xref: /linux/block/blk-settings.c (revision aaa44952bbd1d4db14a4d676bf9595bb5db7e7b0)
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/gcd.h>
11 #include <linux/lcm.h>
12 #include <linux/jiffies.h>
13 #include <linux/gfp.h>
14 #include <linux/dma-mapping.h>
15 
16 #include "blk.h"
17 #include "blk-wbt.h"
18 
19 void blk_queue_rq_timeout(struct request_queue *q, unsigned int timeout)
20 {
21 	q->rq_timeout = timeout;
22 }
23 EXPORT_SYMBOL_GPL(blk_queue_rq_timeout);
24 
25 /**
26  * blk_set_default_limits - reset limits to default values
27  * @lim:  the queue_limits structure to reset
28  *
29  * Description:
30  *   Returns a queue_limit struct to its default state.
31  */
32 void blk_set_default_limits(struct queue_limits *lim)
33 {
34 	lim->max_segments = BLK_MAX_SEGMENTS;
35 	lim->max_discard_segments = 1;
36 	lim->max_integrity_segments = 0;
37 	lim->seg_boundary_mask = BLK_SEG_BOUNDARY_MASK;
38 	lim->virt_boundary_mask = 0;
39 	lim->max_segment_size = BLK_MAX_SEGMENT_SIZE;
40 	lim->max_sectors = lim->max_hw_sectors = BLK_SAFE_MAX_SECTORS;
41 	lim->max_dev_sectors = 0;
42 	lim->chunk_sectors = 0;
43 	lim->max_write_same_sectors = 0;
44 	lim->max_write_zeroes_sectors = 0;
45 	lim->max_zone_append_sectors = 0;
46 	lim->max_discard_sectors = 0;
47 	lim->max_hw_discard_sectors = 0;
48 	lim->discard_granularity = 0;
49 	lim->discard_alignment = 0;
50 	lim->discard_misaligned = 0;
51 	lim->logical_block_size = lim->physical_block_size = lim->io_min = 512;
52 	lim->bounce = BLK_BOUNCE_NONE;
53 	lim->alignment_offset = 0;
54 	lim->io_opt = 0;
55 	lim->misaligned = 0;
56 	lim->zoned = BLK_ZONED_NONE;
57 	lim->zone_write_granularity = 0;
58 }
59 EXPORT_SYMBOL(blk_set_default_limits);
60 
61 /**
62  * blk_set_stacking_limits - set default limits for stacking devices
63  * @lim:  the queue_limits structure to reset
64  *
65  * Description:
66  *   Returns a queue_limit struct to its default state. Should be used
67  *   by stacking drivers like DM that have no internal limits.
68  */
69 void blk_set_stacking_limits(struct queue_limits *lim)
70 {
71 	blk_set_default_limits(lim);
72 
73 	/* Inherit limits from component devices */
74 	lim->max_segments = USHRT_MAX;
75 	lim->max_discard_segments = USHRT_MAX;
76 	lim->max_hw_sectors = UINT_MAX;
77 	lim->max_segment_size = UINT_MAX;
78 	lim->max_sectors = UINT_MAX;
79 	lim->max_dev_sectors = UINT_MAX;
80 	lim->max_write_same_sectors = UINT_MAX;
81 	lim->max_write_zeroes_sectors = UINT_MAX;
82 	lim->max_zone_append_sectors = UINT_MAX;
83 }
84 EXPORT_SYMBOL(blk_set_stacking_limits);
85 
86 /**
87  * blk_queue_bounce_limit - set bounce buffer limit for queue
88  * @q: the request queue for the device
89  * @bounce: bounce limit to enforce
90  *
91  * Description:
92  *    Force bouncing for ISA DMA ranges or highmem.
93  *
94  *    DEPRECATED, don't use in new code.
95  **/
96 void blk_queue_bounce_limit(struct request_queue *q, enum blk_bounce bounce)
97 {
98 	q->limits.bounce = bounce;
99 }
100 EXPORT_SYMBOL(blk_queue_bounce_limit);
101 
102 /**
103  * blk_queue_max_hw_sectors - set max sectors for a request for this queue
104  * @q:  the request queue for the device
105  * @max_hw_sectors:  max hardware sectors in the usual 512b unit
106  *
107  * Description:
108  *    Enables a low level driver to set a hard upper limit,
109  *    max_hw_sectors, on the size of requests.  max_hw_sectors is set by
110  *    the device driver based upon the capabilities of the I/O
111  *    controller.
112  *
113  *    max_dev_sectors is a hard limit imposed by the storage device for
114  *    READ/WRITE requests. It is set by the disk driver.
