xref: /openbmc/linux/block/blk-settings.c (revision 7af6fbdd)
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/memblock.h>	/* for max_pfn/max_low_pfn */
11 #include <linux/gcd.h>
12 #include <linux/lcm.h>
13 #include <linux/jiffies.h>
14 #include <linux/gfp.h>
15 #include <linux/dma-mapping.h>
16 
17 #include "blk.h"
18 #include "blk-wbt.h"
19 
20 unsigned long blk_max_low_pfn;
21 EXPORT_SYMBOL(blk_max_low_pfn);
22 
23 unsigned long blk_max_pfn;
24 
25 void blk_queue_rq_timeout(struct request_queue *q, unsigned int timeout)
26 {
27 	q->rq_timeout = timeout;
28 }
29 EXPORT_SYMBOL_GPL(blk_queue_rq_timeout);
30 
31 /**
32  * blk_set_default_limits - reset limits to default values
33  * @lim:  the queue_limits structure to reset
34  *
35  * Description:
36  *   Returns a queue_limit struct to its default state.
37  */
38 void blk_set_default_limits(struct queue_limits *lim)
39 {
40 	lim->max_segments = BLK_MAX_SEGMENTS;
41 	lim->max_discard_segments = 1;
42 	lim->max_integrity_segments = 0;
43 	lim->seg_boundary_mask = BLK_SEG_BOUNDARY_MASK;
44 	lim->virt_boundary_mask = 0;
45 	lim->max_segment_size = BLK_MAX_SEGMENT_SIZE;
46 	lim->max_sectors = lim->max_hw_sectors = BLK_SAFE_MAX_SECTORS;
47 	lim->max_dev_sectors = 0;
48 	lim->chunk_sectors = 0;
49 	lim->max_write_same_sectors = 0;
50 	lim->max_write_zeroes_sectors = 0;
51 	lim->max_zone_append_sectors = 0;
52 	lim->max_discard_sectors = 0;
53 	lim->max_hw_discard_sectors = 0;
54 	lim->discard_granularity = 0;
55 	lim->discard_alignment = 0;
56 	lim->discard_misaligned = 0;
57 	lim->logical_block_size = lim->physical_block_size = lim->io_min = 512;
58 	lim->bounce_pfn = (unsigned long)(BLK_BOUNCE_ANY >> PAGE_SHIFT);
59 	lim->alignment_offset = 0;
60 	lim->io_opt = 0;
61 	lim->misaligned = 0;
62 	lim->zoned = BLK_ZONED_NONE;
63 }
64 EXPORT_SYMBOL(blk_set_default_limits);
65 
66 /**
67  * blk_set_stacking_limits - set default limits for stacking devices
68  * @lim:  the queue_limits structure to reset
69  *
70  * Description:
71  *   Returns a queue_limit struct to its default state. Should be used
72  *   by stacking drivers like DM that have no internal limits.
73  */
74 void blk_set_stacking_limits(struct queue_limits *lim)
75 {
76 	blk_set_default_limits(lim);
77 
78 	/* Inherit limits from component devices */
79 	lim->max_segments = USHRT_MAX;
80 	lim->max_discard_segments = USHRT_MAX;
81 	lim->max_hw_sectors = UINT_MAX;
82 	lim->max_segment_size = UINT_MAX;
83 	lim->max_sectors = UINT_MAX;
84 	lim->max_dev_sectors = UINT_MAX;
85 	lim->max_write_same_sectors = UINT_MAX;
86 	lim->max_write_zeroes_sectors = UINT_MAX;
87 	lim->max_zone_append_sectors = UINT_MAX;
88 }
89 EXPORT_SYMBOL(blk_set_stacking_limits);
90 
91 /**
92  * blk_queue_bounce_limit - set bounce buffer limit for queue
93  * @q: the request queue for the device
94  * @max_addr: the maximum address the device can handle
95  *
96  * Description:
97  *    Different hardware can have different requirements as to what pages
98  *    it can do I/O directly to. A low level driver can call
99  *    blk_queue_bounce_limit to have lower memory pages allocated as bounce
100  *    buffers for doing I/O to pages residing above @max_addr.
101  **/
102 void blk_queue_bounce_limit(struct request_queue *q, u64 max_addr)
103 {
104 	unsigned long b_pfn = max_addr >> PAGE_SHIFT;
105 	int dma = 0;
106 
107 	q->bounce_gfp = GFP_NOIO;
108 #if BITS_PER_LONG == 64
109 	/*
110 	 * Assume anything <= 4GB can be handled by IOMMU.  Actually
111 	 * some IOMMUs can handle everything, but I don't know of a
112 	 * way to test this here.
