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> | |
57c8a661 | 10 | #include <linux/memblock.h> /* for max_pfn/max_low_pfn */ |
70dd5bf3 | 11 | #include <linux/gcd.h> |
2cda2728 | 12 | #include <linux/lcm.h> |
ad5ebd2f | 13 | #include <linux/jiffies.h> |
5a0e3ad6 | 14 | #include <linux/gfp.h> |
45147fb5 | 15 | #include <linux/dma-mapping.h> |
86db1e29 JA |
16 | |
17 | #include "blk.h" | |
87760e5e | 18 | #include "blk-wbt.h" |
86db1e29 | 19 | |
6728cb0e | 20 | unsigned long blk_max_low_pfn; |
86db1e29 | 21 | EXPORT_SYMBOL(blk_max_low_pfn); |
6728cb0e JA |
22 | |
23 | unsigned long blk_max_pfn; | |
86db1e29 | 24 | |
242f9dcb JA |
25 | void blk_queue_rq_timeout(struct request_queue *q, unsigned int timeout) |
26 | { | |
27 | q->rq_timeout = timeout; | |
28 | } | |
29 | EXPORT_SYMBOL_GPL(blk_queue_rq_timeout); | |
30 | ||
e475bba2 MP |
31 | /** |
32 | * blk_set_default_limits - reset limits to default values | |
f740f5ca | 33 | * @lim: the queue_limits structure to reset |
e475bba2 MP |
34 | * |
35 | * Description: | |
b1bd055d | 36 | * Returns a queue_limit struct to its default state. |
e475bba2 MP |
37 | */ |
38 | void blk_set_default_limits(struct queue_limits *lim) | |
39 | { | |
8a78362c | 40 | lim->max_segments = BLK_MAX_SEGMENTS; |
1e739730 | 41 | lim->max_discard_segments = 1; |
13f05c8d | 42 | lim->max_integrity_segments = 0; |
e475bba2 | 43 | lim->seg_boundary_mask = BLK_SEG_BOUNDARY_MASK; |
03100aad | 44 | lim->virt_boundary_mask = 0; |
eb28d31b | 45 | lim->max_segment_size = BLK_MAX_SEGMENT_SIZE; |
5f009d3f KB |
46 | lim->max_sectors = lim->max_hw_sectors = BLK_SAFE_MAX_SECTORS; |
47 | lim->max_dev_sectors = 0; | |
762380ad | 48 | lim->chunk_sectors = 0; |
4363ac7c | 49 | lim->max_write_same_sectors = 0; |
a6f0788e | 50 | lim->max_write_zeroes_sectors = 0; |
0512a75b | 51 | lim->max_zone_append_sectors = 0; |
86b37281 | 52 | lim->max_discard_sectors = 0; |
0034af03 | 53 | lim->max_hw_discard_sectors = 0; |
86b37281 MP |
54 | lim->discard_granularity = 0; |
55 | lim->discard_alignment = 0; | |
56 | lim->discard_misaligned = 0; | |
e475bba2 | 57 | lim->logical_block_size = lim->physical_block_size = lim->io_min = 512; |
3a02c8e8 | 58 | lim->bounce_pfn = (unsigned long)(BLK_BOUNCE_ANY >> PAGE_SHIFT); |
e475bba2 MP |
59 | lim->alignment_offset = 0; |
60 | lim->io_opt = 0; | |
61 | lim->misaligned = 0; | |
797476b8 | 62 | lim->zoned = BLK_ZONED_NONE; |
e475bba2 MP |
63 | } |
64 | EXPORT_SYMBOL(blk_set_default_limits); | |
65 | ||
b1bd055d MP |
66 | /** |
67 | * blk_set_stacking_limits - set default limits for stacking devices | |
68 | * @lim: the queue_limits structure to reset | |
69 | * | |
70 | * Description: | |
71 | * Returns a queue_limit struct to its default state. Should be used | |
72 | * by stacking drivers like DM that have no internal limits. | |
73 | */ | |
74 | void blk_set_stacking_limits(struct queue_limits *lim) | |
75 | { | |
76 | blk_set_default_limits(lim); | |
77 | ||
78 | /* Inherit limits from component devices */ | |
b1bd055d | 79 | lim->max_segments = USHRT_MAX; |
42c9cdfe | 80 | lim->max_discard_segments = USHRT_MAX; |
b1bd055d | 81 | lim->max_hw_sectors = UINT_MAX; |
d82ae52e | 82 | lim->max_segment_size = UINT_MAX; |
fe86cdce | 83 | lim->max_sectors = UINT_MAX; |
ca369d51 | 84 | lim->max_dev_sectors = UINT_MAX; |
4363ac7c | 85 | lim->max_write_same_sectors = UINT_MAX; |
a6f0788e | 86 | lim->max_write_zeroes_sectors = UINT_MAX; |
0512a75b | 87 | lim->max_zone_append_sectors = UINT_MAX; |
b1bd055d MP |
88 | } |
89 | EXPORT_SYMBOL(blk_set_stacking_limits); | |
90 | ||
86db1e29 JA |
91 | /** |
92 | * blk_queue_bounce_limit - set bounce buffer limit for queue | |
cd0aca2d | 93 | * @q: the request queue for the device |
9f7e45d8 | 94 | * @max_addr: the maximum address the device can handle |
86db1e29 JA |
95 | * |
96 | * Description: | |
97 | * Different hardware can have different requirements as to what pages | |
98 | * it can do I/O directly to. A low level driver can call | |
99 | * blk_queue_bounce_limit to have lower memory pages allocated as bounce | |
9f7e45d8 | 100 | * buffers for doing I/O to pages residing above @max_addr. |
86db1e29 | 101 | **/ |
9f7e45d8 | 102 | void blk_queue_bounce_limit(struct request_queue *q, u64 max_addr) |
86db1e29 | 103 | { |
9f7e45d8 | 104 | unsigned long b_pfn = max_addr >> PAGE_SHIFT; |
86db1e29 JA |
105 | int dma = 0; |
106 | ||
107 | q->bounce_gfp = GFP_NOIO; | |
108 | #if BITS_PER_LONG == 64 | |
cd0aca2d TH |
109 | /* |
110 | * Assume anything <= 4GB can be handled by IOMMU. Actually | |
111 | * some IOMMUs can handle everything, but I don't know of a | |
112 | * way to test this here. | |
113 | */ | |
114 | if (b_pfn < (min_t(u64, 0xffffffffUL, BLK_BOUNCE_HIGH) >> PAGE_SHIFT)) | |
86db1e29 | 115 | dma = 1; |
efb012b3 | 116 | q->limits.bounce_pfn = max(max_low_pfn, b_pfn); |