115  *
116  *    max_sectors is a soft limit imposed by the block layer for
117  *    filesystem type requests.  This value can be overridden on a
118  *    per-device basis in /sys/block/<device>/queue/max_sectors_kb.
119  *    The soft limit can not exceed max_hw_sectors.
120  **/
121 void blk_queue_max_hw_sectors(struct request_queue *q, unsigned int max_hw_sectors)
122 {
123 	struct queue_limits *limits = &q->limits;
124 	unsigned int max_sectors;
125 
126 	if ((max_hw_sectors << 9) < PAGE_SIZE) {
127 		max_hw_sectors = 1 << (PAGE_SHIFT - 9);
128 		printk(KERN_INFO "%s: set to minimum %d\n",
129 		       __func__, max_hw_sectors);
130 	}
131 
132 	max_hw_sectors = round_down(max_hw_sectors,
133 				    limits->logical_block_size >> SECTOR_SHIFT);
134 	limits->max_hw_sectors = max_hw_sectors;
135 
136 	max_sectors = min_not_zero(max_hw_sectors, limits->max_dev_sectors);
137 	max_sectors = min_t(unsigned int, max_sectors, BLK_DEF_MAX_SECTORS);
138 	max_sectors = round_down(max_sectors,
139 				 limits->logical_block_size >> SECTOR_SHIFT);
140 	limits->max_sectors = max_sectors;
141 
142 	q->backing_dev_info->io_pages = max_sectors >> (PAGE_SHIFT - 9);
143 }
144 EXPORT_SYMBOL(blk_queue_max_hw_sectors);
145 
146 /**
147  * blk_queue_chunk_sectors - set size of the chunk for this queue
148  * @q:  the request queue for the device
149  * @chunk_sectors:  chunk sectors in the usual 512b unit
150  *
151  * Description:
152  *    If a driver doesn't want IOs to cross a given chunk size, it can set
153  *    this limit and prevent merging across chunks. Note that the block layer
154  *    must accept a page worth of data at any offset. So if the crossing of
155  *    chunks is a hard limitation in the driver, it must still be prepared
156  *    to split single page bios.
157  **/
158 void blk_queue_chunk_sectors(struct request_queue *q, unsigned int chunk_sectors)
159 {
160 	q->limits.chunk_sectors = chunk_sectors;
161 }
162 EXPORT_SYMBOL(blk_queue_chunk_sectors);
163 
164 /**
165  * blk_queue_max_discard_sectors - set max sectors for a single discard
166  * @q:  the request queue for the device
167  * @max_discard_sectors: maximum number of sectors to discard
168  **/
169 void blk_queue_max_discard_sectors(struct request_queue *q,
170 		unsigned int max_discard_sectors)
171 {
172 	q->limits.max_hw_discard_sectors = max_discard_sectors;
173 	q->limits.max_discard_sectors = max_discard_sectors;
174 }
175 EXPORT_SYMBOL(blk_queue_max_discard_sectors);
176 
177 /**
178  * blk_queue_max_write_same_sectors - set max sectors for a single write same
179  * @q:  the request queue for the device
180  * @max_write_same_sectors: maximum number of sectors to write per command
181  **/
182 void blk_queue_max_write_same_sectors(struct request_queue *q,
183 				      unsigned int max_write_same_sectors)
184 {
185 	q->limits.max_write_same_sectors = max_write_same_sectors;
186 }
187 EXPORT_SYMBOL(blk_queue_max_write_same_sectors);
188 
189 /**
190  * blk_queue_max_write_zeroes_sectors - set max sectors for a single
191  *                                      write zeroes
192  * @q:  the request queue for the device
193  * @max_write_zeroes_sectors: maximum number of sectors to write per command
194  **/
195 void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
196 		unsigned int max_write_zeroes_sectors)
197 {
198 	q->limits.max_write_zeroes_sectors = max_write_zeroes_sectors;
199 }
200 EXPORT_SYMBOL(blk_queue_max_write_zeroes_sectors);
201 
202 /**
203  * blk_queue_max_zone_append_sectors - set max sectors for a single zone append
204  * @q:  the request queue for the device
205  * @max_zone_append_sectors: maximum number of sectors to write per command
206  **/
207 void blk_queue_max_zone_append_sectors(struct request_queue *q,
208 		unsigned int max_zone_append_sectors)
209 {
210 	unsigned int max_sectors;
211 
212 	if (WARN_ON(!blk_queue_is_zoned(q)))
213 		return;
214 
215 	max_sectors = min(q->limits.max_hw_sectors, max_zone_append_sectors);
216 	max_sectors = min(q->limits.chunk_sectors, max_sectors);
217 
218 	/*
219 	 * Signal eventual driver bugs resulting in the max_zone_append sectors limit
220 	 * being 0 due to a 0 argument, the chunk_sectors limit (zone size) not set,
221 	 * or the max_hw_sectors limit not set.
