sector_t start) |
c72758f3 | 558 | { |
e03a72e1 | 559 | unsigned int top, bottom, alignment, ret = 0; |
86b37281 | 560 | |
c72758f3 MP |
561 | t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors); |
562 | t->max_hw_sectors = min_not_zero(t->max_hw_sectors, b->max_hw_sectors); | |
ca369d51 | 563 | t->max_dev_sectors = min_not_zero(t->max_dev_sectors, b->max_dev_sectors); |
a6f0788e CK |
564 | t->max_write_zeroes_sectors = min(t->max_write_zeroes_sectors, |
565 | b->max_write_zeroes_sectors); | |
0512a75b KB |
566 | t->max_zone_append_sectors = min(t->max_zone_append_sectors, |
567 | b->max_zone_append_sectors); | |
9bb33f24 | 568 | t->bounce = max(t->bounce, b->bounce); |
c72758f3 MP |
569 | |
570 | t->seg_boundary_mask = min_not_zero(t->seg_boundary_mask, | |
571 | b->seg_boundary_mask); | |
03100aad KB |
572 | t->virt_boundary_mask = min_not_zero(t->virt_boundary_mask, |
573 | b->virt_boundary_mask); | |
c72758f3 | 574 | |
8a78362c | 575 | t->max_segments = min_not_zero(t->max_segments, b->max_segments); |
1e739730 CH |
576 | t->max_discard_segments = min_not_zero(t->max_discard_segments, |
577 | b->max_discard_segments); | |
13f05c8d MP |
578 | t->max_integrity_segments = min_not_zero(t->max_integrity_segments, |
579 | b->max_integrity_segments); | |
c72758f3 MP |
580 | |
581 | t->max_segment_size = min_not_zero(t->max_segment_size, | |
582 | b->max_segment_size); | |
583 | ||
fe0b393f MP |
584 | t->misaligned |= b->misaligned; |
585 | ||
e03a72e1 | 586 | alignment = queue_limit_alignment_offset(b, start); |
9504e086 | 587 | |
81744ee4 MP |
588 | /* Bottom device has different alignment. Check that it is |
589 | * compatible with the current top alignment. | |
590 | */ | |
9504e086 MP |
591 | if (t->alignment_offset != alignment) { |
592 | ||
593 | top = max(t->physical_block_size, t->io_min) | |
594 | + t->alignment_offset; | |
81744ee4 | 595 | bottom = max(b->physical_block_size, b->io_min) + alignment; |
9504e086 | 596 | |
81744ee4 | 597 | /* Verify that top and bottom intervals line up */ |
b8839b8c | 598 | if (max(top, bottom) % min(top, bottom)) { |
9504e086 | 599 | t->misaligned = 1; |
fe0b393f MP |
600 | ret = -1; |
601 | } | |
9504e086 MP |
602 | } |
603 | ||
c72758f3 MP |
604 | t->logical_block_size = max(t->logical_block_size, |
605 | b->logical_block_size); | |
606 | ||
607 | t->physical_block_size = max(t->physical_block_size, | |
608 | b->physical_block_size); | |
609 | ||
610 | t->io_min = max(t->io_min, b->io_min); | |
e9637415 | 611 | t->io_opt = lcm_not_zero(t->io_opt, b->io_opt); |
c964d62f | 612 | t->dma_alignment = max(t->dma_alignment, b->dma_alignment); |
7e7986f9 MS |
613 | |
614 | /* Set non-power-of-2 compatible chunk_sectors boundary */ | |
615 | if (b->chunk_sectors) | |
616 | t->chunk_sectors = gcd(t->chunk_sectors, b->chunk_sectors); | |
9504e086 | 617 | |
81744ee4 | 618 | /* Physical block size a multiple of the logical block size? */ |
9504e086 MP |
619 | if (t->physical_block_size & (t->logical_block_size - 1)) { |
620 | t->physical_block_size = t->logical_block_size; | |
c72758f3 | 621 | t->misaligned = 1; |
fe0b393f | 622 | ret = -1; |
86b37281 MP |
623 | } |
624 | ||