86db1e29 | 117 | #else |
6728cb0e | 118 | if (b_pfn < blk_max_low_pfn) |
86db1e29 | 119 | dma = 1; |
c49825fa | 120 | q->limits.bounce_pfn = b_pfn; |
260a67a9 | 121 | #endif |
86db1e29 JA |
122 | if (dma) { |
123 | init_emergency_isa_pool(); | |
124 | q->bounce_gfp = GFP_NOIO | GFP_DMA; | |
260a67a9 | 125 | q->limits.bounce_pfn = b_pfn; |
86db1e29 JA |
126 | } |
127 | } | |
86db1e29 JA |
128 | EXPORT_SYMBOL(blk_queue_bounce_limit); |
129 | ||
130 | /** | |
ca369d51 MP |
131 | * blk_queue_max_hw_sectors - set max sectors for a request for this queue |
132 | * @q: the request queue for the device | |
2800aac1 | 133 | * @max_hw_sectors: max hardware sectors in the usual 512b unit |
86db1e29 JA |
134 | * |
135 | * Description: | |
2800aac1 MP |
136 | * Enables a low level driver to set a hard upper limit, |
137 | * max_hw_sectors, on the size of requests. max_hw_sectors is set by | |
4f258a46 MP |
138 | * the device driver based upon the capabilities of the I/O |
139 | * controller. | |
2800aac1 | 140 | * |
ca369d51 MP |
141 | * max_dev_sectors is a hard limit imposed by the storage device for |
142 | * READ/WRITE requests. It is set by the disk driver. | |
143 | * | |
2800aac1 MP |
144 | * max_sectors is a soft limit imposed by the block layer for |
145 | * filesystem type requests. This value can be overridden on a | |
146 | * per-device basis in /sys/block/<device>/queue/max_sectors_kb. | |
147 | * The soft limit can not exceed max_hw_sectors. | |
86db1e29 | 148 | **/ |
ca369d51 | 149 | void blk_queue_max_hw_sectors(struct request_queue *q, unsigned int max_hw_sectors) |
86db1e29 | 150 | { |
ca369d51 MP |
151 | struct queue_limits *limits = &q->limits; |
152 | unsigned int max_sectors; | |
153 | ||
09cbfeaf KS |
154 | if ((max_hw_sectors << 9) < PAGE_SIZE) { |
155 | max_hw_sectors = 1 << (PAGE_SHIFT - 9); | |
24c03d47 | 156 | printk(KERN_INFO "%s: set to minimum %d\n", |
2800aac1 | 157 | __func__, max_hw_sectors); |
86db1e29 JA |
158 | } |
159 | ||
817046ec DLM |
160 | max_hw_sectors = round_down(max_hw_sectors, |
161 | limits->logical_block_size >> SECTOR_SHIFT); | |
30e2bc08 | 162 | limits->max_hw_sectors = max_hw_sectors; |
817046ec | 163 | |
ca369d51 MP |
164 | max_sectors = min_not_zero(max_hw_sectors, limits->max_dev_sectors); |
165 | max_sectors = min_t(unsigned int, max_sectors, BLK_DEF_MAX_SECTORS); | |
817046ec DLM |
166 | max_sectors = round_down(max_sectors, |
167 | limits->logical_block_size >> SECTOR_SHIFT); | |
ca369d51 | 168 | limits->max_sectors = max_sectors; |
817046ec | 169 | |
dc3b17cc | 170 | q->backing_dev_info->io_pages = max_sectors >> (PAGE_SHIFT - 9); |
86db1e29 | 171 | } |
086fa5ff | 172 | EXPORT_SYMBOL(blk_queue_max_hw_sectors); |
86db1e29 | 173 | |
762380ad JA |
174 | /** |
175 | * blk_queue_chunk_sectors - set size of the chunk for this queue | |
176 | * @q: the request queue for the device | |
177 | * @chunk_sectors: chunk sectors in the usual 512b unit | |
178 | * | |
179 | * Description: | |
180 | * If a driver doesn't want IOs to cross a given chunk size, it can set | |
07d098e6 MS |
181 | * this limit and prevent merging across chunks. Note that the block layer |
182 | * must accept a page worth of data at any offset. So if the crossing of | |
183 | * chunks is a hard limitation in the driver, it must still be prepared | |
184 | * to split single page bios. | |
762380ad JA |
185 | **/ |
186 | void blk_queue_chunk_sectors(struct request_queue *q, unsigned int chunk_sectors) | |
187 | { | |
762380ad JA |
188 | q->limits.chunk_sectors = chunk_sectors; |
189 | } | |
190 | EXPORT_SYMBOL(blk_queue_chunk_sectors); | |
191 | ||
67efc925 CH |
192 | /** |
193 | * blk_queue_max_discard_sectors - set max sectors for a single discard | |
194 | * @q: the request queue for the device | |
c7ebf065 | 195 | * @max_discard_sectors: maximum number of sectors to discard |
67efc925 CH |
196 | **/ |
197 | void blk_queue_max_discard_sectors(struct request_queue *q, | |
198 | unsigned int max_discard_sectors) | |
199 | { | |
0034af03 | 200 | q->limits.max_hw_discard_sectors = max_discard_sectors; |
67efc925 CH |
201 | q->limits.max_discard_sectors = max_discard_sectors; |
202 | } | |
203 | EXPORT_SYMBOL(blk_queue_max_discard_sectors); | |
204 | ||
4363ac7c MP |
205 | /** |
206 | * blk_queue_max_write_same_sectors - set max sectors for a single write same | |
207 | * @q: the request queue for the device | |
208 | * @max_write_same_sectors: maximum number of sectors to write per command | |
209 | **/ | |
210 | void blk_queue_max_write_same_sectors(struct request_queue *q, | |
211 | unsigned int max_write_same_sectors) | |
212 | { | |
213 | q->limits.max_write_same_sectors = max_write_same_sectors; | |
214 | } | |
215 | EXPORT_SYMBOL(blk_queue_max_write_same_sectors); | |
216 | ||
a6f0788e CK |
217 | /** |
