]>
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; |
a6f0788e | 45 | lim->max_write_zeroes_sectors = 0; |
0512a75b | 46 | lim->max_zone_append_sectors = 0; |
86b37281 | 47 | lim->max_discard_sectors = 0; |
0034af03 | 48 | lim->max_hw_discard_sectors = 0; |
86b37281 MP |
49 | lim->discard_granularity = 0; |
50 | lim->discard_alignment = 0; | |
51 | lim->discard_misaligned = 0; | |
e475bba2 | 52 | lim->logical_block_size = lim->physical_block_size = lim->io_min = 512; |
9bb33f24 | 53 | lim->bounce = BLK_BOUNCE_NONE; |
e475bba2 MP |
54 | lim->alignment_offset = 0; |
55 | lim->io_opt = 0; | |
56 | lim->misaligned = 0; | |
797476b8 | 57 | lim->zoned = BLK_ZONED_NONE; |
a805a4fa | 58 | lim->zone_write_granularity = 0; |
e475bba2 MP |
59 | } |
60 | EXPORT_SYMBOL(blk_set_default_limits); | |
61 | ||
b1bd055d MP |
62 | /** |
63 | * blk_set_stacking_limits - set default limits for stacking devices | |
64 | * @lim: the queue_limits structure to reset | |
65 | * | |
66 | * Description: | |
67 | * Returns a queue_limit struct to its default state. Should be used | |
68 | * by stacking drivers like DM that have no internal limits. | |
69 | */ | |
70 | void blk_set_stacking_limits(struct queue_limits *lim) | |
71 | { | |
72 | blk_set_default_limits(lim); | |
73 | ||
74 | /* Inherit limits from component devices */ | |
b1bd055d | 75 | lim->max_segments = USHRT_MAX; |
42c9cdfe | 76 | lim->max_discard_segments = USHRT_MAX; |
b1bd055d | 77 | lim->max_hw_sectors = UINT_MAX; |
d82ae52e | 78 | lim->max_segment_size = UINT_MAX; |
fe86cdce | 79 | lim->max_sectors = UINT_MAX; |
ca369d51 | 80 | lim->max_dev_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 | |
d152c682 | 142 | if (!q->disk) |
edb0872f | 143 | return; |
d152c682 | 144 | q->disk->bdi->io_pages = max_sectors >> (PAGE_SHIFT - 9); |
86db1e29 | 145 | } |
086fa5ff | 146 | EXPORT_SYMBOL(blk_queue_max_hw_sectors); |
86db1e29 | 147 | |
762380ad JA |
148 | /** |
149 | * blk_queue_chunk_sectors - set size of the chunk for this queue | |
150 | * @q: the request queue for the device | |
151 | * @chunk_sectors: chunk sectors in the usual 512b unit | |
152 | * | |
153 | * Description: | |
154 | * If a driver doesn't want IOs to cross a given chunk size, it can set | |
07d098e6 MS |
155 | * this limit and prevent merging across chunks. Note that the block layer |
156 | * must accept a page worth of data at any offset. So if the crossing of | |
157 | * chunks is a hard limitation in the driver, it must still be prepared | |
158 | * to split single page bios. | |
762380ad JA |
159 | **/ |
160 | void blk_queue_chunk_sectors(struct request_queue *q, unsigned int chunk_sectors) | |
161 | { | |
762380ad JA |
162 | q->limits.chunk_sectors = chunk_sectors; |
163 | } | |
164 | EXPORT_SYMBOL(blk_queue_chunk_sectors); | |
165 | ||
67efc925 CH |
166 | /** |
167 | * blk_queue_max_discard_sectors - set max sectors for a single discard | |
168 | * @q: the request queue for the device | |
c7ebf065 | 169 | * @max_discard_sectors: maximum number of sectors to discard |
67efc925 CH |
170 | **/ |
171 | void blk_queue_max_discard_sectors(struct request_queue *q, | |
172 | unsigned int max_discard_sectors) | |
173 | { | |
0034af03 | 174 | q->limits.max_hw_discard_sectors = max_discard_sectors; |
67efc925 CH |
175 | q->limits.max_discard_sectors = max_discard_sectors; |
176 | } | |
177 | EXPORT_SYMBOL(blk_queue_max_discard_sectors); | |
178 | ||
a6f0788e CK |
179 | /** |
180 | * blk_queue_max_write_zeroes_sectors - set max sectors for a single | |
181 | * write zeroes | |
182 | * @q: the request queue for the device | |
183 | * @max_write_zeroes_sectors: maximum number of sectors to write per command | |
184 | **/ | |
185 | void blk_queue_max_write_zeroes_sectors(struct request_queue *q, | |
186 | unsigned int max_write_zeroes_sectors) | |
187 | { | |
188 | q->limits.max_write_zeroes_sectors = max_write_zeroes_sectors; | |
189 | } | |
190 | EXPORT_SYMBOL(blk_queue_max_write_zeroes_sectors); | |
191 | ||
0512a75b KB |
192 | /** |
193 | * blk_queue_max_zone_append_sectors - set max sectors for a single zone append | |
194 | * @q: the request queue for the device | |
195 | * @max_zone_append_sectors: maximum number of sectors to write per command | |
196 | **/ | |
197 | void blk_queue_max_zone_append_sectors(struct request_queue *q, | |
198 | unsigned int max_zone_append_sectors) | |
199 | { | |
200 | unsigned int max_sectors; | |
201 | ||
202 | if (WARN_ON(!blk_queue_is_zoned(q))) | |
