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> | |
70dd5bf3 | 10 | #include <linux/gcd.h> |
2cda2728 | 11 | #include <linux/lcm.h> |
ad5ebd2f | 12 | #include <linux/jiffies.h> |
5a0e3ad6 | 13 | #include <linux/gfp.h> |
45147fb5 | 14 | #include <linux/dma-mapping.h> |
86db1e29 JA |
15 | |
16 | #include "blk.h" | |
87760e5e | 17 | #include "blk-wbt.h" |
86db1e29 | 18 | |
242f9dcb JA |
19 | void blk_queue_rq_timeout(struct request_queue *q, unsigned int timeout) |
20 | { | |
21 | q->rq_timeout = timeout; | |
22 | } | |
23 | EXPORT_SYMBOL_GPL(blk_queue_rq_timeout); | |
24 | ||
e475bba2 MP |
25 | /** |
26 | * blk_set_default_limits - reset limits to default values | |
f740f5ca | 27 | * @lim: the queue_limits structure to reset |
e475bba2 MP |
28 | * |
29 | * Description: | |
b1bd055d | 30 | * Returns a queue_limit struct to its default state. |
e475bba2 MP |
31 | */ |
32 | void blk_set_default_limits(struct queue_limits *lim) | |
33 | { | |
8a78362c | 34 | lim->max_segments = BLK_MAX_SEGMENTS; |
1e739730 | 35 | lim->max_discard_segments = 1; |
13f05c8d | 36 | lim->max_integrity_segments = 0; |
e475bba2 | 37 | lim->seg_boundary_mask = BLK_SEG_BOUNDARY_MASK; |
03100aad | 38 | lim->virt_boundary_mask = 0; |
eb28d31b | 39 | lim->max_segment_size = BLK_MAX_SEGMENT_SIZE; |
5f009d3f KB |
40 | lim->max_sectors = lim->max_hw_sectors = BLK_SAFE_MAX_SECTORS; |
41 | lim->max_dev_sectors = 0; | |
762380ad | 42 | lim->chunk_sectors = 0; |
4363ac7c | 43 | lim->max_write_same_sectors = 0; |
a6f0788e | 44 | lim->max_write_zeroes_sectors = 0; |
0512a75b | 45 | lim->max_zone_append_sectors = 0; |
86b37281 | 46 | lim->max_discard_sectors = 0; |
0034af03 | 47 | lim->max_hw_discard_sectors = 0; |
86b37281 MP |
48 | lim->discard_granularity = 0; |
49 | lim->discard_alignment = 0; | |
50 | lim->discard_misaligned = 0; | |
e475bba2 | 51 | lim->logical_block_size = lim->physical_block_size = lim->io_min = 512; |
9bb33f24 | 52 | lim->bounce = BLK_BOUNCE_NONE; |
e475bba2 MP |
53 | lim->alignment_offset = 0; |
54 | lim->io_opt = 0; | |
55 | lim->misaligned = 0; | |
797476b8 | 56 | lim->zoned = BLK_ZONED_NONE; |
a805a4fa | 57 | lim->zone_write_granularity = 0; |
e475bba2 MP |
58 | } |
59 | EXPORT_SYMBOL(blk_set_default_limits); | |
60 | ||
b1bd055d MP |
61 | /** |
62 | * blk_set_stacking_limits - set default limits for stacking devices | |
63 | * @lim: the queue_limits structure to reset | |
64 | * | |
65 | * Description: | |
66 | * Returns a queue_limit struct to its default state. Should be used | |
67 | * by stacking drivers like DM that have no internal limits. | |
68 | */ | |
69 | void blk_set_stacking_limits(struct queue_limits *lim) | |
70 | { | |
71 | blk_set_default_limits(lim); | |
72 | ||
73 | /* Inherit limits from component devices */ | |
b1bd055d | 74 | lim->max_segments = USHRT_MAX; |
42c9cdfe | 75 | lim->max_discard_segments = USHRT_MAX; |
b1bd055d | 76 | lim->max_hw_sectors = UINT_MAX; |
d82ae52e | 77 | lim->max_segment_size = UINT_MAX; |
fe86cdce | 78 | lim->max_sectors = UINT_MAX; |
ca369d51 | 79 | lim->max_dev_sectors = UINT_MAX; |
4363ac7c | 80 | lim->max_write_same_sectors = UINT_MAX; |
a6f0788e | 81 | lim->max_write_zeroes_sectors = UINT_MAX; |
0512a75b | 82 | lim->max_zone_append_sectors = UINT_MAX; |
b1bd055d MP |
83 | } |
84 | EXPORT_SYMBOL(blk_set_stacking_limits); | |
85 | ||
86db1e29 JA |
86 | /** |
87 | * blk_queue_bounce_limit - set bounce buffer limit for queue | |
cd0aca2d | 88 | * @q: the request queue for the device |
9bb33f24 | 89 | * @bounce: bounce limit to enforce |
86db1e29 JA |
90 | * |
91 | * Description: | |
9bb33f24 CH |
92 | * Force bouncing for ISA DMA ranges or highmem. |
93 | * | |
94 | * DEPRECATED, don't use in new code. | |
86db1e29 | 95 | **/ |
9bb33f24 | 96 | void blk_queue_bounce_limit(struct request_queue *q, enum blk_bounce bounce) |
86db1e29 | 97 | { |
9bb33f24 | 98 | q->limits.bounce = bounce; |
86db1e29 | 99 | } |
86db1e29 JA |
100 | EXPORT_SYMBOL(blk_queue_bounce_limit); |
101 | ||
102 | /** | |
ca369d51 MP |
103 | * blk_queue_max_hw_sectors - set max sectors for a request for this queue |
104 | * @q: the request queue for the device | |
2800aac1 | 105 | * @max_hw_sectors: max hardware sectors in the usual 512b unit |
86db1e29 JA |
106 | * |
107 | * Description: | |
2800aac1 MP |
108 | * Enables a low level driver to set a hard upper limit, |
109 | * max_hw_sectors, on the size of requests. max_hw_sectors is set by | |
4f258a46 MP |
110 | * the device driver based upon the capabilities of the I/O |
111 | * controller. | |
2800aac1 | 112 | * |
ca369d51 MP |
113 | * max_dev_sectors is a hard limit imposed by the storage device for |
114 | * READ/WRITE requests. It is set by the disk driver. | |
115 | * | |
2800aac1 MP |
116 | * max_sectors is a soft limit imposed by the block layer for |
117 | * filesystem type requests. This value can be overridden on a | |
118 | * per-device basis in /sys/block/<device>/queue/max_sectors_kb. | |