222 	 */
223 	WARN_ON(!max_sectors);
224 
225 	q->limits.max_zone_append_sectors = max_sectors;
226 }
227 EXPORT_SYMBOL_GPL(blk_queue_max_zone_append_sectors);
228 
229 /**
230  * blk_queue_max_segments - set max hw segments for a request for this queue
231  * @q:  the request queue for the device
232  * @max_segments:  max number of segments
233  *
234  * Description:
235  *    Enables a low level driver to set an upper limit on the number of
236  *    hw data segments in a request.
237  **/
238 void blk_queue_max_segments(struct request_queue *q, unsigned short max_segments)
239 {
240 	if (!max_segments) {
241 		max_segments = 1;
242 		printk(KERN_INFO "%s: set to minimum %d\n",
243 		       __func__, max_segments);
244 	}
245 
246 	q->limits.max_segments = max_segments;
247 }
248 EXPORT_SYMBOL(blk_queue_max_segments);
249 
250 /**
251  * blk_queue_max_discard_segments - set max segments for discard requests
252  * @q:  the request queue for the device
253  * @max_segments:  max number of segments
254  *
255  * Description:
256  *    Enables a low level driver to set an upper limit on the number of
257  *    segments in a discard request.
258  **/
259 void blk_queue_max_discard_segments(struct request_queue *q,
260 		unsigned short max_segments)
261 {
262 	q->limits.max_discard_segments = max_segments;
263 }
264 EXPORT_SYMBOL_GPL(blk_queue_max_discard_segments);
265 
266 /**
267  * blk_queue_max_segment_size - set max segment size for blk_rq_map_sg
268  * @q:  the request queue for the device
269  * @max_size:  max size of segment in bytes
270  *
271  * Description:
272  *    Enables a low level driver to set an upper limit on the size of a
273  *    coalesced segment
274  **/
275 void blk_queue_max_segment_size(struct request_queue *q, unsigned int max_size)
276 {
277 	if (max_size < PAGE_SIZE) {
278 		max_size = PAGE_SIZE;
279 		printk(KERN_INFO "%s: set to minimum %d\n",
280 		       __func__, max_size);
281 	}
282 
283 	/* see blk_queue_virt_boundary() for the explanation */
284 	WARN_ON_ONCE(q->limits.virt_boundary_mask);
285 
286 	q->limits.max_segment_size = max_size;
287 }
288 EXPORT_SYMBOL(blk_queue_max_segment_size);
289 
290 /**
291  * blk_queue_logical_block_size - set logical block size for the queue
292  * @q:  the request queue for the device
293  * @size:  the logical block size, in bytes
294  *
295  * Description:
296  *   This should be set to the lowest possible block size that the
297  *   storage device can address.  The default of 512 covers most
298  *   hardware.
299  **/
300 void blk_queue_logical_block_size(struct request_queue *q, unsigned int size)
301 {
302 	struct queue_limits *limits = &q->limits;
303 
304 	limits->logical_block_size = size;
305 
306 	if (limits->physical_block_size < size)
307 		limits->physical_block_size = size;
308 
309 	if (limits->io_min < limits->physical_block_size)
310 		limits->io_min = limits->physical_block_size;
311 
312 	limits->max_hw_sectors =
313 		round_down(limits->max_hw_sectors, size >> SECTOR_SHIFT);
314 	limits->max_sectors =
315 		round_down(limits->max_sectors, size >> SECTOR_SHIFT);
316 }
317 EXPORT_SYMBOL(blk_queue_logical_block_size);
318 
319 /**
320  * blk_queue_physical_block_size - set physical block size for the queue
321  * @q:  the request queue for the device
322  * @size:  the physical block size, in bytes
323  *
324  * Description:
325  *   This should be set to the lowest possible sector size that the
326  *   hardware can operate on without reverting to read-modify-write
327  *   operations.
