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" | |
87760e5e | 19 | #include "blk-wbt.h" |
86db1e29 | 20 | |
242f9dcb JA |
21 | void blk_queue_rq_timeout(struct request_queue *q, unsigned int timeout) |
22 | { | |
23 | q->rq_timeout = timeout; | |
24 | } | |
25 | EXPORT_SYMBOL_GPL(blk_queue_rq_timeout); | |
26 | ||
e475bba2 MP |
27 | /** |
28 | * blk_set_default_limits - reset limits to default values | |
f740f5ca | 29 | * @lim: the queue_limits structure to reset |
e475bba2 MP |
30 | * |
31 | * Description: | |
b1bd055d | 32 | * Returns a queue_limit struct to its default state. |
e475bba2 MP |
33 | */ |
34 | void blk_set_default_limits(struct queue_limits *lim) | |
35 | { | |
8a78362c | 36 | lim->max_segments = BLK_MAX_SEGMENTS; |
1e739730 | 37 | lim->max_discard_segments = 1; |
13f05c8d | 38 | lim->max_integrity_segments = 0; |
e475bba2 | 39 | lim->seg_boundary_mask = BLK_SEG_BOUNDARY_MASK; |
03100aad | 40 | lim->virt_boundary_mask = 0; |
eb28d31b | 41 | lim->max_segment_size = BLK_MAX_SEGMENT_SIZE; |
5f009d3f KB |
42 | lim->max_sectors = lim->max_hw_sectors = BLK_SAFE_MAX_SECTORS; |
43 | lim->max_dev_sectors = 0; | |
762380ad | 44 | lim->chunk_sectors = 0; |
4363ac7c | 45 | lim->max_write_same_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; |
86b37281 MP |
50 | lim->discard_granularity = 0; |
51 | lim->discard_alignment = 0; | |
52 | lim->discard_misaligned = 0; | |
e475bba2 | 53 | lim->logical_block_size = lim->physical_block_size = lim->io_min = 512; |
9bb33f24 | 54 | lim->bounce = BLK_BOUNCE_NONE; |
e475bba2 MP |
55 | lim->alignment_offset = 0; |
56 | lim->io_opt = 0; | |
57 | lim->misaligned = 0; | |
797476b8 | 58 | lim->zoned = BLK_ZONED_NONE; |
a805a4fa | 59 | lim->zone_write_granularity = 0; |
e475bba2 MP |
60 | } |
61 | EXPORT_SYMBOL(blk_set_default_limits); | |
62 | ||
b1bd055d MP |
63 | /** |
64 | * blk_set_stacking_limits - set default limits for stacking devices | |
65 | * @lim: the queue_limits structure to reset | |
66 | * | |
67 | * Description: | |
68 | * Returns a queue_limit struct to its default state. Should be used | |
69 | * by stacking drivers like DM that have no internal limits. | |
70 | */ | |
71 | void blk_set_stacking_limits(struct queue_limits *lim) | |
72 | { | |
73 | blk_set_default_limits(lim); | |
74 | ||
75 | /* Inherit limits from component devices */ | |
b1bd055d | 76 | lim->max_segments = USHRT_MAX; |
42c9cdfe | 77 | lim->max_discard_segments = USHRT_MAX; |
b1bd055d | 78 | lim->max_hw_sectors = UINT_MAX; |
d82ae52e | 79 | lim->max_segment_size = UINT_MAX; |
fe86cdce | 80 | lim->max_sectors = UINT_MAX; |
ca369d51 | 81 | lim->max_dev_sectors = UINT_MAX; |
4363ac7c | 82 | lim->max_write_same_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); | |
24c03d47 | 130 | printk(KERN_INFO "%s: set to minimum %d\n", |
2800aac1 | 131 | __func__, max_hw_sectors); |
86db1e29 JA |
132 | } |
133 | ||
817046ec DLM |
134 | max_hw_sectors = round_down(max_hw_sectors, |
135 | limits->logical_block_size >> SECTOR_SHIFT); | |
30e2bc08 | 136 | limits->max_hw_sectors = max_hw_sectors; |
817046ec | 137 | |
ca369d51 MP |
138 | max_sectors = min_not_zero(max_hw_sectors, limits->max_dev_sectors); |
139 | max_sectors = min_t(unsigned int, max_sectors, BLK_DEF_MAX_SECTORS); | |
817046ec DLM |
140 | max_sectors = round_down(max_sectors, |
141 | limits->logical_block_size >> SECTOR_SHIFT); | |
ca369d51 | 142 | limits->max_sectors = max_sectors; |
817046ec | 143 | |
d152c682 | 144 | if (!q->disk) |
edb0872f | 145 | return; |
d152c682 | 146 | q->disk->bdi->io_pages = max_sectors >> (PAGE_SHIFT - 9); |
86db1e29 | 147 | } |
086fa5ff | 148 | EXPORT_SYMBOL(blk_queue_max_hw_sectors); |
86db1e29 | 149 | |
762380ad JA |
150 | /** |
151 | * blk_queue_chunk_sectors - set size of the chunk for this queue | |
152 | * @q: the request queue for the device | |
153 | * @chunk_sectors: chunk sectors in the usual 512b unit | |
154 | * | |
155 | * Description: | |
156 | * If a driver doesn't want IOs to cross a given chunk size, it can set | |
07d098e6 MS |
157 | * this limit and prevent merging across chunks. Note that the block layer |
158 | * must accept a page worth of data at any offset. So if the crossing of | |
159 | * chunks is a hard limitation in the driver, it must still be prepared | |
160 | * to split single page bios. | |
762380ad JA |
161 | **/ |
162 | void blk_queue_chunk_sectors(struct request_queue *q, unsigned int chunk_sectors) | |
163 | { | |
762380ad JA |
164 | q->limits.chunk_sectors = chunk_sectors; |
165 | } | |
166 | EXPORT_SYMBOL(blk_queue_chunk_sectors); | |
167 | ||
67efc925 CH |
168 | /** |
169 | * blk_queue_max_discard_sectors - set max sectors for a single discard | |