119 | * The soft limit can not exceed max_hw_sectors. | |
86db1e29 | 120 | **/ |
ca369d51 | 121 | void blk_queue_max_hw_sectors(struct request_queue *q, unsigned int max_hw_sectors) |
86db1e29 | 122 | { |
ca369d51 MP |
123 | struct queue_limits *limits = &q->limits; |
124 | unsigned int max_sectors; | |
125 | ||
09cbfeaf KS |
126 | if ((max_hw_sectors << 9) < PAGE_SIZE) { |
127 | max_hw_sectors = 1 << (PAGE_SHIFT - 9); | |
24c03d47 | 128 | printk(KERN_INFO "%s: set to minimum %d\n", |
2800aac1 | 129 | __func__, max_hw_sectors); |
86db1e29 JA |
130 | } |
131 | ||
817046ec DLM |
132 | max_hw_sectors = round_down(max_hw_sectors, |
133 | limits->logical_block_size >> SECTOR_SHIFT); | |
30e2bc08 | 134 | limits->max_hw_sectors = max_hw_sectors; |
817046ec | 135 | |
ca369d51 MP |
136 | max_sectors = min_not_zero(max_hw_sectors, limits->max_dev_sectors); |
137 | max_sectors = min_t(unsigned int, max_sectors, BLK_DEF_MAX_SECTORS); | |
817046ec DLM |
138 | max_sectors = round_down(max_sectors, |
139 | limits->logical_block_size >> SECTOR_SHIFT); | |
ca369d51 | 140 | limits->max_sectors = max_sectors; |
817046ec | 141 | |
dc3b17cc | 142 | q->backing_dev_info->io_pages = max_sectors >> (PAGE_SHIFT - 9); |
86db1e29 | 143 | } |
086fa5ff | 144 | EXPORT_SYMBOL(blk_queue_max_hw_sectors); |
86db1e29 | 145 | |
762380ad JA |
146 | /** |
147 | * blk_queue_chunk_sectors - set size of the chunk for this queue | |
148 | * @q: the request queue for the device | |
149 | * @chunk_sectors: chunk sectors in the usual 512b unit | |
150 | * | |
151 | * Description: | |
152 | * If a driver doesn't want IOs to cross a given chunk size, it can set | |
07d098e6 MS |
153 | * this limit and prevent merging across chunks. Note that the block layer |
154 | * must accept a page worth of data at any offset. So if the crossing of | |
155 | * chunks is a hard limitation in the driver, it must still be prepared | |
156 | * to split single page bios. | |
762380ad JA |
157 | **/ |
158 | void blk_queue_chunk_sectors(struct request_queue *q, unsigned int chunk_sectors) | |
159 | { | |
762380ad JA |
160 | q->limits.chunk_sectors = chunk_sectors; |
161 | } | |
162 | EXPORT_SYMBOL(blk_queue_chunk_sectors); | |
163 | ||
67efc925 CH |
164 | /** |
165 | * blk_queue_max_discard_sectors - set max sectors for a single discard | |
166 | * @q: the request queue for the device | |
c7ebf065 | 167 | * @max_discard_sectors: maximum number of sectors to discard |
67efc925 CH |
168 | **/ |
169 | void blk_queue_max_discard_sectors(struct request_queue *q, | |
170 | unsigned int max_discard_sectors) | |
171 | { | |
0034af03 | 172 | q->limits.max_hw_discard_sectors = max_discard_sectors; |
67efc925 CH |
173 | q->limits.max_discard_sectors = max_discard_sectors; |
174 | } | |
175 | EXPORT_SYMBOL(blk_queue_max_discard_sectors); | |
176 | ||
4363ac7c MP |
177 | /** |
178 | * blk_queue_max_write_same_sectors - set max sectors for a single write same | |
179 | * @q: the request queue for the device | |
180 | * @max_write_same_sectors: maximum number of sectors to write per command | |
181 | **/ | |
182 | void blk_queue_max_write_same_sectors(struct request_queue *q, | |
183 | unsigned int max_write_same_sectors) | |
184 | { | |
185 | q->limits.max_write_same_sectors = max_write_same_sectors; | |
186 | } | |
187 | EXPORT_SYMBOL(blk_queue_max_write_same_sectors); | |
188 | ||
a6f0788e CK |
189 | /** |
190 | * blk_queue_max_write_zeroes_sectors - set max sectors for a single | |
191 | * write zeroes | |
192 | * @q: the request queue for the device | |
193 | * @max_write_zeroes_sectors: maximum number of sectors to write per command | |
194 | **/ | |
195 | void blk_queue_max_write_zeroes_sectors(struct request_queue *q, | |
196 | unsigned int max_write_zeroes_sectors) | |
197 | { | |
198 | q->limits.max_write_zeroes_sectors = max_write_zeroes_sectors; | |
199 | } | |
200 | EXPORT_SYMBOL(blk_queue_max_write_zeroes_sectors); | |
201 | ||
0512a75b KB |
202 | /** |
203 | * blk_queue_max_zone_append_sectors - set max sectors for a single zone append | |
204 | * @q: the request queue for the device | |
205 | * @max_zone_append_sectors: maximum number of sectors to write per command | |
206 | **/ | |
207 | void blk_queue_max_zone_append_sectors(struct request_queue *q, | |
208 | unsigned int max_zone_append_sectors) | |
209 | { | |
210 | unsigned int max_sectors; | |
211 | ||
212 | if (WARN_ON(!blk_queue_is_zoned(q))) | |
213 | return; | |
214 | ||
215 | max_sectors = min(q->limits.max_hw_sectors, max_zone_append_sectors); | |
216 | max_sectors = min(q->limits.chunk_sectors, max_sectors); | |
217 | ||
218 | /* | |
219 | * Signal eventual driver bugs resulting in the max_zone_append sectors limit | |
220 | * being 0 due to a 0 argument, the chunk_sectors limit (zone size) not set, | |
221 | * or the max_hw_sectors limit not set. | |
222 | */ | |
223 | WARN_ON(!max_sectors); | |
224 | ||