328  */
329 void blk_queue_physical_block_size(struct request_queue *q, unsigned int size)
330 {
331 	q->limits.physical_block_size = size;
332 
333 	if (q->limits.physical_block_size < q->limits.logical_block_size)
334 		q->limits.physical_block_size = q->limits.logical_block_size;
335 
336 	if (q->limits.io_min < q->limits.physical_block_size)
337 		q->limits.io_min = q->limits.physical_block_size;
338 }
339 EXPORT_SYMBOL(blk_queue_physical_block_size);
340 
341 /**
342  * blk_queue_zone_write_granularity - set zone write granularity for the queue
343  * @q:  the request queue for the zoned device
344  * @size:  the zone write granularity size, in bytes
345  *
346  * Description:
347  *   This should be set to the lowest possible size allowing to write in
348  *   sequential zones of a zoned block device.
349  */
350 void blk_queue_zone_write_granularity(struct request_queue *q,
351 				      unsigned int size)
352 {
353 	if (WARN_ON_ONCE(!blk_queue_is_zoned(q)))
354 		return;
355 
356 	q->limits.zone_write_granularity = size;
357 
358 	if (q->limits.zone_write_granularity < q->limits.logical_block_size)
359 		q->limits.zone_write_granularity = q->limits.logical_block_size;
360 }
361 EXPORT_SYMBOL_GPL(blk_queue_zone_write_granularity);
362 
363 /**
364  * blk_queue_alignment_offset - set physical block alignment offset
365  * @q:	the request queue for the device
366  * @offset: alignment offset in bytes
367  *
368  * Description:
369  *   Some devices are naturally misaligned to compensate for things like
370  *   the legacy DOS partition table 63-sector offset.  Low-level drivers
371  *   should call this function for devices whose first sector is not
372  *   naturally aligned.
373  */
374 void blk_queue_alignment_offset(struct request_queue *q, unsigned int offset)
375 {
376 	q->limits.alignment_offset =
377 		offset & (q->limits.physical_block_size - 1);
378 	q->limits.misaligned = 0;
379 }
380 EXPORT_SYMBOL(blk_queue_alignment_offset);
381 
382 void blk_queue_update_readahead(struct request_queue *q)
383 {
384 	/*
385 	 * For read-ahead of large files to be effective, we need to read ahead
386 	 * at least twice the optimal I/O size.
387 	 */
388 	q->backing_dev_info->ra_pages =
389 		max(queue_io_opt(q) * 2 / PAGE_SIZE, VM_READAHEAD_PAGES);
390 	q->backing_dev_info->io_pages =
391 		queue_max_sectors(q) >> (PAGE_SHIFT - 9);
392 }
393 EXPORT_SYMBOL_GPL(blk_queue_update_readahead);
394 
395 /**
396  * blk_limits_io_min - set minimum request size for a device
397  * @limits: the queue limits
398  * @min:  smallest I/O size in bytes
399  *
400  * Description:
401  *   Some devices have an internal block size bigger than the reported
402  *   hardware sector size.  This function can be used to signal the
403  *   smallest I/O the device can perform without incurring a performance
404  *   penalty.
405  */
406 void blk_limits_io_min(struct queue_limits *limits, unsigned int min)
407 {
408 	limits->io_min = min;
409 
410 	if (limits->io_min < limits->logical_block_size)
411 		limits->io_min = limits->logical_block_size;
412 
413 	if (limits->io_min < limits->physical_block_size)
414 		limits->io_min = limits->physical_block_size;
415 }
416 EXPORT_SYMBOL(blk_limits_io_min);
417 
418 /**
419  * blk_queue_io_min - set minimum request size for the queue
420  * @q:	the request queue for the device
421  * @min:  smallest I/O size in bytes
422  *
423  * Description:
424  *   Storage devices may report a granularity or preferred minimum I/O
425  *   size which is the smallest request the device can perform without
426  *   incurring a performance penalty.  For disk drives this is often the
427  *   physical block size.  For RAID arrays it is often the stripe chunk
428  *   size.  A properly aligned multiple of minimum_io_size is the
429  *   preferred request size for workloads where a high number of I/O
430  *   operations is desired.
431  */
432 void blk_queue_io_min(struct request_queue *q, unsigned int min)
433 {
434 	blk_limits_io_min(&q->limits, min);
435 }
436 EXPORT_SYMBOL(blk_queue_io_min);
437 
438 /**
439  * blk_limits_io_opt - set optimal request size for a device
440  * @limits: the queue limits
441  * @opt:  smallest I/O size in bytes
442  *
443  * Description:
444  *   Storage devices may report an optimal I/O size, which is the
445  *   device's preferred unit for sustained I/O.  This is rarely reported
446  *   for disk drives.  For RAID arrays it is usually the stripe width or
447  *   the internal track size.  A properly aligned multiple of
448  *   optimal_io_size is the preferred request size for workloads where
449  *   sustained throughput is desired.