170 | * @q: the request queue for the device | |
c7ebf065 | 171 | * @max_discard_sectors: maximum number of sectors to discard |
67efc925 CH |
172 | **/ |
173 | void blk_queue_max_discard_sectors(struct request_queue *q, | |
174 | unsigned int max_discard_sectors) | |
175 | { | |
0034af03 | 176 | q->limits.max_hw_discard_sectors = max_discard_sectors; |
67efc925 CH |
177 | q->limits.max_discard_sectors = max_discard_sectors; |
178 | } | |
179 | EXPORT_SYMBOL(blk_queue_max_discard_sectors); | |
180 | ||
4363ac7c MP |
181 | /** |
182 | * blk_queue_max_write_same_sectors - set max sectors for a single write same | |
183 | * @q: the request queue for the device | |
184 | * @max_write_same_sectors: maximum number of sectors to write per command | |
185 | **/ | |
186 | void blk_queue_max_write_same_sectors(struct request_queue *q, | |
187 | unsigned int max_write_same_sectors) | |
188 | { | |
189 | q->limits.max_write_same_sectors = max_write_same_sectors; | |
190 | } | |
191 | EXPORT_SYMBOL(blk_queue_max_write_same_sectors); | |
192 | ||
a6f0788e CK |
193 | /** |
194 | * blk_queue_max_write_zeroes_sectors - set max sectors for a single | |
195 | * write zeroes | |
196 | * @q: the request queue for the device | |
197 | * @max_write_zeroes_sectors: maximum number of sectors to write per command | |
198 | **/ | |
199 | void blk_queue_max_write_zeroes_sectors(struct request_queue *q, | |
200 | unsigned int max_write_zeroes_sectors) | |
201 | { | |
202 | q->limits.max_write_zeroes_sectors = max_write_zeroes_sectors; | |
203 | } | |
204 | EXPORT_SYMBOL(blk_queue_max_write_zeroes_sectors); | |
205 | ||
0512a75b KB |
206 | /** |
207 | * blk_queue_max_zone_append_sectors - set max sectors for a single zone append | |
208 | * @q: the request queue for the device | |
209 | * @max_zone_append_sectors: maximum number of sectors to write per command | |
210 | **/ | |
211 | void blk_queue_max_zone_append_sectors(struct request_queue *q, | |
212 | unsigned int max_zone_append_sectors) | |
213 | { | |
214 | unsigned int max_sectors; | |
215 | ||
216 | if (WARN_ON(!blk_queue_is_zoned(q))) | |
217 | return; | |
218 | ||
219 | max_sectors = min(q->limits.max_hw_sectors, max_zone_append_sectors); | |
220 | max_sectors = min(q->limits.chunk_sectors, max_sectors); | |
221 | ||
222 | /* | |
223 | * Signal eventual driver bugs resulting in the max_zone_append sectors limit | |
224 | * being 0 due to a 0 argument, the chunk_sectors limit (zone size) not set, | |
225 | * or the max_hw_sectors limit not set. | |
226 | */ | |
227 | WARN_ON(!max_sectors); | |
228 | ||
229 | q->limits.max_zone_append_sectors = max_sectors; | |
230 | } | |
231 | EXPORT_SYMBOL_GPL(blk_queue_max_zone_append_sectors); | |
232 | ||
86db1e29 | 233 | /** |
8a78362c | 234 | * blk_queue_max_segments - set max hw segments for a request for this queue |
86db1e29 JA |
235 | * @q: the request queue for the device |
236 | * @max_segments: max number of segments | |
237 | * | |
238 | * Description: | |
239 | * Enables a low level driver to set an upper limit on the number of | |
8a78362c | 240 | * hw data segments in a request. |
86db1e29 | 241 | **/ |
8a78362c | 242 | void blk_queue_max_segments(struct request_queue *q, unsigned short max_segments) |
86db1e29 JA |
243 | { |
244 | if (!max_segments) { | |
245 | max_segments = 1; | |
24c03d47 HH |
246 | printk(KERN_INFO "%s: set to minimum %d\n", |
247 | __func__, max_segments); | |
86db1e29 JA |
248 | } |
249 | ||
8a78362c | 250 | q->limits.max_segments = max_segments; |
86db1e29 | 251 | } |
8a78362c | 252 | EXPORT_SYMBOL(blk_queue_max_segments); |
86db1e29 | 253 | |
1e739730 CH |
254 | /** |
255 | * blk_queue_max_discard_segments - set max segments for discard requests | |
256 | * @q: the request queue for the device | |
257 | * @max_segments: max number of segments | |
258 | * | |
259 | * Description: | |
260 | * Enables a low level driver to set an upper limit on the number of | |
261 | * segments in a discard request. | |
262 | **/ | |
263 | void blk_queue_max_discard_segments(struct request_queue *q, | |
264 | unsigned short max_segments) | |
265 | { | |
266 | q->limits.max_discard_segments = max_segments; | |
267 | } | |
268 | EXPORT_SYMBOL_GPL(blk_queue_max_discard_segments); | |
269 | ||
86db1e29 JA |
270 | /** |
271 | * blk_queue_max_segment_size - set max segment size for blk_rq_map_sg | |
272 | * @q: the request queue for the device | |
273 | * @max_size: max size of segment in bytes | |
274 | * | |
275 | * Description: | |
276 | * Enables a low level driver to set an upper limit on the size of a | |
277 | * coalesced segment | |
278 | **/ | |
