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
blk_queue_rq_timeout(struct request_queue * q,unsigned int timeout)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 */
blk_set_default_limits(struct queue_limits * lim)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 */
blk_set_stacking_limits(struct queue_limits * lim)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 **/
blk_queue_bounce_limit(struct request_queue * q,u64 max_addr)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 **/
blk_queue_max_hw_sectors(struct request_queue * q,unsigned int max_hw_sectors)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 **/
blk_queue_chunk_sectors(struct request_queue * q,unsigned int chunk_sectors)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 **/
blk_queue_max_discard_sectors(struct request_queue * q,unsigned int max_discard_sectors)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 **/
blk_queue_max_write_same_sectors(struct request_queue * q,unsigned int max_write_same_sectors)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 **/
blk_queue_max_write_zeroes_sectors(struct request_queue * q,unsigned int max_write_zeroes_sectors)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 **/
blk_queue_max_zone_append_sectors(struct request_queue * q,unsigned int max_zone_append_sectors)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 **/
blk_queue_max_segments(struct request_queue * q,unsigned short max_segments)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 **/
blk_queue_max_discard_segments(struct request_queue * q,unsigned short max_segments)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 **/
blk_queue_max_segment_size(struct request_queue * q,unsigned int max_size)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 **/
blk_queue_logical_block_size(struct request_queue * q,unsigned int size)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 */
blk_queue_physical_block_size(struct request_queue * q,unsigned int size)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 */
blk_queue_alignment_offset(struct request_queue * q,unsigned int offset)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
blk_queue_update_readahead(struct request_queue * q)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 */
blk_limits_io_min(struct queue_limits * limits,unsigned int min)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 */
blk_queue_io_min(struct request_queue * q,unsigned int min)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 */
blk_limits_io_opt(struct queue_limits * limits,unsigned int opt)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 */
blk_queue_io_opt(struct request_queue * q,unsigned int opt)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
blk_round_down_sectors(unsigned int sectors,unsigned int lbs)471 static unsigned int blk_round_down_sectors(unsigned int sectors, unsigned int lbs)
472 {
473 sectors = round_down(sectors, lbs >> SECTOR_SHIFT);
474 if (sectors < PAGE_SIZE >> SECTOR_SHIFT)
475 sectors = PAGE_SIZE >> SECTOR_SHIFT;
476 return sectors;
477 }
478
479 /**
480 * blk_stack_limits - adjust queue_limits for stacked devices
481 * @t: the stacking driver limits (top device)
482 * @b: the underlying queue limits (bottom, component device)
483 * @start: first data sector within component device
484 *
485 * Description:
486 * This function is used by stacking drivers like MD and DM to ensure
487 * that all component devices have compatible block sizes and
488 * alignments. The stacking driver must provide a queue_limits
489 * struct (top) and then iteratively call the stacking function for
490 * all component (bottom) devices. The stacking function will
491 * attempt to combine the values and ensure proper alignment.
492 *
493 * Returns 0 if the top and bottom queue_limits are compatible. The
494 * top device's block sizes and alignment offsets may be adjusted to
495 * ensure alignment with the bottom device. If no compatible sizes
496 * and alignments exist, -1 is returned and the resulting top
497 * queue_limits will have the misaligned flag set to indicate that
498 * the alignment_offset is undefined.
499 */
blk_stack_limits(struct queue_limits * t,struct queue_limits * b,sector_t start)500 int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
501 sector_t start)
502 {
503 unsigned int top, bottom, alignment, ret = 0;
504
505 t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors);
506 t->max_hw_sectors = min_not_zero(t->max_hw_sectors, b->max_hw_sectors);
507 t->max_dev_sectors = min_not_zero(t->max_dev_sectors, b->max_dev_sectors);
508 t->max_write_same_sectors = min(t->max_write_same_sectors,
509 b->max_write_same_sectors);
510 t->max_write_zeroes_sectors = min(t->max_write_zeroes_sectors,
511 b->max_write_zeroes_sectors);
512 t->max_zone_append_sectors = min(t->max_zone_append_sectors,
513 b->max_zone_append_sectors);
514 t->bounce_pfn = min_not_zero(t->bounce_pfn, b->bounce_pfn);
515
516 t->seg_boundary_mask = min_not_zero(t->seg_boundary_mask,
517 b->seg_boundary_mask);
518 t->virt_boundary_mask = min_not_zero(t->virt_boundary_mask,
519 b->virt_boundary_mask);
520
521 t->max_segments = min_not_zero(t->max_segments, b->max_segments);
522 t->max_discard_segments = min_not_zero(t->max_discard_segments,
523 b->max_discard_segments);
524 t->max_integrity_segments = min_not_zero(t->max_integrity_segments,
525 b->max_integrity_segments);
526
527 t->max_segment_size = min_not_zero(t->max_segment_size,
528 b->max_segment_size);
529
530 t->misaligned |= b->misaligned;
531
532 alignment = queue_limit_alignment_offset(b, start);
533
534 /* Bottom device has different alignment. Check that it is
535 * compatible with the current top alignment.