203 | return; | |
204 | ||
205 | max_sectors = min(q->limits.max_hw_sectors, max_zone_append_sectors); | |
206 | max_sectors = min(q->limits.chunk_sectors, max_sectors); | |
207 | ||
208 | /* | |
209 | * Signal eventual driver bugs resulting in the max_zone_append sectors limit | |
210 | * being 0 due to a 0 argument, the chunk_sectors limit (zone size) not set, | |
211 | * or the max_hw_sectors limit not set. | |
212 | */ | |
213 | WARN_ON(!max_sectors); | |
214 | ||
215 | q->limits.max_zone_append_sectors = max_sectors; | |
216 | } | |
217 | EXPORT_SYMBOL_GPL(blk_queue_max_zone_append_sectors); | |
218 | ||
86db1e29 | 219 | /** |
8a78362c | 220 | * blk_queue_max_segments - set max hw segments for a request for this queue |
86db1e29 JA |
221 | * @q: the request queue for the device |
222 | * @max_segments: max number of segments | |
223 | * | |
224 | * Description: | |
225 | * Enables a low level driver to set an upper limit on the number of | |
8a78362c | 226 | * hw data segments in a request. |
86db1e29 | 227 | **/ |
8a78362c | 228 | void blk_queue_max_segments(struct request_queue *q, unsigned short max_segments) |
86db1e29 JA |
229 | { |
230 | if (!max_segments) { | |
231 | max_segments = 1; | |
24c03d47 HH |
232 | printk(KERN_INFO "%s: set to minimum %d\n", |
233 | __func__, max_segments); | |
86db1e29 JA |
234 | } |
235 | ||
8a78362c | 236 | q->limits.max_segments = max_segments; |
86db1e29 | 237 | } |
8a78362c | 238 | EXPORT_SYMBOL(blk_queue_max_segments); |
86db1e29 | 239 | |
1e739730 CH |
240 | /** |
241 | * blk_queue_max_discard_segments - set max segments for discard requests | |
242 | * @q: the request queue for the device | |
243 | * @max_segments: max number of segments | |
244 | * | |
245 | * Description: | |
246 | * Enables a low level driver to set an upper limit on the number of | |
247 | * segments in a discard request. | |
248 | **/ | |
249 | void blk_queue_max_discard_segments(struct request_queue *q, | |
250 | unsigned short max_segments) | |
251 | { | |
252 | q->limits.max_discard_segments = max_segments; | |
253 | } | |
254 | EXPORT_SYMBOL_GPL(blk_queue_max_discard_segments); | |
255 | ||
86db1e29 JA |
256 | /** |
257 | * blk_queue_max_segment_size - set max segment size for blk_rq_map_sg | |
258 | * @q: the request queue for the device | |
259 | * @max_size: max size of segment in bytes | |
260 | * | |
261 | * Description: | |
262 | * Enables a low level driver to set an upper limit on the size of a | |
263 | * coalesced segment | |
264 | **/ | |
265 | void blk_queue_max_segment_size(struct request_queue *q, unsigned int max_size) | |
266 | { | |
09cbfeaf KS |
267 | if (max_size < PAGE_SIZE) { |
268 | max_size = PAGE_SIZE; | |
24c03d47 HH |
269 | printk(KERN_INFO "%s: set to minimum %d\n", |
270 | __func__, max_size); | |
86db1e29 JA |
271 | } |
272 | ||
09324d32 CH |
273 | /* see blk_queue_virt_boundary() for the explanation */ |
274 | WARN_ON_ONCE(q->limits.virt_boundary_mask); | |
275 | ||
025146e1 | 276 | q->limits.max_segment_size = max_size; |
86db1e29 | 277 | } |
86db1e29 JA |
278 | EXPORT_SYMBOL(blk_queue_max_segment_size); |
279 | ||
280 | /** | |
e1defc4f | 281 | * blk_queue_logical_block_size - set logical block size for the queue |
86db1e29 | 282 | * @q: the request queue for the device |
e1defc4f | 283 | * @size: the logical block size, in bytes |
86db1e29 JA |
284 | * |
285 | * Description: | |
e1defc4f MP |
286 | * This should be set to the lowest possible block size that the |
287 | * storage device can address. The default of 512 covers most | |
288 | * hardware. | |
86db1e29 | 289 | **/ |
ad6bf88a | 290 | void blk_queue_logical_block_size(struct request_queue *q, unsigned int size) |
86db1e29 | 291 | { |
817046ec DLM |
292 | struct queue_limits *limits = &q->limits; |
293 | ||
294 | limits->logical_block_size = size; | |
295 | ||
296 | if (limits->physical_block_size < size) | |
297 | limits->physical_block_size = size; | |
c72758f3 | 298 | |
817046ec DLM |
299 | if (limits->io_min < limits->physical_block_size) |
300 | limits->io_min = limits->physical_block_size; | |
c72758f3 | 301 | |
817046ec DLM |
302 | limits->max_hw_sectors = |
303 | round_down(limits->max_hw_sectors, size >> SECTOR_SHIFT); | |
304 | limits->max_sectors = | |
305 | round_down(limits->max_sectors, size >> SECTOR_SHIFT); | |
86db1e29 | 306 | } |
e1defc4f | 307 | EXPORT_SYMBOL(blk_queue_logical_block_size); |
86db1e29 | 308 | |
c72758f3 MP |