279 | void blk_queue_max_segment_size(struct request_queue *q, unsigned int max_size) | |
280 | { | |
09cbfeaf KS |
281 | if (max_size < PAGE_SIZE) { |
282 | max_size = PAGE_SIZE; | |
24c03d47 HH |
283 | printk(KERN_INFO "%s: set to minimum %d\n", |
284 | __func__, max_size); | |
86db1e29 JA |
285 | } |
286 | ||
09324d32 CH |
287 | /* see blk_queue_virt_boundary() for the explanation */ |
288 | WARN_ON_ONCE(q->limits.virt_boundary_mask); | |
289 | ||
025146e1 | 290 | q->limits.max_segment_size = max_size; |
86db1e29 | 291 | } |
86db1e29 JA |
292 | EXPORT_SYMBOL(blk_queue_max_segment_size); |
293 | ||
294 | /** | |
e1defc4f | 295 | * blk_queue_logical_block_size - set logical block size for the queue |
86db1e29 | 296 | * @q: the request queue for the device |
e1defc4f | 297 | * @size: the logical block size, in bytes |
86db1e29 JA |
298 | * |
299 | * Description: | |
e1defc4f MP |
300 | * This should be set to the lowest possible block size that the |
301 | * storage device can address. The default of 512 covers most | |
302 | * hardware. | |
86db1e29 | 303 | **/ |
ad6bf88a | 304 | void blk_queue_logical_block_size(struct request_queue *q, unsigned int size) |
86db1e29 | 305 | { |
817046ec DLM |
306 | struct queue_limits *limits = &q->limits; |
307 | ||
308 | limits->logical_block_size = size; | |
309 | ||
310 | if (limits->physical_block_size < size) | |
311 | limits->physical_block_size = size; | |
c72758f3 | 312 | |
817046ec DLM |
313 | if (limits->io_min < limits->physical_block_size) |
314 | limits->io_min = limits->physical_block_size; | |
c72758f3 | 315 | |
817046ec DLM |
316 | limits->max_hw_sectors = |
317 | round_down(limits->max_hw_sectors, size >> SECTOR_SHIFT); | |
318 | limits->max_sectors = | |
319 | round_down(limits->max_sectors, size >> SECTOR_SHIFT); | |
86db1e29 | 320 | } |
e1defc4f | 321 | EXPORT_SYMBOL(blk_queue_logical_block_size); |
86db1e29 | 322 | |
c72758f3 MP |
323 | /** |
324 | * blk_queue_physical_block_size - set physical block size for the queue | |
325 | * @q: the request queue for the device | |
326 | * @size: the physical block size, in bytes | |
327 | * | |
328 | * Description: | |
329 | * This should be set to the lowest possible sector size that the | |
330 | * hardware can operate on without reverting to read-modify-write | |
331 | * operations. | |
332 | */ | |
892b6f90 | 333 | void blk_queue_physical_block_size(struct request_queue *q, unsigned int size) |
c72758f3 MP |
334 | { |
335 | q->limits.physical_block_size = size; | |
336 | ||
337 | if (q->limits.physical_block_size < q->limits.logical_block_size) | |
338 | q->limits.physical_block_size = q->limits.logical_block_size; | |
339 | ||
340 | if (q->limits.io_min < q->limits.physical_block_size) | |
341 | q->limits.io_min = q->limits.physical_block_size; | |
342 | } | |
343 | EXPORT_SYMBOL(blk_queue_physical_block_size); | |
344 | ||
a805a4fa DLM |
345 | /** |
346 | * blk_queue_zone_write_granularity - set zone write granularity for the queue | |
347 | * @q: the request queue for the zoned device | |
348 | * @size: the zone write granularity size, in bytes | |
349 | * | |
350 | * Description: | |
351 | * This should be set to the lowest possible size allowing to write in | |
352 | * sequential zones of a zoned block device. | |
353 | */ | |
354 | void blk_queue_zone_write_granularity(struct request_queue *q, | |
355 | unsigned int size) | |
356 | { | |
357 | if (WARN_ON_ONCE(!blk_queue_is_zoned(q))) | |
358 | return; | |
359 | ||
360 | q->limits.zone_write_granularity = size; | |
361 | ||
362 | if (q->limits.zone_write_granularity < q->limits.logical_block_size) | |
363 | q->limits.zone_write_granularity = q->limits.logical_block_size; | |
364 | } | |
365 | EXPORT_SYMBOL_GPL(blk_queue_zone_write_granularity); | |
366 | ||
c72758f3 MP |
367 | /** |
368 | * blk_queue_alignment_offset - set physical block alignment offset | |
369 | * @q: the request queue for the device | |
8ebf9756 | 370 | * @offset: alignment offset in bytes |
c72758f3 MP |
371 | * |
372 | * Description: | |
373 | * Some devices are naturally misaligned to compensate for things like | |
374 | * the legacy DOS partition table 63-sector offset. Low-level drivers | |
375 | * should call this function for devices whose first sector is not | |
376 | * naturally aligned. | |
377 | */ | |
378 | void blk_queue_alignment_offset(struct request_queue *q, unsigned int offset) | |
379 | { | |
380 | q->limits.alignment_offset = | |
381 | offset & (q->limits.physical_block_size - 1); | |
382 | q->limits.misaligned = 0; | |
383 | } | |
384 | EXPORT_SYMBOL(blk_queue_alignment_offset); | |
385 | ||