113 	 */
114 	if (b_pfn < (min_t(u64, 0xffffffffUL, BLK_BOUNCE_HIGH) >> PAGE_SHIFT))
115 		dma = 1;
116 	q->limits.bounce_pfn = max(max_low_pfn, b_pfn);
117 #else
118 	if (b_pfn < blk_max_low_pfn)
119 		dma = 1;
120 	q->limits.bounce_pfn = b_pfn;
121 #endif
122 	if (dma) {
123 		init_emergency_isa_pool();
124 		q->bounce_gfp = GFP_NOIO | GFP_DMA;
125 		q->limits.bounce_pfn = b_pfn;
126 	}
127 }
128 EXPORT_SYMBOL(blk_queue_bounce_limit);
129 
130 /**
131  * blk_queue_max_hw_sectors - set max sectors for a request for this queue
132  * @q:  the request queue for the device
133  * @max_hw_sectors:  max hardware sectors in the usual 512b unit
134  *
135  * Description:
136  *    Enables a low level driver to set a hard upper limit,
137  *    max_hw_sectors, on the size of requests.  max_hw_sectors is set by
138  *    the device driver based upon the capabilities of the I/O
139  *    controller.
140  *
141  *    max_dev_sectors is a hard limit imposed by the storage device for
142  *    READ/WRITE requests. It is set by the disk driver.
143  *
144  *    max_sectors is a soft limit imposed by the block layer for
145  *    filesystem type requests.  This value can be overridden on a
146  *    per-device basis in /sys/block/<device>/queue/max_sectors_kb.
147  *    The soft limit can not exceed max_hw_sectors.
148  **/
149 void blk_queue_max_hw_sectors(struct request_queue *q, unsigned int max_hw_sectors)
150 {
151 	struct queue_limits *limits = &q->limits;
152 	unsigned int max_sectors;
153 
154 	if ((max_hw_sectors << 9) < PAGE_SIZE) {
155 		max_hw_sectors = 1 << (PAGE_SHIFT - 9);
156 		printk(KERN_INFO "%s: set to minimum %d\n",
157 		       __func__, max_hw_sectors);
158 	}
159 
160 	limits->max_hw_sectors = max_hw_sectors;
161 	max_sectors = min_not_zero(max_hw_sectors, limits->max_dev_sectors);
162 	max_sectors = min_t(unsigned int, max_sectors, BLK_DEF_MAX_SECTORS);
163 	limits->max_sectors = max_sectors;
164 	q->backing_dev_info->io_pages = max_sectors >> (PAGE_SHIFT - 9);
165 }
166 EXPORT_SYMBOL(blk_queue_max_hw_sectors);
167 
168 /**
169  * blk_queue_chunk_sectors - set size of the chunk for this queue
170  * @q:  the request queue for the device
171  * @chunk_sectors:  chunk sectors in the usual 512b unit
172  *
173  * Description:
174  *    If a driver doesn't want IOs to cross a given chunk size, it can set
175  *    this limit and prevent merging across chunks. Note that the block layer
176  *    must accept a page worth of data at any offset. So if the crossing of
177  *    chunks is a hard limitation in the driver, it must still be prepared
178  *    to split single page bios.