450  */
451 void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt)
452 {
453 	limits->io_opt = opt;
454 }
455 EXPORT_SYMBOL(blk_limits_io_opt);
456 
457 /**
458  * blk_queue_io_opt - set optimal request size for the queue
459  * @q:	the request queue for the device
460  * @opt:  optimal request size in bytes
461  *
462  * Description:
463  *   Storage devices may report an optimal I/O size, which is the
464  *   device's preferred unit for sustained I/O.  This is rarely reported
465  *   for disk drives.  For RAID arrays it is usually the stripe width or
466  *   the internal track size.  A properly aligned multiple of
467  *   optimal_io_size is the preferred request size for workloads where
468  *   sustained throughput is desired.
469  */
470 void blk_queue_io_opt(struct request_queue *q, unsigned int opt)
471 {
472 	blk_limits_io_opt(&q->limits, opt);
473 	q->backing_dev_info->ra_pages =
474 		max(queue_io_opt(q) * 2 / PAGE_SIZE, VM_READAHEAD_PAGES);
475 }
476 EXPORT_SYMBOL(blk_queue_io_opt);
477 
478 static unsigned int blk_round_down_sectors(unsigned int sectors, unsigned int lbs)
479 {
480 	sectors = round_down(sectors, lbs >> SECTOR_SHIFT);
481 	if (sectors < PAGE_SIZE >> SECTOR_SHIFT)
482 		sectors = PAGE_SIZE >> SECTOR_SHIFT;
483 	return sectors;
484 }
485 
486 /**
487  * blk_stack_limits - adjust queue_limits for stacked devices
488  * @t:	the stacking driver limits (top device)
489  * @b:  the underlying queue limits (bottom, component device)
490  * @start:  first data sector within component device
491  *
492  * Description:
493  *    This function is used by stacking drivers like MD and DM to ensure
494  *    that all component devices have compatible block sizes and
495  *    alignments.  The stacking driver must provide a queue_limits
496  *    struct (top) and then iteratively call the stacking function for
497  *    all component (bottom) devices.  The stacking function will
498  *    attempt to combine the values and ensure proper alignment.
499  *
500  *    Returns 0 if the top and bottom queue_limits are compatible.  The
501  *    top device's block sizes and alignment offsets may be adjusted to
502  *    ensure alignment with the bottom device. If no compatible sizes
503  *    and alignments exist, -1 is returned and the resulting top
504  *    queue_limits will have the misaligned flag set to indicate that
505  *    the alignment_offset is undefined.
506  */
507 int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
508 		     sector_t start)
509 {
510 	unsigned int top, bottom, alignment, ret = 0;
511 
512 	t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors);
513 	t->max_hw_sectors = min_not_zero(t->max_hw_sectors, b->max_hw_sectors);
514 	t->max_dev_sectors = min_not_zero(t->max_dev_sectors, b->max_dev_sectors);
515 	t->max_write_same_sectors = min(t->max_write_same_sectors,
516 					b->max_write_same_sectors);
517 	t->max_write_zeroes_sectors = min(t->max_write_zeroes_sectors,
518 					b->max_write_zeroes_sectors);
519 	t->max_zone_append_sectors = min(t->max_zone_append_sectors,
520 					b->max_zone_append_sectors);
521 	t->bounce = max(t->bounce, b->bounce);
522 
523 	t->seg_boundary_mask = min_not_zero(t->seg_boundary_mask,
524 					    b->seg_boundary_mask);
525 	t->virt_boundary_mask = min_not_zero(t->virt_boundary_mask,
526 					    b->virt_boundary_mask);
527 
528 	t->max_segments = min_not_zero(t->max_segments, b->max_segments);
529 	t->max_discard_segments = min_not_zero(t->max_discard_segments,
530 					       b->max_discard_segments);
531 	t->max_integrity_segments = min_not_zero(t->max_integrity_segments,
532 						 b->max_integrity_segments);
533 
534 	t->max_segment_size = min_not_zero(t->max_segment_size,
535 					   b->max_segment_size);
536 
537 	t->misaligned |= b->misaligned;
538 
539 	alignment = queue_limit_alignment_offset(b, start);
540 
541 	/* Bottom device has different alignment.  Check that it is
542 	 * compatible with the current top alignment.
