1 // SPDX-License-Identifier: GPL-2.0
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>
10 #include <linux/pagemap.h>
11 #include <linux/backing-dev-defs.h>
12 #include <linux/gcd.h>
13 #include <linux/lcm.h>
14 #include <linux/jiffies.h>
15 #include <linux/gfp.h>
16 #include <linux/dma-mapping.h>
17
18 #include "blk.h"
19 #include "blk-rq-qos.h"
20 #include "blk-wbt.h"
21
blk_queue_rq_timeout(struct request_queue * q,unsigned int timeout)22 void blk_queue_rq_timeout(struct request_queue *q, unsigned int timeout)
23 {
24 q->rq_timeout = timeout;
25 }
26 EXPORT_SYMBOL_GPL(blk_queue_rq_timeout);
27
28 /**
29 * blk_set_default_limits - reset limits to default values
30 * @lim: the queue_limits structure to reset
31 *
32 * Description:
33 * Returns a queue_limit struct to its default state.
34 */
blk_set_default_limits(struct queue_limits * lim)35 void blk_set_default_limits(struct queue_limits *lim)
36 {
37 lim->max_segments = BLK_MAX_SEGMENTS;
38 lim->max_discard_segments = 1;
39 lim->max_integrity_segments = 0;
40 lim->seg_boundary_mask = BLK_SEG_BOUNDARY_MASK;
41 lim->virt_boundary_mask = 0;
42 lim->max_segment_size = BLK_MAX_SEGMENT_SIZE;
43 lim->max_sectors = lim->max_hw_sectors = BLK_SAFE_MAX_SECTORS;
44 lim->max_user_sectors = lim->max_dev_sectors = 0;
45 lim->chunk_sectors = 0;
46 lim->max_write_zeroes_sectors = 0;
47 lim->max_zone_append_sectors = 0;
48 lim->max_discard_sectors = 0;
49 lim->max_hw_discard_sectors = 0;
50 lim->max_secure_erase_sectors = 0;
51 lim->discard_granularity = 512;
52 lim->discard_alignment = 0;
53 lim->discard_misaligned = 0;
54 lim->logical_block_size = lim->physical_block_size = lim->io_min = 512;
55 lim->bounce = BLK_BOUNCE_NONE;
56 lim->alignment_offset = 0;
57 lim->io_opt = 0;
58 lim->misaligned = 0;
59 lim->zoned = false;
60 lim->zone_write_granularity = 0;
61 lim->dma_alignment = 511;
62 }
63
64 /**
65 * blk_set_stacking_limits - set default limits for stacking devices
66 * @lim: the queue_limits structure to reset
67 *
68 * Description:
69 * Returns a queue_limit struct to its default state. Should be used
70 * by stacking drivers like DM that have no internal limits.
71 */
blk_set_stacking_limits(struct queue_limits * lim)72 void blk_set_stacking_limits(struct queue_limits *lim)
73 {
74 blk_set_default_limits(lim);
75
76 /* Inherit limits from component devices */
77 lim->max_segments = USHRT_MAX;
78 lim->max_discard_segments = USHRT_MAX;
79 lim->max_hw_sectors = UINT_MAX;
80 lim->max_segment_size = UINT_MAX;
81 lim->max_sectors = UINT_MAX;
82 lim->max_dev_sectors = UINT_MAX;
83 lim->max_write_zeroes_sectors = UINT_MAX;
84 lim->max_zone_append_sectors = UINT_MAX;
85 }
86 EXPORT_SYMBOL(blk_set_stacking_limits);
87
88 /**
89 * blk_queue_bounce_limit - set bounce buffer limit for queue
90 * @q: the request queue for the device
91 * @bounce: bounce limit to enforce
92 *
93 * Description:
94 * Force bouncing for ISA DMA ranges or highmem.
95 *
96 * DEPRECATED, don't use in new code.
97 **/
blk_queue_bounce_limit(struct request_queue * q,enum blk_bounce bounce)98 void blk_queue_bounce_limit(struct request_queue *q, enum blk_bounce bounce)
99 {
100 q->limits.bounce = bounce;
101 }
102 EXPORT_SYMBOL(blk_queue_bounce_limit);
103
104 /**
105 * blk_queue_max_hw_sectors - set max sectors for a request for this queue
106 * @q: the request queue for the device
107 * @max_hw_sectors: max hardware sectors in the usual 512b unit
108 *
109 * Description:
110 * Enables a low level driver to set a hard upper limit,
111 * max_hw_sectors, on the size of requests. max_hw_sectors is set by
112 * the device driver based upon the capabilities of the I/O
113 * controller.
114 *
115 * max_dev_sectors is a hard limit imposed by the storage device for
116 * READ/WRITE requests. It is set by the disk driver.