179  **/
180 void blk_queue_chunk_sectors(struct request_queue *q, unsigned int chunk_sectors)
181 {
182 	q->limits.chunk_sectors = chunk_sectors;
183 }
184 EXPORT_SYMBOL(blk_queue_chunk_sectors);
185 
186 /**
187  * blk_queue_max_discard_sectors - set max sectors for a single discard
188  * @q:  the request queue for the device
189  * @max_discard_sectors: maximum number of sectors to discard
190  **/
191 void blk_queue_max_discard_sectors(struct request_queue *q,
192 		unsigned int max_discard_sectors)
193 {
194 	q->limits.max_hw_discard_sectors = max_discard_sectors;
195 	q->limits.max_discard_sectors = max_discard_sectors;
196 }
197 EXPORT_SYMBOL(blk_queue_max_discard_sectors);
198 
199 /**
200  * blk_queue_max_write_same_sectors - set max sectors for a single write same
201  * @q:  the request queue for the device
202  * @max_write_same_sectors: maximum number of sectors to write per command
203  **/
204 void blk_queue_max_write_same_sectors(struct request_queue *q,
205 				      unsigned int max_write_same_sectors)
206 {
207 	q->limits.max_write_same_sectors = max_write_same_sectors;
208 }
209 EXPORT_SYMBOL(blk_queue_max_write_same_sectors);
210 
211 /**
212  * blk_queue_max_write_zeroes_sectors - set max sectors for a single
213  *                                      write zeroes
214  * @q:  the request queue for the device
215  * @max_write_zeroes_sectors: maximum number of sectors to write per command
216  **/
217 void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
218 		unsigned int max_write_zeroes_sectors)
219 {
220 	q->limits.max_write_zeroes_sectors = max_write_zeroes_sectors;
221 }
222 EXPORT_SYMBOL(blk_queue_max_write_zeroes_sectors);
223 
224 /**
225  * blk_queue_max_zone_append_sectors - set max sectors for a single zone append
226  * @q:  the request queue for the device
227  * @max_zone_append_sectors: maximum number of sectors to write per command
228  **/
229 void blk_queue_max_zone_append_sectors(struct request_queue *q,
230 		unsigned int max_zone_append_sectors)
231 {
232 	unsigned int max_sectors;
233 
234 	if (WARN_ON(!blk_queue_is_zoned(q)))
235 		return;
236 
237 	max_sectors = min(q->limits.max_hw_sectors, max_zone_append_sectors);
238 	max_sectors = min(q->limits.chunk_sectors, max_sectors);
239 
240 	/*
241 	 * Signal eventual driver bugs resulting in the max_zone_append sectors limit
242 	 * being 0 due to a 0 argument, the chunk_sectors limit (zone size) not set,
243 	 * or the max_hw_sectors limit not set.
244 	 */
245 	WARN_ON(!max_sectors);
246 
247 	q->limits.max_zone_append_sectors = max_sectors;
248 }
249 EXPORT_SYMBOL_GPL(blk_queue_max_zone_append_sectors);
250 
251 /**
252  * blk_queue_max_segments - set max hw segments for a request for this queue
253  * @q:  the request queue for the device
254  * @max_segments:  max number of segments
255  *
256  * Description:
257  *    Enables a low level driver to set an upper limit on the number of
258  *    hw data segments in a request.
259  **/
260 void blk_queue_max_segments(struct request_queue *q, unsigned short max_segments)
261 {
262 	if (!max_segments) {
263 		max_segments = 1;
264 		printk(KERN_INFO "%s: set to minimum %d\n",
265 		       __func__, max_segments);
266 	}
267 
268 	q->limits.max_segments = max_segments;
269 }
270 EXPORT_SYMBOL(blk_queue_max_segments);
271 
272 /**
273  * blk_queue_max_discard_segments - set max segments for discard requests
274  * @q:  the request queue for the device
275  * @max_segments:  max number of segments
276  *
277  * Description:
278  *    Enables a low level driver to set an upper limit on the number of
279  *    segments in a discard request.
280  **/
281 void blk_queue_max_discard_segments(struct request_queue *q,
282 		unsigned short max_segments)
283 {
284 	q->limits.max_discard_segments = max_segments;
285 }
286 EXPORT_SYMBOL_GPL(blk_queue_max_discard_segments);
287 
288 /**
289  * blk_queue_max_segment_size - set max segment size for blk_rq_map_sg
290  * @q:  the request queue for the device
291  * @max_size:  max size of segment in bytes
292  *
293  * Description:
294  *    Enables a low level driver to set an upper limit on the size of a
295  *    coalesced segment
296  **/
297 void blk_queue_max_segment_size(struct request_queue *q, unsigned int max_size)
298 {
299 	if (max_size < PAGE_SIZE) {
300 		max_size = PAGE_SIZE;
301 		printk(KERN_INFO "%s: set to minimum %d\n",
302 		       __func__, max_size);
303 	}
304 
305 	/* see blk_queue_virt_boundary() for the explanation */
306 	WARN_ON_ONCE(q->limits.virt_boundary_mask);
307 
308 	q->limits.max_segment_size = max_size;
309 }
310 EXPORT_SYMBOL(blk_queue_max_segment_size);
311 
312 /**
313  * blk_queue_logical_block_size - set logical block size for the queue
314  * @q:  the request queue for the device
315  * @size:  the logical block size, in bytes
316  *
317  * Description:
318  *   This should be set to the lowest possible block size that the
319  *   storage device can address.  The default of 512 covers most
320  *   hardware.