225 | q->limits.max_zone_append_sectors = max_sectors; | |
226 | } | |
227 | EXPORT_SYMBOL_GPL(blk_queue_max_zone_append_sectors); | |
228 | ||
86db1e29 | 229 | /** |
8a78362c | 230 | * blk_queue_max_segments - set max hw segments for a request for this queue |
86db1e29 JA |
231 | * @q: the request queue for the device |
232 | * @max_segments: max number of segments | |
233 | * | |
234 | * Description: | |
235 | * Enables a low level driver to set an upper limit on the number of | |
8a78362c | 236 | * hw data segments in a request. |
86db1e29 | 237 | **/ |
8a78362c | 238 | void blk_queue_max_segments(struct request_queue *q, unsigned short max_segments) |
86db1e29 JA |
239 | { |
240 | if (!max_segments) { | |
241 | max_segments = 1; | |
24c03d47 HH |
242 | printk(KERN_INFO "%s: set to minimum %d\n", |
243 | __func__, max_segments); | |
86db1e29 JA |
244 | } |
245 | ||
8a78362c | 246 | q->limits.max_segments = max_segments; |
86db1e29 | 247 | } |
8a78362c | 248 | EXPORT_SYMBOL(blk_queue_max_segments); |
86db1e29 | 249 | |
1e739730 CH |
250 | /** |
251 | * blk_queue_max_discard_segments - set max segments for discard requests | |
252 | * @q: the request queue for the device | |
253 | * @max_segments: max number of segments | |
254 | * | |
255 | * Description: | |
256 | * Enables a low level driver to set an upper limit on the number of | |
257 | * segments in a discard request. | |
258 | **/ | |
259 | void blk_queue_max_discard_segments(struct request_queue *q, | |
260 | unsigned short max_segments) | |
261 | { | |
262 | q->limits.max_discard_segments = max_segments; | |
263 | } | |
264 | EXPORT_SYMBOL_GPL(blk_queue_max_discard_segments); | |
265 | ||
86db1e29 JA |
266 | /** |
267 | * blk_queue_max_segment_size - set max segment size for blk_rq_map_sg | |
268 | * @q: the request queue for the device | |
269 | * @max_size: max size of segment in bytes | |
270 | * | |
271 | * Description: | |
272 | * Enables a low level driver to set an upper limit on the size of a | |
273 | * coalesced segment | |
274 | **/ | |
275 | void blk_queue_max_segment_size(struct request_queue *q, unsigned int max_size) | |
276 | { | |
09cbfeaf KS |
277 | if (max_size < PAGE_SIZE) { |
278 | max_size = PAGE_SIZE; | |
24c03d47 HH |
279 | printk(KERN_INFO "%s: set to minimum %d\n", |
280 | __func__, max_size); | |
86db1e29 JA |
281 | } |
282 | ||
09324d32 CH |
283 | /* see blk_queue_virt_boundary() for the explanation */ |
284 | WARN_ON_ONCE(q->limits.virt_boundary_mask); | |
285 | ||
025146e1 | 286 | q->limits.max_segment_size = max_size; |
86db1e29 | 287 | } |
86db1e29 JA |
288 | EXPORT_SYMBOL(blk_queue_max_segment_size); |
289 | ||
290 | /** | |
e1defc4f | 291 | * blk_queue_logical_block_size - set logical block size for the queue |
86db1e29 | 292 | * @q: the request queue for the device |
e1defc4f | 293 | * @size: the logical block size, in bytes |
86db1e29 JA |
294 | * |
295 | * Description: | |
e1defc4f MP |
296 | * This should be set to the lowest possible block size that the |
297 | * storage device can address. The default of 512 covers most | |
298 | * hardware. | |
86db1e29 | 299 | **/ |
ad6bf88a | 300 | void blk_queue_logical_block_size(struct request_queue *q, unsigned int size) |
86db1e29 | 301 | { |
817046ec DLM |
302 | struct queue_limits *limits = &q->limits; |
303 | ||
304 | limits->logical_block_size = size; | |
305 | ||
306 | if (limits->physical_block_size < size) | |
307 | limits->physical_block_size = size; | |
c72758f3 | 308 | |
817046ec DLM |
309 | if (limits->io_min < limits->physical_block_size) |
310 | limits->io_min = limits->physical_block_size; | |
c72758f3 | 311 | |
817046ec DLM |
312 | limits->max_hw_sectors = |
313 | round_down(limits->max_hw_sectors, size >> SECTOR_SHIFT); | |
314 | limits->max_sectors = | |
315 | round_down(limits->max_sectors, size >> SECTOR_SHIFT); | |
86db1e29 | 316 | } |
e1defc4f | 317 | EXPORT_SYMBOL(blk_queue_logical_block_size); |
86db1e29 | 318 | |
c72758f3 MP |
319 | /** |
320 | * blk_queue_physical_block_size - set physical block size for the queue | |
321 | * @q: the request queue for the device | |
322 | * @size: the physical block size, in bytes | |
323 | * | |
324 | * Description: | |
325 | * This should be set to the lowest possible sector size that the | |
326 | * hardware can operate on without reverting to read-modify-write | |
327 | * operations. | |
328 | */ | |
892b6f90 | 329 | void blk_queue_physical_block_size(struct request_queue *q, unsigned int size) |
c72758f3 MP |
330 | { |
331 | q->limits.physical_block_size = size; | |
332 | ||
333 | if (q->limits.physical_block_size < q->limits.logical_block_size) | |
334 | q->limits.physical_block_size = q->limits.logical_block_size; | |
335 | ||
336 | if (q->limits.io_min < q->limits.physical_block_size) | |
337 | q->limits.io_min = q->limits.physical_block_size; | |
338 | } | |
339 | EXPORT_SYMBOL(blk_queue_physical_block_size); | |
340 | ||