81744ee4 | 625 | /* Minimum I/O a multiple of the physical block size? */ |
9504e086 MP |
626 | if (t->io_min & (t->physical_block_size - 1)) { |
627 | t->io_min = t->physical_block_size; | |
628 | t->misaligned = 1; | |
fe0b393f | 629 | ret = -1; |
c72758f3 MP |
630 | } |
631 | ||
81744ee4 | 632 | /* Optimal I/O a multiple of the physical block size? */ |
9504e086 MP |
633 | if (t->io_opt & (t->physical_block_size - 1)) { |
634 | t->io_opt = 0; | |
635 | t->misaligned = 1; | |
fe0b393f | 636 | ret = -1; |
9504e086 | 637 | } |
c72758f3 | 638 | |
22ada802 MS |
639 | /* chunk_sectors a multiple of the physical block size? */ |
640 | if ((t->chunk_sectors << 9) & (t->physical_block_size - 1)) { | |
641 | t->chunk_sectors = 0; | |
642 | t->misaligned = 1; | |
643 | ret = -1; | |
644 | } | |
645 | ||
c78afc62 KO |
646 | t->raid_partial_stripes_expensive = |
647 | max(t->raid_partial_stripes_expensive, | |
648 | b->raid_partial_stripes_expensive); | |
649 | ||
81744ee4 | 650 | /* Find lowest common alignment_offset */ |
e9637415 | 651 | t->alignment_offset = lcm_not_zero(t->alignment_offset, alignment) |
b8839b8c | 652 | % max(t->physical_block_size, t->io_min); |
86b37281 | 653 | |
81744ee4 | 654 | /* Verify that new alignment_offset is on a logical block boundary */ |
fe0b393f | 655 | if (t->alignment_offset & (t->logical_block_size - 1)) { |
c72758f3 | 656 | t->misaligned = 1; |
fe0b393f MP |
657 | ret = -1; |
658 | } | |
c72758f3 | 659 | |
97f433c3 MP |
660 | t->max_sectors = blk_round_down_sectors(t->max_sectors, t->logical_block_size); |
661 | t->max_hw_sectors = blk_round_down_sectors(t->max_hw_sectors, t->logical_block_size); | |
662 | t->max_dev_sectors = blk_round_down_sectors(t->max_dev_sectors, t->logical_block_size); | |
663 | ||
9504e086 MP |
664 | /* Discard alignment and granularity */ |
665 | if (b->discard_granularity) { | |
e03a72e1 | 666 | alignment = queue_limit_discard_alignment(b, start); |
9504e086 MP |
667 | |
668 | if (t->discard_granularity != 0 && | |
669 | t->discard_alignment != alignment) { | |
670 | top = t->discard_granularity + t->discard_alignment; | |
671 | bottom = b->discard_granularity + alignment; | |
70dd5bf3 | 672 | |
9504e086 | 673 | /* Verify that top and bottom intervals line up */ |
8dd2cb7e | 674 | if ((max(top, bottom) % min(top, bottom)) != 0) |
9504e086 MP |
675 | t->discard_misaligned = 1; |
676 | } | |
677 | ||
81744ee4 MP |
678 | t->max_discard_sectors = min_not_zero(t->max_discard_sectors, |
679 | b->max_discard_sectors); | |
0034af03 JA |
680 | t->max_hw_discard_sectors = min_not_zero(t->max_hw_discard_sectors, |
681 | b->max_hw_discard_sectors); | |
9504e086 MP |
682 | t->discard_granularity = max(t->discard_granularity, |
683 | b->discard_granularity); | |
e9637415 | 684 | t->discard_alignment = lcm_not_zero(t->discard_alignment, alignment) % |
8dd2cb7e | 685 | t->discard_granularity; |
9504e086 | 686 | } |
44abff2c CH |
687 | t->max_secure_erase_sectors = min_not_zero(t->max_secure_erase_sectors, |
688 | b->max_secure_erase_sectors); | |
a805a4fa DLM |
689 | t->zone_write_granularity = max(t->zone_write_granularity, |
690 | b->zone_write_granularity); | |
3093a479 | 691 | t->zoned = max(t->zoned, b->zoned); |