321  **/
322 void blk_queue_logical_block_size(struct request_queue *q, unsigned int size)
323 {
324 	q->limits.logical_block_size = size;
325 
326 	if (q->limits.physical_block_size < size)
327 		q->limits.physical_block_size = size;
328 
329 	if (q->limits.io_min < q->limits.physical_block_size)
330 		q->limits.io_min = q->limits.physical_block_size;
331 }
332 EXPORT_SYMBOL(blk_queue_logical_block_size);
333 
334 /**
335  * blk_queue_physical_block_size - set physical block size for the queue
336  * @q:  the request queue for the device
337  * @size:  the physical block size, in bytes
338  *
339  * Description:
340  *   This should be set to the lowest possible sector size that the
341  *   hardware can operate on without reverting to read-modify-write
342  *   operations.
343  */
344 void blk_queue_physical_block_size(struct request_queue *q, unsigned int size)
345 {
346 	q->limits.physical_block_size = size;
347 
348 	if (q->limits.physical_block_size < q->limits.logical_block_size)
349 		q->limits.physical_block_size = q->limits.logical_block_size;
350 
351 	if (q->limits.io_min < q->limits.physical_block_size)
352 		q->limits.io_min = q->limits.physical_block_size;
353 }
354 EXPORT_SYMBOL(blk_queue_physical_block_size);
355 
356 /**
357  * blk_queue_alignment_offset - set physical block alignment offset
358  * @q:	the request queue for the device
359  * @offset: alignment offset in bytes
360  *
361  * Description:
362  *   Some devices are naturally misaligned to compensate for things like
363  *   the legacy DOS partition table 63-sector offset.  Low-level drivers
364  *   should call this function for devices whose first sector is not
365  *   naturally aligned.
366  */
367 void blk_queue_alignment_offset(struct request_queue *q, unsigned int offset)
368 {
369 	q->limits.alignment_offset =
370 		offset & (q->limits.physical_block_size - 1);
371 	q->limits.misaligned = 0;
372 }
373 EXPORT_SYMBOL(blk_queue_alignment_offset);
374 
375 void blk_queue_update_readahead(struct request_queue *q)
376 {
377 	/*
378 	 * For read-ahead of large files to be effective, we need to read ahead
379 	 * at least twice the optimal I/O size.
380 	 */
381 	q->backing_dev_info->ra_pages =
382 		max(queue_io_opt(q) * 2 / PAGE_SIZE, VM_READAHEAD_PAGES);
383 	q->backing_dev_info->io_pages =
384 		queue_max_sectors(q) >> (PAGE_SHIFT - 9);
385 }
386 EXPORT_SYMBOL_GPL(blk_queue_update_readahead);
387 
388 /**
389  * blk_limits_io_min - set minimum request size for a device
390  * @limits: the queue limits
391  * @min:  smallest I/O size in bytes
392  *
393  * Description:
394  *   Some devices have an internal block size bigger than the reported
395  *   hardware sector size.  This function can be used to signal the
396  *   smallest I/O the device can perform without incurring a performance
397  *   penalty.
398  */
399 void blk_limits_io_min(struct queue_limits *limits, unsigned int min)
400 {
401 	limits->io_min = min;
402 
403 	if (limits->io_min < limits->logical_block_size)
404 		limits->io_min = limits->logical_block_size;
405 
406 	if (limits->io_min < limits->physical_block_size)
407 		limits->io_min = limits->physical_block_size;
408 }
409 EXPORT_SYMBOL(blk_limits_io_min);
410 
411 /**
412  * blk_queue_io_min - set minimum request size for the queue
413  * @q:	the request queue for the device
414  * @min:  smallest I/O size in bytes
415  *
416  * Description:
417  *   Storage devices may report a granularity or preferred minimum I/O
418  *   size which is the smallest request the device can perform without
419  *   incurring a performance penalty.  For disk drives this is often the
420  *   physical block size.  For RAID arrays it is often the stripe chunk
421  *   size.  A properly aligned multiple of minimum_io_size is the
422  *   preferred request size for workloads where a high number of I/O
423  *   operations is desired.