218 | * blk_queue_max_write_zeroes_sectors - set max sectors for a single | |
219 | * write zeroes | |
220 | * @q: the request queue for the device | |
221 | * @max_write_zeroes_sectors: maximum number of sectors to write per command | |
222 | **/ | |
223 | void blk_queue_max_write_zeroes_sectors(struct request_queue *q, | |
224 | unsigned int max_write_zeroes_sectors) | |
225 | { | |
226 | q->limits.max_write_zeroes_sectors = max_write_zeroes_sectors; | |
227 | } | |
228 | EXPORT_SYMBOL(blk_queue_max_write_zeroes_sectors); | |
229 | ||
0512a75b KB |
230 | /** |
231 | * blk_queue_max_zone_append_sectors - set max sectors for a single zone append | |
232 | * @q: the request queue for the device | |
233 | * @max_zone_append_sectors: maximum number of sectors to write per command | |
234 | **/ | |
235 | void blk_queue_max_zone_append_sectors(struct request_queue *q, | |
236 | unsigned int max_zone_append_sectors) | |
237 | { | |
238 | unsigned int max_sectors; | |
239 | ||
240 | if (WARN_ON(!blk_queue_is_zoned(q))) | |
241 | return; | |
242 | ||
243 | max_sectors = min(q->limits.max_hw_sectors, max_zone_append_sectors); | |
244 | max_sectors = min(q->limits.chunk_sectors, max_sectors); | |
245 | ||
246 | /* | |
247 | * Signal eventual driver bugs resulting in the max_zone_append sectors limit | |
248 | * being 0 due to a 0 argument, the chunk_sectors limit (zone size) not set, | |
249 | * or the max_hw_sectors limit not set. | |
250 | */ | |
251 | WARN_ON(!max_sectors); | |
252 | ||
253 | q->limits.max_zone_append_sectors = max_sectors; | |
254 | } | |
255 | EXPORT_SYMBOL_GPL(blk_queue_max_zone_append_sectors); | |
256 | ||
86db1e29 | 257 | /** |
8a78362c | 258 | * blk_queue_max_segments - set max hw segments for a request for this queue |
86db1e29 JA |
259 | * @q: the request queue for the device |
260 | * @max_segments: max number of segments | |
261 | * | |
262 | * Description: | |
263 | * Enables a low level driver to set an upper limit on the number of | |
8a78362c | 264 | * hw data segments in a request. |
86db1e29 | 265 | **/ |
8a78362c | 266 | void blk_queue_max_segments(struct request_queue *q, unsigned short max_segments) |
86db1e29 JA |
267 | { |
268 | if (!max_segments) { | |
269 | max_segments = 1; | |
24c03d47 HH |
270 | printk(KERN_INFO "%s: set to minimum %d\n", |
271 | __func__, max_segments); | |
86db1e29 JA |
272 | } |
273 | ||
8a78362c | 274 | q->limits.max_segments = max_segments; |
86db1e29 | 275 | } |
8a78362c | 276 | EXPORT_SYMBOL(blk_queue_max_segments); |
86db1e29 | 277 | |
1e739730 CH |
278 | /** |
279 | * blk_queue_max_discard_segments - set max segments for discard requests | |
280 | * @q: the request queue for the device | |
281 | * @max_segments: max number of segments | |
282 | * | |
283 | * Description: | |
284 | * Enables a low level driver to set an upper limit on the number of | |
285 | * segments in a discard request. | |
286 | **/ | |
287 | void blk_queue_max_discard_segments(struct request_queue *q, | |
288 | unsigned short max_segments) | |
289 | { | |
290 | q->limits.max_discard_segments = max_segments; | |
291 | } | |
292 | EXPORT_SYMBOL_GPL(blk_queue_max_discard_segments); | |
293 | ||
86db1e29 JA |
294 | /** |
295 | * blk_queue_max_segment_size - set max segment size for blk_rq_map_sg | |
296 | * @q: the request queue for the device | |
297 | * @max_size: max size of segment in bytes | |
298 | * | |
299 | * Description: | |
300 | * Enables a low level driver to set an upper limit on the size of a | |
301 | * coalesced segment | |
302 | **/ | |
303 | void blk_queue_max_segment_size(struct request_queue *q, unsigned int max_size) | |
304 | { | |
09cbfeaf KS |
305 | if (max_size < PAGE_SIZE) { |
306 | max_size = PAGE_SIZE; | |
24c03d47 HH |
307 | printk(KERN_INFO "%s: set to minimum %d\n", |
308 | __func__, max_size); | |
86db1e29 JA |
309 | } |
310 | ||
09324d32 CH |
311 | /* see blk_queue_virt_boundary() for the explanation */ |
312 | WARN_ON_ONCE(q->limits.virt_boundary_mask); | |
313 | ||
025146e1 | 314 | q->limits.max_segment_size = max_size; |
86db1e29 | 315 | } |
86db1e29 JA |
316 | EXPORT_SYMBOL(blk_queue_max_segment_size); |
317 | ||
318 | /** | |
e1defc4f | 319 | * blk_queue_logical_block_size - set logical block size for the queue |
86db1e29 | 320 | * @q: the request queue for the device |
e1defc4f | 321 | * @size: the logical block size, in bytes |
86db1e29 JA |
322 | * |
323 | * Description: | |
e1defc4f MP |
324 | * This should be set to the lowest possible block size that the |
325 | * storage device can address. The default of 512 covers most | |
326 | * hardware. | |
86db1e29 | 327 | **/ |
ad6bf88a | 328 | void blk_queue_logical_block_size(struct request_queue *q, unsigned int size) |
86db1e29 | 329 | { |
817046ec DLM |
330 | struct queue_limits *limits = &q->limits; |
331 | ||
332 | limits->logical_block_size = size; | |
