1 // SPDX-License-Identifier: GPL-2.0
3 * Functions related to setting various queue properties from drivers
5 #include <linux/kernel.h>
6 #include <linux/module.h>
7 #include <linux/init.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>
21 void blk_queue_rq_timeout(struct request_queue *q, unsigned int timeout)
23 q->rq_timeout = timeout;
25 EXPORT_SYMBOL_GPL(blk_queue_rq_timeout);
28 * blk_set_default_limits - reset limits to default values
29 * @lim: the queue_limits structure to reset
32 * Returns a queue_limit struct to its default state.
34 void blk_set_default_limits(struct queue_limits *lim)
36 lim->max_segments = BLK_MAX_SEGMENTS;
37 lim->max_discard_segments = 1;
38 lim->max_integrity_segments = 0;
39 lim->seg_boundary_mask = BLK_SEG_BOUNDARY_MASK;
40 lim->virt_boundary_mask = 0;
41 lim->max_segment_size = BLK_MAX_SEGMENT_SIZE;
42 lim->max_sectors = lim->max_hw_sectors = BLK_SAFE_MAX_SECTORS;
43 lim->max_dev_sectors = 0;
44 lim->chunk_sectors = 0;
45 lim->max_write_zeroes_sectors = 0;
46 lim->max_zone_append_sectors = 0;
47 lim->max_discard_sectors = 0;
48 lim->max_hw_discard_sectors = 0;
49 lim->max_secure_erase_sectors = 0;
50 lim->discard_granularity = 0;
51 lim->discard_alignment = 0;
52 lim->discard_misaligned = 0;
53 lim->logical_block_size = lim->physical_block_size = lim->io_min = 512;
54 lim->bounce = BLK_BOUNCE_NONE;
55 lim->alignment_offset = 0;
58 lim->zoned = BLK_ZONED_NONE;
59 lim->zone_write_granularity = 0;
60 lim->dma_alignment = 511;
64 * blk_set_stacking_limits - set default limits for stacking devices
65 * @lim: the queue_limits structure to reset
68 * Returns a queue_limit struct to its default state. Should be used
69 * by stacking drivers like DM that have no internal limits.
71 void blk_set_stacking_limits(struct queue_limits *lim)
73 blk_set_default_limits(lim);
75 /* Inherit limits from component devices */
76 lim->max_segments = USHRT_MAX;
77 lim->max_discard_segments = USHRT_MAX;
78 lim->max_hw_sectors = UINT_MAX;
79 lim->max_segment_size = UINT_MAX;
80 lim->max_sectors = UINT_MAX;
81 lim->max_dev_sectors = UINT_MAX;
82 lim->max_write_zeroes_sectors = UINT_MAX;
83 lim->max_zone_append_sectors = UINT_MAX;
85 EXPORT_SYMBOL(blk_set_stacking_limits);
88 * blk_queue_bounce_limit - set bounce buffer limit for queue
89 * @q: the request queue for the device
90 * @bounce: bounce limit to enforce
93 * Force bouncing for ISA DMA ranges or highmem.
95 * DEPRECATED, don't use in new code.
97 void blk_queue_bounce_limit(struct request_queue *q, enum blk_bounce bounce)
99 q->limits.bounce = bounce;
101 EXPORT_SYMBOL(blk_queue_bounce_limit);
104 * blk_queue_max_hw_sectors - set max sectors for a request for this queue
105 * @q: the request queue for the device
106 * @max_hw_sectors: max hardware sectors in the usual 512b unit
109 * Enables a low level driver to set a hard upper limit,
110 * max_hw_sectors, on the size of requests. max_hw_sectors is set by
111 * the device driver based upon the capabilities of the I/O
114 * max_dev_sectors is a hard limit imposed by the storage device for
115 * READ/WRITE requests. It is set by the disk driver.
117 * max_sectors is a soft limit imposed by the block layer for
118 * filesystem type requests. This value can be overridden on a
119 * per-device basis in /sys/block/<device>/queue/max_sectors_kb.
120 * The soft limit can not exceed max_hw_sectors.