424  */
425 void blk_queue_io_min(struct request_queue *q, unsigned int min)
426 {
427 	blk_limits_io_min(&q->limits, min);
428 }
429 EXPORT_SYMBOL(blk_queue_io_min);
430 
431 /**
432  * blk_limits_io_opt - set optimal request size for a device
433  * @limits: the queue limits
434  * @opt:  smallest I/O size in bytes
435  *
436  * Description:
437  *   Storage devices may report an optimal I/O size, which is the
438  *   device's preferred unit for sustained I/O.  This is rarely reported
439  *   for disk drives.  For RAID arrays it is usually the stripe width or
440  *   the internal track size.  A properly aligned multiple of
441  *   optimal_io_size is the preferred request size for workloads where
442  *   sustained throughput is desired.
443  */
444 void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt)
445 {
446 	limits->io_opt = opt;
447 }
448 EXPORT_SYMBOL(blk_limits_io_opt);
449 
450 /**
451  * blk_queue_io_opt - set optimal request size for the queue
452  * @q:	the request queue for the device
453  * @opt:  optimal request size in bytes
454  *
455  * Description:
456  *   Storage devices may report an optimal I/O size, which is the
457  *   device's preferred unit for sustained I/O.  This is rarely reported
458  *   for disk drives.  For RAID arrays it is usually the stripe width or
459  *   the internal track size.  A properly aligned multiple of
460  *   optimal_io_size is the preferred request size for workloads where
461  *   sustained throughput is desired.
462  */
463 void blk_queue_io_opt(struct request_queue *q, unsigned int opt)
464 {
465 	blk_limits_io_opt(&q->limits, opt);
466 	q->backing_dev_info->ra_pages =
467 		max(queue_io_opt(q) * 2 / PAGE_SIZE, VM_READAHEAD_PAGES);
468 }
469 EXPORT_SYMBOL(blk_queue_io_opt);
470 
471 /**
472  * blk_stack_limits - adjust queue_limits for stacked devices
473  * @t:	the stacking driver limits (top device)
474  * @b:  the underlying queue limits (bottom, component device)
475  * @start:  first data sector within component device
476  *
477  * Description:
478  *    This function is used by stacking drivers like MD and DM to ensure
479  *    that all component devices have compatible block sizes and
480  *    alignments.  The stacking driver must provide a queue_limits
481  *    struct (top) and then iteratively call the stacking function for
482  *    all component (bottom) devices.  The stacking function will
483  *    attempt to combine the values and ensure proper alignment.
484  *
485  *    Returns 0 if the top and bottom queue_limits are compatible.  The
486  *    top device's block sizes and alignment offsets may be adjusted to
487  *    ensure alignment with the bottom device. If no compatible sizes
488  *    and alignments exist, -1 is returned and the resulting top
489  *    queue_limits will have the misaligned flag set to indicate that
490  *    the alignment_offset is undefined.
491  */
492 int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
493 		     sector_t start)
494 {
495 	unsigned int top, bottom, alignment, ret = 0;
496 
497 	t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors);
498 	t->max_hw_sectors = min_not_zero(t->max_hw_sectors, b->max_hw_sectors);
499 	t->max_dev_sectors = min_not_zero(t->max_dev_sectors, b->max_dev_sectors);
500 	t->max_write_same_sectors = min(t->max_write_same_sectors,
501 					b->max_write_same_sectors);
502 	t->max_write_zeroes_sectors = min(t->max_write_zeroes_sectors,
503 					b->max_write_zeroes_sectors);
504 	t->max_zone_append_sectors = min(t->max_zone_append_sectors,
505 					b->max_zone_append_sectors);
506 	t->bounce_pfn = min_not_zero(t->bounce_pfn, b->bounce_pfn);
507 
508 	t->seg_boundary_mask = min_not_zero(t->seg_boundary_mask,
509 					    b->seg_boundary_mask);
510 	t->virt_boundary_mask = min_not_zero(t->virt_boundary_mask,
511 					    b->virt_boundary_mask);
512 
513 	t->max_segments = min_not_zero(t->max_segments, b->max_segments);
514 	t->max_discard_segments = min_not_zero(t->max_discard_segments,
515 					       b->max_discard_segments);
516 	t->max_integrity_segments = min_not_zero(t->max_integrity_segments,
517 						 b->max_integrity_segments);
518 
519 	t->max_segment_size = min_not_zero(t->max_segment_size,
520 					   b->max_segment_size);
521 
522 	t->misaligned |= b->misaligned;
523 
524 	alignment = queue_limit_alignment_offset(b, start);
525 
526 	/* Bottom device has different alignment.  Check that it is
527 	 * compatible with the current top alignment.