309 | /** |
310 | * blk_queue_physical_block_size - set physical block size for the queue | |
311 | * @q: the request queue for the device | |
312 | * @size: the physical block size, in bytes | |
313 | * | |
314 | * Description: | |
315 | * This should be set to the lowest possible sector size that the | |
316 | * hardware can operate on without reverting to read-modify-write | |
317 | * operations. | |
318 | */ | |
892b6f90 | 319 | void blk_queue_physical_block_size(struct request_queue *q, unsigned int size) |
c72758f3 MP |
320 | { |
321 | q->limits.physical_block_size = size; | |
322 | ||
323 | if (q->limits.physical_block_size < q->limits.logical_block_size) | |
324 | q->limits.physical_block_size = q->limits.logical_block_size; | |
325 | ||
326 | if (q->limits.io_min < q->limits.physical_block_size) | |
327 | q->limits.io_min = q->limits.physical_block_size; | |
328 | } | |
329 | EXPORT_SYMBOL(blk_queue_physical_block_size); | |
330 | ||
a805a4fa DLM |
331 | /** |
332 | * blk_queue_zone_write_granularity - set zone write granularity for the queue | |
333 | * @q: the request queue for the zoned device | |
334 | * @size: the zone write granularity size, in bytes | |
335 | * | |
336 | * Description: | |
337 | * This should be set to the lowest possible size allowing to write in | |
338 | * sequential zones of a zoned block device. | |
339 | */ | |
340 | void blk_queue_zone_write_granularity(struct request_queue *q, | |
341 | unsigned int size) | |
342 | { | |
343 | if (WARN_ON_ONCE(!blk_queue_is_zoned(q))) | |
344 | return; | |
345 | ||
346 | q->limits.zone_write_granularity = size; | |
347 | ||
348 | if (q->limits.zone_write_granularity < q->limits.logical_block_size) | |
349 | q->limits.zone_write_granularity = q->limits.logical_block_size; | |
350 | } | |
351 | EXPORT_SYMBOL_GPL(blk_queue_zone_write_granularity); | |
352 | ||
c72758f3 MP |
353 | /** |
354 | * blk_queue_alignment_offset - set physical block alignment offset | |
355 | * @q: the request queue for the device | |
8ebf9756 | 356 | * @offset: alignment offset in bytes |
c72758f3 MP |
357 | * |
358 | * Description: | |
359 | * Some devices are naturally misaligned to compensate for things like | |
360 | * the legacy DOS partition table 63-sector offset. Low-level drivers | |
361 | * should call this function for devices whose first sector is not | |
362 | * naturally aligned. | |
363 | */ | |
364 | void blk_queue_alignment_offset(struct request_queue *q, unsigned int offset) | |
365 | { | |
366 | q->limits.alignment_offset = | |
367 | offset & (q->limits.physical_block_size - 1); | |
368 | q->limits.misaligned = 0; | |
369 | } | |
370 | EXPORT_SYMBOL(blk_queue_alignment_offset); | |
371 | ||
471aa704 | 372 | void disk_update_readahead(struct gendisk *disk) |
c2e4cd57 | 373 | { |
471aa704 CH |
374 | struct request_queue *q = disk->queue; |
375 | ||
c2e4cd57 CH |
376 | /* |
377 | * For read-ahead of large files to be effective, we need to read ahead | |
378 | * at least twice the optimal I/O size. | |
379 | */ | |
edb0872f | 380 | disk->bdi->ra_pages = |
c2e4cd57 | 381 | max(queue_io_opt(q) * 2 / PAGE_SIZE, VM_READAHEAD_PAGES); |
edb0872f | 382 | disk->bdi->io_pages = queue_max_sectors(q) >> (PAGE_SHIFT - 9); |
c2e4cd57 | 383 | } |
471aa704 | 384 | EXPORT_SYMBOL_GPL(disk_update_readahead); |
c2e4cd57 | 385 | |
7c958e32 MP |
386 | /** |
387 | * blk_limits_io_min - set minimum request size for a device | |
388 | * @limits: the queue limits | |
389 | * @min: smallest I/O size in bytes | |
390 | * | |
391 | * Description: | |
392 | * Some devices have an internal block size bigger than the reported | |
393 | * hardware sector size. This function can be used to signal the | |
394 | * smallest I/O the device can perform without incurring a performance | |
395 | * penalty. | |
396 | */ | |
397 | void blk_limits_io_min(struct queue_limits *limits, unsigned int min) | |
398 | { | |
399 | limits->io_min = min; | |
400 | ||
401 | if (limits->io_min < limits->logical_block_size) | |
402 | limits->io_min = limits->logical_block_size; | |
403 | ||
404 | if (limits->io_min < limits->physical_block_size) | |
405 | limits->io_min = limits->physical_block_size; | |
406 | } | |
407 | EXPORT_SYMBOL(blk_limits_io_min); | |
408 | ||
c72758f3 MP |
409 | /** |
410 | * blk_queue_io_min - set minimum request size for the queue | |
411 | * @q: the request queue for the device | |
8ebf9756 | 412 | * @min: smallest I/O size in bytes |
c72758f3 MP |