117 *
118 * max_sectors is a soft limit imposed by the block layer for
119 * filesystem type requests. This value can be overridden on a
120 * per-device basis in /sys/block/<device>/queue/max_sectors_kb.
121 * The soft limit can not exceed max_hw_sectors.
122 **/
blk_queue_max_hw_sectors(struct request_queue * q,unsigned int max_hw_sectors)123 void blk_queue_max_hw_sectors(struct request_queue *q, unsigned int max_hw_sectors)
124 {
125 struct queue_limits *limits = &q->limits;
126 unsigned int max_sectors;
127
128 if ((max_hw_sectors << 9) < PAGE_SIZE) {
129 max_hw_sectors = 1 << (PAGE_SHIFT - 9);
130 pr_info("%s: set to minimum %u\n", __func__, max_hw_sectors);
131 }
132
133 max_hw_sectors = round_down(max_hw_sectors,
134 limits->logical_block_size >> SECTOR_SHIFT);
135 limits->max_hw_sectors = max_hw_sectors;
136
137 max_sectors = min_not_zero(max_hw_sectors, limits->max_dev_sectors);
138
139 if (limits->max_user_sectors)
140 max_sectors = min(max_sectors, limits->max_user_sectors);
141 else
142 max_sectors = min(max_sectors, BLK_DEF_MAX_SECTORS_CAP);
143
144 max_sectors = round_down(max_sectors,
145 limits->logical_block_size >> SECTOR_SHIFT);
146 limits->max_sectors = max_sectors;
147
148 if (!q->disk)
149 return;
150 q->disk->bdi->io_pages = max_sectors >> (PAGE_SHIFT - 9);
151 }
152 EXPORT_SYMBOL(blk_queue_max_hw_sectors);
153
154 /**
155 * blk_queue_chunk_sectors - set size of the chunk for this queue
156 * @q: the request queue for the device
157 * @chunk_sectors: chunk sectors in the usual 512b unit
158 *
159 * Description:
160 * If a driver doesn't want IOs to cross a given chunk size, it can set
161 * this limit and prevent merging across chunks. Note that the block layer
162 * must accept a page worth of data at any offset. So if the crossing of
163 * chunks is a hard limitation in the driver, it must still be prepared
164 * to split single page bios.
165 **/
blk_queue_chunk_sectors(struct request_queue * q,unsigned int chunk_sectors)166 void blk_queue_chunk_sectors(struct request_queue *q, unsigned int chunk_sectors)
167 {
168 q->limits.chunk_sectors = chunk_sectors;
169 }
170 EXPORT_SYMBOL(blk_queue_chunk_sectors);
171
172 /**
173 * blk_queue_max_discard_sectors - set max sectors for a single discard
174 * @q: the request queue for the device
175 * @max_discard_sectors: maximum number of sectors to discard
176 **/
blk_queue_max_discard_sectors(struct request_queue * q,unsigned int max_discard_sectors)177 void blk_queue_max_discard_sectors(struct request_queue *q,
178 unsigned int max_discard_sectors)
179 {
180 q->limits.max_hw_discard_sectors = max_discard_sectors;
181 q->limits.max_discard_sectors = max_discard_sectors;
182 }
183 EXPORT_SYMBOL(blk_queue_max_discard_sectors);
184
185 /**
186 * blk_queue_max_secure_erase_sectors - set max sectors for a secure erase
187 * @q: the request queue for the device
188 * @max_sectors: maximum number of sectors to secure_erase
189 **/
blk_queue_max_secure_erase_sectors(struct request_queue * q,unsigned int max_sectors)190 void blk_queue_max_secure_erase_sectors(struct request_queue *q,
191 unsigned int max_sectors)
192 {
193 q->limits.max_secure_erase_sectors = max_sectors;
194 }
195 EXPORT_SYMBOL(blk_queue_max_secure_erase_sectors);
196
197 /**
198 * blk_queue_max_write_zeroes_sectors - set max sectors for a single
199 * write zeroes
200 * @q: the request queue for the device
201 * @max_write_zeroes_sectors: maximum number of sectors to write per command
202 **/
blk_queue_max_write_zeroes_sectors(struct request_queue * q,unsigned int max_write_zeroes_sectors)203 void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
204 unsigned int max_write_zeroes_sectors)
205 {
206 q->limits.max_write_zeroes_sectors = max_write_zeroes_sectors;
207 }
208 EXPORT_SYMBOL(blk_queue_max_write_zeroes_sectors);
209
210 /**
211 * blk_queue_max_zone_append_sectors - set max sectors for a single zone append
212 * @q: the request queue for the device
213 * @max_zone_append_sectors: maximum number of sectors to write per command
214 **/
blk_queue_max_zone_append_sectors(struct request_queue * q,unsigned int max_zone_append_sectors)215 void blk_queue_max_zone_append_sectors(struct request_queue *q,
216 unsigned int max_zone_append_sectors)
217 {
218 unsigned int max_sectors;
219
220 if (WARN_ON(!blk_queue_is_zoned(q)))
221 return;
222
223 max_sectors = min(q->limits.max_hw_sectors, max_zone_append_sectors);
224 max_sectors = min(q->limits.chunk_sectors, max_sectors);
225
226 /*
227 * Signal eventual driver bugs resulting in the max_zone_append sectors limit
228 * being 0 due to a 0 argument, the chunk_sectors limit (zone size) not set,
229 * or the max_hw_sectors limit not set.