528 	 */
529 	if (t->alignment_offset != alignment) {
530 
531 		top = max(t->physical_block_size, t->io_min)
532 			+ t->alignment_offset;
533 		bottom = max(b->physical_block_size, b->io_min) + alignment;
534 
535 		/* Verify that top and bottom intervals line up */
536 		if (max(top, bottom) % min(top, bottom)) {
537 			t->misaligned = 1;
538 			ret = -1;
539 		}
540 	}
541 
542 	t->logical_block_size = max(t->logical_block_size,
543 				    b->logical_block_size);
544 
545 	t->physical_block_size = max(t->physical_block_size,
546 				     b->physical_block_size);
547 
548 	t->io_min = max(t->io_min, b->io_min);
549 	t->io_opt = lcm_not_zero(t->io_opt, b->io_opt);
550 	t->chunk_sectors = lcm_not_zero(t->chunk_sectors, b->chunk_sectors);
551 
552 	/* Physical block size a multiple of the logical block size? */
553 	if (t->physical_block_size & (t->logical_block_size - 1)) {
554 		t->physical_block_size = t->logical_block_size;
555 		t->misaligned = 1;
556 		ret = -1;
557 	}
558 
559 	/* Minimum I/O a multiple of the physical block size? */
560 	if (t->io_min & (t->physical_block_size - 1)) {
561 		t->io_min = t->physical_block_size;
562 		t->misaligned = 1;
563 		ret = -1;
564 	}
565 
566 	/* Optimal I/O a multiple of the physical block size? */
567 	if (t->io_opt & (t->physical_block_size - 1)) {
568 		t->io_opt = 0;
569 		t->misaligned = 1;
570 		ret = -1;
571 	}
572 
573 	/* chunk_sectors a multiple of the physical block size? */
574 	if ((t->chunk_sectors << 9) & (t->physical_block_size - 1)) {
575 		t->chunk_sectors = 0;
576 		t->misaligned = 1;
577 		ret = -1;
578 	}
579 
580 	t->raid_partial_stripes_expensive =
581 		max(t->raid_partial_stripes_expensive,
582 		    b->raid_partial_stripes_expensive);
583 
584 	/* Find lowest common alignment_offset */
585 	t->alignment_offset = lcm_not_zero(t->alignment_offset, alignment)
586 		% max(t->physical_block_size, t->io_min);
587 
588 	/* Verify that new alignment_offset is on a logical block boundary */
589 	if (t->alignment_offset & (t->logical_block_size - 1)) {
590 		t->misaligned = 1;
591 		ret = -1;
592 	}
593 
594 	/* Discard alignment and granularity */
595 	if (b->discard_granularity) {
596 		alignment = queue_limit_discard_alignment(b, start);
597 
598 		if (t->discard_granularity != 0 &&
599 		    t->discard_alignment != alignment) {
600 			top = t->discard_granularity + t->discard_alignment;
601 			bottom = b->discard_granularity + alignment;
602 
603 			/* Verify that top and bottom intervals line up */
604 			if ((max(top, bottom) % min(top, bottom)) != 0)
605 				t->discard_misaligned = 1;
606 		}
607 
608 		t->max_discard_sectors = min_not_zero(t->max_discard_sectors,
609 						      b->max_discard_sectors);
610 		t->max_hw_discard_sectors = min_not_zero(t->max_hw_discard_sectors,
611 							 b->max_hw_discard_sectors);
612 		t->discard_granularity = max(t->discard_granularity,
613 					     b->discard_granularity);
614 		t->discard_alignment = lcm_not_zero(t->discard_alignment, alignment) %
615 			t->discard_granularity;
616 	}
617 
618 	t->zoned = max(t->zoned, b->zoned);
619 	return ret;
620 }
621 EXPORT_SYMBOL(blk_stack_limits);
622 
623 /**
624  * disk_stack_limits - adjust queue limits for stacked drivers
625  * @disk:  MD/DM gendisk (top)
626  * @bdev:  the underlying block device (bottom)
627  * @offset:  offset to beginning of data within component device
628  *
629  * Description:
630  *    Merges the limits for a top level gendisk and a bottom level
631  *    block_device.
