block: don't hold the queue_lock over blk_abort_request
[linux-block.git] / block / blk-settings.c
CommitLineData
86db1e29
JA
1/*
2 * Functions related to setting various queue properties from drivers
3 */
4#include <linux/kernel.h>
5#include <linux/module.h>
6#include <linux/init.h>
7#include <linux/bio.h>
8#include <linux/blkdev.h>
57c8a661 9#include <linux/memblock.h> /* for max_pfn/max_low_pfn */
70dd5bf3 10#include <linux/gcd.h>
2cda2728 11#include <linux/lcm.h>
ad5ebd2f 12#include <linux/jiffies.h>
5a0e3ad6 13#include <linux/gfp.h>
86db1e29
JA
14
15#include "blk.h"
87760e5e 16#include "blk-wbt.h"
86db1e29 17
6728cb0e 18unsigned long blk_max_low_pfn;
86db1e29 19EXPORT_SYMBOL(blk_max_low_pfn);
6728cb0e
JA
20
21unsigned long blk_max_pfn;
86db1e29 22
242f9dcb
JA
23void blk_queue_rq_timeout(struct request_queue *q, unsigned int timeout)
24{
25 q->rq_timeout = timeout;
26}
27EXPORT_SYMBOL_GPL(blk_queue_rq_timeout);
28
e475bba2
MP
29/**
30 * blk_set_default_limits - reset limits to default values
f740f5ca 31 * @lim: the queue_limits structure to reset
e475bba2
MP
32 *
33 * Description:
b1bd055d 34 * Returns a queue_limit struct to its default state.
e475bba2
MP
35 */
36void blk_set_default_limits(struct queue_limits *lim)
37{
8a78362c 38 lim->max_segments = BLK_MAX_SEGMENTS;
1e739730 39 lim->max_discard_segments = 1;
13f05c8d 40 lim->max_integrity_segments = 0;
e475bba2 41 lim->seg_boundary_mask = BLK_SEG_BOUNDARY_MASK;
03100aad 42 lim->virt_boundary_mask = 0;
eb28d31b 43 lim->max_segment_size = BLK_MAX_SEGMENT_SIZE;
5f009d3f
KB
44 lim->max_sectors = lim->max_hw_sectors = BLK_SAFE_MAX_SECTORS;
45 lim->max_dev_sectors = 0;
762380ad 46 lim->chunk_sectors = 0;
4363ac7c 47 lim->max_write_same_sectors = 0;
a6f0788e 48 lim->max_write_zeroes_sectors = 0;
86b37281 49 lim->max_discard_sectors = 0;
0034af03 50 lim->max_hw_discard_sectors = 0;
86b37281
MP
51 lim->discard_granularity = 0;
52 lim->discard_alignment = 0;
53 lim->discard_misaligned = 0;
e475bba2 54 lim->logical_block_size = lim->physical_block_size = lim->io_min = 512;
3a02c8e8 55 lim->bounce_pfn = (unsigned long)(BLK_BOUNCE_ANY >> PAGE_SHIFT);
e475bba2
MP
56 lim->alignment_offset = 0;
57 lim->io_opt = 0;
58 lim->misaligned = 0;
e692cb66 59 lim->cluster = 1;
797476b8 60 lim->zoned = BLK_ZONED_NONE;
e475bba2
MP
61}
62EXPORT_SYMBOL(blk_set_default_limits);
63
b1bd055d
MP
64/**
65 * blk_set_stacking_limits - set default limits for stacking devices
66 * @lim: the queue_limits structure to reset
67 *
68 * Description:
69 * Returns a queue_limit struct to its default state. Should be used
70 * by stacking drivers like DM that have no internal limits.