fe0b393f | 692 | return ret; |
c72758f3 | 693 | } |
5d85d324 | 694 | EXPORT_SYMBOL(blk_stack_limits); |
c72758f3 MP |
695 | |
696 | /** | |
697 | * disk_stack_limits - adjust queue limits for stacked drivers | |
77634f33 | 698 | * @disk: MD/DM gendisk (top) |
c72758f3 MP |
699 | * @bdev: the underlying block device (bottom) |
700 | * @offset: offset to beginning of data within component device | |
701 | * | |
702 | * Description: | |
e03a72e1 MP |
703 | * Merges the limits for a top level gendisk and a bottom level |
704 | * block_device. | |
c72758f3 MP |
705 | */ |
706 | void disk_stack_limits(struct gendisk *disk, struct block_device *bdev, | |
707 | sector_t offset) | |
708 | { | |
709 | struct request_queue *t = disk->queue; | |
c72758f3 | 710 | |
9efa82ef | 711 | if (blk_stack_limits(&t->limits, &bdev_get_queue(bdev)->limits, |
453b8ab6 CH |
712 | get_start_sect(bdev) + (offset >> 9)) < 0) |
713 | pr_notice("%s: Warning: Device %pg is misaligned\n", | |
714 | disk->disk_name, bdev); | |
e74d93e9 | 715 | |
471aa704 | 716 | disk_update_readahead(disk); |
c72758f3 MP |
717 | } |
718 | EXPORT_SYMBOL(disk_stack_limits); | |
719 | ||
27f8221a FT |
720 | /** |
721 | * blk_queue_update_dma_pad - update pad mask | |
722 | * @q: the request queue for the device | |
723 | * @mask: pad mask | |
724 | * | |
725 | * Update dma pad mask. | |
726 | * | |
727 | * Appending pad buffer to a request modifies the last entry of a | |
728 | * scatter list such that it includes the pad buffer. | |
729 | **/ | |
730 | void blk_queue_update_dma_pad(struct request_queue *q, unsigned int mask) | |
731 | { | |
732 | if (mask > q->dma_pad_mask) | |
733 | q->dma_pad_mask = mask; | |
734 | } | |
735 | EXPORT_SYMBOL(blk_queue_update_dma_pad); | |
736 | ||
86db1e29 JA |
737 | /** |
738 | * blk_queue_segment_boundary - set boundary rules for segment merging | |
739 | * @q: the request queue for the device | |
740 | * @mask: the memory boundary mask | |
741 | **/ | |
742 | void blk_queue_segment_boundary(struct request_queue *q, unsigned long mask) | |
743 | { | |
09cbfeaf KS |
744 | if (mask < PAGE_SIZE - 1) { |
745 | mask = PAGE_SIZE - 1; | |
f19d1e3b | 746 | pr_info("%s: set to minimum %lx\n", __func__, mask); |
86db1e29 JA |
747 | } |
748 | ||
025146e1 | 749 | q->limits.seg_boundary_mask = mask; |
86db1e29 | 750 | } |
86db1e29 JA |
751 | EXPORT_SYMBOL(blk_queue_segment_boundary); |
752 | ||
03100aad KB |
753 | /** |
754 | * blk_queue_virt_boundary - set boundary rules for bio merging | |
755 | * @q: the request queue for the device | |
756 | * @mask: the memory boundary mask | |
757 | **/ | |
758 | void blk_queue_virt_boundary(struct request_queue *q, unsigned long mask) | |
759 | { | |
760 | q->limits.virt_boundary_mask = mask; | |
09324d32 CH |
761 | |
762 | /* | |
763 | * Devices that require a virtual boundary do not support scatter/gather | |
764 | * I/O natively, but instead require a descriptor list entry for each | |
765 | * page (which might not be idential to the Linux PAGE_SIZE). Because | |
766 | * of that they are not limited by our notion of "segment size". | |
767 | */ | |
c6c84f78 CH |
768 | if (mask) |
769 | q->limits.max_segment_size = UINT_MAX; | |