471aa704 | 386 | void disk_update_readahead(struct gendisk *disk) |
c2e4cd57 | 387 | { |
471aa704 CH |
388 | struct request_queue *q = disk->queue; |
389 | ||
c2e4cd57 CH |
390 | /* |
391 | * For read-ahead of large files to be effective, we need to read ahead | |
392 | * at least twice the optimal I/O size. | |
393 | */ | |
edb0872f | 394 | disk->bdi->ra_pages = |
c2e4cd57 | 395 | max(queue_io_opt(q) * 2 / PAGE_SIZE, VM_READAHEAD_PAGES); |
edb0872f | 396 | disk->bdi->io_pages = queue_max_sectors(q) >> (PAGE_SHIFT - 9); |
c2e4cd57 | 397 | } |
471aa704 | 398 | EXPORT_SYMBOL_GPL(disk_update_readahead); |
c2e4cd57 | 399 | |
7c958e32 MP |
400 | /** |
401 | * blk_limits_io_min - set minimum request size for a device | |
402 | * @limits: the queue limits | |
403 | * @min: smallest I/O size in bytes | |
404 | * | |
405 | * Description: | |
406 | * Some devices have an internal block size bigger than the reported | |
407 | * hardware sector size. This function can be used to signal the | |
408 | * smallest I/O the device can perform without incurring a performance | |
409 | * penalty. | |
410 | */ | |
411 | void blk_limits_io_min(struct queue_limits *limits, unsigned int min) | |
412 | { | |
413 | limits->io_min = min; | |
414 | ||
415 | if (limits->io_min < limits->logical_block_size) | |
416 | limits->io_min = limits->logical_block_size; | |
417 | ||
418 | if (limits->io_min < limits->physical_block_size) | |
419 | limits->io_min = limits->physical_block_size; | |
420 | } | |
421 | EXPORT_SYMBOL(blk_limits_io_min); | |
422 | ||
c72758f3 MP |
423 | /** |
424 | * blk_queue_io_min - set minimum request size for the queue | |
425 | * @q: the request queue for the device | |
8ebf9756 | 426 | * @min: smallest I/O size in bytes |
c72758f3 MP |
427 | * |
428 | * Description: | |
7e5f5fb0 MP |
429 | * Storage devices may report a granularity or preferred minimum I/O |
430 | * size which is the smallest request the device can perform without | |
431 | * incurring a performance penalty. For disk drives this is often the | |
432 | * physical block size. For RAID arrays it is often the stripe chunk | |
433 | * size. A properly aligned multiple of minimum_io_size is the | |
434 | * preferred request size for workloads where a high number of I/O | |
435 | * operations is desired. | |
c72758f3 MP |
436 | */ |
437 | void blk_queue_io_min(struct request_queue *q, unsigned int min) | |
438 | { | |
7c958e32 | 439 | blk_limits_io_min(&q->limits, min); |
c72758f3 MP |
440 | } |
441 | EXPORT_SYMBOL(blk_queue_io_min); | |
442 | ||
3c5820c7 MP |
443 | /** |
444 | * blk_limits_io_opt - set optimal request size for a device | |
445 | * @limits: the queue limits | |
446 | * @opt: smallest I/O size in bytes | |
447 | * | |
448 | * Description: | |
449 | * Storage devices may report an optimal I/O size, which is the | |
450 | * device's preferred unit for sustained I/O. This is rarely reported | |
451 | * for disk drives. For RAID arrays it is usually the stripe width or | |
452 | * the internal track size. A properly aligned multiple of | |
453 | * optimal_io_size is the preferred request size for workloads where | |
454 | * sustained throughput is desired. | |
455 | */ | |
456 | void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt) | |
457 | { | |
458 | limits->io_opt = opt; | |
459 | } | |
460 | EXPORT_SYMBOL(blk_limits_io_opt); | |
461 | ||
c72758f3 MP |
462 | /** |
463 | * blk_queue_io_opt - set optimal request size for the queue | |
464 | * @q: the request queue for the device | |
8ebf9756 | 465 | * @opt: optimal request size in bytes |
c72758f3 MP |
466 | * |
467 | * Description: | |
7e5f5fb0 MP |
468 | * Storage devices may report an optimal I/O size, which is the |
469 | * device's preferred unit for sustained I/O. This is rarely reported | |
470 | * for disk drives. For RAID arrays it is usually the stripe width or | |
471 | * the internal track size. A properly aligned multiple of | |
472 | * optimal_io_size is the preferred request size for workloads where | |
473 | * sustained throughput is desired. | |
c72758f3 MP |
474 | */ |
475 | void blk_queue_io_opt(struct request_queue *q, unsigned int opt) | |
476 | { | |
3c5820c7 | 477 | blk_limits_io_opt(&q->limits, opt); |
d152c682 | 478 | if (!q->disk) |
edb0872f | 479 | return; |
d152c682 | 480 | q->disk->bdi->ra_pages = |
c2e4cd57 | 481 | max(queue_io_opt(q) * 2 / PAGE_SIZE, VM_READAHEAD_PAGES); |
c72758f3 MP |
482 | } |
483 | EXPORT_SYMBOL(blk_queue_io_opt); | |
484 | ||
97f433c3 MP |
485 | static unsigned int blk_round_down_sectors(unsigned int sectors, unsigned int lbs) |
486 | { | |
487 | sectors = round_down(sectors, lbs >> SECTOR_SHIFT); | |