333 | ||
334 | if (limits->physical_block_size < size) | |
335 | limits->physical_block_size = size; | |
c72758f3 | 336 | |
817046ec DLM |
337 | if (limits->io_min < limits->physical_block_size) |
338 | limits->io_min = limits->physical_block_size; | |
c72758f3 | 339 | |
817046ec DLM |
340 | limits->max_hw_sectors = |
341 | round_down(limits->max_hw_sectors, size >> SECTOR_SHIFT); | |
342 | limits->max_sectors = | |
343 | round_down(limits->max_sectors, size >> SECTOR_SHIFT); | |
86db1e29 | 344 | } |
e1defc4f | 345 | EXPORT_SYMBOL(blk_queue_logical_block_size); |
86db1e29 | 346 | |
c72758f3 MP |
347 | /** |
348 | * blk_queue_physical_block_size - set physical block size for the queue | |
349 | * @q: the request queue for the device | |
350 | * @size: the physical block size, in bytes | |
351 | * | |
352 | * Description: | |
353 | * This should be set to the lowest possible sector size that the | |
354 | * hardware can operate on without reverting to read-modify-write | |
355 | * operations. | |
356 | */ | |
892b6f90 | 357 | void blk_queue_physical_block_size(struct request_queue *q, unsigned int size) |
c72758f3 MP |
358 | { |
359 | q->limits.physical_block_size = size; | |
360 | ||
361 | if (q->limits.physical_block_size < q->limits.logical_block_size) | |
362 | q->limits.physical_block_size = q->limits.logical_block_size; | |
363 | ||
364 | if (q->limits.io_min < q->limits.physical_block_size) | |
365 | q->limits.io_min = q->limits.physical_block_size; | |
366 | } | |
367 | EXPORT_SYMBOL(blk_queue_physical_block_size); | |
368 | ||
369 | /** | |
370 | * blk_queue_alignment_offset - set physical block alignment offset | |
371 | * @q: the request queue for the device | |
8ebf9756 | 372 | * @offset: alignment offset in bytes |
c72758f3 MP |
373 | * |
374 | * Description: | |
375 | * Some devices are naturally misaligned to compensate for things like | |
376 | * the legacy DOS partition table 63-sector offset. Low-level drivers | |
377 | * should call this function for devices whose first sector is not | |
378 | * naturally aligned. | |
379 | */ | |
380 | void blk_queue_alignment_offset(struct request_queue *q, unsigned int offset) | |
381 | { | |
382 | q->limits.alignment_offset = | |
383 | offset & (q->limits.physical_block_size - 1); | |
384 | q->limits.misaligned = 0; | |
385 | } | |
386 | EXPORT_SYMBOL(blk_queue_alignment_offset); | |
387 | ||
c2e4cd57 CH |
388 | void blk_queue_update_readahead(struct request_queue *q) |
389 | { | |
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 | */ | |
394 | q->backing_dev_info->ra_pages = | |
395 | max(queue_io_opt(q) * 2 / PAGE_SIZE, VM_READAHEAD_PAGES); | |
396 | q->backing_dev_info->io_pages = | |
397 | queue_max_sectors(q) >> (PAGE_SHIFT - 9); | |
398 | } | |
399 | EXPORT_SYMBOL_GPL(blk_queue_update_readahead); | |
400 | ||
7c958e32 MP |
401 | /** |
402 | * blk_limits_io_min - set minimum request size for a device | |
403 | * @limits: the queue limits | |
404 | * @min: smallest I/O size in bytes | |
405 | * | |
406 | * Description: | |
407 | * Some devices have an internal block size bigger than the reported | |
408 | * hardware sector size. This function can be used to signal the | |
409 | * smallest I/O the device can perform without incurring a performance | |
410 | * penalty. | |
411 | */ | |
412 | void blk_limits_io_min(struct queue_limits *limits, unsigned int min) | |
413 | { | |
414 | limits->io_min = min; | |
415 | ||
416 | if (limits->io_min < limits->logical_block_size) | |
417 | limits->io_min = limits->logical_block_size; | |
418 | ||
419 | if (limits->io_min < limits->physical_block_size) | |
420 | limits->io_min = limits->physical_block_size; | |
421 | } | |
422 | EXPORT_SYMBOL(blk_limits_io_min); | |
423 | ||
c72758f3 MP |
424 | /** |
425 | * blk_queue_io_min - set minimum request size for the queue | |
426 | * @q: the request queue for the device | |
8ebf9756 | 427 | * @min: smallest I/O size in bytes |
c72758f3 MP |
428 | * |
429 | * Description: | |
7e5f5fb0 MP |
430 | * Storage devices may report a granularity or preferred minimum I/O |
431 | * size which is the smallest request the device can perform without | |
432 | * incurring a performance penalty. For disk drives this is often the | |
433 | * physical block size. For RAID arrays it is often the stripe chunk | |
434 | * size. A properly aligned multiple of minimum_io_size is the | |
435 | * preferred request size for workloads where a high number of I/O | |
436 | * operations is desired. | |
c72758f3 MP |
437 | */ |
438 | void blk_queue_io_min(struct request_queue *q, unsigned int min) | |
439 | { | |
7c958e32 | 440 | blk_limits_io_min(&q->limits, min); |
c72758f3 MP |
441 | } |
442 | EXPORT_SYMBOL(blk_queue_io_min); | |
443 | ||
3c5820c7 MP |
444 | /** |
445 | * blk_limits_io_opt - set optimal request size for a device | |