230 */
231 WARN_ON(!max_sectors);
232
233 q->limits.max_zone_append_sectors = max_sectors;
234 }
235 EXPORT_SYMBOL_GPL(blk_queue_max_zone_append_sectors);
236
237 /**
238 * blk_queue_max_segments - set max hw segments for a request for this queue
239 * @q: the request queue for the device
240 * @max_segments: max number of segments
241 *
242 * Description:
243 * Enables a low level driver to set an upper limit on the number of
244 * hw data segments in a request.
245 **/
blk_queue_max_segments(struct request_queue * q,unsigned short max_segments)246 void blk_queue_max_segments(struct request_queue *q, unsigned short max_segments)
247 {
248 if (!max_segments) {
249 max_segments = 1;
250 pr_info("%s: set to minimum %u\n", __func__, max_segments);
251 }
252
253 q->limits.max_segments = max_segments;
254 }
255 EXPORT_SYMBOL(blk_queue_max_segments);
256
257 /**
258 * blk_queue_max_discard_segments - set max segments for discard requests
259 * @q: the request queue for the device
260 * @max_segments: max number of segments
261 *
262 * Description:
263 * Enables a low level driver to set an upper limit on the number of
264 * segments in a discard request.
265 **/
blk_queue_max_discard_segments(struct request_queue * q,unsigned short max_segments)266 void blk_queue_max_discard_segments(struct request_queue *q,
267 unsigned short max_segments)
268 {
269 q->limits.max_discard_segments = max_segments;
270 }
271 EXPORT_SYMBOL_GPL(blk_queue_max_discard_segments);
272
273 /**
274 * blk_queue_max_segment_size - set max segment size for blk_rq_map_sg
275 * @q: the request queue for the device
276 * @max_size: max size of segment in bytes
277 *
278 * Description:
279 * Enables a low level driver to set an upper limit on the size of a
280 * coalesced segment
281 **/
blk_queue_max_segment_size(struct request_queue * q,unsigned int max_size)282 void blk_queue_max_segment_size(struct request_queue *q, unsigned int max_size)
283 {
284 if (max_size < PAGE_SIZE) {
285 max_size = PAGE_SIZE;
286 pr_info("%s: set to minimum %u\n", __func__, max_size);
287 }
288
289 /* see blk_queue_virt_boundary() for the explanation */
290 WARN_ON_ONCE(q->limits.virt_boundary_mask);
291
292 q->limits.max_segment_size = max_size;
293 }
294 EXPORT_SYMBOL(blk_queue_max_segment_size);
295
296 /**
297 * blk_queue_logical_block_size - set logical block size for the queue
298 * @q: the request queue for the device
299 * @size: the logical block size, in bytes
300 *
301 * Description:
302 * This should be set to the lowest possible block size that the
303 * storage device can address. The default of 512 covers most
304 * hardware.
305 **/
blk_queue_logical_block_size(struct request_queue * q,unsigned int size)306 void blk_queue_logical_block_size(struct request_queue *q, unsigned int size)
307 {
308 struct queue_limits *limits = &q->limits;
309
310 limits->logical_block_size = size;
311
312 if (limits->discard_granularity < limits->logical_block_size)
313 limits->discard_granularity = limits->logical_block_size;
314
315 if (limits->physical_block_size < size)
316 limits->physical_block_size = size;
317
318 if (limits->io_min < limits->physical_block_size)
319 limits->io_min = limits->physical_block_size;
320
321 limits->max_hw_sectors =
322 round_down(limits->max_hw_sectors, size >> SECTOR_SHIFT);
323 limits->max_sectors =
324 round_down(limits->max_sectors, size >> SECTOR_SHIFT);
325 }
326 EXPORT_SYMBOL(blk_queue_logical_block_size);
327
328 /**
329 * blk_queue_physical_block_size - set physical block size for the queue
330 * @q: the request queue for the device
331 * @size: the physical block size, in bytes
332 *
333 * Description:
334 * This should be set to the lowest possible sector size that the
335 * hardware can operate on without reverting to read-modify-write
336 * operations.