413 | * |
414 | * Description: | |
7e5f5fb0 MP |
415 | * Storage devices may report a granularity or preferred minimum I/O |
416 | * size which is the smallest request the device can perform without | |
417 | * incurring a performance penalty. For disk drives this is often the | |
418 | * physical block size. For RAID arrays it is often the stripe chunk | |
419 | * size. A properly aligned multiple of minimum_io_size is the | |
420 | * preferred request size for workloads where a high number of I/O | |
421 | * operations is desired. | |
c72758f3 MP |
422 | */ |
423 | void blk_queue_io_min(struct request_queue *q, unsigned int min) | |
424 | { | |
7c958e32 | 425 | blk_limits_io_min(&q->limits, min); |
c72758f3 MP |
426 | } |
427 | EXPORT_SYMBOL(blk_queue_io_min); | |
428 | ||
3c5820c7 MP |
429 | /** |
430 | * blk_limits_io_opt - set optimal request size for a device | |
431 | * @limits: the queue limits | |
432 | * @opt: smallest I/O size in bytes | |
433 | * | |
434 | * Description: | |
435 | * Storage devices may report an optimal I/O size, which is the | |
436 | * device's preferred unit for sustained I/O. This is rarely reported | |
437 | * for disk drives. For RAID arrays it is usually the stripe width or | |
438 | * the internal track size. A properly aligned multiple of | |
439 | * optimal_io_size is the preferred request size for workloads where | |
440 | * sustained throughput is desired. | |
441 | */ | |
442 | void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt) | |
443 | { | |
444 | limits->io_opt = opt; | |
445 | } | |
446 | EXPORT_SYMBOL(blk_limits_io_opt); | |
447 | ||
c72758f3 MP |
448 | /** |
449 | * blk_queue_io_opt - set optimal request size for the queue | |
450 | * @q: the request queue for the device | |
8ebf9756 | 451 | * @opt: optimal request size in bytes |
c72758f3 MP |
452 | * |
453 | * Description: | |
7e5f5fb0 MP |
454 | * Storage devices may report an optimal I/O size, which is the |
455 | * device's preferred unit for sustained I/O. This is rarely reported | |
456 | * for disk drives. For RAID arrays it is usually the stripe width or | |
457 | * the internal track size. A properly aligned multiple of | |
458 | * optimal_io_size is the preferred request size for workloads where | |
459 | * sustained throughput is desired. | |
c72758f3 MP |
460 | */ |
461 | void blk_queue_io_opt(struct request_queue *q, unsigned int opt) | |
462 | { | |
3c5820c7 | 463 | blk_limits_io_opt(&q->limits, opt); |
d152c682 | 464 | if (!q->disk) |
edb0872f | 465 | return; |
d152c682 | 466 | q->disk->bdi->ra_pages = |
c2e4cd57 | 467 | max(queue_io_opt(q) * 2 / PAGE_SIZE, VM_READAHEAD_PAGES); |
c72758f3 MP |
468 | } |
469 | EXPORT_SYMBOL(blk_queue_io_opt); | |
470 | ||
89098b07 CH |
471 | static int queue_limit_alignment_offset(struct queue_limits *lim, |
472 | sector_t sector) | |
473 | { | |
474 | unsigned int granularity = max(lim->physical_block_size, lim->io_min); | |
475 | unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT) | |
476 | << SECTOR_SHIFT; | |
477 | ||
478 | return (granularity + lim->alignment_offset - alignment) % granularity; | |
479 | } | |
480 | ||
97f433c3 MP |
481 | static unsigned int blk_round_down_sectors(unsigned int sectors, unsigned int lbs) |
482 | { | |
483 | sectors = round_down(sectors, lbs >> SECTOR_SHIFT); | |
484 | if (sectors < PAGE_SIZE >> SECTOR_SHIFT) | |
485 | sectors = PAGE_SIZE >> SECTOR_SHIFT; | |
486 | return sectors; | |
487 | } | |
488 | ||
c72758f3 MP |
489 | /** |
490 | * blk_stack_limits - adjust queue_limits for stacked devices | |
81744ee4 MP |
491 | * @t: the stacking driver limits (top device) |
492 | * @b: the underlying queue limits (bottom, component device) | |
e03a72e1 | 493 | * @start: first data sector within component device |
c72758f3 MP |
494 | * |
495 | * Description: | |
81744ee4 MP |
496 | * This function is used by stacking drivers like MD and DM to ensure |
497 | * that all component devices have compatible block sizes and | |
498 | * alignments. The stacking driver must provide a queue_limits | |
499 | * struct (top) and then iteratively call the stacking function for | |
500 | * all component (bottom) devices. The stacking function will | |
501 | * attempt to combine the values and ensure proper alignment. | |
502 | * | |
503 | * Returns 0 if the top and bottom queue_limits are compatible. The | |
504 | * top device's block sizes and alignment offsets may be adjusted to | |