446 | * @limits: the queue limits | |
447 | * @opt: smallest I/O size in bytes | |
448 | * | |
449 | * Description: | |
450 | * Storage devices may report an optimal I/O size, which is the | |
451 | * device's preferred unit for sustained I/O. This is rarely reported | |
452 | * for disk drives. For RAID arrays it is usually the stripe width or | |
453 | * the internal track size. A properly aligned multiple of | |
454 | * optimal_io_size is the preferred request size for workloads where | |
455 | * sustained throughput is desired. | |
456 | */ | |
457 | void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt) | |
458 | { | |
459 | limits->io_opt = opt; | |
460 | } | |
461 | EXPORT_SYMBOL(blk_limits_io_opt); | |
462 | ||
c72758f3 MP |
463 | /** |
464 | * blk_queue_io_opt - set optimal request size for the queue | |
465 | * @q: the request queue for the device | |
8ebf9756 | 466 | * @opt: optimal request size in bytes |
c72758f3 MP |
467 | * |
468 | * Description: | |
7e5f5fb0 MP |
469 | * Storage devices may report an optimal I/O size, which is the |
470 | * device's preferred unit for sustained I/O. This is rarely reported | |
471 | * for disk drives. For RAID arrays it is usually the stripe width or | |
472 | * the internal track size. A properly aligned multiple of | |
473 | * optimal_io_size is the preferred request size for workloads where | |
474 | * sustained throughput is desired. | |
c72758f3 MP |
475 | */ |
476 | void blk_queue_io_opt(struct request_queue *q, unsigned int opt) | |
477 | { | |
3c5820c7 | 478 | blk_limits_io_opt(&q->limits, opt); |
c2e4cd57 CH |
479 | q->backing_dev_info->ra_pages = |
480 | max(queue_io_opt(q) * 2 / PAGE_SIZE, VM_READAHEAD_PAGES); | |
c72758f3 MP |
481 | } |
482 | EXPORT_SYMBOL(blk_queue_io_opt); | |
483 | ||
c72758f3 MP |
484 | /** |
485 | * blk_stack_limits - adjust queue_limits for stacked devices | |
81744ee4 MP |
486 | * @t: the stacking driver limits (top device) |
487 | * @b: the underlying queue limits (bottom, component device) | |
e03a72e1 | 488 | * @start: first data sector within component device |
c72758f3 MP |
489 | * |
490 | * Description: | |
81744ee4 MP |
491 | * This function is used by stacking drivers like MD and DM to ensure |
492 | * that all component devices have compatible block sizes and | |
493 | * alignments. The stacking driver must provide a queue_limits | |
494 | * struct (top) and then iteratively call the stacking function for | |
495 | * all component (bottom) devices. The stacking function will | |
496 | * attempt to combine the values and ensure proper alignment. | |
497 | * | |
498 | * Returns 0 if the top and bottom queue_limits are compatible. The | |
499 | * top device's block sizes and alignment offsets may be adjusted to | |
500 | * ensure alignment with the bottom device. If no compatible sizes | |
501 | * and alignments exist, -1 is returned and the resulting top | |
502 | * queue_limits will have the misaligned flag set to indicate that | |
503 | * the alignment_offset is undefined. | |
c72758f3 MP |
504 | */ |
505 | int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, | |
e03a72e1 | 506 | sector_t start) |
c72758f3 | 507 | { |
e03a72e1 | 508 | unsigned int top, bottom, alignment, ret = 0; |
86b37281 | 509 | |
c72758f3 MP |
510 | t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors); |
511 | t->max_hw_sectors = min_not_zero(t->max_hw_sectors, b->max_hw_sectors); | |
ca369d51 | 512 | t->max_dev_sectors = min_not_zero(t->max_dev_sectors, b->max_dev_sectors); |
4363ac7c MP |
513 | t->max_write_same_sectors = min(t->max_write_same_sectors, |
514 | b->max_write_same_sectors); | |
a6f0788e CK |
515 | t->max_write_zeroes_sectors = min(t->max_write_zeroes_sectors, |
516 | b->max_write_zeroes_sectors); | |
0512a75b KB |
517 | t->max_zone_append_sectors = min(t->max_zone_append_sectors, |
518 | b->max_zone_append_sectors); | |
77634f33 | 519 | t->bounce_pfn = min_not_zero(t->bounce_pfn, b->bounce_pfn); |
c72758f3 MP |
520 | |
521 | t->seg_boundary_mask = min_not_zero(t->seg_boundary_mask, | |
522 | b->seg_boundary_mask); | |
03100aad KB |
523 | t->virt_boundary_mask = min_not_zero(t->virt_boundary_mask, |
524 | b->virt_boundary_mask); | |
c72758f3 | 525 | |
8a78362c | 526 | t->max_segments = min_not_zero(t->max_segments, b->max_segments); |
1e739730 CH |
527 | t->max_discard_segments = min_not_zero(t->max_discard_segments, |
528 | b->max_discard_segments); | |
13f05c8d MP |
529 | t->max_integrity_segments = min_not_zero(t->max_integrity_segments, |
530 | b->max_integrity_segments); | |
c72758f3 MP |
531 | |
532 | t->max_segment_size = min_not_zero(t->max_segment_size, | |
533 | b->max_segment_size); | |
534 | ||
fe0b393f MP |
535 | t->misaligned |= b->misaligned; |
536 | ||
e03a72e1 | 537 | alignment = queue_limit_alignment_offset(b, start); |
9504e086 | 538 | |