a805a4fa DLM |
341 | /** |
342 | * blk_queue_zone_write_granularity - set zone write granularity for the queue | |
343 | * @q: the request queue for the zoned device | |
344 | * @size: the zone write granularity size, in bytes | |
345 | * | |
346 | * Description: | |
347 | * This should be set to the lowest possible size allowing to write in | |
348 | * sequential zones of a zoned block device. | |
349 | */ | |
350 | void blk_queue_zone_write_granularity(struct request_queue *q, | |
351 | unsigned int size) | |
352 | { | |
353 | if (WARN_ON_ONCE(!blk_queue_is_zoned(q))) | |
354 | return; | |
355 | ||
356 | q->limits.zone_write_granularity = size; | |
357 | ||
358 | if (q->limits.zone_write_granularity < q->limits.logical_block_size) | |
359 | q->limits.zone_write_granularity = q->limits.logical_block_size; | |
360 | } | |
361 | EXPORT_SYMBOL_GPL(blk_queue_zone_write_granularity); | |
362 | ||
c72758f3 MP |
363 | /** |
364 | * blk_queue_alignment_offset - set physical block alignment offset | |
365 | * @q: the request queue for the device | |
8ebf9756 | 366 | * @offset: alignment offset in bytes |
c72758f3 MP |
367 | * |
368 | * Description: | |
369 | * Some devices are naturally misaligned to compensate for things like | |
370 | * the legacy DOS partition table 63-sector offset. Low-level drivers | |
371 | * should call this function for devices whose first sector is not | |
372 | * naturally aligned. | |
373 | */ | |
374 | void blk_queue_alignment_offset(struct request_queue *q, unsigned int offset) | |
375 | { | |
376 | q->limits.alignment_offset = | |
377 | offset & (q->limits.physical_block_size - 1); | |
378 | q->limits.misaligned = 0; | |
379 | } | |
380 | EXPORT_SYMBOL(blk_queue_alignment_offset); | |
381 | ||
c2e4cd57 CH |
382 | void blk_queue_update_readahead(struct request_queue *q) |
383 | { | |
384 | /* | |
385 | * For read-ahead of large files to be effective, we need to read ahead | |
386 | * at least twice the optimal I/O size. | |
387 | */ | |
388 | q->backing_dev_info->ra_pages = | |
389 | max(queue_io_opt(q) * 2 / PAGE_SIZE, VM_READAHEAD_PAGES); | |
390 | q->backing_dev_info->io_pages = | |
391 | queue_max_sectors(q) >> (PAGE_SHIFT - 9); | |
392 | } | |
393 | EXPORT_SYMBOL_GPL(blk_queue_update_readahead); | |
394 | ||
7c958e32 MP |
395 | /** |
396 | * blk_limits_io_min - set minimum request size for a device | |
397 | * @limits: the queue limits | |
398 | * @min: smallest I/O size in bytes | |
399 | * | |
400 | * Description: | |
401 | * Some devices have an internal block size bigger than the reported | |
402 | * hardware sector size. This function can be used to signal the | |
403 | * smallest I/O the device can perform without incurring a performance | |
404 | * penalty. | |
405 | */ | |
406 | void blk_limits_io_min(struct queue_limits *limits, unsigned int min) | |
407 | { | |
408 | limits->io_min = min; | |
409 | ||
410 | if (limits->io_min < limits->logical_block_size) | |
411 | limits->io_min = limits->logical_block_size; | |
412 | ||
413 | if (limits->io_min < limits->physical_block_size) | |
414 | limits->io_min = limits->physical_block_size; | |
415 | } | |
416 | EXPORT_SYMBOL(blk_limits_io_min); | |
417 | ||
c72758f3 MP |
418 | /** |
419 | * blk_queue_io_min - set minimum request size for the queue | |
420 | * @q: the request queue for the device | |
8ebf9756 | 421 | * @min: smallest I/O size in bytes |
c72758f3 MP |
422 | * |
423 | * Description: | |
7e5f5fb0 MP |
424 | * Storage devices may report a granularity or preferred minimum I/O |
425 | * size which is the smallest request the device can perform without | |
426 | * incurring a performance penalty. For disk drives this is often the | |
427 | * physical block size. For RAID arrays it is often the stripe chunk | |
428 | * size. A properly aligned multiple of minimum_io_size is the | |
429 | * preferred request size for workloads where a high number of I/O | |
430 | * operations is desired. | |
c72758f3 MP |
431 | */ |
432 | void blk_queue_io_min(struct request_queue *q, unsigned int min) | |
433 | { | |
7c958e32 | 434 | blk_limits_io_min(&q->limits, min); |
c72758f3 MP |
435 | } |
436 | EXPORT_SYMBOL(blk_queue_io_min); | |
437 | ||
3c5820c7 MP |
438 | /** |
439 | * blk_limits_io_opt - set optimal request size for a device | |
440 | * @limits: the queue limits | |
441 | * @opt: smallest I/O size in bytes | |
442 | * | |
443 | * Description: | |
444 | * Storage devices may report an optimal I/O size, which is the | |
445 | * device's preferred unit for sustained I/O. This is rarely reported | |
446 | * for disk drives. For RAID arrays it is usually the stripe width or | |
447 | * the internal track size. A properly aligned multiple of | |
448 | * optimal_io_size is the preferred request size for workloads where | |
449 | * sustained throughput is desired. | |
450 | */ | |
451 | void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt) | |
452 | { | |