03100aad KB |
770 | } |
771 | EXPORT_SYMBOL(blk_queue_virt_boundary); | |
772 | ||
86db1e29 JA |
773 | /** |
774 | * blk_queue_dma_alignment - set dma length and memory alignment | |
775 | * @q: the request queue for the device | |
776 | * @mask: alignment mask | |
777 | * | |
778 | * description: | |
710027a4 | 779 | * set required memory and length alignment for direct dma transactions. |
8feb4d20 | 780 | * this is used when building direct io requests for the queue. |
86db1e29 JA |
781 | * |
782 | **/ | |
783 | void blk_queue_dma_alignment(struct request_queue *q, int mask) | |
784 | { | |
c964d62f | 785 | q->limits.dma_alignment = mask; |
86db1e29 | 786 | } |
86db1e29 JA |
787 | EXPORT_SYMBOL(blk_queue_dma_alignment); |
788 | ||
789 | /** | |
790 | * blk_queue_update_dma_alignment - update dma length and memory alignment | |
791 | * @q: the request queue for the device | |
792 | * @mask: alignment mask | |
793 | * | |
794 | * description: | |
710027a4 | 795 | * update required memory and length alignment for direct dma transactions. |
86db1e29 JA |
796 | * If the requested alignment is larger than the current alignment, then |
797 | * the current queue alignment is updated to the new value, otherwise it | |
798 | * is left alone. The design of this is to allow multiple objects | |
799 | * (driver, device, transport etc) to set their respective | |
800 | * alignments without having them interfere. | |
801 | * | |
802 | **/ | |
803 | void blk_queue_update_dma_alignment(struct request_queue *q, int mask) | |
804 | { | |
805 | BUG_ON(mask > PAGE_SIZE); | |
806 | ||
c964d62f KB |
807 | if (mask > q->limits.dma_alignment) |
808 | q->limits.dma_alignment = mask; | |
86db1e29 | 809 | } |
86db1e29 JA |
810 | EXPORT_SYMBOL(blk_queue_update_dma_alignment); |
811 | ||
d278d4a8 JA |
812 | /** |
813 | * blk_set_queue_depth - tell the block layer about the device queue depth | |
814 | * @q: the request queue for the device | |
815 | * @depth: queue depth | |
816 | * | |
817 | */ | |
818 | void blk_set_queue_depth(struct request_queue *q, unsigned int depth) | |
819 | { | |
820 | q->queue_depth = depth; | |
9677a3e0 | 821 | rq_qos_queue_depth_changed(q); |
d278d4a8 JA |
822 | } |
823 | EXPORT_SYMBOL(blk_set_queue_depth); | |
824 | ||
93e9d8e8 JA |
825 | /** |
826 | * blk_queue_write_cache - configure queue's write cache | |
827 | * @q: the request queue for the device | |
828 | * @wc: write back cache on or off | |
829 | * @fua: device supports FUA writes, if true | |
830 | * | |
831 | * Tell the block layer about the write cache of @q. | |
832 | */ | |
833 | void blk_queue_write_cache(struct request_queue *q, bool wc, bool fua) | |
834 | { | |
43c9835b CH |
835 | if (wc) { |
836 | blk_queue_flag_set(QUEUE_FLAG_HW_WC, q); | |
57d74df9 | 837 | blk_queue_flag_set(QUEUE_FLAG_WC, q); |
43c9835b CH |
838 | } else { |
839 | blk_queue_flag_clear(QUEUE_FLAG_HW_WC, q); | |
57d74df9 | 840 | blk_queue_flag_clear(QUEUE_FLAG_WC, q); |
43c9835b | 841 | } |
c888a8f9 | 842 | if (fua) |
57d74df9 | 843 | blk_queue_flag_set(QUEUE_FLAG_FUA, q); |
c888a8f9 | 844 | else |
57d74df9 | 845 | blk_queue_flag_clear(QUEUE_FLAG_FUA, q); |
93e9d8e8 JA |
846 | } |