71 */
72void blk_set_stacking_limits(struct queue_limits *lim)
73{
74 blk_set_default_limits(lim);
75
76 /* Inherit limits from component devices */
b1bd055d 77 lim->max_segments = USHRT_MAX;
42c9cdfe 78 lim->max_discard_segments = USHRT_MAX;
b1bd055d 79 lim->max_hw_sectors = UINT_MAX;
d82ae52e 80 lim->max_segment_size = UINT_MAX;
fe86cdce 81 lim->max_sectors = UINT_MAX;
ca369d51 82 lim->max_dev_sectors = UINT_MAX;
4363ac7c 83 lim->max_write_same_sectors = UINT_MAX;
a6f0788e 84 lim->max_write_zeroes_sectors = UINT_MAX;
b1bd055d
MP
85}
86EXPORT_SYMBOL(blk_set_stacking_limits);
87
86db1e29
JA
88/**
89 * blk_queue_make_request - define an alternate make_request function for a device
90 * @q: the request queue for the device to be affected
91 * @mfn: the alternate make_request function
92 *
93 * Description:
94 * The normal way for &struct bios to be passed to a device
95 * driver is for them to be collected into requests on a request
96 * queue, and then to allow the device driver to select requests
97 * off that queue when it is ready. This works well for many block
98 * devices. However some block devices (typically virtual devices
99 * such as md or lvm) do not benefit from the processing on the
100 * request queue, and are served best by having the requests passed
101 * directly to them. This can be achieved by providing a function
102 * to blk_queue_make_request().
103 *
104 * Caveat:
105 * The driver that does this *must* be able to deal appropriately
106 * with buffers in "highmemory". This can be accomplished by either calling
d004a5e7 107 * kmap_atomic() to get a temporary kernel mapping, or by calling
86db1e29
JA
108 * blk_queue_bounce() to create a buffer in normal memory.
109 **/
6728cb0e 110void blk_queue_make_request(struct request_queue *q, make_request_fn *mfn)
86db1e29
JA
111{
112 /*
113 * set defaults
114 */
115 q->nr_requests = BLKDEV_MAX_RQ;
0e435ac2 116
86db1e29 117 q->make_request_fn = mfn;
86db1e29 118 blk_queue_dma_alignment(q, 511);
86db1e29 119
e475bba2 120 blk_set_default_limits(&q->limits);
86db1e29 121}
86db1e29
JA
122EXPORT_SYMBOL(blk_queue_make_request);
123
124/**
125 * blk_queue_bounce_limit - set bounce buffer limit for queue
cd0aca2d 126 * @q: the request queue for the device
9f7e45d8 127 * @max_addr: the maximum address the device can handle
86db1e29
JA
128 *
129 * Description:
130 * Different hardware can have different requirements as to what pages
131 * it can do I/O directly to. A low level driver can call
132 * blk_queue_bounce_limit to have lower memory pages allocated as bounce
9f7e45d8 133 * buffers for doing I/O to pages residing above @max_addr.
86db1e29 134 **/
9f7e45d8 135void blk_queue_bounce_limit(struct request_queue *q, u64 max_addr)
86db1e29 136{
9f7e45d8 137 unsigned long b_pfn = max_addr >> PAGE_SHIFT;
86db1e29
JA
138 int dma = 0;
139
140 q->bounce_gfp = GFP_NOIO;
141#if BITS_PER_LONG == 64
cd0aca2d
TH
142 /*
143 * Assume anything <= 4GB can be handled by IOMMU. Actually
144 * some IOMMUs can handle everything, but I don't know of a
145 * way to test this here.