505 | * ensure alignment with the bottom device. If no compatible sizes | |
506 | * and alignments exist, -1 is returned and the resulting top | |
507 | * queue_limits will have the misaligned flag set to indicate that | |
508 | * the alignment_offset is undefined. | |
c72758f3 MP |
509 | */ |
510 | int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, | |
e03a72e1 | 511 | sector_t start) |
c72758f3 | 512 | { |
e03a72e1 | 513 | unsigned int top, bottom, alignment, ret = 0; |
86b37281 | 514 | |
c72758f3 MP |
515 | t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors); |
516 | t->max_hw_sectors = min_not_zero(t->max_hw_sectors, b->max_hw_sectors); | |
ca369d51 | 517 | t->max_dev_sectors = min_not_zero(t->max_dev_sectors, b->max_dev_sectors); |
a6f0788e CK |
518 | t->max_write_zeroes_sectors = min(t->max_write_zeroes_sectors, |
519 | b->max_write_zeroes_sectors); | |
0512a75b KB |
520 | t->max_zone_append_sectors = min(t->max_zone_append_sectors, |
521 | b->max_zone_append_sectors); | |
9bb33f24 | 522 | t->bounce = max(t->bounce, b->bounce); |
c72758f3 MP |
523 | |
524 | t->seg_boundary_mask = min_not_zero(t->seg_boundary_mask, | |
525 | b->seg_boundary_mask); | |
03100aad KB |
526 | t->virt_boundary_mask = min_not_zero(t->virt_boundary_mask, |
527 | b->virt_boundary_mask); | |
c72758f3 | 528 | |
8a78362c | 529 | t->max_segments = min_not_zero(t->max_segments, b->max_segments); |
1e739730 CH |
530 | t->max_discard_segments = min_not_zero(t->max_discard_segments, |
531 | b->max_discard_segments); | |
13f05c8d MP |
532 | t->max_integrity_segments = min_not_zero(t->max_integrity_segments, |
533 | b->max_integrity_segments); | |
c72758f3 MP |
534 | |
535 | t->max_segment_size = min_not_zero(t->max_segment_size, | |
536 | b->max_segment_size); | |
537 | ||
fe0b393f MP |
538 | t->misaligned |= b->misaligned; |
539 | ||
e03a72e1 | 540 | alignment = queue_limit_alignment_offset(b, start); |
9504e086 | 541 | |
81744ee4 MP |
542 | /* Bottom device has different alignment. Check that it is |
543 | * compatible with the current top alignment. | |
544 | */ | |
9504e086 MP |
545 | if (t->alignment_offset != alignment) { |
546 | ||
547 | top = max(t->physical_block_size, t->io_min) | |
548 | + t->alignment_offset; | |
81744ee4 | 549 | bottom = max(b->physical_block_size, b->io_min) + alignment; |
9504e086 | 550 | |
81744ee4 | 551 | /* Verify that top and bottom intervals line up */ |
b8839b8c | 552 | if (max(top, bottom) % min(top, bottom)) { |
9504e086 | 553 | t->misaligned = 1; |
fe0b393f MP |
554 | ret = -1; |
555 | } | |
9504e086 MP |
556 | } |
557 | ||
c72758f3 MP |
558 | t->logical_block_size = max(t->logical_block_size, |
559 | b->logical_block_size); | |
560 | ||
561 | t->physical_block_size = max(t->physical_block_size, | |
562 | b->physical_block_size); | |
563 | ||
564 | t->io_min = max(t->io_min, b->io_min); | |
e9637415 | 565 | t->io_opt = lcm_not_zero(t->io_opt, b->io_opt); |
7e7986f9 MS |
566 | |
567 | /* Set non-power-of-2 compatible chunk_sectors boundary */ | |
568 | if (b->chunk_sectors) | |
569 | t->chunk_sectors = gcd(t->chunk_sectors, b->chunk_sectors); | |
9504e086 | 570 | |
81744ee4 | 571 | /* Physical block size a multiple of the logical block size? */ |
9504e086 MP |
572 | if (t->physical_block_size & (t->logical_block_size - 1)) { |
573 | t->physical_block_size = t->logical_block_size; | |
c72758f3 | 574 | t->misaligned = 1; |
fe0b393f | 575 | ret = -1; |
86b37281 MP |
576 | } |
577 | ||
81744ee4 | 578 | /* Minimum I/O a multiple of the physical block size? */ |
9504e086 MP |
579 | if (t->io_min & (t->physical_block_size - 1)) { |
580 | t->io_min = t->physical_block_size; | |
581 | t->misaligned = 1; | |
fe0b393f | 582 | ret = -1; |
c72758f3 MP |
583 | } |
584 | ||
81744ee4 | 585 | /* Optimal I/O a multiple of the physical block size? */ |
9504e086 MP |
586 | if (t->io_opt & (t->physical_block_size - 1)) { |
587 | t->io_opt = 0; | |
588 | t->misaligned = 1; | |
fe0b393f | 589 | ret = -1; |
9504e086 | 590 | } |
c72758f3 | 591 | |
22ada802 MS |
592 | /* chunk_sectors a multiple of the physical block size? */ |
593 | if ((t->chunk_sectors << 9) & (t->physical_block_size - 1)) { | |
594 | t->chunk_sectors = 0; | |
595 | t->misaligned = 1; | |
596 | ret = -1; | |
597 | } | |
598 | ||
c78afc62 KO |
599 | t->raid_partial_stripes_expensive = |
600 | max(t->raid_partial_stripes_expensive, | |
601 | b->raid_partial_stripes_expensive); | |
602 | ||
81744ee4 | 603 | /* Find lowest common alignment_offset */ |