543 	 */
544 	if (t->alignment_offset != alignment) {
545 
546 		top = max(t->physical_block_size, t->io_min)
547 			+ t->alignment_offset;
548 		bottom = max(b->physical_block_size, b->io_min) + alignment;
549 
550 		/* Verify that top and bottom intervals line up */
551 		if (max(top, bottom) % min(top, bottom)) {
552 			t->misaligned = 1;
553 			ret = -1;
554 		}
555 	}
556 
557 	t->logical_block_size = max(t->logical_block_size,
558 				    b->logical_block_size);
559 
560 	t->physical_block_size = max(t->physical_block_size,
561 				     b->physical_block_size);
562 
563 	t->io_min = max(t->io_min, b->io_min);
564 	t->io_opt = lcm_not_zero(t->io_opt, b->io_opt);
565 
566 	/* Set non-power-of-2 compatible chunk_sectors boundary */
567 	if (b->chunk_sectors)
568 		t->chunk_sectors = gcd(t->chunk_sectors, b->chunk_sectors);
569 
570 	/* Physical block size a multiple of the logical block size? */
571 	if (t->physical_block_size & (t->logical_block_size - 1)) {
572 		t->physical_block_size = t->logical_block_size;
573 		t->misaligned = 1;
574 		ret = -1;
575 	}
576 
577 	/* Minimum I/O a multiple of the physical block size? */
578 	if (t->io_min & (t->physical_block_size - 1)) {
579 		t->io_min = t->physical_block_size;
580 		t->misaligned = 1;
581 		ret = -1;
582 	}
583 
584 	/* Optimal I/O a multiple of the physical block size? */
585 	if (t->io_opt & (t->physical_block_size - 1)) {
586 		t->io_opt = 0;
587 		t->misaligned = 1;
588 		ret = -1;
589 	}
590 
591 	/* chunk_sectors a multiple of the physical block size? */
592 	if ((t->chunk_sectors << 9) & (t->physical_block_size - 1)) {
593 		t->chunk_sectors = 0;
594 		t->misaligned = 1;
595 		ret = -1;
596 	}
597 
598 	t->raid_partial_stripes_expensive =
599 		max(t->raid_partial_stripes_expensive,
600 		    b->raid_partial_stripes_expensive);
601 
602 	/* Find lowest common alignment_offset */
603 	t->alignment_offset = lcm_not_zero(t->alignment_offset, alignment)
604 		% max(t->physical_block_size, t->io_min);
605 
606 	/* Verify that new alignment_offset is on a logical block boundary */
607 	if (t->alignment_offset & (t->logical_block_size - 1)) {
608 		t->misaligned = 1;
609 		ret = -1;
610 	}
611 
612 	t->max_sectors = blk_round_down_sectors(t->max_sectors, t->logical_block_size);
613 	t->max_hw_sectors = blk_round_down_sectors(t->max_hw_sectors, t->logical_block_size);
614 	t->max_dev_sectors = blk_round_down_sectors(t->max_dev_sectors, t->logical_block_size);
615 
616 	/* Discard alignment and granularity */
617 	if (b->discard_granularity) {
618 		alignment = queue_limit_discard_alignment(b, start);
619 
620 		if (t->discard_granularity != 0 &&
621 		    t->discard_alignment != alignment) {
622 			top = t->discard_granularity + t->discard_alignment;
623 			bottom = b->discard_granularity + alignment;
624 
625 			/* Verify that top and bottom intervals line up */
626 			if ((max(top, bottom) % min(top, bottom)) != 0)
627 				t->discard_misaligned = 1;
628 		}
629 
630 		t->max_discard_sectors = min_not_zero(t->max_discard_sectors,
631 						      b->max_discard_sectors);
632 		t->max_hw_discard_sectors = min_not_zero(t->max_hw_discard_sectors,
633 							 b->max_hw_discard_sectors);
634 		t->discard_granularity = max(t->discard_granularity,
635 					     b->discard_granularity);
636 		t->discard_alignment = lcm_not_zero(t->discard_alignment, alignment) %
637 			t->discard_granularity;
638 	}
639 
640 	t->zone_write_granularity = max(t->zone_write_granularity,
641 					b->zone_write_granularity);
642 	t->zoned = max(t->zoned, b->zoned);
643 	return ret;
644 }
645 EXPORT_SYMBOL(blk_stack_limits);
646 
647 /**
648  * disk_stack_limits - adjust queue limits for stacked drivers
649  * @disk:  MD/DM gendisk (top)
650  * @bdev:  the underlying block device (bottom)
651  * @offset:  offset to beginning of data within component device
652  *
653  * Description:
654  *    Merges the limits for a top level gendisk and a bottom level
655  *    block_device.