146 */
147 if (b_pfn < (min_t(u64, 0xffffffffUL, BLK_BOUNCE_HIGH) >> PAGE_SHIFT))
86db1e29 148 dma = 1;
efb012b3 149 q->limits.bounce_pfn = max(max_low_pfn, b_pfn);
86db1e29 150#else
6728cb0e 151 if (b_pfn < blk_max_low_pfn)
86db1e29 152 dma = 1;
c49825fa 153 q->limits.bounce_pfn = b_pfn;
260a67a9 154#endif
86db1e29
JA
155 if (dma) {
156 init_emergency_isa_pool();
157 q->bounce_gfp = GFP_NOIO | GFP_DMA;
260a67a9 158 q->limits.bounce_pfn = b_pfn;
86db1e29
JA
159 }
160}
86db1e29
JA
161EXPORT_SYMBOL(blk_queue_bounce_limit);
162
163/**
ca369d51
MP
164 * blk_queue_max_hw_sectors - set max sectors for a request for this queue
165 * @q: the request queue for the device
2800aac1 166 * @max_hw_sectors: max hardware sectors in the usual 512b unit
86db1e29
JA
167 *
168 * Description:
2800aac1
MP
169 * Enables a low level driver to set a hard upper limit,
170 * max_hw_sectors, on the size of requests. max_hw_sectors is set by
4f258a46
MP
171 * the device driver based upon the capabilities of the I/O
172 * controller.
2800aac1 173 *
ca369d51
MP
174 * max_dev_sectors is a hard limit imposed by the storage device for
175 * READ/WRITE requests. It is set by the disk driver.
176 *
2800aac1
MP
177 * max_sectors is a soft limit imposed by the block layer for
178 * filesystem type requests. This value can be overridden on a
179 * per-device basis in /sys/block/<device>/queue/max_sectors_kb.
180 * The soft limit can not exceed max_hw_sectors.
86db1e29 181 **/
ca369d51 182void blk_queue_max_hw_sectors(struct request_queue *q, unsigned int max_hw_sectors)
86db1e29 183{
ca369d51
MP
184 struct queue_limits *limits = &q->limits;
185 unsigned int max_sectors;
186
09cbfeaf
KS
187 if ((max_hw_sectors << 9) < PAGE_SIZE) {
188 max_hw_sectors = 1 << (PAGE_SHIFT - 9);
24c03d47 189 printk(KERN_INFO "%s: set to minimum %d\n",
2800aac1 190 __func__, max_hw_sectors);
86db1e29
JA
191 }
192
30e2bc08 193 limits->max_hw_sectors = max_hw_sectors;
ca369d51
MP
194 max_sectors = min_not_zero(max_hw_sectors, limits->max_dev_sectors);
195 max_sectors = min_t(unsigned int, max_sectors, BLK_DEF_MAX_SECTORS);
196 limits->max_sectors = max_sectors;
dc3b17cc 197 q->backing_dev_info->io_pages = max_sectors >> (PAGE_SHIFT - 9);
86db1e29 198}
086fa5ff 199EXPORT_SYMBOL(blk_queue_max_hw_sectors);
86db1e29 200
762380ad
JA
201/**
202 * blk_queue_chunk_sectors - set size of the chunk for this queue
203 * @q: the request queue for the device
204 * @chunk_sectors: chunk sectors in the usual 512b unit
205 *
206 * Description:
207 * If a driver doesn't want IOs to cross a given chunk size, it can set
208 * this limit and prevent merging across chunks. Note that the chunk size
58a4915a
JA
209 * must currently be a power-of-2 in sectors. Also note that the block
210 * layer must accept a page worth of data at any offset. So if the
211 * crossing of chunks is a hard limitation in the driver, it must still be
212 * prepared to split single page bios.