337 */
blk_queue_physical_block_size(struct request_queue * q,unsigned int size)338 void blk_queue_physical_block_size(struct request_queue *q, unsigned int size)
339 {
340 q->limits.physical_block_size = size;
341
342 if (q->limits.physical_block_size < q->limits.logical_block_size)
343 q->limits.physical_block_size = q->limits.logical_block_size;
344
345 if (q->limits.discard_granularity < q->limits.physical_block_size)
346 q->limits.discard_granularity = q->limits.physical_block_size;
347
348 if (q->limits.io_min < q->limits.physical_block_size)
349 q->limits.io_min = q->limits.physical_block_size;
350 }
351 EXPORT_SYMBOL(blk_queue_physical_block_size);
352
353 /**
354 * blk_queue_zone_write_granularity - set zone write granularity for the queue
355 * @q: the request queue for the zoned device
356 * @size: the zone write granularity size, in bytes
357 *
358 * Description:
359 * This should be set to the lowest possible size allowing to write in
360 * sequential zones of a zoned block device.
361 */
blk_queue_zone_write_granularity(struct request_queue * q,unsigned int size)362 void blk_queue_zone_write_granularity(struct request_queue *q,
363 unsigned int size)
364 {
365 if (WARN_ON_ONCE(!blk_queue_is_zoned(q)))
366 return;
367
368 q->limits.zone_write_granularity = size;
369
370 if (q->limits.zone_write_granularity < q->limits.logical_block_size)
371 q->limits.zone_write_granularity = q->limits.logical_block_size;
372 }
373 EXPORT_SYMBOL_GPL(blk_queue_zone_write_granularity);
374
375 /**
376 * blk_queue_alignment_offset - set physical block alignment offset
377 * @q: the request queue for the device
378 * @offset: alignment offset in bytes
379 *
380 * Description:
381 * Some devices are naturally misaligned to compensate for things like
382 * the legacy DOS partition table 63-sector offset. Low-level drivers
383 * should call this function for devices whose first sector is not
384 * naturally aligned.
385 */
blk_queue_alignment_offset(struct request_queue * q,unsigned int offset)386 void blk_queue_alignment_offset(struct request_queue *q, unsigned int offset)
387 {
388 q->limits.alignment_offset =
389 offset & (q->limits.physical_block_size - 1);
390 q->limits.misaligned = 0;
391 }
392 EXPORT_SYMBOL(blk_queue_alignment_offset);
393
disk_update_readahead(struct gendisk * disk)394 void disk_update_readahead(struct gendisk *disk)
395 {
396 struct request_queue *q = disk->queue;
397
398 /*
399 * For read-ahead of large files to be effective, we need to read ahead
400 * at least twice the optimal I/O size.
401 */
402 disk->bdi->ra_pages =
403 max(queue_io_opt(q) * 2 / PAGE_SIZE, VM_READAHEAD_PAGES);
404 disk->bdi->io_pages = queue_max_sectors(q) >> (PAGE_SHIFT - 9);
405 }
406 EXPORT_SYMBOL_GPL(disk_update_readahead);
407
408 /**
409 * blk_limits_io_min - set minimum request size for a device
410 * @limits: the queue limits
411 * @min: smallest I/O size in bytes
412 *
413 * Description:
414 * Some devices have an internal block size bigger than the reported
415 * hardware sector size. This function can be used to signal the
416 * smallest I/O the device can perform without incurring a performance
417 * penalty.
418 */
blk_limits_io_min(struct queue_limits * limits,unsigned int min)419 void blk_limits_io_min(struct queue_limits *limits, unsigned int min)
420 {
421 limits->io_min = min;
422
423 if (limits->io_min < limits->logical_block_size)
424 limits->io_min = limits->logical_block_size;
425
426 if (limits->io_min < limits->physical_block_size)
427 limits->io_min = limits->physical_block_size;
428 }
429 EXPORT_SYMBOL(blk_limits_io_min);
430
431 /**
432 * blk_queue_io_min - set minimum request size for the queue
433 * @q: the request queue for the device
434 * @min: smallest I/O size in bytes
435 *
436 * Description:
437 * Storage devices may report a granularity or preferred minimum I/O
438 * size which is the smallest request the device can perform without
439 * incurring a performance penalty. For disk drives this is often the
440 * physical block size. For RAID arrays it is often the stripe chunk
441 * size. A properly aligned multiple of minimum_io_size is the
442 * preferred request size for workloads where a high number of I/O
443 * operations is desired.
444 */
blk_queue_io_min(struct request_queue * q,unsigned int min)445 void blk_queue_io_min(struct request_queue *q, unsigned int min)
446 {
447 blk_limits_io_min(&q->limits, min);
448 }
449 EXPORT_SYMBOL(blk_queue_io_min);
450
451 /**
452 * blk_limits_io_opt - set optimal request size for a device
453 * @limits: the queue limits
454 * @opt: smallest I/O size in bytes
455 *
456 * Description:
457 * Storage devices may report an optimal I/O size, which is the
458 * device's preferred unit for sustained I/O. This is rarely reported
459 * for disk drives. For RAID arrays it is usually the stripe width or
460 * the internal track size. A properly aligned multiple of
461 * optimal_io_size is the preferred request size for workloads where
462 * sustained throughput is desired.