e9637415 | 604 | t->alignment_offset = lcm_not_zero(t->alignment_offset, alignment) |
b8839b8c | 605 | % max(t->physical_block_size, t->io_min); |
86b37281 | 606 | |
81744ee4 | 607 | /* Verify that new alignment_offset is on a logical block boundary */ |
fe0b393f | 608 | if (t->alignment_offset & (t->logical_block_size - 1)) { |
c72758f3 | 609 | t->misaligned = 1; |
fe0b393f MP |
610 | ret = -1; |
611 | } | |
c72758f3 | 612 | |
97f433c3 MP |
613 | t->max_sectors = blk_round_down_sectors(t->max_sectors, t->logical_block_size); |
614 | t->max_hw_sectors = blk_round_down_sectors(t->max_hw_sectors, t->logical_block_size); | |
615 | t->max_dev_sectors = blk_round_down_sectors(t->max_dev_sectors, t->logical_block_size); | |
616 | ||
9504e086 MP |
617 | /* Discard alignment and granularity */ |
618 | if (b->discard_granularity) { | |
e03a72e1 | 619 | alignment = queue_limit_discard_alignment(b, start); |
9504e086 MP |
620 | |
621 | if (t->discard_granularity != 0 && | |
622 | t->discard_alignment != alignment) { | |
623 | top = t->discard_granularity + t->discard_alignment; | |
624 | bottom = b->discard_granularity + alignment; | |
70dd5bf3 | 625 | |
9504e086 | 626 | /* Verify that top and bottom intervals line up */ |
8dd2cb7e | 627 | if ((max(top, bottom) % min(top, bottom)) != 0) |
9504e086 MP |
628 | t->discard_misaligned = 1; |
629 | } | |
630 | ||
81744ee4 MP |
631 | t->max_discard_sectors = min_not_zero(t->max_discard_sectors, |
632 | b->max_discard_sectors); | |
0034af03 JA |
633 | t->max_hw_discard_sectors = min_not_zero(t->max_hw_discard_sectors, |
634 | b->max_hw_discard_sectors); | |
9504e086 MP |
635 | t->discard_granularity = max(t->discard_granularity, |
636 | b->discard_granularity); | |
e9637415 | 637 | t->discard_alignment = lcm_not_zero(t->discard_alignment, alignment) % |
8dd2cb7e | 638 | t->discard_granularity; |
9504e086 | 639 | } |
70dd5bf3 | 640 | |
a805a4fa DLM |
641 | t->zone_write_granularity = max(t->zone_write_granularity, |
642 | b->zone_write_granularity); | |
3093a479 | 643 | t->zoned = max(t->zoned, b->zoned); |
fe0b393f | 644 | return ret; |
c72758f3 | 645 | } |
5d85d324 | 646 | EXPORT_SYMBOL(blk_stack_limits); |
c72758f3 MP |
647 | |
648 | /** | |
649 | * disk_stack_limits - adjust queue limits for stacked drivers | |
77634f33 | 650 | * @disk: MD/DM gendisk (top) |
c72758f3 MP |
651 | * @bdev: the underlying block device (bottom) |
652 | * @offset: offset to beginning of data within component device | |
653 | * | |
654 | * Description: | |
e03a72e1 MP |
655 | * Merges the limits for a top level gendisk and a bottom level |
656 | * block_device. | |
c72758f3 MP |
657 | */ |
658 | void disk_stack_limits(struct gendisk *disk, struct block_device *bdev, | |
659 | sector_t offset) | |
660 | { | |
661 | struct request_queue *t = disk->queue; | |
c72758f3 | 662 | |
9efa82ef | 663 | if (blk_stack_limits(&t->limits, &bdev_get_queue(bdev)->limits, |
453b8ab6 CH |
664 | get_start_sect(bdev) + (offset >> 9)) < 0) |
665 | pr_notice("%s: Warning: Device %pg is misaligned\n", | |
666 | disk->disk_name, bdev); | |
e74d93e9 | 667 | |
471aa704 | 668 | disk_update_readahead(disk); |
c72758f3 MP |
669 | } |
670 | EXPORT_SYMBOL(disk_stack_limits); | |
671 | ||
27f8221a FT |
672 | /** |
673 | * blk_queue_update_dma_pad - update pad mask | |
674 | * @q: the request queue for the device | |
675 | * @mask: pad mask | |
676 | * | |
677 | * Update dma pad mask. | |
678 | * | |
679 | * Appending pad buffer to a request modifies the last entry of a | |
680 | * scatter list such that it includes the pad buffer. | |
681 | **/ | |
682 | void blk_queue_update_dma_pad(struct request_queue *q, unsigned int mask) | |
683 | { | |
684 | if (mask > q->dma_pad_mask) | |
685 | q->dma_pad_mask = mask; | |
686 | } | |
687 | EXPORT_SYMBOL(blk_queue_update_dma_pad); | |
688 | ||
86db1e29 JA |
689 | /** |
690 | * blk_queue_segment_boundary - set boundary rules for segment merging | |
691 | * @q: the request queue for the device | |
692 | * @mask: the memory boundary mask | |
693 | **/ | |
694 | void blk_queue_segment_boundary(struct request_queue *q, unsigned long mask) | |
695 | { | |
09cbfeaf KS |
696 | if (mask < PAGE_SIZE - 1) { |
697 | mask = PAGE_SIZE - 1; | |
24c03d47 HH |
698 | printk(KERN_INFO "%s: set to minimum %lx\n", |
699 | __func__, mask); | |
86db1e29 JA |
700 | } |
701 | ||
025146e1 | 702 | q->limits.seg_boundary_mask = mask; |
86db1e29 | 703 | } |
86db1e29 JA |