122 void blk_queue_max_hw_sectors(struct request_queue *q, unsigned int max_hw_sectors)
124 struct queue_limits *limits = &q->limits;
125 unsigned int max_sectors;
127 if ((max_hw_sectors << 9) < PAGE_SIZE) {
128 max_hw_sectors = 1 << (PAGE_SHIFT - 9);
129 printk(KERN_INFO "%s: set to minimum %d\n",
130 __func__, max_hw_sectors);
133 max_hw_sectors = round_down(max_hw_sectors,
134 limits->logical_block_size >> SECTOR_SHIFT);
135 limits->max_hw_sectors = max_hw_sectors;
137 max_sectors = min_not_zero(max_hw_sectors, limits->max_dev_sectors);
138 max_sectors = min_t(unsigned int, max_sectors, BLK_DEF_MAX_SECTORS);
139 max_sectors = round_down(max_sectors,
140 limits->logical_block_size >> SECTOR_SHIFT);
141 limits->max_sectors = max_sectors;
145 q->disk->bdi->io_pages = max_sectors >> (PAGE_SHIFT - 9);
147 EXPORT_SYMBOL(blk_queue_max_hw_sectors);
150 * blk_queue_chunk_sectors - set size of the chunk for this queue
151 * @q: the request queue for the device
152 * @chunk_sectors: chunk sectors in the usual 512b unit
155 * If a driver doesn't want IOs to cross a given chunk size, it can set
156 * this limit and prevent merging across chunks. Note that the block layer
157 * must accept a page worth of data at any offset. So if the crossing of
158 * chunks is a hard limitation in the driver, it must still be prepared
159 * to split single page bios.
161 void blk_queue_chunk_sectors(struct request_queue *q, unsigned int chunk_sectors)
163 q->limits.chunk_sectors = chunk_sectors;
165 EXPORT_SYMBOL(blk_queue_chunk_sectors);
168 * blk_queue_max_discard_sectors - set max sectors for a single discard
169 * @q: the request queue for the device
170 * @max_discard_sectors: maximum number of sectors to discard
172 void blk_queue_max_discard_sectors(struct request_queue *q,
173 unsigned int max_discard_sectors)
175 q->limits.max_hw_discard_sectors = max_discard_sectors;
176 q->limits.max_discard_sectors = max_discard_sectors;
178 EXPORT_SYMBOL(blk_queue_max_discard_sectors);
181 * blk_queue_max_secure_erase_sectors - set max sectors for a secure erase
182 * @q: the request queue for the device
183 * @max_sectors: maximum number of sectors to secure_erase
185 void blk_queue_max_secure_erase_sectors(struct request_queue *q,
186 unsigned int max_sectors)
188 q->limits.max_secure_erase_sectors = max_sectors;
190 EXPORT_SYMBOL(blk_queue_max_secure_erase_sectors);
193 * blk_queue_max_write_zeroes_sectors - set max sectors for a single
195 * @q: the request queue for the device
196 * @max_write_zeroes_sectors: maximum number of sectors to write per command
198 void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
199 unsigned int max_write_zeroes_sectors)
201 q->limits.max_write_zeroes_sectors = max_write_zeroes_sectors;
203 EXPORT_SYMBOL(blk_queue_max_write_zeroes_sectors);
206 * blk_queue_max_zone_append_sectors - set max sectors for a single zone append
207 * @q: the request queue for the device
208 * @max_zone_append_sectors: maximum number of sectors to write per command
210 void blk_queue_max_zone_append_sectors(struct request_queue *q,
211 unsigned int max_zone_append_sectors)
213 unsigned int max_sectors;
215 if (WARN_ON(!blk_queue_is_zoned(q)))
218 max_sectors = min(q->limits.max_hw_sectors, max_zone_append_sectors);
219 max_sectors = min(q->limits.chunk_sectors, max_sectors);
222 * Signal eventual driver bugs resulting in the max_zone_append sectors limit
223 * being 0 due to a 0 argument, the chunk_sectors limit (zone size) not set,
224 * or the max_hw_sectors limit not set.