453 | limits->io_opt = opt; | |
454 | } | |
455 | EXPORT_SYMBOL(blk_limits_io_opt); | |
456 | ||
c72758f3 MP |
457 | /** |
458 | * blk_queue_io_opt - set optimal request size for the queue | |
459 | * @q: the request queue for the device | |
8ebf9756 | 460 | * @opt: optimal request size in bytes |
c72758f3 MP |
461 | * |
462 | * Description: | |
7e5f5fb0 MP |
463 | * Storage devices may report an optimal I/O size, which is the |
464 | * device's preferred unit for sustained I/O. This is rarely reported | |
465 | * for disk drives. For RAID arrays it is usually the stripe width or | |
466 | * the internal track size. A properly aligned multiple of | |
467 | * optimal_io_size is the preferred request size for workloads where | |
468 | * sustained throughput is desired. | |
c72758f3 MP |
469 | */ |
470 | void blk_queue_io_opt(struct request_queue *q, unsigned int opt) | |
471 | { | |
3c5820c7 | 472 | blk_limits_io_opt(&q->limits, opt); |
c2e4cd57 CH |
473 | q->backing_dev_info->ra_pages = |
474 | max(queue_io_opt(q) * 2 / PAGE_SIZE, VM_READAHEAD_PAGES); | |
c72758f3 MP |
475 | } |
476 | EXPORT_SYMBOL(blk_queue_io_opt); | |
477 | ||
97f433c3 MP |
478 | static unsigned int blk_round_down_sectors(unsigned int sectors, unsigned int lbs) |
479 | { | |
480 | sectors = round_down(sectors, lbs >> SECTOR_SHIFT); | |
481 | if (sectors < PAGE_SIZE >> SECTOR_SHIFT) | |
482 | sectors = PAGE_SIZE >> SECTOR_SHIFT; | |
483 | return sectors; | |
484 | } | |
485 | ||
c72758f3 MP |
486 | /** |
487 | * blk_stack_limits - adjust queue_limits for stacked devices | |
81744ee4 MP |
488 | * @t: the stacking driver limits (top device) |
489 | * @b: the underlying queue limits (bottom, component device) | |
e03a72e1 | 490 | * @start: first data sector within component device |
c72758f3 MP |
491 | * |
492 | * Description: | |
81744ee4 MP |
493 | * This function is used by stacking drivers like MD and DM to ensure |
494 | * that all component devices have compatible block sizes and | |
495 | * alignments. The stacking driver must provide a queue_limits | |
496 | * struct (top) and then iteratively call the stacking function for | |
497 | * all component (bottom) devices. The stacking function will | |
498 | * attempt to combine the values and ensure proper alignment. | |
499 | * | |
500 | * Returns 0 if the top and bottom queue_limits are compatible. The | |
501 | * top device's block sizes and alignment offsets may be adjusted to | |
502 | * ensure alignment with the bottom device. If no compatible sizes | |
503 | * and alignments exist, -1 is returned and the resulting top | |
504 | * queue_limits will have the misaligned flag set to indicate that | |
505 | * the alignment_offset is undefined. | |
c72758f3 MP |
506 | */ |
507 | int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, | |
e03a72e1 | 508 | sector_t start) |
c72758f3 | 509 | { |
e03a72e1 | 510 | unsigned int top, bottom, alignment, ret = 0; |
86b37281 | 511 | |
c72758f3 MP |
512 | t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors); |
513 | t->max_hw_sectors = min_not_zero(t->max_hw_sectors, b->max_hw_sectors); | |
ca369d51 | 514 | t->max_dev_sectors = min_not_zero(t->max_dev_sectors, b->max_dev_sectors); |
4363ac7c MP |
515 | t->max_write_same_sectors = min(t->max_write_same_sectors, |
516 | b->max_write_same_sectors); | |
a6f0788e CK |
517 | t->max_write_zeroes_sectors = min(t->max_write_zeroes_sectors, |
518 | b->max_write_zeroes_sectors); | |
0512a75b KB |
519 | t->max_zone_append_sectors = min(t->max_zone_append_sectors, |
520 | b->max_zone_append_sectors); | |
9bb33f24 | 521 | t->bounce = max(t->bounce, b->bounce); |
c72758f3 MP |
522 | |
523 | t->seg_boundary_mask = min_not_zero(t->seg_boundary_mask, | |
524 | b->seg_boundary_mask); | |
03100aad KB |
525 | t->virt_boundary_mask = min_not_zero(t->virt_boundary_mask, |
526 | b->virt_boundary_mask); | |
c72758f3 | 527 | |
8a78362c | 528 | t->max_segments = min_not_zero(t->max_segments, b->max_segments); |
1e739730 CH |
529 | t->max_discard_segments = min_not_zero(t->max_discard_segments, |
530 | b->max_discard_segments); | |
13f05c8d MP |
531 | t->max_integrity_segments = min_not_zero(t->max_integrity_segments, |
532 | b->max_integrity_segments); | |
c72758f3 MP |
533 | |
534 | t->max_segment_size = min_not_zero(t->max_segment_size, | |
535 | b->max_segment_size); | |
536 | ||
fe0b393f MP |
537 | t->misaligned |= b->misaligned; |
538 | ||
e03a72e1 | 539 | alignment = queue_limit_alignment_offset(b, start); |
9504e086 | 540 | |
81744ee4 MP |
541 | /* Bottom device has different alignment. Check that it is |
542 | * compatible with the current top alignment. | |
543 | */ | |
9504e086 MP |
544 | if (t->alignment_offset != alignment) { |
545 | ||
546 | top = max(t->physical_block_size, t->io_min) | |
547 | + t->alignment_offset; | |
81744ee4 | 548 | bottom = max(b->physical_block_size, b->io_min) + alignment; |
9504e086 | 549 | |