704 | EXPORT_SYMBOL(blk_queue_segment_boundary); |
705 | ||
03100aad KB |
706 | /** |
707 | * blk_queue_virt_boundary - set boundary rules for bio merging | |
708 | * @q: the request queue for the device | |
709 | * @mask: the memory boundary mask | |
710 | **/ | |
711 | void blk_queue_virt_boundary(struct request_queue *q, unsigned long mask) | |
712 | { | |
713 | q->limits.virt_boundary_mask = mask; | |
09324d32 CH |
714 | |
715 | /* | |
716 | * Devices that require a virtual boundary do not support scatter/gather | |
717 | * I/O natively, but instead require a descriptor list entry for each | |
718 | * page (which might not be idential to the Linux PAGE_SIZE). Because | |
719 | * of that they are not limited by our notion of "segment size". | |
720 | */ | |
c6c84f78 CH |
721 | if (mask) |
722 | q->limits.max_segment_size = UINT_MAX; | |
03100aad KB |
723 | } |
724 | EXPORT_SYMBOL(blk_queue_virt_boundary); | |
725 | ||
86db1e29 JA |
726 | /** |
727 | * blk_queue_dma_alignment - set dma length and memory alignment | |
728 | * @q: the request queue for the device | |
729 | * @mask: alignment mask | |
730 | * | |
731 | * description: | |
710027a4 | 732 | * set required memory and length alignment for direct dma transactions. |
8feb4d20 | 733 | * this is used when building direct io requests for the queue. |
86db1e29 JA |
734 | * |
735 | **/ | |
736 | void blk_queue_dma_alignment(struct request_queue *q, int mask) | |
737 | { | |
738 | q->dma_alignment = mask; | |
739 | } | |
86db1e29 JA |
740 | EXPORT_SYMBOL(blk_queue_dma_alignment); |
741 | ||
742 | /** | |
743 | * blk_queue_update_dma_alignment - update dma length and memory alignment | |
744 | * @q: the request queue for the device | |
745 | * @mask: alignment mask | |
746 | * | |
747 | * description: | |
710027a4 | 748 | * update required memory and length alignment for direct dma transactions. |
86db1e29 JA |
749 | * If the requested alignment is larger than the current alignment, then |
750 | * the current queue alignment is updated to the new value, otherwise it | |
751 | * is left alone. The design of this is to allow multiple objects | |
752 | * (driver, device, transport etc) to set their respective | |
753 | * alignments without having them interfere. | |
754 | * | |
755 | **/ | |
756 | void blk_queue_update_dma_alignment(struct request_queue *q, int mask) | |
757 | { | |
758 | BUG_ON(mask > PAGE_SIZE); | |
759 | ||
760 | if (mask > q->dma_alignment) | |
761 | q->dma_alignment = mask; | |
762 | } | |
86db1e29 JA |
763 | EXPORT_SYMBOL(blk_queue_update_dma_alignment); |
764 | ||
d278d4a8 JA |
765 | /** |
766 | * blk_set_queue_depth - tell the block layer about the device queue depth | |
767 | * @q: the request queue for the device | |
768 | * @depth: queue depth | |
769 | * | |
770 | */ | |
771 | void blk_set_queue_depth(struct request_queue *q, unsigned int depth) | |
772 | { | |
773 | q->queue_depth = depth; | |
9677a3e0 | 774 | rq_qos_queue_depth_changed(q); |
d278d4a8 JA |
775 | } |
776 | EXPORT_SYMBOL(blk_set_queue_depth); | |
777 | ||
93e9d8e8 JA |
778 | /** |
779 | * blk_queue_write_cache - configure queue's write cache | |
780 | * @q: the request queue for the device | |
781 | * @wc: write back cache on or off | |
782 | * @fua: device supports FUA writes, if true | |
783 | * | |
784 | * Tell the block layer about the write cache of @q. | |
785 | */ | |
786 | void blk_queue_write_cache(struct request_queue *q, bool wc, bool fua) | |
787 | { | |
c888a8f9 | 788 | if (wc) |
57d74df9 | 789 | blk_queue_flag_set(QUEUE_FLAG_WC, q); |
c888a8f9 | 790 | else |
57d74df9 | 791 | blk_queue_flag_clear(QUEUE_FLAG_WC, q); |
c888a8f9 | 792 | if (fua) |
57d74df9 | 793 | blk_queue_flag_set(QUEUE_FLAG_FUA, q); |
c888a8f9 | 794 | else |
57d74df9 | 795 | blk_queue_flag_clear(QUEUE_FLAG_FUA, q); |
87760e5e | 796 | |
a7905043 | 797 | wbt_set_write_cache(q, test_bit(QUEUE_FLAG_WC, &q->queue_flags)); |
93e9d8e8 JA |
798 | } |
799 | EXPORT_SYMBOL_GPL(blk_queue_write_cache); | |
800 | ||
68c43f13 DLM |
801 | /** |
802 | * blk_queue_required_elevator_features - Set a queue required elevator features | |
803 | * @q: the request queue for the target device | |
804 | * @features: Required elevator features OR'ed together | |
805 | * | |
806 | * Tell the block layer that for the device controlled through @q, only the | |
807 | * only elevators that can be used are those that implement at least the set of | |
808 | * features specified by @features. | |