463 */
blk_limits_io_opt(struct queue_limits * limits,unsigned int opt)464 void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt)
465 {
466 limits->io_opt = opt;
467 }
468 EXPORT_SYMBOL(blk_limits_io_opt);
469
470 /**
471 * blk_queue_io_opt - set optimal request size for the queue
472 * @q: the request queue for the device
473 * @opt: optimal request size in bytes
474 *
475 * Description:
476 * Storage devices may report an optimal I/O size, which is the
477 * device's preferred unit for sustained I/O. This is rarely reported
478 * for disk drives. For RAID arrays it is usually the stripe width or
479 * the internal track size. A properly aligned multiple of
480 * optimal_io_size is the preferred request size for workloads where
481 * sustained throughput is desired.
482 */
blk_queue_io_opt(struct request_queue * q,unsigned int opt)483 void blk_queue_io_opt(struct request_queue *q, unsigned int opt)
484 {
485 blk_limits_io_opt(&q->limits, opt);
486 if (!q->disk)
487 return;
488 q->disk->bdi->ra_pages =
489 max(queue_io_opt(q) * 2 / PAGE_SIZE, VM_READAHEAD_PAGES);
490 }
491 EXPORT_SYMBOL(blk_queue_io_opt);
492
queue_limit_alignment_offset(const struct queue_limits * lim,sector_t sector)493 static int queue_limit_alignment_offset(const struct queue_limits *lim,
494 sector_t sector)
495 {
496 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
497 unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT)
498 << SECTOR_SHIFT;
499
500 return (granularity + lim->alignment_offset - alignment) % granularity;
501 }
502
queue_limit_discard_alignment(const struct queue_limits * lim,sector_t sector)503 static unsigned int queue_limit_discard_alignment(
504 const struct queue_limits *lim, sector_t sector)
505 {
506 unsigned int alignment, granularity, offset;
507
508 if (!lim->max_discard_sectors)
509 return 0;
510
511 /* Why are these in bytes, not sectors? */
512 alignment = lim->discard_alignment >> SECTOR_SHIFT;
513 granularity = lim->discard_granularity >> SECTOR_SHIFT;
514 if (!granularity)
515 return 0;
516
517 /* Offset of the partition start in 'granularity' sectors */
518 offset = sector_div(sector, granularity);
519
520 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
521 offset = (granularity + alignment - offset) % granularity;
522
523 /* Turn it back into bytes, gaah */
524 return offset << SECTOR_SHIFT;
525 }
526
blk_round_down_sectors(unsigned int sectors,unsigned int lbs)527 static unsigned int blk_round_down_sectors(unsigned int sectors, unsigned int lbs)
528 {
529 sectors = round_down(sectors, lbs >> SECTOR_SHIFT);
530 if (sectors < PAGE_SIZE >> SECTOR_SHIFT)
531 sectors = PAGE_SIZE >> SECTOR_SHIFT;
532 return sectors;
533 }
534
535 /**
536 * blk_stack_limits - adjust queue_limits for stacked devices
537 * @t: the stacking driver limits (top device)
538 * @b: the underlying queue limits (bottom, component device)
539 * @start: first data sector within component device
540 *
541 * Description:
542 * This function is used by stacking drivers like MD and DM to ensure
543 * that all component devices have compatible block sizes and
544 * alignments. The stacking driver must provide a queue_limits
545 * struct (top) and then iteratively call the stacking function for
546 * all component (bottom) devices. The stacking function will
547 * attempt to combine the values and ensure proper alignment.
548 *
549 * Returns 0 if the top and bottom queue_limits are compatible. The
550 * top device's block sizes and alignment offsets may be adjusted to
551 * ensure alignment with the bottom device. If no compatible sizes
552 * and alignments exist, -1 is returned and the resulting top
553 * queue_limits will have the misaligned flag set to indicate that
554 * the alignment_offset is undefined.