488 | if (sectors < PAGE_SIZE >> SECTOR_SHIFT) | |
489 | sectors = PAGE_SIZE >> SECTOR_SHIFT; | |
490 | return sectors; | |
491 | } | |
492 | ||
c72758f3 MP |
493 | /** |
494 | * blk_stack_limits - adjust queue_limits for stacked devices | |
81744ee4 MP |
495 | * @t: the stacking driver limits (top device) |
496 | * @b: the underlying queue limits (bottom, component device) | |
e03a72e1 | 497 | * @start: first data sector within component device |
c72758f3 MP |
498 | * |
499 | * Description: | |
81744ee4 MP |
500 | * This function is used by stacking drivers like MD and DM to ensure |
501 | * that all component devices have compatible block sizes and | |
502 | * alignments. The stacking driver must provide a queue_limits | |
503 | * struct (top) and then iteratively call the stacking function for | |
504 | * all component (bottom) devices. The stacking function will | |
505 | * attempt to combine the values and ensure proper alignment. | |
506 | * | |
507 | * Returns 0 if the top and bottom queue_limits are compatible. The | |
508 | * top device's block sizes and alignment offsets may be adjusted to | |
509 | * ensure alignment with the bottom device. If no compatible sizes | |
510 | * and alignments exist, -1 is returned and the resulting top | |
511 | * queue_limits will have the misaligned flag set to indicate that | |
512 | * the alignment_offset is undefined. | |
c72758f3 MP |
513 | */ |
514 | int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, | |
e03a72e1 | 515 | sector_t start) |
c72758f3 | 516 | { |
e03a72e1 | 517 | unsigned int top, bottom, alignment, ret = 0; |
86b37281 | 518 | |
c72758f3 MP |
519 | t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors); |
520 | t->max_hw_sectors = min_not_zero(t->max_hw_sectors, b->max_hw_sectors); | |
ca369d51 | 521 | t->max_dev_sectors = min_not_zero(t->max_dev_sectors, b->max_dev_sectors); |
4363ac7c MP |
522 | t->max_write_same_sectors = min(t->max_write_same_sectors, |
523 | b->max_write_same_sectors); | |
a6f0788e CK |
524 | t->max_write_zeroes_sectors = min(t->max_write_zeroes_sectors, |
525 | b->max_write_zeroes_sectors); | |
0512a75b KB |
526 | t->max_zone_append_sectors = min(t->max_zone_append_sectors, |
527 | b->max_zone_append_sectors); | |
9bb33f24 | 528 | t->bounce = max(t->bounce, b->bounce); |
c72758f3 MP |
529 | |
530 | t->seg_boundary_mask = min_not_zero(t->seg_boundary_mask, | |
531 | b->seg_boundary_mask); | |
03100aad KB |
532 | t->virt_boundary_mask = min_not_zero(t->virt_boundary_mask, |
533 | b->virt_boundary_mask); | |
c72758f3 | 534 | |
8a78362c | 535 | t->max_segments = min_not_zero(t->max_segments, b->max_segments); |
1e739730 CH |
536 | t->max_discard_segments = min_not_zero(t->max_discard_segments, |
537 | b->max_discard_segments); | |
13f05c8d MP |
538 | t->max_integrity_segments = min_not_zero(t->max_integrity_segments, |
539 | b->max_integrity_segments); | |
c72758f3 MP |
540 | |
541 | t->max_segment_size = min_not_zero(t->max_segment_size, | |
542 | b->max_segment_size); | |
543 | ||
fe0b393f MP |
544 | t->misaligned |= b->misaligned; |
545 | ||
e03a72e1 | 546 | alignment = queue_limit_alignment_offset(b, start); |
9504e086 | 547 | |
81744ee4 MP |
548 | /* Bottom device has different alignment. Check that it is |
549 | * compatible with the current top alignment. | |
550 | */ | |
9504e086 MP |
551 | if (t->alignment_offset != alignment) { |
552 | ||
553 | top = max(t->physical_block_size, t->io_min) | |
554 | + t->alignment_offset; | |
81744ee4 | 555 | bottom = max(b->physical_block_size, b->io_min) + alignment; |
9504e086 | 556 | |
81744ee4 | 557 | /* Verify that top and bottom intervals line up */ |
b8839b8c | 558 | if (max(top, bottom) % min(top, bottom)) { |
9504e086 | 559 | t->misaligned = 1; |
fe0b393f MP |
560 | ret = -1; |
561 | } | |
9504e086 MP |
562 | } |
563 | ||
c72758f3 MP |
564 | t->logical_block_size = max(t->logical_block_size, |
565 | b->logical_block_size); | |
566 | ||
567 | t->physical_block_size = max(t->physical_block_size, | |
568 | b->physical_block_size); | |
569 | ||
570 | t->io_min = max(t->io_min, b->io_min); | |
e9637415 | 571 | t->io_opt = lcm_not_zero(t->io_opt, b->io_opt); |
7e7986f9 MS |
572 | |
573 | /* Set non-power-of-2 compatible chunk_sectors boundary */ | |
574 | if (b->chunk_sectors) | |
575 | t->chunk_sectors = gcd(t->chunk_sectors, b->chunk_sectors); | |
9504e086 | 576 | |
81744ee4 | 577 | /* Physical block size a multiple of the logical block size? */ |
9504e086 MP |
578 | if (t->physical_block_size & (t->logical_block_size - 1)) { |
579 | t->physical_block_size = t->logical_block_size; | |
c72758f3 | 580 | t->misaligned = 1; |
fe0b393f | 581 | ret = -1; |