809 | */ | |
810 | void blk_queue_required_elevator_features(struct request_queue *q, | |
811 | unsigned int features) | |
812 | { | |
813 | q->required_elevator_features = features; | |
814 | } | |
815 | EXPORT_SYMBOL_GPL(blk_queue_required_elevator_features); | |
816 | ||
45147fb5 YS |
817 | /** |
818 | * blk_queue_can_use_dma_map_merging - configure queue for merging segments. | |
819 | * @q: the request queue for the device | |
820 | * @dev: the device pointer for dma | |
821 | * | |
822 | * Tell the block layer about merging the segments by dma map of @q. | |
823 | */ | |
824 | bool blk_queue_can_use_dma_map_merging(struct request_queue *q, | |
825 | struct device *dev) | |
826 | { | |
827 | unsigned long boundary = dma_get_merge_boundary(dev); | |
828 | ||
829 | if (!boundary) | |
830 | return false; | |
831 | ||
832 | /* No need to update max_segment_size. see blk_queue_virt_boundary() */ | |
833 | blk_queue_virt_boundary(q, boundary); | |
834 | ||
835 | return true; | |
836 | } | |
837 | EXPORT_SYMBOL_GPL(blk_queue_can_use_dma_map_merging); | |
838 | ||
e0c60d01 SK |
839 | static bool disk_has_partitions(struct gendisk *disk) |
840 | { | |
841 | unsigned long idx; | |
842 | struct block_device *part; | |
843 | bool ret = false; | |
844 | ||
845 | rcu_read_lock(); | |
846 | xa_for_each(&disk->part_tbl, idx, part) { | |
847 | if (bdev_is_partition(part)) { | |
848 | ret = true; | |
849 | break; | |
850 | } | |
851 | } | |
852 | rcu_read_unlock(); | |
853 | ||
854 | return ret; | |
855 | } | |
856 | ||
27ba3e8f DLM |
857 | /** |
858 | * blk_queue_set_zoned - configure a disk queue zoned model. | |
859 | * @disk: the gendisk of the queue to configure | |
860 | * @model: the zoned model to set | |
861 | * | |
862 | * Set the zoned model of the request queue of @disk according to @model. | |
863 | * When @model is BLK_ZONED_HM (host managed), this should be called only | |
864 | * if zoned block device support is enabled (CONFIG_BLK_DEV_ZONED option). | |
865 | * If @model specifies BLK_ZONED_HA (host aware), the effective model used | |
866 | * depends on CONFIG_BLK_DEV_ZONED settings and on the existence of partitions | |
867 | * on the disk. | |
868 | */ | |
869 | void blk_queue_set_zoned(struct gendisk *disk, enum blk_zoned_model model) | |
870 | { | |
a805a4fa DLM |
871 | struct request_queue *q = disk->queue; |
872 | ||
27ba3e8f DLM |
873 | switch (model) { |
874 | case BLK_ZONED_HM: | |
875 | /* | |
876 | * Host managed devices are supported only if | |
877 | * CONFIG_BLK_DEV_ZONED is enabled. | |
878 | */ | |
879 | WARN_ON_ONCE(!IS_ENABLED(CONFIG_BLK_DEV_ZONED)); | |
880 | break; | |
881 | case BLK_ZONED_HA: | |
882 | /* | |
883 | * Host aware devices can be treated either as regular block | |
884 | * devices (similar to drive managed devices) or as zoned block | |
885 | * devices to take advantage of the zone command set, similarly | |
886 | * to host managed devices. We try the latter if there are no | |
887 | * partitions and zoned block device support is enabled, else | |
888 | * we do nothing special as far as the block layer is concerned. | |
889 | */ | |
890 | if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) || | |
e0c60d01 | 891 | disk_has_partitions(disk)) |
27ba3e8f DLM |
892 | model = BLK_ZONED_NONE; |
893 | break; | |
894 | case BLK_ZONED_NONE: | |
895 | default: | |
896 | if (WARN_ON_ONCE(model != BLK_ZONED_NONE)) | |
897 | model = BLK_ZONED_NONE; | |
898 | break; | |
899 | } | |
900 | ||
a805a4fa DLM |
901 | q->limits.zoned = model; |
902 | if (model != BLK_ZONED_NONE) { | |
903 | /* | |
904 | * Set the zone write granularity to the device logical block | |
905 | * size by default. The driver can change this value if needed. | |
906 | */ | |
907 | blk_queue_zone_write_granularity(q, | |
908 | queue_logical_block_size(q)); | |
508aebb8 DLM |
909 | } else { |
910 | blk_queue_clear_zone_settings(q); | |
a805a4fa | 911 | } |
27ba3e8f DLM |
912 | } |
913 | EXPORT_SYMBOL_GPL(blk_queue_set_zoned); | |
89098b07 CH |
914 | |
915 | int bdev_alignment_offset(struct block_device *bdev) | |
916 | { | |
917 | struct request_queue *q = bdev_get_queue(bdev); | |
918 | ||
919 | if (q->limits.misaligned) | |
920 | return -1; | |
921 | if (bdev_is_partition(bdev)) | |
922 | return queue_limit_alignment_offset(&q->limits, | |
923 | bdev->bd_start_sect); | |
924 | return q->limits.alignment_offset; | |
925 | } | |
926 | EXPORT_SYMBOL_GPL(bdev_alignment_offset); |