226 WARN_ON(!max_sectors);
228 q->limits.max_zone_append_sectors = max_sectors;
230 EXPORT_SYMBOL_GPL(blk_queue_max_zone_append_sectors);
233 * blk_queue_max_segments - set max hw segments for a request for this queue
234 * @q: the request queue for the device
235 * @max_segments: max number of segments
238 * Enables a low level driver to set an upper limit on the number of
239 * hw data segments in a request.
241 void blk_queue_max_segments(struct request_queue *q, unsigned short max_segments)
245 printk(KERN_INFO "%s: set to minimum %d\n",
246 __func__, max_segments);
249 q->limits.max_segments = max_segments;
251 EXPORT_SYMBOL(blk_queue_max_segments);
254 * blk_queue_max_discard_segments - set max segments for discard requests
255 * @q: the request queue for the device
256 * @max_segments: max number of segments
259 * Enables a low level driver to set an upper limit on the number of
260 * segments in a discard request.
262 void blk_queue_max_discard_segments(struct request_queue *q,
263 unsigned short max_segments)
265 q->limits.max_discard_segments = max_segments;
267 EXPORT_SYMBOL_GPL(blk_queue_max_discard_segments);
270 * blk_queue_max_segment_size - set max segment size for blk_rq_map_sg
271 * @q: the request queue for the device
272 * @max_size: max size of segment in bytes
275 * Enables a low level driver to set an upper limit on the size of a
278 void blk_queue_max_segment_size(struct request_queue *q, unsigned int max_size)
280 if (max_size < PAGE_SIZE) {
281 max_size = PAGE_SIZE;
282 printk(KERN_INFO "%s: set to minimum %d\n",
286 /* see blk_queue_virt_boundary() for the explanation */
287 WARN_ON_ONCE(q->limits.virt_boundary_mask);
289 q->limits.max_segment_size = max_size;
291 EXPORT_SYMBOL(blk_queue_max_segment_size);
294 * blk_queue_logical_block_size - set logical block size for the queue
295 * @q: the request queue for the device
296 * @size: the logical block size, in bytes
299 * This should be set to the lowest possible block size that the
300 * storage device can address. The default of 512 covers most
303 void blk_queue_logical_block_size(struct request_queue *q, unsigned int size)
305 struct queue_limits *limits = &q->limits;
307 limits->logical_block_size = size;
309 if (limits->physical_block_size < size)
310 limits->physical_block_size = size;
312 if (limits->io_min < limits->physical_block_size)
313 limits->io_min = limits->physical_block_size;
315 limits->max_hw_sectors =
316 round_down(limits->max_hw_sectors, size >> SECTOR_SHIFT);
317 limits->max_sectors =
318 round_down(limits->max_sectors, size >> SECTOR_SHIFT);
320 EXPORT_SYMBOL(blk_queue_logical_block_size);
323 * blk_queue_physical_block_size - set physical block size for the queue
324 * @q: the request queue for the device
325 * @size: the physical block size, in bytes
328 * This should be set to the lowest possible sector size that the
329 * hardware can operate on without reverting to read-modify-write
332 void blk_queue_physical_block_size(struct request_queue *q, unsigned int size)
334 q->limits.physical_block_size = size;
336 if (q->limits.physical_block_size < q->limits.logical_block_size)
337 q->limits.physical_block_size = q->limits.logical_block_size;
339 if (q->limits.io_min < q->limits.physical_block_size)
340 q->limits.io_min = q->limits.physical_block_size;
342 EXPORT_SYMBOL(blk_queue_physical_block_size);
345 * blk_queue_zone_write_granularity - set zone write granularity for the queue
346 * @q: the request queue for the zoned device
347 * @size: the zone write granularity size, in bytes
350 * This should be set to the lowest possible size allowing to write in
351 * sequential zones of a zoned block device.