81744ee4 | 550 | /* Verify that top and bottom intervals line up */ |
b8839b8c | 551 | if (max(top, bottom) % min(top, bottom)) { |
9504e086 | 552 | t->misaligned = 1; |
fe0b393f MP |
553 | ret = -1; |
554 | } | |
9504e086 MP |
555 | } |
556 | ||
c72758f3 MP |
557 | t->logical_block_size = max(t->logical_block_size, |
558 | b->logical_block_size); | |
559 | ||
560 | t->physical_block_size = max(t->physical_block_size, | |
561 | b->physical_block_size); | |
562 | ||
563 | t->io_min = max(t->io_min, b->io_min); | |
e9637415 | 564 | t->io_opt = lcm_not_zero(t->io_opt, b->io_opt); |
7e7986f9 MS |
565 | |
566 | /* Set non-power-of-2 compatible chunk_sectors boundary */ | |
567 | if (b->chunk_sectors) | |
568 | t->chunk_sectors = gcd(t->chunk_sectors, b->chunk_sectors); | |
9504e086 | 569 | |
81744ee4 | 570 | /* Physical block size a multiple of the logical block size? */ |
9504e086 MP |
571 | if (t->physical_block_size & (t->logical_block_size - 1)) { |
572 | t->physical_block_size = t->logical_block_size; | |
c72758f3 | 573 | t->misaligned = 1; |
fe0b393f | 574 | ret = -1; |
86b37281 MP |
575 | } |
576 | ||
81744ee4 | 577 | /* Minimum I/O a multiple of the physical block size? */ |
9504e086 MP |
578 | if (t->io_min & (t->physical_block_size - 1)) { |
579 | t->io_min = t->physical_block_size; | |
580 | t->misaligned = 1; | |
fe0b393f | 581 | ret = -1; |
c72758f3 MP |
582 | } |
583 | ||
81744ee4 | 584 | /* Optimal I/O a multiple of the physical block size? */ |
9504e086 MP |
585 | if (t->io_opt & (t->physical_block_size - 1)) { |
586 | t->io_opt = 0; | |
587 | t->misaligned = 1; | |
fe0b393f | 588 | ret = -1; |
9504e086 | 589 | } |
c72758f3 | 590 | |
22ada802 MS |
591 | /* chunk_sectors a multiple of the physical block size? */ |
592 | if ((t->chunk_sectors << 9) & (t->physical_block_size - 1)) { | |
593 | t->chunk_sectors = 0; | |
594 | t->misaligned = 1; | |
595 | ret = -1; | |
596 | } | |
597 | ||
c78afc62 KO |
598 | t->raid_partial_stripes_expensive = |
599 | max(t->raid_partial_stripes_expensive, | |
600 | b->raid_partial_stripes_expensive); | |
601 | ||
81744ee4 | 602 | /* Find lowest common alignment_offset */ |
e9637415 | 603 | t->alignment_offset = lcm_not_zero(t->alignment_offset, alignment) |
b8839b8c | 604 | % max(t->physical_block_size, t->io_min); |
86b37281 | 605 | |
81744ee4 | 606 | /* Verify that new alignment_offset is on a logical block boundary */ |
fe0b393f | 607 | if (t->alignment_offset & (t->logical_block_size - 1)) { |
c72758f3 | 608 | t->misaligned = 1; |
fe0b393f MP |
609 | ret = -1; |
610 | } | |
c72758f3 | 611 | |
97f433c3 MP |
612 | t->max_sectors = blk_round_down_sectors(t->max_sectors, t->logical_block_size); |
613 | t->max_hw_sectors = blk_round_down_sectors(t->max_hw_sectors, t->logical_block_size); | |
614 | t->max_dev_sectors = blk_round_down_sectors(t->max_dev_sectors, t->logical_block_size); | |
615 | ||
9504e086 MP |
616 | /* Discard alignment and granularity */ |
617 | if (b->discard_granularity) { | |
e03a72e1 | 618 | alignment = queue_limit_discard_alignment(b, start); |
9504e086 MP |
619 | |
620 | if (t->discard_granularity != 0 && | |
621 | t->discard_alignment != alignment) { | |
622 | top = t->discard_granularity + t->discard_alignment; | |
623 | bottom = b->discard_granularity + alignment; | |
70dd5bf3 | 624 | |
9504e086 | 625 | /* Verify that top and bottom intervals line up */ |
8dd2cb7e | 626 | if ((max(top, bottom) % min(top, bottom)) != 0) |
9504e086 MP |
627 | t->discard_misaligned = 1; |
628 | } | |
629 | ||
81744ee4 MP |
630 | t->max_discard_sectors = min_not_zero(t->max_discard_sectors, |
631 | b->max_discard_sectors); | |
0034af03 JA |
632 | t->max_hw_discard_sectors = min_not_zero(t->max_hw_discard_sectors, |
633 | b->max_hw_discard_sectors); | |
9504e086 MP |
634 | t->discard_granularity = max(t->discard_granularity, |
635 | b->discard_granularity); | |
e9637415 | 636 | t->discard_alignment = lcm_not_zero(t->discard_alignment, alignment) % |
8dd2cb7e | 637 | t->discard_granularity; |
9504e086 | 638 | } |
70dd5bf3 | 639 | |
a805a4fa DLM |
640 | t->zone_write_granularity = max(t->zone_write_granularity, |
641 | b->zone_write_granularity); | |
3093a479 | 642 | t->zoned = max(t->zoned, b->zoned); |
fe0b393f | 643 | return ret; |
c72758f3 | 644 | } |
5d85d324 | 645 | EXPORT_SYMBOL(blk_stack_limits); |
c72758f3 MP |
646 | |
647 | /** | |
648 | * disk_stack_limits - adjust queue limits for stacked drivers | |
77634f33 | 649 | * @disk: MD/DM gendisk (top) |
c72758f3 MP |
650 | * @bdev: the underlying block device (bottom) |
651 | * @offset: offset to beginning of data within component device | |
652 | * | |
653 | * Description: | |
e03a72e1 MP |
654 | * Merges the limits for a top level gendisk and a bottom level |
655 | * block_device. | |
c72758f3 MP |
656 | */ |
657 | void disk_stack_limits(struct gendisk *disk, struct block_device *bdev, | |
658 | sector_t offset) | |
659 | { | |