81744ee4 MP |
539 | /* Bottom device has different alignment. Check that it is |
540 | * compatible with the current top alignment. | |
541 | */ | |
9504e086 MP |
542 | if (t->alignment_offset != alignment) { |
543 | ||
544 | top = max(t->physical_block_size, t->io_min) | |
545 | + t->alignment_offset; | |
81744ee4 | 546 | bottom = max(b->physical_block_size, b->io_min) + alignment; |
9504e086 | 547 | |
81744ee4 | 548 | /* Verify that top and bottom intervals line up */ |
b8839b8c | 549 | if (max(top, bottom) % min(top, bottom)) { |
9504e086 | 550 | t->misaligned = 1; |
fe0b393f MP |
551 | ret = -1; |
552 | } | |
9504e086 MP |
553 | } |
554 | ||
c72758f3 MP |
555 | t->logical_block_size = max(t->logical_block_size, |
556 | b->logical_block_size); | |
557 | ||
558 | t->physical_block_size = max(t->physical_block_size, | |
559 | b->physical_block_size); | |
560 | ||
561 | t->io_min = max(t->io_min, b->io_min); | |
e9637415 | 562 | t->io_opt = lcm_not_zero(t->io_opt, b->io_opt); |
7e7986f9 MS |
563 | |
564 | /* Set non-power-of-2 compatible chunk_sectors boundary */ | |
565 | if (b->chunk_sectors) | |
566 | t->chunk_sectors = gcd(t->chunk_sectors, b->chunk_sectors); | |
9504e086 | 567 | |
81744ee4 | 568 | /* Physical block size a multiple of the logical block size? */ |
9504e086 MP |
569 | if (t->physical_block_size & (t->logical_block_size - 1)) { |
570 | t->physical_block_size = t->logical_block_size; | |
c72758f3 | 571 | t->misaligned = 1; |
fe0b393f | 572 | ret = -1; |
86b37281 MP |
573 | } |
574 | ||
81744ee4 | 575 | /* Minimum I/O a multiple of the physical block size? */ |
9504e086 MP |
576 | if (t->io_min & (t->physical_block_size - 1)) { |
577 | t->io_min = t->physical_block_size; | |
578 | t->misaligned = 1; | |
fe0b393f | 579 | ret = -1; |
c72758f3 MP |
580 | } |
581 | ||
81744ee4 | 582 | /* Optimal I/O a multiple of the physical block size? */ |
9504e086 MP |
583 | if (t->io_opt & (t->physical_block_size - 1)) { |
584 | t->io_opt = 0; | |
585 | t->misaligned = 1; | |
fe0b393f | 586 | ret = -1; |
9504e086 | 587 | } |
c72758f3 | 588 | |
22ada802 MS |
589 | /* chunk_sectors a multiple of the physical block size? */ |
590 | if ((t->chunk_sectors << 9) & (t->physical_block_size - 1)) { | |
591 | t->chunk_sectors = 0; | |
592 | t->misaligned = 1; | |
593 | ret = -1; | |
594 | } | |
595 | ||
c78afc62 KO |
596 | t->raid_partial_stripes_expensive = |
597 | max(t->raid_partial_stripes_expensive, | |
598 | b->raid_partial_stripes_expensive); | |
599 | ||
81744ee4 | 600 | /* Find lowest common alignment_offset */ |
e9637415 | 601 | t->alignment_offset = lcm_not_zero(t->alignment_offset, alignment) |
b8839b8c | 602 | % max(t->physical_block_size, t->io_min); |
86b37281 | 603 | |
81744ee4 | 604 | /* Verify that new alignment_offset is on a logical block boundary */ |
fe0b393f | 605 | if (t->alignment_offset & (t->logical_block_size - 1)) { |
c72758f3 | 606 | t->misaligned = 1; |
fe0b393f MP |
607 | ret = -1; |
608 | } | |
c72758f3 | 609 | |
9504e086 MP |
610 | /* Discard alignment and granularity */ |
611 | if (b->discard_granularity) { | |
e03a72e1 | 612 | alignment = queue_limit_discard_alignment(b, start); |
9504e086 MP |
613 | |
614 | if (t->discard_granularity != 0 && | |
615 | t->discard_alignment != alignment) { | |
616 | top = t->discard_granularity + t->discard_alignment; | |
617 | bottom = b->discard_granularity + alignment; | |
70dd5bf3 | 618 | |
9504e086 | 619 | /* Verify that top and bottom intervals line up */ |
8dd2cb7e | 620 | if ((max(top, bottom) % min(top, bottom)) != 0) |
9504e086 MP |
621 | t->discard_misaligned = 1; |
622 | } | |
623 | ||
81744ee4 MP |
624 | t->max_discard_sectors = min_not_zero(t->max_discard_sectors, |
625 | b->max_discard_sectors); | |
0034af03 JA |
626 | t->max_hw_discard_sectors = min_not_zero(t->max_hw_discard_sectors, |
627 | b->max_hw_discard_sectors); | |
9504e086 MP |
628 | t->discard_granularity = max(t->discard_granularity, |
629 | b->discard_granularity); | |
e9637415 | 630 | t->discard_alignment = lcm_not_zero(t->discard_alignment, alignment) % |
8dd2cb7e | 631 | t->discard_granularity; |
9504e086 | 632 | } |
70dd5bf3 | 633 | |
3093a479 | 634 | t->zoned = max(t->zoned, b->zoned); |
fe0b393f | 635 | return ret; |
c72758f3 | 636 | } |
5d85d324 | 637 | EXPORT_SYMBOL(blk_stack_limits); |
c72758f3 MP |
638 | |
639 | /** | |
640 | * disk_stack_limits - adjust queue limits for stacked drivers | |
77634f33 | 641 | * @disk: MD/DM gendisk (top) |
c72758f3 MP |
642 | * @bdev: the underlying block device (bottom) |
643 | * @offset: offset to beginning of data within component device | |
644 | * | |
645 | * Description: | |
e03a72e1 MP |
646 | * Merges the limits for a top level gendisk and a bottom level |
647 | * block_device. | |
c72758f3 MP |
648 | */ |
649 | void disk_stack_limits(struct gendisk *disk, struct block_device *bdev, | |