536 */
537 if (t->alignment_offset != alignment) {
538
539 top = max(t->physical_block_size, t->io_min)
540 + t->alignment_offset;
541 bottom = max(b->physical_block_size, b->io_min) + alignment;
542
543 /* Verify that top and bottom intervals line up */
544 if (max(top, bottom) % min(top, bottom)) {
545 t->misaligned = 1;
546 ret = -1;
547 }
548 }
549
550 t->logical_block_size = max(t->logical_block_size,
551 b->logical_block_size);
552
553 t->physical_block_size = max(t->physical_block_size,
554 b->physical_block_size);
555
556 t->io_min = max(t->io_min, b->io_min);
557 t->io_opt = lcm_not_zero(t->io_opt, b->io_opt);
558
559 /* Set non-power-of-2 compatible chunk_sectors boundary */
560 if (b->chunk_sectors)
561 t->chunk_sectors = gcd(t->chunk_sectors, b->chunk_sectors);
562
563 /* Physical block size a multiple of the logical block size? */
564 if (t->physical_block_size & (t->logical_block_size - 1)) {
565 t->physical_block_size = t->logical_block_size;
566 t->misaligned = 1;
567 ret = -1;
568 }
569
570 /* Minimum I/O a multiple of the physical block size? */
571 if (t->io_min & (t->physical_block_size - 1)) {
572 t->io_min = t->physical_block_size;
573 t->misaligned = 1;
574 ret = -1;
575 }
576
577 /* Optimal I/O a multiple of the physical block size? */
578 if (t->io_opt & (t->physical_block_size - 1)) {
579 t->io_opt = 0;
580 t->misaligned = 1;
581 ret = -1;
582 }
583
584 /* chunk_sectors a multiple of the physical block size? */
585 if ((t->chunk_sectors << 9) & (t->physical_block_size - 1)) {
586 t->chunk_sectors = 0;
587 t->misaligned = 1;
588 ret = -1;
589 }
590
591 t->raid_partial_stripes_expensive =
592 max(t->raid_partial_stripes_expensive,
593 b->raid_partial_stripes_expensive);
594
595 /* Find lowest common alignment_offset */
596 t->alignment_offset = lcm_not_zero(t->alignment_offset, alignment)
597 % max(t->physical_block_size, t->io_min);
598
599 /* Verify that new alignment_offset is on a logical block boundary */
600 if (t->alignment_offset & (t->logical_block_size - 1)) {
601 t->misaligned = 1;
602 ret = -1;
603 }
604
605 t->max_sectors = blk_round_down_sectors(t->max_sectors, t->logical_block_size);
606 t->max_hw_sectors = blk_round_down_sectors(t->max_hw_sectors, t->logical_block_size);
607 t->max_dev_sectors = blk_round_down_sectors(t->max_dev_sectors, t->logical_block_size);
608
609 /* Discard alignment and granularity */
610 if (b->discard_granularity) {
611 alignment = queue_limit_discard_alignment(b, start);
612
613 if (t->discard_granularity != 0 &&
614 t->discard_alignment != alignment) {
615 top = t->discard_granularity + t->discard_alignment;
616 bottom = b->discard_granularity + alignment;
617
618 /* Verify that top and bottom intervals line up */
619 if ((max(top, bottom) % min(top, bottom)) != 0)
620 t->discard_misaligned = 1;
621 }
622
623 t->max_discard_sectors = min_not_zero(t->max_discard_sectors,
624 b->max_discard_sectors);
625 t->max_hw_discard_sectors = min_not_zero(t->max_hw_discard_sectors,
626 b->max_hw_discard_sectors);
627 t->discard_granularity = max(t->discard_granularity,
628 b->discard_granularity);
629 t->discard_alignment = lcm_not_zero(t->discard_alignment, alignment) %
630 t->discard_granularity;
631 }
632
633 t->zoned = max(t->zoned, b->zoned);
634 return ret;
635 }
636 EXPORT_SYMBOL(blk_stack_limits);
637
638 /**
639 * disk_stack_limits - adjust queue limits for stacked drivers
640 * @disk: MD/DM gendisk (top)
641 * @bdev: the underlying block device (bottom)
642 * @offset: offset to beginning of data within component device
643 *
644 * Description:
645 * Merges the limits for a top level gendisk and a bottom level
646 * block_device.