660 | struct request_queue *t = disk->queue; | |
c72758f3 | 661 | |
9efa82ef CH |
662 | if (blk_stack_limits(&t->limits, &bdev_get_queue(bdev)->limits, |
663 | get_start_sect(bdev) + (offset >> 9)) < 0) { | |
c72758f3 MP |
664 | char top[BDEVNAME_SIZE], bottom[BDEVNAME_SIZE]; |
665 | ||
666 | disk_name(disk, 0, top); | |
667 | bdevname(bdev, bottom); | |
668 | ||
669 | printk(KERN_NOTICE "%s: Warning: Device %s is misaligned\n", | |
670 | top, bottom); | |
671 | } | |
e74d93e9 | 672 | |
c2e4cd57 | 673 | blk_queue_update_readahead(disk->queue); |
c72758f3 MP |
674 | } |
675 | EXPORT_SYMBOL(disk_stack_limits); | |
676 | ||
27f8221a FT |
677 | /** |
678 | * blk_queue_update_dma_pad - update pad mask | |
679 | * @q: the request queue for the device | |
680 | * @mask: pad mask | |
681 | * | |
682 | * Update dma pad mask. | |
683 | * | |
684 | * Appending pad buffer to a request modifies the last entry of a | |
685 | * scatter list such that it includes the pad buffer. | |
686 | **/ | |
687 | void blk_queue_update_dma_pad(struct request_queue *q, unsigned int mask) | |
688 | { | |
689 | if (mask > q->dma_pad_mask) | |
690 | q->dma_pad_mask = mask; | |
691 | } | |
692 | EXPORT_SYMBOL(blk_queue_update_dma_pad); | |
693 | ||
86db1e29 JA |
694 | /** |
695 | * blk_queue_segment_boundary - set boundary rules for segment merging | |
696 | * @q: the request queue for the device | |
697 | * @mask: the memory boundary mask | |
698 | **/ | |
699 | void blk_queue_segment_boundary(struct request_queue *q, unsigned long mask) | |
700 | { | |
09cbfeaf KS |
701 | if (mask < PAGE_SIZE - 1) { |
702 | mask = PAGE_SIZE - 1; | |
24c03d47 HH |
703 | printk(KERN_INFO "%s: set to minimum %lx\n", |
704 | __func__, mask); | |
86db1e29 JA |
705 | } |
706 | ||
025146e1 | 707 | q->limits.seg_boundary_mask = mask; |
86db1e29 | 708 | } |
86db1e29 JA |
709 | EXPORT_SYMBOL(blk_queue_segment_boundary); |
710 | ||
03100aad KB |
711 | /** |
712 | * blk_queue_virt_boundary - set boundary rules for bio merging | |
713 | * @q: the request queue for the device | |
714 | * @mask: the memory boundary mask | |
715 | **/ | |
716 | void blk_queue_virt_boundary(struct request_queue *q, unsigned long mask) | |
717 | { | |
718 | q->limits.virt_boundary_mask = mask; | |
09324d32 CH |
719 | |
720 | /* | |
721 | * Devices that require a virtual boundary do not support scatter/gather | |
722 | * I/O natively, but instead require a descriptor list entry for each | |
723 | * page (which might not be idential to the Linux PAGE_SIZE). Because | |
724 | * of that they are not limited by our notion of "segment size". | |
725 | */ | |
c6c84f78 CH |
726 | if (mask) |
727 | q->limits.max_segment_size = UINT_MAX; | |
03100aad KB |
728 | } |
729 | EXPORT_SYMBOL(blk_queue_virt_boundary); | |
730 | ||
86db1e29 JA |
731 | /** |
732 | * blk_queue_dma_alignment - set dma length and memory alignment | |
733 | * @q: the request queue for the device | |
734 | * @mask: alignment mask | |
735 | * | |
736 | * description: | |
710027a4 | 737 | * set required memory and length alignment for direct dma transactions. |
8feb4d20 | 738 | * this is used when building direct io requests for the queue. |
86db1e29 JA |
739 | * |
740 | **/ | |
741 | void blk_queue_dma_alignment(struct request_queue *q, int mask) | |
742 | { | |
743 | q->dma_alignment = mask; | |
744 | } | |
86db1e29 JA |
745 | EXPORT_SYMBOL(blk_queue_dma_alignment); |
746 | ||
747 | /** | |
748 | * blk_queue_update_dma_alignment - update dma length and memory alignment | |
749 | * @q: the request queue for the device | |
750 | * @mask: alignment mask | |
751 | * | |
752 | * description: | |
710027a4 | 753 | * update required memory and length alignment for direct dma transactions. |
86db1e29 JA |
754 | * If the requested alignment is larger than the current alignment, then |
755 | * the current queue alignment is updated to the new value, otherwise it | |
756 | * is left alone. The design of this is to allow multiple objects | |
757 | * (driver, device, transport etc) to set their respective | |
758 | * alignments without having them interfere. | |
759 | * | |
760 | **/ | |
761 | void blk_queue_update_dma_alignment(struct request_queue *q, int mask) | |
762 | { | |
763 | BUG_ON(mask > PAGE_SIZE); | |
764 | ||
765 | if (mask > q->dma_alignment) | |
766 | q->dma_alignment = mask; | |
767 | } | |
86db1e29 JA |
768 | EXPORT_SYMBOL(blk_queue_update_dma_alignment); |
769 | ||
d278d4a8 JA |
770 | /** |
771 | * blk_set_queue_depth - tell the block layer about the device queue depth | |
772 | * @q: the request queue for the device | |
773 | * @depth: queue depth | |
774 | * | |
775 | */ | |
776 | void blk_set_queue_depth(struct request_queue *q, unsigned int depth) | |
777 | { | |
778 | q->queue_depth = depth; | |
9677a3e0 | 779 | rq_qos_queue_depth_changed(q); |
d278d4a8 JA |
780 | } |
781 | EXPORT_SYMBOL(blk_set_queue_depth); | |
782 | ||
93e9d8e8 JA |
783 | /** |