86b37281 MP |
582 | } |
583 | ||
81744ee4 | 584 | /* Minimum I/O a multiple of the physical block size? */ |
9504e086 MP |
585 | if (t->io_min & (t->physical_block_size - 1)) { |
586 | t->io_min = t->physical_block_size; | |
587 | t->misaligned = 1; | |
fe0b393f | 588 | ret = -1; |
c72758f3 MP |
589 | } |
590 | ||
81744ee4 | 591 | /* Optimal I/O a multiple of the physical block size? */ |
9504e086 MP |
592 | if (t->io_opt & (t->physical_block_size - 1)) { |
593 | t->io_opt = 0; | |
594 | t->misaligned = 1; | |
fe0b393f | 595 | ret = -1; |
9504e086 | 596 | } |
c72758f3 | 597 | |
22ada802 MS |
598 | /* chunk_sectors a multiple of the physical block size? */ |
599 | if ((t->chunk_sectors << 9) & (t->physical_block_size - 1)) { | |
600 | t->chunk_sectors = 0; | |
601 | t->misaligned = 1; | |
602 | ret = -1; | |
603 | } | |
604 | ||
c78afc62 KO |
605 | t->raid_partial_stripes_expensive = |
606 | max(t->raid_partial_stripes_expensive, | |
607 | b->raid_partial_stripes_expensive); | |
608 | ||
81744ee4 | 609 | /* Find lowest common alignment_offset */ |
e9637415 | 610 | t->alignment_offset = lcm_not_zero(t->alignment_offset, alignment) |
b8839b8c | 611 | % max(t->physical_block_size, t->io_min); |
86b37281 | 612 | |
81744ee4 | 613 | /* Verify that new alignment_offset is on a logical block boundary */ |
fe0b393f | 614 | if (t->alignment_offset & (t->logical_block_size - 1)) { |
c72758f3 | 615 | t->misaligned = 1; |
fe0b393f MP |
616 | ret = -1; |
617 | } | |
c72758f3 | 618 | |
97f433c3 MP |
619 | t->max_sectors = blk_round_down_sectors(t->max_sectors, t->logical_block_size); |
620 | t->max_hw_sectors = blk_round_down_sectors(t->max_hw_sectors, t->logical_block_size); | |
621 | t->max_dev_sectors = blk_round_down_sectors(t->max_dev_sectors, t->logical_block_size); | |
622 | ||
9504e086 MP |
623 | /* Discard alignment and granularity */ |
624 | if (b->discard_granularity) { | |
e03a72e1 | 625 | alignment = queue_limit_discard_alignment(b, start); |
9504e086 MP |
626 | |
627 | if (t->discard_granularity != 0 && | |
628 | t->discard_alignment != alignment) { | |
629 | top = t->discard_granularity + t->discard_alignment; | |
630 | bottom = b->discard_granularity + alignment; | |
70dd5bf3 | 631 | |
9504e086 | 632 | /* Verify that top and bottom intervals line up */ |
8dd2cb7e | 633 | if ((max(top, bottom) % min(top, bottom)) != 0) |
9504e086 MP |
634 | t->discard_misaligned = 1; |
635 | } | |
636 | ||
81744ee4 MP |
637 | t->max_discard_sectors = min_not_zero(t->max_discard_sectors, |
638 | b->max_discard_sectors); | |
0034af03 JA |
639 | t->max_hw_discard_sectors = min_not_zero(t->max_hw_discard_sectors, |
640 | b->max_hw_discard_sectors); | |
9504e086 MP |
641 | t->discard_granularity = max(t->discard_granularity, |
642 | b->discard_granularity); | |
e9637415 | 643 | t->discard_alignment = lcm_not_zero(t->discard_alignment, alignment) % |
8dd2cb7e | 644 | t->discard_granularity; |
9504e086 | 645 | } |
70dd5bf3 | 646 | |
a805a4fa DLM |
647 | t->zone_write_granularity = max(t->zone_write_granularity, |
648 | b->zone_write_granularity); | |
3093a479 | 649 | t->zoned = max(t->zoned, b->zoned); |
fe0b393f | 650 | return ret; |
c72758f3 | 651 | } |
5d85d324 | 652 | EXPORT_SYMBOL(blk_stack_limits); |
c72758f3 MP |
653 | |
654 | /** | |
655 | * disk_stack_limits - adjust queue limits for stacked drivers | |
77634f33 | 656 | * @disk: MD/DM gendisk (top) |
c72758f3 MP |
657 | * @bdev: the underlying block device (bottom) |
658 | * @offset: offset to beginning of data within component device | |
659 | * | |
660 | * Description: | |
e03a72e1 MP |
661 | * Merges the limits for a top level gendisk and a bottom level |
662 | * block_device. | |
c72758f3 MP |
663 | */ |
664 | void disk_stack_limits(struct gendisk *disk, struct block_device *bdev, | |
665 | sector_t offset) | |
666 | { | |
667 | struct request_queue *t = disk->queue; | |
c72758f3 | 668 | |
9efa82ef | 669 | if (blk_stack_limits(&t->limits, &bdev_get_queue(bdev)->limits, |
453b8ab6 CH |
670 | get_start_sect(bdev) + (offset >> 9)) < 0) |
671 | pr_notice("%s: Warning: Device %pg is misaligned\n", | |
672 | disk->disk_name, bdev); | |
e74d93e9 | 673 | |
471aa704 | 674 | disk_update_readahead(disk); |
c72758f3 MP |
675 | } |
676 | EXPORT_SYMBOL(disk_stack_limits); | |
677 | ||
27f8221a FT |
678 | /** |
679 | * blk_queue_update_dma_pad - update pad mask | |
680 | * @q: the request queue for the device | |
681 | * @mask: pad mask | |
682 | * | |
683 | * Update dma pad mask. | |
684 | * | |
685 | * Appending pad buffer to a request modifies the last entry of a | |
686 | * scatter list such that it includes the pad buffer. | |
687 | **/ | |