647 */
disk_stack_limits(struct gendisk * disk,struct block_device * bdev,sector_t offset)648 void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
649 sector_t offset)
650 {
651 struct request_queue *t = disk->queue;
652
653 if (blk_stack_limits(&t->limits, &bdev_get_queue(bdev)->limits,
654 get_start_sect(bdev) + (offset >> 9)) < 0) {
655 char top[BDEVNAME_SIZE], bottom[BDEVNAME_SIZE];
656
657 disk_name(disk, 0, top);
658 bdevname(bdev, bottom);
659
660 printk(KERN_NOTICE "%s: Warning: Device %s is misaligned\n",
661 top, bottom);
662 }
663
664 blk_queue_update_readahead(disk->queue);
665 }
666 EXPORT_SYMBOL(disk_stack_limits);
667
668 /**
669 * blk_queue_update_dma_pad - update pad mask
670 * @q: the request queue for the device
671 * @mask: pad mask
672 *
673 * Update dma pad mask.
674 *
675 * Appending pad buffer to a request modifies the last entry of a
676 * scatter list such that it includes the pad buffer.
677 **/
blk_queue_update_dma_pad(struct request_queue * q,unsigned int mask)678 void blk_queue_update_dma_pad(struct request_queue *q, unsigned int mask)
679 {
680 if (mask > q->dma_pad_mask)
681 q->dma_pad_mask = mask;
682 }
683 EXPORT_SYMBOL(blk_queue_update_dma_pad);
684
685 /**
686 * blk_queue_segment_boundary - set boundary rules for segment merging
687 * @q: the request queue for the device
688 * @mask: the memory boundary mask
689 **/
blk_queue_segment_boundary(struct request_queue * q,unsigned long mask)690 void blk_queue_segment_boundary(struct request_queue *q, unsigned long mask)
691 {
692 if (mask < PAGE_SIZE - 1) {
693 mask = PAGE_SIZE - 1;
694 printk(KERN_INFO "%s: set to minimum %lx\n",
695 __func__, mask);
696 }
697
698 q->limits.seg_boundary_mask = mask;
699 }
700 EXPORT_SYMBOL(blk_queue_segment_boundary);
701
702 /**
703 * blk_queue_virt_boundary - set boundary rules for bio merging
704 * @q: the request queue for the device
705 * @mask: the memory boundary mask
706 **/
blk_queue_virt_boundary(struct request_queue * q,unsigned long mask)707 void blk_queue_virt_boundary(struct request_queue *q, unsigned long mask)
708 {
709 q->limits.virt_boundary_mask = mask;
710
711 /*
712 * Devices that require a virtual boundary do not support scatter/gather
713 * I/O natively, but instead require a descriptor list entry for each
714 * page (which might not be idential to the Linux PAGE_SIZE). Because
715 * of that they are not limited by our notion of "segment size".
716 */
717 if (mask)
718 q->limits.max_segment_size = UINT_MAX;
719 }
720 EXPORT_SYMBOL(blk_queue_virt_boundary);
721
722 /**
723 * blk_queue_dma_alignment - set dma length and memory alignment
724 * @q: the request queue for the device
725 * @mask: alignment mask
726 *
727 * description:
728 * set required memory and length alignment for direct dma transactions.
729 * this is used when building direct io requests for the queue.
730 *
731 **/
blk_queue_dma_alignment(struct request_queue * q,int mask)732 void blk_queue_dma_alignment(struct request_queue *q, int mask)
733 {
734 q->dma_alignment = mask;
735 }
736 EXPORT_SYMBOL(blk_queue_dma_alignment);
737
738 /**
739 * blk_queue_update_dma_alignment - update dma length and memory alignment
740 * @q: the request queue for the device
741 * @mask: alignment mask
742 *
743 * description:
744 * update required memory and length alignment for direct dma transactions.
745 * If the requested alignment is larger than the current alignment, then
746 * the current queue alignment is updated to the new value, otherwise it
747 * is left alone. The design of this is to allow multiple objects
748 * (driver, device, transport etc) to set their respective
749 * alignments without having them interfere.
750 *
751 **/
blk_queue_update_dma_alignment(struct request_queue * q,int mask)752 void blk_queue_update_dma_alignment(struct request_queue *q, int mask)
753 {
754 BUG_ON(mask > PAGE_SIZE);
755
756 if (mask > q->dma_alignment)
757 q->dma_alignment = mask;
758 }
759 EXPORT_SYMBOL(blk_queue_update_dma_alignment);
760
761 /**
762 * blk_set_queue_depth - tell the block layer about the device queue depth
763 * @q: the request queue for the device
764 * @depth: queue depth
765 *
766 */
blk_set_queue_depth(struct request_queue * q,unsigned int depth)767 void blk_set_queue_depth(struct request_queue *q, unsigned int depth)
768 {
769 q->queue_depth = depth;
770 rq_qos_queue_depth_changed(q);
771 }
772 EXPORT_SYMBOL(blk_set_queue_depth);
773
774 /**
775 * blk_queue_write_cache - configure queue's write cache
776 * @q: the request queue for the device
777 * @wc: write back cache on or off
778 * @fua: device supports FUA writes, if true
779 *
780 * Tell the block layer about the write cache of @q.