632  */
633 void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
634 		       sector_t offset)
635 {
636 	struct request_queue *t = disk->queue;
637 
638 	if (blk_stack_limits(&t->limits, &bdev_get_queue(bdev)->limits,
639 			get_start_sect(bdev) + (offset >> 9)) < 0) {
640 		char top[BDEVNAME_SIZE], bottom[BDEVNAME_SIZE];
641 
642 		disk_name(disk, 0, top);
643 		bdevname(bdev, bottom);
644 
645 		printk(KERN_NOTICE "%s: Warning: Device %s is misaligned\n",
646 		       top, bottom);
647 	}
648 
649 	blk_queue_update_readahead(disk->queue);
650 }
651 EXPORT_SYMBOL(disk_stack_limits);
652 
653 /**
654  * blk_queue_update_dma_pad - update pad mask
655  * @q:     the request queue for the device
656  * @mask:  pad mask
657  *
658  * Update dma pad mask.
659  *
660  * Appending pad buffer to a request modifies the last entry of a
661  * scatter list such that it includes the pad buffer.
662  **/
663 void blk_queue_update_dma_pad(struct request_queue *q, unsigned int mask)
664 {
665 	if (mask > q->dma_pad_mask)
666 		q->dma_pad_mask = mask;
667 }
668 EXPORT_SYMBOL(blk_queue_update_dma_pad);
669 
670 /**
671  * blk_queue_segment_boundary - set boundary rules for segment merging
672  * @q:  the request queue for the device
673  * @mask:  the memory boundary mask
674  **/
675 void blk_queue_segment_boundary(struct request_queue *q, unsigned long mask)
676 {
677 	if (mask < PAGE_SIZE - 1) {
678 		mask = PAGE_SIZE - 1;
679 		printk(KERN_INFO "%s: set to minimum %lx\n",
680 		       __func__, mask);
681 	}
682 
683 	q->limits.seg_boundary_mask = mask;
684 }
685 EXPORT_SYMBOL(blk_queue_segment_boundary);
686 
687 /**
688  * blk_queue_virt_boundary - set boundary rules for bio merging
689  * @q:  the request queue for the device
690  * @mask:  the memory boundary mask
691  **/
692 void blk_queue_virt_boundary(struct request_queue *q, unsigned long mask)
693 {
694 	q->limits.virt_boundary_mask = mask;
695 
696 	/*
697 	 * Devices that require a virtual boundary do not support scatter/gather
698 	 * I/O natively, but instead require a descriptor list entry for each
699 	 * page (which might not be idential to the Linux PAGE_SIZE).  Because
700 	 * of that they are not limited by our notion of "segment size".
701 	 */
702 	if (mask)
703 		q->limits.max_segment_size = UINT_MAX;
704 }
705 EXPORT_SYMBOL(blk_queue_virt_boundary);
706 
707 /**
708  * blk_queue_dma_alignment - set dma length and memory alignment
709  * @q:     the request queue for the device
710  * @mask:  alignment mask
711  *
712  * description:
713  *    set required memory and length alignment for direct dma transactions.
714  *    this is used when building direct io requests for the queue.
715  *
716  **/
717 void blk_queue_dma_alignment(struct request_queue *q, int mask)
718 {
719 	q->dma_alignment = mask;
720 }
721 EXPORT_SYMBOL(blk_queue_dma_alignment);
722 
723 /**
724  * blk_queue_update_dma_alignment - update dma length and memory alignment
725  * @q:     the request queue for the device
726  * @mask:  alignment mask
727  *
728  * description:
729  *    update required memory and length alignment for direct dma transactions.
730  *    If the requested alignment is larger than the current alignment, then
731  *    the current queue alignment is updated to the new value, otherwise it
732  *    is left alone.  The design of this is to allow multiple objects
733  *    (driver, device, transport etc) to set their respective
734  *    alignments without having them interfere.
735  *
736  **/
737 void blk_queue_update_dma_alignment(struct request_queue *q, int mask)
738 {
739 	BUG_ON(mask > PAGE_SIZE);
740 
741 	if (mask > q->dma_alignment)
742 		q->dma_alignment = mask;
743 }
744 EXPORT_SYMBOL(blk_queue_update_dma_alignment);
745 
746 /**
747  * blk_set_queue_depth - tell the block layer about the device queue depth
748  * @q:		the request queue for the device
749  * @depth:		queue depth
750  *
751  */
752 void blk_set_queue_depth(struct request_queue *q, unsigned int depth)
753 {
754 	q->queue_depth = depth;
755 	rq_qos_queue_depth_changed(q);
756 }
757 EXPORT_SYMBOL(blk_set_queue_depth);
758 
759 /**
760  * blk_queue_write_cache - configure queue's write cache
761  * @q:		the request queue for the device
762  * @wc:		write back cache on or off
763  * @fua:	device supports FUA writes, if true
764  *
765  * Tell the block layer about the write cache of @q.