555 */
blk_stack_limits(struct queue_limits * t,struct queue_limits * b,sector_t start)556 int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
557 sector_t start)
558 {
559 unsigned int top, bottom, alignment, ret = 0;
560
561 t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors);
562 t->max_hw_sectors = min_not_zero(t->max_hw_sectors, b->max_hw_sectors);
563 t->max_dev_sectors = min_not_zero(t->max_dev_sectors, b->max_dev_sectors);
564 t->max_write_zeroes_sectors = min(t->max_write_zeroes_sectors,
565 b->max_write_zeroes_sectors);
566 t->max_zone_append_sectors = min(t->max_zone_append_sectors,
567 b->max_zone_append_sectors);
568 t->bounce = max(t->bounce, b->bounce);
569
570 t->seg_boundary_mask = min_not_zero(t->seg_boundary_mask,
571 b->seg_boundary_mask);
572 t->virt_boundary_mask = min_not_zero(t->virt_boundary_mask,
573 b->virt_boundary_mask);
574
575 t->max_segments = min_not_zero(t->max_segments, b->max_segments);
576 t->max_discard_segments = min_not_zero(t->max_discard_segments,
577 b->max_discard_segments);
578 t->max_integrity_segments = min_not_zero(t->max_integrity_segments,
579 b->max_integrity_segments);
580
581 t->max_segment_size = min_not_zero(t->max_segment_size,
582 b->max_segment_size);
583
584 t->misaligned |= b->misaligned;
585
586 alignment = queue_limit_alignment_offset(b, start);
587
588 /* Bottom device has different alignment. Check that it is
589 * compatible with the current top alignment.
590 */
591 if (t->alignment_offset != alignment) {
592
593 top = max(t->physical_block_size, t->io_min)
594 + t->alignment_offset;
595 bottom = max(b->physical_block_size, b->io_min) + alignment;
596
597 /* Verify that top and bottom intervals line up */
598 if (max(top, bottom) % min(top, bottom)) {
599 t->misaligned = 1;
600 ret = -1;
601 }
602 }
603
604 t->logical_block_size = max(t->logical_block_size,
605 b->logical_block_size);
606
607 t->physical_block_size = max(t->physical_block_size,
608 b->physical_block_size);
609
610 t->io_min = max(t->io_min, b->io_min);
611 t->io_opt = lcm_not_zero(t->io_opt, b->io_opt);
612 t->dma_alignment = max(t->dma_alignment, b->dma_alignment);
613
614 /* Set non-power-of-2 compatible chunk_sectors boundary */
615 if (b->chunk_sectors)
616 t->chunk_sectors = gcd(t->chunk_sectors, b->chunk_sectors);
617
618 /* Physical block size a multiple of the logical block size? */
619 if (t->physical_block_size & (t->logical_block_size - 1)) {
620 t->physical_block_size = t->logical_block_size;
621 t->misaligned = 1;
622 ret = -1;
623 }
624
625 /* Minimum I/O a multiple of the physical block size? */
626 if (t->io_min & (t->physical_block_size - 1)) {
627 t->io_min = t->physical_block_size;
628 t->misaligned = 1;
629 ret = -1;
630 }
631
632 /* Optimal I/O a multiple of the physical block size? */
633 if (t->io_opt & (t->physical_block_size - 1)) {
634 t->io_opt = 0;
635 t->misaligned = 1;
636 ret = -1;
637 }
638
639 /* chunk_sectors a multiple of the physical block size? */
640 if ((t->chunk_sectors << 9) & (t->physical_block_size - 1)) {
641 t->chunk_sectors = 0;
642 t->misaligned = 1;
643 ret = -1;
644 }
645
646 t->raid_partial_stripes_expensive =
647 max(t->raid_partial_stripes_expensive,
648 b->raid_partial_stripes_expensive);
649
650 /* Find lowest common alignment_offset */
651 t->alignment_offset = lcm_not_zero(t->alignment_offset, alignment)
652 % max(t->physical_block_size, t->io_min);
653
654 /* Verify that new alignment_offset is on a logical block boundary */
655 if (t->alignment_offset & (t->logical_block_size - 1)) {
656 t->misaligned = 1;
657 ret = -1;
658 }
659
660 t->max_sectors = blk_round_down_sectors(t->max_sectors, t->logical_block_size);
661 t->max_hw_sectors = blk_round_down_sectors(t->max_hw_sectors, t->logical_block_size);
662 t->max_dev_sectors = blk_round_down_sectors(t->max_dev_sectors, t->logical_block_size);
663
664 /* Discard alignment and granularity */
665 if (b->discard_granularity) {
666 alignment = queue_limit_discard_alignment(b, start);
667
668 if (t->discard_granularity != 0 &&
669 t->discard_alignment != alignment) {
670 top = t->discard_granularity + t->discard_alignment;
671 bottom = b->discard_granularity + alignment;
672
673 /* Verify that top and bottom intervals line up */
674 if ((max(top, bottom) % min(top, bottom)) != 0)
675 t->discard_misaligned = 1;
676 }
677
678 t->max_discard_sectors = min_not_zero(t->max_discard_sectors,
679 b->max_discard_sectors);
680 t->max_hw_discard_sectors = min_not_zero(t->max_hw_discard_sectors,
681 b->max_hw_discard_sectors);
682 t->discard_granularity = max(t->discard_granularity,
683 b->discard_granularity);
684 t->discard_alignment = lcm_not_zero(t->discard_alignment, alignment) %
685 t->discard_granularity;
686 }
687 t->max_secure_erase_sectors = min_not_zero(t->max_secure_erase_sectors,
688 b->max_secure_erase_sectors);
689 t->zone_write_granularity = max(t->zone_write_granularity,
690 b->zone_write_granularity);
691 t->zoned = max(t->zoned, b->zoned);
692 return ret;
693 }
694 EXPORT_SYMBOL(blk_stack_limits);
695
696 /**
697 * disk_stack_limits - adjust queue limits for stacked drivers
698 * @disk: MD/DM gendisk (top)
699 * @bdev: the underlying block device (bottom)
700 * @offset: offset to beginning of data within component device
701 *
702 * Description:
703 * Merges the limits for a top level gendisk and a bottom level
704 * block_device.