784 | * blk_queue_write_cache - configure queue's write cache | |
785 | * @q: the request queue for the device | |
786 | * @wc: write back cache on or off | |
787 | * @fua: device supports FUA writes, if true | |
788 | * | |
789 | * Tell the block layer about the write cache of @q. | |
790 | */ | |
791 | void blk_queue_write_cache(struct request_queue *q, bool wc, bool fua) | |
792 | { | |
c888a8f9 | 793 | if (wc) |
57d74df9 | 794 | blk_queue_flag_set(QUEUE_FLAG_WC, q); |
c888a8f9 | 795 | else |
57d74df9 | 796 | blk_queue_flag_clear(QUEUE_FLAG_WC, q); |
c888a8f9 | 797 | if (fua) |
57d74df9 | 798 | blk_queue_flag_set(QUEUE_FLAG_FUA, q); |
c888a8f9 | 799 | else |
57d74df9 | 800 | blk_queue_flag_clear(QUEUE_FLAG_FUA, q); |
87760e5e | 801 | |
a7905043 | 802 | wbt_set_write_cache(q, test_bit(QUEUE_FLAG_WC, &q->queue_flags)); |
93e9d8e8 JA |
803 | } |
804 | EXPORT_SYMBOL_GPL(blk_queue_write_cache); | |
805 | ||
68c43f13 DLM |
806 | /** |
807 | * blk_queue_required_elevator_features - Set a queue required elevator features | |
808 | * @q: the request queue for the target device | |
809 | * @features: Required elevator features OR'ed together | |
810 | * | |
811 | * Tell the block layer that for the device controlled through @q, only the | |
812 | * only elevators that can be used are those that implement at least the set of | |
813 | * features specified by @features. | |
814 | */ | |
815 | void blk_queue_required_elevator_features(struct request_queue *q, | |
816 | unsigned int features) | |
817 | { | |
818 | q->required_elevator_features = features; | |
819 | } | |
820 | EXPORT_SYMBOL_GPL(blk_queue_required_elevator_features); | |
821 | ||
45147fb5 YS |
822 | /** |
823 | * blk_queue_can_use_dma_map_merging - configure queue for merging segments. | |
824 | * @q: the request queue for the device | |
825 | * @dev: the device pointer for dma | |
826 | * | |
827 | * Tell the block layer about merging the segments by dma map of @q. | |
828 | */ | |
829 | bool blk_queue_can_use_dma_map_merging(struct request_queue *q, | |
830 | struct device *dev) | |
831 | { | |
832 | unsigned long boundary = dma_get_merge_boundary(dev); | |
833 | ||
834 | if (!boundary) | |
835 | return false; | |
836 | ||
837 | /* No need to update max_segment_size. see blk_queue_virt_boundary() */ | |
838 | blk_queue_virt_boundary(q, boundary); | |
839 | ||
840 | return true; | |
841 | } | |
842 | EXPORT_SYMBOL_GPL(blk_queue_can_use_dma_map_merging); | |
843 | ||
27ba3e8f DLM |
844 | /** |
845 | * blk_queue_set_zoned - configure a disk queue zoned model. | |
846 | * @disk: the gendisk of the queue to configure | |
847 | * @model: the zoned model to set | |
848 | * | |
849 | * Set the zoned model of the request queue of @disk according to @model. | |
850 | * When @model is BLK_ZONED_HM (host managed), this should be called only | |
851 | * if zoned block device support is enabled (CONFIG_BLK_DEV_ZONED option). | |
852 | * If @model specifies BLK_ZONED_HA (host aware), the effective model used | |
853 | * depends on CONFIG_BLK_DEV_ZONED settings and on the existence of partitions | |
854 | * on the disk. | |
855 | */ | |
856 | void blk_queue_set_zoned(struct gendisk *disk, enum blk_zoned_model model) | |
857 | { | |
a805a4fa DLM |
858 | struct request_queue *q = disk->queue; |
859 | ||
27ba3e8f DLM |
860 | switch (model) { |
861 | case BLK_ZONED_HM: | |
862 | /* | |
863 | * Host managed devices are supported only if | |
864 | * CONFIG_BLK_DEV_ZONED is enabled. | |
865 | */ | |
866 | WARN_ON_ONCE(!IS_ENABLED(CONFIG_BLK_DEV_ZONED)); | |
867 | break; | |
868 | case BLK_ZONED_HA: | |
869 | /* | |
870 | * Host aware devices can be treated either as regular block | |
871 | * devices (similar to drive managed devices) or as zoned block | |
872 | * devices to take advantage of the zone command set, similarly | |
873 | * to host managed devices. We try the latter if there are no | |
874 | * partitions and zoned block device support is enabled, else | |
875 | * we do nothing special as far as the block layer is concerned. | |
876 | */ | |
877 | if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) || | |
a33df75c | 878 | !xa_empty(&disk->part_tbl)) |
27ba3e8f DLM |
879 | model = BLK_ZONED_NONE; |
880 | break; | |
881 | case BLK_ZONED_NONE: | |
882 | default: | |
883 | if (WARN_ON_ONCE(model != BLK_ZONED_NONE)) | |
884 | model = BLK_ZONED_NONE; | |
885 | break; | |
886 | } | |
887 | ||
a805a4fa DLM |
888 | q->limits.zoned = model; |
889 | if (model != BLK_ZONED_NONE) { | |
890 | /* | |
891 | * Set the zone write granularity to the device logical block | |
892 | * size by default. The driver can change this value if needed. | |
893 | */ | |
894 | blk_queue_zone_write_granularity(q, | |
895 | queue_logical_block_size(q)); | |
508aebb8 DLM |
896 | } else { |
897 | blk_queue_clear_zone_settings(q); | |
a805a4fa | 898 | } |
27ba3e8f DLM |
899 | } |
900 | EXPORT_SYMBOL_GPL(blk_queue_set_zoned); |