353 void blk_queue_zone_write_granularity(struct request_queue *q,
356 if (WARN_ON_ONCE(!blk_queue_is_zoned(q)))
359 q->limits.zone_write_granularity = size;
361 if (q->limits.zone_write_granularity < q->limits.logical_block_size)
362 q->limits.zone_write_granularity = q->limits.logical_block_size;
364 EXPORT_SYMBOL_GPL(blk_queue_zone_write_granularity);
367 * blk_queue_alignment_offset - set physical block alignment offset
368 * @q: the request queue for the device
369 * @offset: alignment offset in bytes
372 * Some devices are naturally misaligned to compensate for things like
373 * the legacy DOS partition table 63-sector offset. Low-level drivers
374 * should call this function for devices whose first sector is not
377 void blk_queue_alignment_offset(struct request_queue *q, unsigned int offset)
379 q->limits.alignment_offset =
380 offset & (q->limits.physical_block_size - 1);
381 q->limits.misaligned = 0;
383 EXPORT_SYMBOL(blk_queue_alignment_offset);
385 void disk_update_readahead(struct gendisk *disk)
387 struct request_queue *q = disk->queue;
390 * For read-ahead of large files to be effective, we need to read ahead
391 * at least twice the optimal I/O size.
393 disk->bdi->ra_pages =
394 max(queue_io_opt(q) * 2 / PAGE_SIZE, VM_READAHEAD_PAGES);
395 disk->bdi->io_pages = queue_max_sectors(q) >> (PAGE_SHIFT - 9);
397 EXPORT_SYMBOL_GPL(disk_update_readahead);
400 * blk_limits_io_min - set minimum request size for a device
401 * @limits: the queue limits
402 * @min: smallest I/O size in bytes
405 * Some devices have an internal block size bigger than the reported
406 * hardware sector size. This function can be used to signal the
407 * smallest I/O the device can perform without incurring a performance
410 void blk_limits_io_min(struct queue_limits *limits, unsigned int min)
412 limits->io_min = min;
414 if (limits->io_min < limits->logical_block_size)
415 limits->io_min = limits->logical_block_size;
417 if (limits->io_min < limits->physical_block_size)
418 limits->io_min = limits->physical_block_size;
420 EXPORT_SYMBOL(blk_limits_io_min);
423 * blk_queue_io_min - set minimum request size for the queue
424 * @q: the request queue for the device
425 * @min: smallest I/O size in bytes
428 * Storage devices may report a granularity or preferred minimum I/O
429 * size which is the smallest request the device can perform without
430 * incurring a performance penalty. For disk drives this is often the
431 * physical block size. For RAID arrays it is often the stripe chunk
432 * size. A properly aligned multiple of minimum_io_size is the
433 * preferred request size for workloads where a high number of I/O
434 * operations is desired.
436 void blk_queue_io_min(struct request_queue *q, unsigned int min)
438 blk_limits_io_min(&q->limits, min);
440 EXPORT_SYMBOL(blk_queue_io_min);
443 * blk_limits_io_opt - set optimal request size for a device
444 * @limits: the queue limits
445 * @opt: smallest I/O size in bytes
448 * Storage devices may report an optimal I/O size, which is the
449 * device's preferred unit for sustained I/O. This is rarely reported
450 * for disk drives. For RAID arrays it is usually the stripe width or
451 * the internal track size. A properly aligned multiple of
452 * optimal_io_size is the preferred request size for workloads where
453 * sustained throughput is desired.
455 void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt)
457 limits->io_opt = opt;
459 EXPORT_SYMBOL(blk_limits_io_opt);
462 * blk_queue_io_opt - set optimal request size for the queue
463 * @q: the request queue for the device
464 * @opt: optimal request size in bytes
467 * Storage devices may report an optimal I/O size, which is the
468 * device's preferred unit for sustained I/O. This is rarely reported
469 * for disk drives. For RAID arrays it is usually the stripe width or
470 * the internal track size. A properly aligned multiple of
471 * optimal_io_size is the preferred request size for workloads where
472 * sustained throughput is desired.