705 */
disk_stack_limits(struct gendisk * disk,struct block_device * bdev,sector_t offset)706 void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
707 sector_t offset)
708 {
709 struct request_queue *t = disk->queue;
710
711 if (blk_stack_limits(&t->limits, &bdev_get_queue(bdev)->limits,
712 get_start_sect(bdev) + (offset >> 9)) < 0)
713 pr_notice("%s: Warning: Device %pg is misaligned\n",
714 disk->disk_name, bdev);
715
716 disk_update_readahead(disk);
717 }
718 EXPORT_SYMBOL(disk_stack_limits);
719
720 /**
721 * blk_queue_update_dma_pad - update pad mask
722 * @q: the request queue for the device
723 * @mask: pad mask
724 *
725 * Update dma pad mask.
726 *
727 * Appending pad buffer to a request modifies the last entry of a
728 * scatter list such that it includes the pad buffer.
729 **/
blk_queue_update_dma_pad(struct request_queue * q,unsigned int mask)730 void blk_queue_update_dma_pad(struct request_queue *q, unsigned int mask)
731 {
732 if (mask > q->dma_pad_mask)
733 q->dma_pad_mask = mask;
734 }
735 EXPORT_SYMBOL(blk_queue_update_dma_pad);
736
737 /**
738 * blk_queue_segment_boundary - set boundary rules for segment merging
739 * @q: the request queue for the device
740 * @mask: the memory boundary mask
741 **/
blk_queue_segment_boundary(struct request_queue * q,unsigned long mask)742 void blk_queue_segment_boundary(struct request_queue *q, unsigned long mask)
743 {
744 if (mask < PAGE_SIZE - 1) {
745 mask = PAGE_SIZE - 1;
746 pr_info("%s: set to minimum %lx\n", __func__, mask);
747 }
748
749 q->limits.seg_boundary_mask = mask;
750 }
751 EXPORT_SYMBOL(blk_queue_segment_boundary);
752
753 /**
754 * blk_queue_virt_boundary - set boundary rules for bio merging
755 * @q: the request queue for the device
756 * @mask: the memory boundary mask
757 **/
blk_queue_virt_boundary(struct request_queue * q,unsigned long mask)758 void blk_queue_virt_boundary(struct request_queue *q, unsigned long mask)
759 {
760 q->limits.virt_boundary_mask = mask;
761
762 /*
763 * Devices that require a virtual boundary do not support scatter/gather
764 * I/O natively, but instead require a descriptor list entry for each
765 * page (which might not be idential to the Linux PAGE_SIZE). Because
766 * of that they are not limited by our notion of "segment size".
767 */
768 if (mask)
769 q->limits.max_segment_size = UINT_MAX;
770 }
771 EXPORT_SYMBOL(blk_queue_virt_boundary);
772
773 /**
774 * blk_queue_dma_alignment - set dma length and memory alignment
775 * @q: the request queue for the device
776 * @mask: alignment mask
777 *
778 * description:
779 * set required memory and length alignment for direct dma transactions.
780 * this is used when building direct io requests for the queue.
781 *
782 **/
blk_queue_dma_alignment(struct request_queue * q,int mask)783 void blk_queue_dma_alignment(struct request_queue *q, int mask)
784 {
785 q->limits.dma_alignment = mask;
786 }
787 EXPORT_SYMBOL(blk_queue_dma_alignment);
788
789 /**
790 * blk_queue_update_dma_alignment - update dma length and memory alignment
791 * @q: the request queue for the device
792 * @mask: alignment mask
793 *
794 * description:
795 * update required memory and length alignment for direct dma transactions.
796 * If the requested alignment is larger than the current alignment, then
797 * the current queue alignment is updated to the new value, otherwise it
798 * is left alone. The design of this is to allow multiple objects
799 * (driver, device, transport etc) to set their respective
800 * alignments without having them interfere.