781 */
blk_queue_write_cache(struct request_queue * q,bool wc,bool fua)782 void blk_queue_write_cache(struct request_queue *q, bool wc, bool fua)
783 {
784 if (wc)
785 blk_queue_flag_set(QUEUE_FLAG_WC, q);
786 else
787 blk_queue_flag_clear(QUEUE_FLAG_WC, q);
788 if (fua)
789 blk_queue_flag_set(QUEUE_FLAG_FUA, q);
790 else
791 blk_queue_flag_clear(QUEUE_FLAG_FUA, q);
792
793 wbt_set_write_cache(q, test_bit(QUEUE_FLAG_WC, &q->queue_flags));
794 }
795 EXPORT_SYMBOL_GPL(blk_queue_write_cache);
796
797 /**
798 * blk_queue_required_elevator_features - Set a queue required elevator features
799 * @q: the request queue for the target device
800 * @features: Required elevator features OR'ed together
801 *
802 * Tell the block layer that for the device controlled through @q, only the
803 * only elevators that can be used are those that implement at least the set of
804 * features specified by @features.
805 */
blk_queue_required_elevator_features(struct request_queue * q,unsigned int features)806 void blk_queue_required_elevator_features(struct request_queue *q,
807 unsigned int features)
808 {
809 q->required_elevator_features = features;
810 }
811 EXPORT_SYMBOL_GPL(blk_queue_required_elevator_features);
812
813 /**
814 * blk_queue_can_use_dma_map_merging - configure queue for merging segments.
815 * @q: the request queue for the device
816 * @dev: the device pointer for dma
817 *
818 * Tell the block layer about merging the segments by dma map of @q.
819 */
blk_queue_can_use_dma_map_merging(struct request_queue * q,struct device * dev)820 bool blk_queue_can_use_dma_map_merging(struct request_queue *q,
821 struct device *dev)
822 {
823 unsigned long boundary = dma_get_merge_boundary(dev);
824
825 if (!boundary)
826 return false;
827
828 /* No need to update max_segment_size. see blk_queue_virt_boundary() */
829 blk_queue_virt_boundary(q, boundary);
830
831 return true;
832 }
833 EXPORT_SYMBOL_GPL(blk_queue_can_use_dma_map_merging);
834
835 /**
836 * blk_queue_set_zoned - configure a disk queue zoned model.
837 * @disk: the gendisk of the queue to configure
838 * @model: the zoned model to set
839 *
840 * Set the zoned model of the request queue of @disk according to @model.
841 * When @model is BLK_ZONED_HM (host managed), this should be called only
842 * if zoned block device support is enabled (CONFIG_BLK_DEV_ZONED option).
843 * If @model specifies BLK_ZONED_HA (host aware), the effective model used
844 * depends on CONFIG_BLK_DEV_ZONED settings and on the existence of partitions
845 * on the disk.
846 */
blk_queue_set_zoned(struct gendisk * disk,enum blk_zoned_model model)847 void blk_queue_set_zoned(struct gendisk *disk, enum blk_zoned_model model)
848 {
849 switch (model) {
850 case BLK_ZONED_HM:
851 /*
852 * Host managed devices are supported only if
853 * CONFIG_BLK_DEV_ZONED is enabled.
854 */
855 WARN_ON_ONCE(!IS_ENABLED(CONFIG_BLK_DEV_ZONED));
856 break;
857 case BLK_ZONED_HA:
858 /*
859 * Host aware devices can be treated either as regular block
860 * devices (similar to drive managed devices) or as zoned block
861 * devices to take advantage of the zone command set, similarly
862 * to host managed devices. We try the latter if there are no
863 * partitions and zoned block device support is enabled, else
864 * we do nothing special as far as the block layer is concerned.
865 */
866 if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) ||
867 disk_has_partitions(disk))
868 model = BLK_ZONED_NONE;
869 break;
870 case BLK_ZONED_NONE:
871 default:
872 if (WARN_ON_ONCE(model != BLK_ZONED_NONE))
873 model = BLK_ZONED_NONE;
874 break;
875 }
876
877 disk->queue->limits.zoned = model;
878 }
879 EXPORT_SYMBOL_GPL(blk_queue_set_zoned);
880
blk_settings_init(void)881 static int __init blk_settings_init(void)
882 {
883 blk_max_low_pfn = max_low_pfn - 1;
884 blk_max_pfn = max_pfn - 1;
885 return 0;
886 }
887 subsys_initcall(blk_settings_init);
888