474 void blk_queue_io_opt(struct request_queue *q, unsigned int opt)
476 blk_limits_io_opt(&q->limits, opt);
479 q->disk->bdi->ra_pages =
480 max(queue_io_opt(q) * 2 / PAGE_SIZE, VM_READAHEAD_PAGES);
482 EXPORT_SYMBOL(blk_queue_io_opt);
484 static int queue_limit_alignment_offset(const struct queue_limits *lim,
487 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
488 unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT)
491 return (granularity + lim->alignment_offset - alignment) % granularity;
494 static unsigned int queue_limit_discard_alignment(
495 const struct queue_limits *lim, sector_t sector)
497 unsigned int alignment, granularity, offset;
499 if (!lim->max_discard_sectors)
502 /* Why are these in bytes, not sectors? */
503 alignment = lim->discard_alignment >> SECTOR_SHIFT;
504 granularity = lim->discard_granularity >> SECTOR_SHIFT;
508 /* Offset of the partition start in 'granularity' sectors */
509 offset = sector_div(sector, granularity);
511 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
512 offset = (granularity + alignment - offset) % granularity;
514 /* Turn it back into bytes, gaah */
515 return offset << SECTOR_SHIFT;
518 static unsigned int blk_round_down_sectors(unsigned int sectors, unsigned int lbs)
520 sectors = round_down(sectors, lbs >> SECTOR_SHIFT);
521 if (sectors < PAGE_SIZE >> SECTOR_SHIFT)
522 sectors = PAGE_SIZE >> SECTOR_SHIFT;
527 * blk_stack_limits - adjust queue_limits for stacked devices
528 * @t: the stacking driver limits (top device)
529 * @b: the underlying queue limits (bottom, component device)
530 * @start: first data sector within component device
533 * This function is used by stacking drivers like MD and DM to ensure
534 * that all component devices have compatible block sizes and
535 * alignments. The stacking driver must provide a queue_limits
536 * struct (top) and then iteratively call the stacking function for
537 * all component (bottom) devices. The stacking function will
538 * attempt to combine the values and ensure proper alignment.
540 * Returns 0 if the top and bottom queue_limits are compatible. The
541 * top device's block sizes and alignment offsets may be adjusted to
542 * ensure alignment with the bottom device. If no compatible sizes
543 * and alignments exist, -1 is returned and the resulting top
544 * queue_limits will have the misaligned flag set to indicate that
545 * the alignment_offset is undefined.
547 int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
550 unsigned int top, bottom, alignment, ret = 0;
552 t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors);
553 t->max_hw_sectors = min_not_zero(t->max_hw_sectors, b->max_hw_sectors);
554 t->max_dev_sectors = min_not_zero(t->max_dev_sectors, b->max_dev_sectors);
555 t->max_write_zeroes_sectors = min(t->max_write_zeroes_sectors,
556 b->max_write_zeroes_sectors);
557 t->max_zone_append_sectors = min(t->max_zone_append_sectors,
558 b->max_zone_append_sectors);
559 t->bounce = max(t->bounce, b->bounce);
561 t->seg_boundary_mask = min_not_zero(t->seg_boundary_mask,
562 b->seg_boundary_mask);
563 t->virt_boundary_mask = min_not_zero(t->virt_boundary_mask,
564 b->virt_boundary_mask);
566 t->max_segments = min_not_zero(t->max_segments, b->max_segments);
567 t->max_discard_segments = min_not_zero(t->max_discard_segments,
568 b->max_discard_segments);
569 t->max_integrity_segments = min_not_zero(t->max_integrity_segments,
570 b->max_integrity_segments);
572 t->max_segment_size = min_not_zero(t->max_segment_size,
573 b->max_segment_size);
575 t->misaligned |= b->misaligned;
577 alignment = queue_limit_alignment_offset(b, start);
579 /* Bottom device has different alignment. Check that it is
580 * compatible with the current top alignment.