688 | void blk_queue_update_dma_pad(struct request_queue *q, unsigned int mask) | |
689 | { | |
690 | if (mask > q->dma_pad_mask) | |
691 | q->dma_pad_mask = mask; | |
692 | } | |
693 | EXPORT_SYMBOL(blk_queue_update_dma_pad); | |
694 | ||
86db1e29 JA |
695 | /** |
696 | * blk_queue_segment_boundary - set boundary rules for segment merging | |
697 | * @q: the request queue for the device | |
698 | * @mask: the memory boundary mask | |
699 | **/ | |
700 | void blk_queue_segment_boundary(struct request_queue *q, unsigned long mask) | |
701 | { | |
09cbfeaf KS |
702 | if (mask < PAGE_SIZE - 1) { |
703 | mask = PAGE_SIZE - 1; | |
24c03d47 HH |
704 | printk(KERN_INFO "%s: set to minimum %lx\n", |
705 | __func__, mask); | |
86db1e29 JA |
706 | } |
707 | ||
025146e1 | 708 | q->limits.seg_boundary_mask = mask; |
86db1e29 | 709 | } |
86db1e29 JA |
710 | EXPORT_SYMBOL(blk_queue_segment_boundary); |
711 | ||
03100aad KB |
712 | /** |
713 | * blk_queue_virt_boundary - set boundary rules for bio merging | |
714 | * @q: the request queue for the device | |
715 | * @mask: the memory boundary mask | |
716 | **/ | |
717 | void blk_queue_virt_boundary(struct request_queue *q, unsigned long mask) | |
718 | { | |
719 | q->limits.virt_boundary_mask = mask; | |
09324d32 CH |
720 | |
721 | /* | |
722 | * Devices that require a virtual boundary do not support scatter/gather | |
723 | * I/O natively, but instead require a descriptor list entry for each | |
724 | * page (which might not be idential to the Linux PAGE_SIZE). Because | |
725 | * of that they are not limited by our notion of "segment size". | |
726 | */ | |
c6c84f78 CH |
727 | if (mask) |
728 | q->limits.max_segment_size = UINT_MAX; | |
03100aad KB |
729 | } |
730 | EXPORT_SYMBOL(blk_queue_virt_boundary); | |
731 | ||
86db1e29 JA |
732 | /** |
733 | * blk_queue_dma_alignment - set dma length and memory alignment | |
734 | * @q: the request queue for the device | |
735 | * @mask: alignment mask | |
736 | * | |
737 | * description: | |
710027a4 | 738 | * set required memory and length alignment for direct dma transactions. |
8feb4d20 | 739 | * this is used when building direct io requests for the queue. |
86db1e29 JA |
740 | * |
741 | **/ | |
742 | void blk_queue_dma_alignment(struct request_queue *q, int mask) | |
743 | { | |
744 | q->dma_alignment = mask; | |
745 | } | |
86db1e29 JA |
746 | EXPORT_SYMBOL(blk_queue_dma_alignment); |
747 | ||
748 | /** | |
749 | * blk_queue_update_dma_alignment - update dma length and memory alignment | |
750 | * @q: the request queue for the device | |
751 | * @mask: alignment mask | |
752 | * | |
753 | * description: | |
710027a4 | 754 | * update required memory and length alignment for direct dma transactions. |
86db1e29 JA |
755 | * If the requested alignment is larger than the current alignment, then |
756 | * the current queue alignment is updated to the new value, otherwise it | |
757 | * is left alone. The design of this is to allow multiple objects | |
758 | * (driver, device, transport etc) to set their respective | |
759 | * alignments without having them interfere. | |
760 | * | |
761 | **/ | |
762 | void blk_queue_update_dma_alignment(struct request_queue *q, int mask) | |
763 | { | |
764 | BUG_ON(mask > PAGE_SIZE); | |
765 | ||
766 | if (mask > q->dma_alignment) | |
767 | q->dma_alignment = mask; | |
768 | } | |
86db1e29 JA |
769 | EXPORT_SYMBOL(blk_queue_update_dma_alignment); |
770 | ||
d278d4a8 JA |
771 | /** |
772 | * blk_set_queue_depth - tell the block layer about the device queue depth | |
773 | * @q: the request queue for the device | |
774 | * @depth: queue depth | |
775 | * | |
776 | */ | |
777 | void blk_set_queue_depth(struct request_queue *q, unsigned int depth) | |
778 | { | |
779 | q->queue_depth = depth; | |
9677a3e0 | 780 | rq_qos_queue_depth_changed(q); |
d278d4a8 JA |
781 | } |
782 | EXPORT_SYMBOL(blk_set_queue_depth); | |
783 | ||
93e9d8e8 JA |
784 | /** |
785 | * blk_queue_write_cache - configure queue's write cache | |
786 | * @q: the request queue for the device | |
787 | * @wc: write back cache on or off | |
788 | * @fua: device supports FUA writes, if true | |
789 | * | |
790 | * Tell the block layer about the write cache of @q. | |
791 | */ | |
792 | void blk_queue_write_cache(struct request_queue *q, bool wc, bool fua) | |
793 | { | |
c888a8f9 | 794 | if (wc) |
57d74df9 | 795 | blk_queue_flag_set(QUEUE_FLAG_WC, q); |
c888a8f9 | 796 | else |
57d74df9 | 797 | blk_queue_flag_clear(QUEUE_FLAG_WC, q); |
c888a8f9 | 798 | if (fua) |
57d74df9 | 799 | blk_queue_flag_set(QUEUE_FLAG_FUA, q); |
c888a8f9 | 800 | else |
57d74df9 | 801 | blk_queue_flag_clear(QUEUE_FLAG_FUA, q); |