656  */
657 void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
658 		       sector_t offset)
659 {
660 	struct request_queue *t = disk->queue;
661 
662 	if (blk_stack_limits(&t->limits, &bdev_get_queue(bdev)->limits,
663 			get_start_sect(bdev) + (offset >> 9)) < 0) {
664 		char top[BDEVNAME_SIZE], bottom[BDEVNAME_SIZE];
665 
666 		disk_name(disk, 0, top);
667 		bdevname(bdev, bottom);
668 
669 		printk(KERN_NOTICE "%s: Warning: Device %s is misaligned\n",
670 		       top, bottom);
671 	}
672 
673 	blk_queue_update_readahead(disk->queue);
674 }
675 EXPORT_SYMBOL(disk_stack_limits);
676 
677 /**
678  * blk_queue_update_dma_pad - update pad mask
679  * @q:     the request queue for the device
680  * @mask:  pad mask
681  *
682  * Update dma pad mask.
683  *
684  * Appending pad buffer to a request modifies the last entry of a
685  * scatter list such that it includes the pad buffer.
686  **/
687 void blk_queue_update_dma_pad(struct request_queue *q, unsigned int mask)
688 {
689 	if (mask > q->dma_pad_mask)
690 		q->dma_pad_mask = mask;
691 }
692 EXPORT_SYMBOL(blk_queue_update_dma_pad);
693 
694 /**
695  * blk_queue_segment_boundary - set boundary rules for segment merging
696  * @q:  the request queue for the device
697  * @mask:  the memory boundary mask
698  **/
699 void blk_queue_segment_boundary(struct request_queue *q, unsigned long mask)
700 {
701 	if (mask < PAGE_SIZE - 1) {
702 		mask = PAGE_SIZE - 1;
703 		printk(KERN_INFO "%s: set to minimum %lx\n",
704 		       __func__, mask);
705 	}
706 
707 	q->limits.seg_boundary_mask = mask;
708 }
709 EXPORT_SYMBOL(blk_queue_segment_boundary);
710 
711 /**
712  * blk_queue_virt_boundary - set boundary rules for bio merging
713  * @q:  the request queue for the device
714  * @mask:  the memory boundary mask
715  **/
716 void blk_queue_virt_boundary(struct request_queue *q, unsigned long mask)
717 {
718 	q->limits.virt_boundary_mask = mask;
719 
720 	/*
721 	 * Devices that require a virtual boundary do not support scatter/gather
722 	 * I/O natively, but instead require a descriptor list entry for each
723 	 * page (which might not be idential to the Linux PAGE_SIZE).  Because
724 	 * of that they are not limited by our notion of "segment size".
725 	 */
726 	if (mask)
727 		q->limits.max_segment_size = UINT_MAX;
728 }
729 EXPORT_SYMBOL(blk_queue_virt_boundary);
730 
731 /**
732  * blk_queue_dma_alignment - set dma length and memory alignment
733  * @q:     the request queue for the device
734  * @mask:  alignment mask
735  *
736  * description:
737  *    set required memory and length alignment for direct dma transactions.
738  *    this is used when building direct io requests for the queue.
739  *
740  **/
741 void blk_queue_dma_alignment(struct request_queue *q, int mask)
742 {
743 	q->dma_alignment = mask;
744 }
745 EXPORT_SYMBOL(blk_queue_dma_alignment);
746 
747 /**
748  * blk_queue_update_dma_alignment - update dma length and memory alignment
749  * @q:     the request queue for the device
750  * @mask:  alignment mask
751  *
752  * description:
753  *    update required memory and length alignment for direct dma transactions.
754  *    If the requested alignment is larger than the current alignment, then
755  *    the current queue alignment is updated to the new value, otherwise it
756  *    is left alone.  The design of this is to allow multiple objects
757  *    (driver, device, transport etc) to set their respective
758  *    alignments without having them interfere.
759  *
760  **/
761 void blk_queue_update_dma_alignment(struct request_queue *q, int mask)
762 {
763 	BUG_ON(mask > PAGE_SIZE);
764 
765 	if (mask > q->dma_alignment)
766 		q->dma_alignment = mask;
767 }
768 EXPORT_SYMBOL(blk_queue_update_dma_alignment);
769 
770 /**
771  * blk_set_queue_depth - tell the block layer about the device queue depth
772  * @q:		the request queue for the device
773  * @depth:		queue depth
774  *
775  */
776 void blk_set_queue_depth(struct request_queue *q, unsigned int depth)
777 {
778 	q->queue_depth = depth;
779 	rq_qos_queue_depth_changed(q);
780 }
781 EXPORT_SYMBOL(blk_set_queue_depth);
782 
783 /**
784  * blk_queue_write_cache - configure queue's write cache
785  * @q:		the request queue for the device
786  * @wc:		write back cache on or off
787  * @fua:	device supports FUA writes, if true
788  *
789  * Tell the block layer about the write cache of @q.