582 if (t->alignment_offset != alignment) {
584 top = max(t->physical_block_size, t->io_min)
585 + t->alignment_offset;
586 bottom = max(b->physical_block_size, b->io_min) + alignment;
588 /* Verify that top and bottom intervals line up */
589 if (max(top, bottom) % min(top, bottom)) {
595 t->logical_block_size = max(t->logical_block_size,
596 b->logical_block_size);
598 t->physical_block_size = max(t->physical_block_size,
599 b->physical_block_size);
601 t->io_min = max(t->io_min, b->io_min);
602 t->io_opt = lcm_not_zero(t->io_opt, b->io_opt);
603 t->dma_alignment = max(t->dma_alignment, b->dma_alignment);
605 /* Set non-power-of-2 compatible chunk_sectors boundary */
606 if (b->chunk_sectors)
607 t->chunk_sectors = gcd(t->chunk_sectors, b->chunk_sectors);
609 /* Physical block size a multiple of the logical block size? */
610 if (t->physical_block_size & (t->logical_block_size - 1)) {
611 t->physical_block_size = t->logical_block_size;
616 /* Minimum I/O a multiple of the physical block size? */
617 if (t->io_min & (t->physical_block_size - 1)) {
618 t->io_min = t->physical_block_size;
623 /* Optimal I/O a multiple of the physical block size? */
624 if (t->io_opt & (t->physical_block_size - 1)) {
630 /* chunk_sectors a multiple of the physical block size? */
631 if ((t->chunk_sectors << 9) & (t->physical_block_size - 1)) {
632 t->chunk_sectors = 0;
637 t->raid_partial_stripes_expensive =
638 max(t->raid_partial_stripes_expensive,
639 b->raid_partial_stripes_expensive);
641 /* Find lowest common alignment_offset */
642 t->alignment_offset = lcm_not_zero(t->alignment_offset, alignment)
643 % max(t->physical_block_size, t->io_min);
645 /* Verify that new alignment_offset is on a logical block boundary */
646 if (t->alignment_offset & (t->logical_block_size - 1)) {
651 t->max_sectors = blk_round_down_sectors(t->max_sectors, t->logical_block_size);
652 t->max_hw_sectors = blk_round_down_sectors(t->max_hw_sectors, t->logical_block_size);
653 t->max_dev_sectors = blk_round_down_sectors(t->max_dev_sectors, t->logical_block_size);
655 /* Discard alignment and granularity */
656 if (b->discard_granularity) {
657 alignment = queue_limit_discard_alignment(b, start);
659 if (t->discard_granularity != 0 &&
660 t->discard_alignment != alignment) {
661 top = t->discard_granularity + t->discard_alignment;
662 bottom = b->discard_granularity + alignment;
664 /* Verify that top and bottom intervals line up */
665 if ((max(top, bottom) % min(top, bottom)) != 0)
666 t->discard_misaligned = 1;
669 t->max_discard_sectors = min_not_zero(t->max_discard_sectors,
670 b->max_discard_sectors);
671 t->max_hw_discard_sectors = min_not_zero(t->max_hw_discard_sectors,
672 b->max_hw_discard_sectors);
673 t->discard_granularity = max(t->discard_granularity,
674 b->discard_granularity);
675 t->discard_alignment = lcm_not_zero(t->discard_alignment, alignment) %
676 t->discard_granularity;
678 t->max_secure_erase_sectors = min_not_zero(t->max_secure_erase_sectors,
679 b->max_secure_erase_sectors);
680 t->zone_write_granularity = max(t->zone_write_granularity,
681 b->zone_write_granularity);
682 t->zoned = max(t->zoned, b->zoned);
685 EXPORT_SYMBOL(blk_stack_limits);
688 * disk_stack_limits - adjust queue limits for stacked drivers
689 * @disk: MD/DM gendisk (top)
690 * @bdev: the underlying block device (bottom)
691 * @offset: offset to beginning of data within component device
694 * Merges the limits for a top level gendisk and a bottom level
697 void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
700 struct request_queue *t = disk->queue;
702 if (blk_stack_limits(&t->limits, &bdev_get_queue(bdev)->limits,
703 get_start_sect(bdev) + (offset >> 9)) < 0)
704 pr_notice("%s: Warning: Device %pg is misaligned\n",
705 disk->disk_name, bdev);
707 disk_update_readahead(disk);
709 EXPORT_SYMBOL(disk_stack_limits);
712 * blk_queue_update_dma_pad - update pad mask
713 * @q: the request queue for the device
716 * Update dma pad mask.