87760e5e | 802 | |
a7905043 | 803 | wbt_set_write_cache(q, test_bit(QUEUE_FLAG_WC, &q->queue_flags)); |
93e9d8e8 JA |
804 | } |
805 | EXPORT_SYMBOL_GPL(blk_queue_write_cache); | |
806 | ||
68c43f13 DLM |
807 | /** |
808 | * blk_queue_required_elevator_features - Set a queue required elevator features | |
809 | * @q: the request queue for the target device | |
810 | * @features: Required elevator features OR'ed together | |
811 | * | |
812 | * Tell the block layer that for the device controlled through @q, only the | |
813 | * only elevators that can be used are those that implement at least the set of | |
814 | * features specified by @features. | |
815 | */ | |
816 | void blk_queue_required_elevator_features(struct request_queue *q, | |
817 | unsigned int features) | |
818 | { | |
819 | q->required_elevator_features = features; | |
820 | } | |
821 | EXPORT_SYMBOL_GPL(blk_queue_required_elevator_features); | |
822 | ||
45147fb5 YS |
823 | /** |
824 | * blk_queue_can_use_dma_map_merging - configure queue for merging segments. | |
825 | * @q: the request queue for the device | |
826 | * @dev: the device pointer for dma | |
827 | * | |
828 | * Tell the block layer about merging the segments by dma map of @q. | |
829 | */ | |
830 | bool blk_queue_can_use_dma_map_merging(struct request_queue *q, | |
831 | struct device *dev) | |
832 | { | |
833 | unsigned long boundary = dma_get_merge_boundary(dev); | |
834 | ||
835 | if (!boundary) | |
836 | return false; | |
837 | ||
838 | /* No need to update max_segment_size. see blk_queue_virt_boundary() */ | |
839 | blk_queue_virt_boundary(q, boundary); | |
840 | ||
841 | return true; | |
842 | } | |
843 | EXPORT_SYMBOL_GPL(blk_queue_can_use_dma_map_merging); | |
844 | ||
e0c60d01 SK |
845 | static bool disk_has_partitions(struct gendisk *disk) |
846 | { | |
847 | unsigned long idx; | |
848 | struct block_device *part; | |
849 | bool ret = false; | |
850 | ||
851 | rcu_read_lock(); | |
852 | xa_for_each(&disk->part_tbl, idx, part) { | |
853 | if (bdev_is_partition(part)) { | |
854 | ret = true; | |
855 | break; | |
856 | } | |
857 | } | |
858 | rcu_read_unlock(); | |
859 | ||
860 | return ret; | |
861 | } | |
862 | ||
27ba3e8f DLM |
863 | /** |
864 | * blk_queue_set_zoned - configure a disk queue zoned model. | |
865 | * @disk: the gendisk of the queue to configure | |
866 | * @model: the zoned model to set | |
867 | * | |
868 | * Set the zoned model of the request queue of @disk according to @model. | |
869 | * When @model is BLK_ZONED_HM (host managed), this should be called only | |
870 | * if zoned block device support is enabled (CONFIG_BLK_DEV_ZONED option). | |
871 | * If @model specifies BLK_ZONED_HA (host aware), the effective model used | |
872 | * depends on CONFIG_BLK_DEV_ZONED settings and on the existence of partitions | |
873 | * on the disk. | |
874 | */ | |
875 | void blk_queue_set_zoned(struct gendisk *disk, enum blk_zoned_model model) | |
876 | { | |
a805a4fa DLM |
877 | struct request_queue *q = disk->queue; |
878 | ||
27ba3e8f DLM |
879 | switch (model) { |
880 | case BLK_ZONED_HM: | |
881 | /* | |
882 | * Host managed devices are supported only if | |
883 | * CONFIG_BLK_DEV_ZONED is enabled. | |
884 | */ | |
885 | WARN_ON_ONCE(!IS_ENABLED(CONFIG_BLK_DEV_ZONED)); | |
886 | break; | |
887 | case BLK_ZONED_HA: | |
888 | /* | |
889 | * Host aware devices can be treated either as regular block | |
890 | * devices (similar to drive managed devices) or as zoned block | |
891 | * devices to take advantage of the zone command set, similarly | |
892 | * to host managed devices. We try the latter if there are no | |
893 | * partitions and zoned block device support is enabled, else | |
894 | * we do nothing special as far as the block layer is concerned. | |
895 | */ | |
896 | if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) || | |
e0c60d01 | 897 | disk_has_partitions(disk)) |
27ba3e8f DLM |
898 | model = BLK_ZONED_NONE; |
899 | break; | |
900 | case BLK_ZONED_NONE: | |
901 | default: | |
902 | if (WARN_ON_ONCE(model != BLK_ZONED_NONE)) | |
903 | model = BLK_ZONED_NONE; | |
904 | break; | |
905 | } | |
906 | ||
a805a4fa DLM |
907 | q->limits.zoned = model; |
908 | if (model != BLK_ZONED_NONE) { | |
909 | /* | |
910 | * Set the zone write granularity to the device logical block | |
911 | * size by default. The driver can change this value if needed. | |
912 | */ | |
913 | blk_queue_zone_write_granularity(q, | |
914 | queue_logical_block_size(q)); | |
508aebb8 DLM |
915 | } else { |
916 | blk_queue_clear_zone_settings(q); | |
a805a4fa | 917 | } |
27ba3e8f DLM |
918 | } |
919 | EXPORT_SYMBOL_GPL(blk_queue_set_zoned); |