790  */
791 void blk_queue_write_cache(struct request_queue *q, bool wc, bool fua)
792 {
793 	if (wc)
794 		blk_queue_flag_set(QUEUE_FLAG_WC, q);
795 	else
796 		blk_queue_flag_clear(QUEUE_FLAG_WC, q);
797 	if (fua)
798 		blk_queue_flag_set(QUEUE_FLAG_FUA, q);
799 	else
800 		blk_queue_flag_clear(QUEUE_FLAG_FUA, q);
801 
802 	wbt_set_write_cache(q, test_bit(QUEUE_FLAG_WC, &q->queue_flags));
803 }
804 EXPORT_SYMBOL_GPL(blk_queue_write_cache);
805 
806 /**
807  * blk_queue_required_elevator_features - Set a queue required elevator features
808  * @q:		the request queue for the target device
809  * @features:	Required elevator features OR'ed together
810  *
811  * Tell the block layer that for the device controlled through @q, only the
812  * only elevators that can be used are those that implement at least the set of
813  * features specified by @features.
814  */
815 void blk_queue_required_elevator_features(struct request_queue *q,
816 					  unsigned int features)
817 {
818 	q->required_elevator_features = features;
819 }
820 EXPORT_SYMBOL_GPL(blk_queue_required_elevator_features);
821 
822 /**
823  * blk_queue_can_use_dma_map_merging - configure queue for merging segments.
824  * @q:		the request queue for the device
825  * @dev:	the device pointer for dma
826  *
827  * Tell the block layer about merging the segments by dma map of @q.
828  */
829 bool blk_queue_can_use_dma_map_merging(struct request_queue *q,
830 				       struct device *dev)
831 {
832 	unsigned long boundary = dma_get_merge_boundary(dev);
833 
834 	if (!boundary)
835 		return false;
836 
837 	/* No need to update max_segment_size. see blk_queue_virt_boundary() */
838 	blk_queue_virt_boundary(q, boundary);
839 
840 	return true;
841 }
842 EXPORT_SYMBOL_GPL(blk_queue_can_use_dma_map_merging);
843 
844 /**
845  * blk_queue_set_zoned - configure a disk queue zoned model.
846  * @disk:	the gendisk of the queue to configure
847  * @model:	the zoned model to set
848  *
849  * Set the zoned model of the request queue of @disk according to @model.
850  * When @model is BLK_ZONED_HM (host managed), this should be called only
851  * if zoned block device support is enabled (CONFIG_BLK_DEV_ZONED option).
852  * If @model specifies BLK_ZONED_HA (host aware), the effective model used
853  * depends on CONFIG_BLK_DEV_ZONED settings and on the existence of partitions
854  * on the disk.
855  */
856 void blk_queue_set_zoned(struct gendisk *disk, enum blk_zoned_model model)
857 {
858 	struct request_queue *q = disk->queue;
859 
860 	switch (model) {
861 	case BLK_ZONED_HM:
862 		/*
863 		 * Host managed devices are supported only if
864 		 * CONFIG_BLK_DEV_ZONED is enabled.
865 		 */
866 		WARN_ON_ONCE(!IS_ENABLED(CONFIG_BLK_DEV_ZONED));
867 		break;
868 	case BLK_ZONED_HA:
869 		/*
870 		 * Host aware devices can be treated either as regular block
871 		 * devices (similar to drive managed devices) or as zoned block
872 		 * devices to take advantage of the zone command set, similarly
873 		 * to host managed devices. We try the latter if there are no
874 		 * partitions and zoned block device support is enabled, else
875 		 * we do nothing special as far as the block layer is concerned.
876 		 */
877 		if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) ||
878 		    !xa_empty(&disk->part_tbl))
879 			model = BLK_ZONED_NONE;
880 		break;
881 	case BLK_ZONED_NONE:
882 	default:
883 		if (WARN_ON_ONCE(model != BLK_ZONED_NONE))
884 			model = BLK_ZONED_NONE;
885 		break;
886 	}
887 
888 	q->limits.zoned = model;
889 	if (model != BLK_ZONED_NONE) {
890 		/*
891 		 * Set the zone write granularity to the device logical block
892 		 * size by default. The driver can change this value if needed.
893 		 */
894 		blk_queue_zone_write_granularity(q,
895 						queue_logical_block_size(q));
896 	} else {
897 		blk_queue_clear_zone_settings(q);
898 	}
899 }
900 EXPORT_SYMBOL_GPL(blk_queue_set_zoned);
901