650 | sector_t offset) | |
651 | { | |
652 | struct request_queue *t = disk->queue; | |
c72758f3 | 653 | |
9efa82ef CH |
654 | if (blk_stack_limits(&t->limits, &bdev_get_queue(bdev)->limits, |
655 | get_start_sect(bdev) + (offset >> 9)) < 0) { | |
c72758f3 MP |
656 | char top[BDEVNAME_SIZE], bottom[BDEVNAME_SIZE]; |
657 | ||
658 | disk_name(disk, 0, top); | |
659 | bdevname(bdev, bottom); | |
660 | ||
661 | printk(KERN_NOTICE "%s: Warning: Device %s is misaligned\n", | |
662 | top, bottom); | |
663 | } | |
e74d93e9 | 664 | |
c2e4cd57 | 665 | blk_queue_update_readahead(disk->queue); |
c72758f3 MP |
666 | } |
667 | EXPORT_SYMBOL(disk_stack_limits); | |
668 | ||
27f8221a FT |
669 | /** |
670 | * blk_queue_update_dma_pad - update pad mask | |
671 | * @q: the request queue for the device | |
672 | * @mask: pad mask | |
673 | * | |
674 | * Update dma pad mask. | |
675 | * | |
676 | * Appending pad buffer to a request modifies the last entry of a | |
677 | * scatter list such that it includes the pad buffer. | |
678 | **/ | |
679 | void blk_queue_update_dma_pad(struct request_queue *q, unsigned int mask) | |
680 | { | |
681 | if (mask > q->dma_pad_mask) | |
682 | q->dma_pad_mask = mask; | |
683 | } | |
684 | EXPORT_SYMBOL(blk_queue_update_dma_pad); | |
685 | ||
86db1e29 JA |
686 | /** |
687 | * blk_queue_segment_boundary - set boundary rules for segment merging | |
688 | * @q: the request queue for the device | |
689 | * @mask: the memory boundary mask | |
690 | **/ | |
691 | void blk_queue_segment_boundary(struct request_queue *q, unsigned long mask) | |
692 | { | |
09cbfeaf KS |
693 | if (mask < PAGE_SIZE - 1) { |
694 | mask = PAGE_SIZE - 1; | |
24c03d47 HH |
695 | printk(KERN_INFO "%s: set to minimum %lx\n", |
696 | __func__, mask); | |
86db1e29 JA |
697 | } |
698 | ||
025146e1 | 699 | q->limits.seg_boundary_mask = mask; |
86db1e29 | 700 | } |
86db1e29 JA |
701 | EXPORT_SYMBOL(blk_queue_segment_boundary); |
702 | ||
03100aad KB |
703 | /** |
704 | * blk_queue_virt_boundary - set boundary rules for bio merging | |
705 | * @q: the request queue for the device | |
706 | * @mask: the memory boundary mask | |
707 | **/ | |
708 | void blk_queue_virt_boundary(struct request_queue *q, unsigned long mask) | |
709 | { | |
710 | q->limits.virt_boundary_mask = mask; | |
09324d32 CH |
711 | |
712 | /* | |
713 | * Devices that require a virtual boundary do not support scatter/gather | |
714 | * I/O natively, but instead require a descriptor list entry for each | |
715 | * page (which might not be idential to the Linux PAGE_SIZE). Because | |
716 | * of that they are not limited by our notion of "segment size". | |
717 | */ | |
c6c84f78 CH |
718 | if (mask) |
719 | q->limits.max_segment_size = UINT_MAX; | |
03100aad KB |
720 | } |
721 | EXPORT_SYMBOL(blk_queue_virt_boundary); | |
722 | ||
86db1e29 JA |
723 | /** |
724 | * blk_queue_dma_alignment - set dma length and memory alignment | |
725 | * @q: the request queue for the device | |
726 | * @mask: alignment mask | |
727 | * | |
728 | * description: | |
710027a4 | 729 | * set required memory and length alignment for direct dma transactions. |
8feb4d20 | 730 | * this is used when building direct io requests for the queue. |
86db1e29 JA |
731 | * |
732 | **/ | |
733 | void blk_queue_dma_alignment(struct request_queue *q, int mask) | |
734 | { | |
735 | q->dma_alignment = mask; | |
736 | } | |
86db1e29 JA |
737 | EXPORT_SYMBOL(blk_queue_dma_alignment); |
738 | ||
739 | /** | |
740 | * blk_queue_update_dma_alignment - update dma length and memory alignment | |
741 | * @q: the request queue for the device | |
742 | * @mask: alignment mask | |
743 | * | |
744 | * description: | |
710027a4 | 745 | * update required memory and length alignment for direct dma transactions. |
86db1e29 JA |
746 | * If the requested alignment is larger than the current alignment, then |
747 | * the current queue alignment is updated to the new value, otherwise it | |
748 | * is left alone. The design of this is to allow multiple objects | |
749 | * (driver, device, transport etc) to set their respective | |
750 | * alignments without having them interfere. | |
751 | * | |
752 | **/ | |
753 | void blk_queue_update_dma_alignment(struct request_queue *q, int mask) | |
754 | { | |
755 | BUG_ON(mask > PAGE_SIZE); | |
756 | ||
757 | if (mask > q->dma_alignment) | |
758 | q->dma_alignment = mask; | |
759 | } | |
86db1e29 JA |
760 | EXPORT_SYMBOL(blk_queue_update_dma_alignment); |
761 | ||
d278d4a8 JA |
762 | /** |
763 | * blk_set_queue_depth - tell the block layer about the device queue depth | |
764 | * @q: the request queue for the device | |
765 | * @depth: queue depth | |
766 | * | |
767 | */ | |
768 | void blk_set_queue_depth(struct request_queue *q, unsigned int depth) | |
769 | { | |
770 | q->queue_depth = depth; | |