718 * Appending pad buffer to a request modifies the last entry of a
719 * scatter list such that it includes the pad buffer.
721 void blk_queue_update_dma_pad(struct request_queue *q, unsigned int mask)
723 if (mask > q->dma_pad_mask)
724 q->dma_pad_mask = mask;
726 EXPORT_SYMBOL(blk_queue_update_dma_pad);
729 * blk_queue_segment_boundary - set boundary rules for segment merging
730 * @q: the request queue for the device
731 * @mask: the memory boundary mask
733 void blk_queue_segment_boundary(struct request_queue *q, unsigned long mask)
735 if (mask < PAGE_SIZE - 1) {
736 mask = PAGE_SIZE - 1;
737 printk(KERN_INFO "%s: set to minimum %lx\n",
741 q->limits.seg_boundary_mask = mask;
743 EXPORT_SYMBOL(blk_queue_segment_boundary);
746 * blk_queue_virt_boundary - set boundary rules for bio merging
747 * @q: the request queue for the device
748 * @mask: the memory boundary mask
750 void blk_queue_virt_boundary(struct request_queue *q, unsigned long mask)
752 q->limits.virt_boundary_mask = mask;
755 * Devices that require a virtual boundary do not support scatter/gather
756 * I/O natively, but instead require a descriptor list entry for each
757 * page (which might not be idential to the Linux PAGE_SIZE). Because
758 * of that they are not limited by our notion of "segment size".
761 q->limits.max_segment_size = UINT_MAX;
763 EXPORT_SYMBOL(blk_queue_virt_boundary);
766 * blk_queue_dma_alignment - set dma length and memory alignment
767 * @q: the request queue for the device
768 * @mask: alignment mask
771 * set required memory and length alignment for direct dma transactions.
772 * this is used when building direct io requests for the queue.
775 void blk_queue_dma_alignment(struct request_queue *q, int mask)
777 q->limits.dma_alignment = mask;
779 EXPORT_SYMBOL(blk_queue_dma_alignment);
782 * blk_queue_update_dma_alignment - update dma length and memory alignment
783 * @q: the request queue for the device
784 * @mask: alignment mask
787 * update required memory and length alignment for direct dma transactions.
788 * If the requested alignment is larger than the current alignment, then
789 * the current queue alignment is updated to the new value, otherwise it
790 * is left alone. The design of this is to allow multiple objects
791 * (driver, device, transport etc) to set their respective
792 * alignments without having them interfere.
795 void blk_queue_update_dma_alignment(struct request_queue *q, int mask)
797 BUG_ON(mask > PAGE_SIZE);
799 if (mask > q->limits.dma_alignment)
800 q->limits.dma_alignment = mask;
802 EXPORT_SYMBOL(blk_queue_update_dma_alignment);
805 * blk_set_queue_depth - tell the block layer about the device queue depth
806 * @q: the request queue for the device
807 * @depth: queue depth
810 void blk_set_queue_depth(struct request_queue *q, unsigned int depth)
812 q->queue_depth = depth;
813 rq_qos_queue_depth_changed(q);
815 EXPORT_SYMBOL(blk_set_queue_depth);
818 * blk_queue_write_cache - configure queue's write cache
819 * @q: the request queue for the device
820 * @wc: write back cache on or off
821 * @fua: device supports FUA writes, if true
823 * Tell the block layer about the write cache of @q.
825 void blk_queue_write_cache(struct request_queue *q, bool wc, bool fua)
828 blk_queue_flag_set(QUEUE_FLAG_WC, q);
830 blk_queue_flag_clear(QUEUE_FLAG_WC, q);
832 blk_queue_flag_set(QUEUE_FLAG_FUA, q);
834 blk_queue_flag_clear(QUEUE_FLAG_FUA, q);
836 wbt_set_write_cache(q, test_bit(QUEUE_FLAG_WC, &q->queue_flags));
838 EXPORT_SYMBOL_GPL(blk_queue_write_cache);
841 * blk_queue_required_elevator_features - Set a queue required elevator features
842 * @q: the request queue for the target device
843 * @features: Required elevator features OR'ed together
845 * Tell the block layer that for the device controlled through @q, only the
846 * only elevators that can be used are those that implement at least the set of
847 * features specified by @features.