766  */
767 void blk_queue_write_cache(struct request_queue *q, bool wc, bool fua)
768 {
769 	if (wc)
770 		blk_queue_flag_set(QUEUE_FLAG_WC, q);
771 	else
772 		blk_queue_flag_clear(QUEUE_FLAG_WC, q);
773 	if (fua)
774 		blk_queue_flag_set(QUEUE_FLAG_FUA, q);
775 	else
776 		blk_queue_flag_clear(QUEUE_FLAG_FUA, q);
777 
778 	wbt_set_write_cache(q, test_bit(QUEUE_FLAG_WC, &q->queue_flags));
779 }
780 EXPORT_SYMBOL_GPL(blk_queue_write_cache);
781 
782 /**
783  * blk_queue_required_elevator_features - Set a queue required elevator features
784  * @q:		the request queue for the target device
785  * @features:	Required elevator features OR'ed together
786  *
787  * Tell the block layer that for the device controlled through @q, only the
788  * only elevators that can be used are those that implement at least the set of
789  * features specified by @features.
790  */
791 void blk_queue_required_elevator_features(struct request_queue *q,
792 					  unsigned int features)
793 {
794 	q->required_elevator_features = features;
795 }
796 EXPORT_SYMBOL_GPL(blk_queue_required_elevator_features);
797 
798 /**
799  * blk_queue_can_use_dma_map_merging - configure queue for merging segments.
800  * @q:		the request queue for the device
801  * @dev:	the device pointer for dma
802  *
803  * Tell the block layer about merging the segments by dma map of @q.
804  */
805 bool blk_queue_can_use_dma_map_merging(struct request_queue *q,
806 				       struct device *dev)
807 {
808 	unsigned long boundary = dma_get_merge_boundary(dev);
809 
810 	if (!boundary)
811 		return false;
812 
813 	/* No need to update max_segment_size. see blk_queue_virt_boundary() */
814 	blk_queue_virt_boundary(q, boundary);
815 
816 	return true;
817 }
818 EXPORT_SYMBOL_GPL(blk_queue_can_use_dma_map_merging);
819 
820 /**
821  * blk_queue_set_zoned - configure a disk queue zoned model.
822  * @disk:	the gendisk of the queue to configure
823  * @model:	the zoned model to set
824  *
825  * Set the zoned model of the request queue of @disk according to @model.
826  * When @model is BLK_ZONED_HM (host managed), this should be called only
827  * if zoned block device support is enabled (CONFIG_BLK_DEV_ZONED option).
828  * If @model specifies BLK_ZONED_HA (host aware), the effective model used
829  * depends on CONFIG_BLK_DEV_ZONED settings and on the existence of partitions
830  * on the disk.
831  */
832 void blk_queue_set_zoned(struct gendisk *disk, enum blk_zoned_model model)
833 {
834 	switch (model) {
835 	case BLK_ZONED_HM:
836 		/*
837 		 * Host managed devices are supported only if
838 		 * CONFIG_BLK_DEV_ZONED is enabled.
839 		 */
840 		WARN_ON_ONCE(!IS_ENABLED(CONFIG_BLK_DEV_ZONED));
841 		break;
842 	case BLK_ZONED_HA:
843 		/*
844 		 * Host aware devices can be treated either as regular block
845 		 * devices (similar to drive managed devices) or as zoned block
846 		 * devices to take advantage of the zone command set, similarly
847 		 * to host managed devices. We try the latter if there are no
848 		 * partitions and zoned block device support is enabled, else
849 		 * we do nothing special as far as the block layer is concerned.
850 		 */
851 		if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) ||
852 		    disk_has_partitions(disk))
853 			model = BLK_ZONED_NONE;
854 		break;
855 	case BLK_ZONED_NONE:
856 	default:
857 		if (WARN_ON_ONCE(model != BLK_ZONED_NONE))
858 			model = BLK_ZONED_NONE;
859 		break;
860 	}
861 
862 	disk->queue->limits.zoned = model;
863 }
864 EXPORT_SYMBOL_GPL(blk_queue_set_zoned);
865 
866 static int __init blk_settings_init(void)
867 {
868 	blk_max_low_pfn = max_low_pfn - 1;
869 	blk_max_pfn = max_pfn - 1;
870 	return 0;
871 }
872 subsys_initcall(blk_settings_init);
873