9677a3e0 | 771 | rq_qos_queue_depth_changed(q); |
d278d4a8 JA |
772 | } |
773 | EXPORT_SYMBOL(blk_set_queue_depth); | |
774 | ||
93e9d8e8 JA |
775 | /** |
776 | * blk_queue_write_cache - configure queue's write cache | |
777 | * @q: the request queue for the device | |
778 | * @wc: write back cache on or off | |
779 | * @fua: device supports FUA writes, if true | |
780 | * | |
781 | * Tell the block layer about the write cache of @q. | |
782 | */ | |
783 | void blk_queue_write_cache(struct request_queue *q, bool wc, bool fua) | |
784 | { | |
c888a8f9 | 785 | if (wc) |
57d74df9 | 786 | blk_queue_flag_set(QUEUE_FLAG_WC, q); |
c888a8f9 | 787 | else |
57d74df9 | 788 | blk_queue_flag_clear(QUEUE_FLAG_WC, q); |
c888a8f9 | 789 | if (fua) |
57d74df9 | 790 | blk_queue_flag_set(QUEUE_FLAG_FUA, q); |
c888a8f9 | 791 | else |
57d74df9 | 792 | blk_queue_flag_clear(QUEUE_FLAG_FUA, q); |
87760e5e | 793 | |
a7905043 | 794 | wbt_set_write_cache(q, test_bit(QUEUE_FLAG_WC, &q->queue_flags)); |
93e9d8e8 JA |
795 | } |
796 | EXPORT_SYMBOL_GPL(blk_queue_write_cache); | |
797 | ||
68c43f13 DLM |
798 | /** |
799 | * blk_queue_required_elevator_features - Set a queue required elevator features | |
800 | * @q: the request queue for the target device | |
801 | * @features: Required elevator features OR'ed together | |
802 | * | |
803 | * Tell the block layer that for the device controlled through @q, only the | |
804 | * only elevators that can be used are those that implement at least the set of | |
805 | * features specified by @features. | |
806 | */ | |
807 | void blk_queue_required_elevator_features(struct request_queue *q, | |
808 | unsigned int features) | |
809 | { | |
810 | q->required_elevator_features = features; | |
811 | } | |
812 | EXPORT_SYMBOL_GPL(blk_queue_required_elevator_features); | |
813 | ||
45147fb5 YS |
814 | /** |
815 | * blk_queue_can_use_dma_map_merging - configure queue for merging segments. | |
816 | * @q: the request queue for the device | |
817 | * @dev: the device pointer for dma | |
818 | * | |
819 | * Tell the block layer about merging the segments by dma map of @q. | |
820 | */ | |
821 | bool blk_queue_can_use_dma_map_merging(struct request_queue *q, | |
822 | struct device *dev) | |
823 | { | |
824 | unsigned long boundary = dma_get_merge_boundary(dev); | |
825 | ||
826 | if (!boundary) | |
827 | return false; | |
828 | ||
829 | /* No need to update max_segment_size. see blk_queue_virt_boundary() */ | |
830 | blk_queue_virt_boundary(q, boundary); | |
831 | ||
832 | return true; | |
833 | } | |
834 | EXPORT_SYMBOL_GPL(blk_queue_can_use_dma_map_merging); | |
835 | ||
27ba3e8f DLM |
836 | /** |
837 | * blk_queue_set_zoned - configure a disk queue zoned model. | |
838 | * @disk: the gendisk of the queue to configure | |
839 | * @model: the zoned model to set | |
840 | * | |
841 | * Set the zoned model of the request queue of @disk according to @model. | |
842 | * When @model is BLK_ZONED_HM (host managed), this should be called only | |
843 | * if zoned block device support is enabled (CONFIG_BLK_DEV_ZONED option). | |
844 | * If @model specifies BLK_ZONED_HA (host aware), the effective model used | |
845 | * depends on CONFIG_BLK_DEV_ZONED settings and on the existence of partitions | |
846 | * on the disk. | |
847 | */ | |
848 | void blk_queue_set_zoned(struct gendisk *disk, enum blk_zoned_model model) | |
849 | { | |
850 | switch (model) { | |
851 | case BLK_ZONED_HM: | |
852 | /* | |
853 | * Host managed devices are supported only if | |
854 | * CONFIG_BLK_DEV_ZONED is enabled. | |
855 | */ | |
856 | WARN_ON_ONCE(!IS_ENABLED(CONFIG_BLK_DEV_ZONED)); | |
857 | break; | |
858 | case BLK_ZONED_HA: | |
859 | /* | |
860 | * Host aware devices can be treated either as regular block | |
861 | * devices (similar to drive managed devices) or as zoned block | |
862 | * devices to take advantage of the zone command set, similarly | |
863 | * to host managed devices. We try the latter if there are no | |
864 | * partitions and zoned block device support is enabled, else | |
865 | * we do nothing special as far as the block layer is concerned. | |
866 | */ | |
867 | if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) || | |
a33df75c | 868 | !xa_empty(&disk->part_tbl)) |
27ba3e8f DLM |
869 | model = BLK_ZONED_NONE; |
870 | break; | |
871 | case BLK_ZONED_NONE: | |
872 | default: | |
873 | if (WARN_ON_ONCE(model != BLK_ZONED_NONE)) | |
874 | model = BLK_ZONED_NONE; | |
875 | break; | |
876 | } | |
877 | ||
878 | disk->queue->limits.zoned = model; | |
879 | } | |
880 | EXPORT_SYMBOL_GPL(blk_queue_set_zoned); | |
881 | ||
aeb3d3a8 | 882 | static int __init blk_settings_init(void) |
86db1e29 JA |
883 | { |
884 | blk_max_low_pfn = max_low_pfn - 1; | |
885 | blk_max_pfn = max_pfn - 1; | |
886 | return 0; | |
887 | } | |
888 | subsys_initcall(blk_settings_init); |