849 void blk_queue_required_elevator_features(struct request_queue *q,
850 unsigned int features)
852 q->required_elevator_features = features;
854 EXPORT_SYMBOL_GPL(blk_queue_required_elevator_features);
857 * blk_queue_can_use_dma_map_merging - configure queue for merging segments.
858 * @q: the request queue for the device
859 * @dev: the device pointer for dma
861 * Tell the block layer about merging the segments by dma map of @q.
863 bool blk_queue_can_use_dma_map_merging(struct request_queue *q,
866 unsigned long boundary = dma_get_merge_boundary(dev);
871 /* No need to update max_segment_size. see blk_queue_virt_boundary() */
872 blk_queue_virt_boundary(q, boundary);
876 EXPORT_SYMBOL_GPL(blk_queue_can_use_dma_map_merging);
878 static bool disk_has_partitions(struct gendisk *disk)
881 struct block_device *part;
885 xa_for_each(&disk->part_tbl, idx, part) {
886 if (bdev_is_partition(part)) {
897 * disk_set_zoned - configure the zoned model for a disk
898 * @disk: the gendisk of the queue to configure
899 * @model: the zoned model to set
901 * Set the zoned model of @disk to @model.
903 * When @model is BLK_ZONED_HM (host managed), this should be called only
904 * if zoned block device support is enabled (CONFIG_BLK_DEV_ZONED option).
905 * If @model specifies BLK_ZONED_HA (host aware), the effective model used
906 * depends on CONFIG_BLK_DEV_ZONED settings and on the existence of partitions
909 void disk_set_zoned(struct gendisk *disk, enum blk_zoned_model model)
911 struct request_queue *q = disk->queue;
916 * Host managed devices are supported only if
917 * CONFIG_BLK_DEV_ZONED is enabled.
919 WARN_ON_ONCE(!IS_ENABLED(CONFIG_BLK_DEV_ZONED));
923 * Host aware devices can be treated either as regular block
924 * devices (similar to drive managed devices) or as zoned block
925 * devices to take advantage of the zone command set, similarly
926 * to host managed devices. We try the latter if there are no
927 * partitions and zoned block device support is enabled, else
928 * we do nothing special as far as the block layer is concerned.
930 if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) ||
931 disk_has_partitions(disk))
932 model = BLK_ZONED_NONE;
936 if (WARN_ON_ONCE(model != BLK_ZONED_NONE))
937 model = BLK_ZONED_NONE;
941 q->limits.zoned = model;
942 if (model != BLK_ZONED_NONE) {
944 * Set the zone write granularity to the device logical block
945 * size by default. The driver can change this value if needed.
947 blk_queue_zone_write_granularity(q,
948 queue_logical_block_size(q));
950 disk_clear_zone_settings(disk);
953 EXPORT_SYMBOL_GPL(disk_set_zoned);
955 int bdev_alignment_offset(struct block_device *bdev)
957 struct request_queue *q = bdev_get_queue(bdev);
959 if (q->limits.misaligned)
961 if (bdev_is_partition(bdev))
962 return queue_limit_alignment_offset(&q->limits,
963 bdev->bd_start_sect);
964 return q->limits.alignment_offset;
966 EXPORT_SYMBOL_GPL(bdev_alignment_offset);
968 unsigned int bdev_discard_alignment(struct block_device *bdev)
970 struct request_queue *q = bdev_get_queue(bdev);
972 if (bdev_is_partition(bdev))
973 return queue_limit_discard_alignment(&q->limits,
974 bdev->bd_start_sect);
975 return q->limits.discard_alignment;
977 EXPORT_SYMBOL_GPL(bdev_discard_alignment);