847 | EXPORT_SYMBOL_GPL(blk_queue_write_cache); | |
848 | ||
68c43f13 DLM |
849 | /** |
850 | * blk_queue_required_elevator_features - Set a queue required elevator features | |
851 | * @q: the request queue for the target device | |
852 | * @features: Required elevator features OR'ed together | |
853 | * | |
854 | * Tell the block layer that for the device controlled through @q, only the | |
855 | * only elevators that can be used are those that implement at least the set of | |
856 | * features specified by @features. | |
857 | */ | |
858 | void blk_queue_required_elevator_features(struct request_queue *q, | |
859 | unsigned int features) | |
860 | { | |
861 | q->required_elevator_features = features; | |
862 | } | |
863 | EXPORT_SYMBOL_GPL(blk_queue_required_elevator_features); | |
864 | ||
45147fb5 YS |
865 | /** |
866 | * blk_queue_can_use_dma_map_merging - configure queue for merging segments. | |
867 | * @q: the request queue for the device | |
868 | * @dev: the device pointer for dma | |
869 | * | |
870 | * Tell the block layer about merging the segments by dma map of @q. | |
871 | */ | |
872 | bool blk_queue_can_use_dma_map_merging(struct request_queue *q, | |
873 | struct device *dev) | |
874 | { | |
875 | unsigned long boundary = dma_get_merge_boundary(dev); | |
876 | ||
877 | if (!boundary) | |
878 | return false; | |
879 | ||
880 | /* No need to update max_segment_size. see blk_queue_virt_boundary() */ | |
881 | blk_queue_virt_boundary(q, boundary); | |
882 | ||
883 | return true; | |
884 | } | |
885 | EXPORT_SYMBOL_GPL(blk_queue_can_use_dma_map_merging); | |
886 | ||
27ba3e8f | 887 | /** |
d73e93b4 CH |
888 | * disk_set_zoned - inidicate a zoned device |
889 | * @disk: gendisk to configure | |
27ba3e8f | 890 | */ |
d73e93b4 | 891 | void disk_set_zoned(struct gendisk *disk) |
27ba3e8f | 892 | { |
a805a4fa DLM |
893 | struct request_queue *q = disk->queue; |
894 | ||
d73e93b4 CH |
895 | WARN_ON_ONCE(!IS_ENABLED(CONFIG_BLK_DEV_ZONED)); |
896 | ||
897 | /* | |
898 | * Set the zone write granularity to the device logical block | |
899 | * size by default. The driver can change this value if needed. | |
900 | */ | |
901 | q->limits.zoned = true; | |
902 | blk_queue_zone_write_granularity(q, queue_logical_block_size(q)); | |
27ba3e8f | 903 | } |
6b2bd274 | 904 | EXPORT_SYMBOL_GPL(disk_set_zoned); |
89098b07 CH |
905 | |
906 | int bdev_alignment_offset(struct block_device *bdev) | |
907 | { | |
908 | struct request_queue *q = bdev_get_queue(bdev); | |
909 | ||
910 | if (q->limits.misaligned) | |
911 | return -1; | |
912 | if (bdev_is_partition(bdev)) | |
913 | return queue_limit_alignment_offset(&q->limits, | |
914 | bdev->bd_start_sect); | |
915 | return q->limits.alignment_offset; | |
916 | } | |
917 | EXPORT_SYMBOL_GPL(bdev_alignment_offset); | |
5c4b4a5c CH |
918 | |
919 | unsigned int bdev_discard_alignment(struct block_device *bdev) | |
920 | { | |
921 | struct request_queue *q = bdev_get_queue(bdev); | |
922 | ||
923 | if (bdev_is_partition(bdev)) | |
924 | return queue_limit_discard_alignment(&q->limits, | |
925 | bdev->bd_start_sect); | |
926 | return q->limits.discard_alignment; | |
927 | } | |
928 | EXPORT_SYMBOL_GPL(bdev_discard_alignment); |