801 *
802 **/
blk_queue_update_dma_alignment(struct request_queue * q,int mask)803 void blk_queue_update_dma_alignment(struct request_queue *q, int mask)
804 {
805 BUG_ON(mask > PAGE_SIZE);
806
807 if (mask > q->limits.dma_alignment)
808 q->limits.dma_alignment = mask;
809 }
810 EXPORT_SYMBOL(blk_queue_update_dma_alignment);
811
812 /**
813 * blk_set_queue_depth - tell the block layer about the device queue depth
814 * @q: the request queue for the device
815 * @depth: queue depth
816 *
817 */
blk_set_queue_depth(struct request_queue * q,unsigned int depth)818 void blk_set_queue_depth(struct request_queue *q, unsigned int depth)
819 {
820 q->queue_depth = depth;
821 rq_qos_queue_depth_changed(q);
822 }
823 EXPORT_SYMBOL(blk_set_queue_depth);
824
825 /**
826 * blk_queue_write_cache - configure queue's write cache
827 * @q: the request queue for the device
828 * @wc: write back cache on or off
829 * @fua: device supports FUA writes, if true
830 *
831 * Tell the block layer about the write cache of @q.
832 */
blk_queue_write_cache(struct request_queue * q,bool wc,bool fua)833 void blk_queue_write_cache(struct request_queue *q, bool wc, bool fua)
834 {
835 if (wc) {
836 blk_queue_flag_set(QUEUE_FLAG_HW_WC, q);
837 blk_queue_flag_set(QUEUE_FLAG_WC, q);
838 } else {
839 blk_queue_flag_clear(QUEUE_FLAG_HW_WC, q);
840 blk_queue_flag_clear(QUEUE_FLAG_WC, q);
841 }
842 if (fua)
843 blk_queue_flag_set(QUEUE_FLAG_FUA, q);
844 else
845 blk_queue_flag_clear(QUEUE_FLAG_FUA, q);
846 }
847 EXPORT_SYMBOL_GPL(blk_queue_write_cache);
848
849 /**
850 * blk_queue_required_elevator_features - Set a queue required elevator features
851 * @q: the request queue for the target device
852 * @features: Required elevator features OR'ed together
853 *
854 * Tell the block layer that for the device controlled through @q, only the
855 * only elevators that can be used are those that implement at least the set of
856 * features specified by @features.
857 */
blk_queue_required_elevator_features(struct request_queue * q,unsigned int features)858 void blk_queue_required_elevator_features(struct request_queue *q,
859 unsigned int features)
860 {
861 q->required_elevator_features = features;
862 }
863 EXPORT_SYMBOL_GPL(blk_queue_required_elevator_features);
864
865 /**
866 * blk_queue_can_use_dma_map_merging - configure queue for merging segments.
867 * @q: the request queue for the device
868 * @dev: the device pointer for dma
869 *
870 * Tell the block layer about merging the segments by dma map of @q.
871 */
blk_queue_can_use_dma_map_merging(struct request_queue * q,struct device * dev)872 bool blk_queue_can_use_dma_map_merging(struct request_queue *q,
873 struct device *dev)
874 {
875 unsigned long boundary = dma_get_merge_boundary(dev);
876
877 if (!boundary)
878 return false;
879
880 /* No need to update max_segment_size. see blk_queue_virt_boundary() */
881 blk_queue_virt_boundary(q, boundary);
882
883 return true;
884 }
885 EXPORT_SYMBOL_GPL(blk_queue_can_use_dma_map_merging);
886
887 /**
888 * disk_set_zoned - inidicate a zoned device
889 * @disk: gendisk to configure
890 */
disk_set_zoned(struct gendisk * disk)891 void disk_set_zoned(struct gendisk *disk)
892 {
893 struct request_queue *q = disk->queue;
894
895 WARN_ON_ONCE(!IS_ENABLED(CONFIG_BLK_DEV_ZONED));
896
897 /*
898 * Set the zone write granularity to the device logical block
899 * size by default. The driver can change this value if needed.
900 */
901 q->limits.zoned = true;
902 blk_queue_zone_write_granularity(q, queue_logical_block_size(q));
903 }
904 EXPORT_SYMBOL_GPL(disk_set_zoned);
905
bdev_alignment_offset(struct block_device * bdev)906 int bdev_alignment_offset(struct block_device *bdev)
907 {
908 struct request_queue *q = bdev_get_queue(bdev);
909
910 if (q->limits.misaligned)
911 return -1;
912 if (bdev_is_partition(bdev))
913 return queue_limit_alignment_offset(&q->limits,
914 bdev->bd_start_sect);
915 return q->limits.alignment_offset;
916 }
917 EXPORT_SYMBOL_GPL(bdev_alignment_offset);
918
bdev_discard_alignment(struct block_device * bdev)919 unsigned int bdev_discard_alignment(struct block_device *bdev)
920 {
921 struct request_queue *q = bdev_get_queue(bdev);
922
923 if (bdev_is_partition(bdev))
924 return queue_limit_discard_alignment(&q->limits,
925 bdev->bd_start_sect);
926 return q->limits.discard_alignment;
927 }
928 EXPORT_SYMBOL_GPL(bdev_discard_alignment);
929