762380ad
JA
213 **/
214void blk_queue_chunk_sectors(struct request_queue *q, unsigned int chunk_sectors)
215{
216 BUG_ON(!is_power_of_2(chunk_sectors));
217 q->limits.chunk_sectors = chunk_sectors;
218}
219EXPORT_SYMBOL(blk_queue_chunk_sectors);
220
67efc925
CH
221/**
222 * blk_queue_max_discard_sectors - set max sectors for a single discard
223 * @q: the request queue for the device
c7ebf065 224 * @max_discard_sectors: maximum number of sectors to discard
67efc925
CH
225 **/
226void blk_queue_max_discard_sectors(struct request_queue *q,
227 unsigned int max_discard_sectors)
228{
0034af03 229 q->limits.max_hw_discard_sectors = max_discard_sectors;
67efc925
CH
230 q->limits.max_discard_sectors = max_discard_sectors;
231}
232EXPORT_SYMBOL(blk_queue_max_discard_sectors);
233
4363ac7c
MP
234/**
235 * blk_queue_max_write_same_sectors - set max sectors for a single write same
236 * @q: the request queue for the device
237 * @max_write_same_sectors: maximum number of sectors to write per command
238 **/
239void blk_queue_max_write_same_sectors(struct request_queue *q,
240 unsigned int max_write_same_sectors)
241{
242 q->limits.max_write_same_sectors = max_write_same_sectors;
243}
244EXPORT_SYMBOL(blk_queue_max_write_same_sectors);
245
a6f0788e
CK
246/**
247 * blk_queue_max_write_zeroes_sectors - set max sectors for a single
248 * write zeroes
249 * @q: the request queue for the device
250 * @max_write_zeroes_sectors: maximum number of sectors to write per command
251 **/
252void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
253 unsigned int max_write_zeroes_sectors)
254{
255 q->limits.max_write_zeroes_sectors = max_write_zeroes_sectors;
256}
257EXPORT_SYMBOL(blk_queue_max_write_zeroes_sectors);
258
86db1e29 259/**
8a78362c 260 * blk_queue_max_segments - set max hw segments for a request for this queue
86db1e29
JA
261 * @q: the request queue for the device
262 * @max_segments: max number of segments
263 *
264 * Description:
265 * Enables a low level driver to set an upper limit on the number of
8a78362c 266 * hw data segments in a request.
86db1e29 267 **/
8a78362c 268void blk_queue_max_segments(struct request_queue *q, unsigned short max_segments)
86db1e29
JA
269{
270 if (!max_segments) {
271 max_segments = 1;
24c03d47
HH
272 printk(KERN_INFO "%s: set to minimum %d\n",
273 __func__, max_segments);
86db1e29
JA
274 }
275
8a78362c 276 q->limits.max_segments = max_segments;
86db1e29 277}
8a78362c 278EXPORT_SYMBOL(blk_queue_max_segments);
86db1e29 279
1e739730
CH
280/**
281 * blk_queue_max_discard_segments - set max segments for discard requests
282 * @q: the request queue for the device
283 * @max_segments: max number of segments
284 *
285 * Description:
286 * Enables a low level driver to set an upper limit on the number of
287 * segments in a discard request.
288 **/
289void blk_queue_max_discard_segments(struct request_queue *q,
290 unsigned short max_segments)
291{
292 q->limits.max_discard_segments = max_segments;
293}
294EXPORT_SYMBOL_GPL(blk_queue_max_discard_segments);
295
86db1e29
JA
296/**
297 * blk_queue_max_segment_size - set max segment size for blk_rq_map_sg
298 * @q: the request queue for the device
299 * @max_size: max size of segment in bytes
300 *
301 * Description:
302 * Enables a low level driver to set an upper limit on the size of a
303 * coalesced segment
304 **/
305void blk_queue_max_segment_size(struct request_queue *q, unsigned int max_size)
306{
09cbfeaf
KS
307 if (max_size < PAGE_SIZE) {
308 max_size = PAGE_SIZE;
24c03d47
HH
309 printk(KERN_INFO "%s: set to minimum %d\n",
310 __func__, max_size);
86db1e29
JA
311 }
312
025146e1 313 q->limits.max_segment_size = max_size;
86db1e29 314}
86db1e29
JA
315EXPORT_SYMBOL(blk_queue_max_segment_size);
316
317/**
e1defc4f 318 * blk_queue_logical_block_size - set logical block size for the queue
86db1e29 319 * @q: the request queue for the device
e1defc4f 320 * @size: the logical block size, in bytes
86db1e29
JA
321 *
322 * Description:
e1defc4f
MP
323 * This should be set to the lowest possible block size that the
324 * storage device can address. The default of 512 covers most
325 * hardware.
86db1e29 326 **/
e1defc4f 327void blk_queue_logical_block_size(struct request_queue *q, unsigned short size)
86db1e29 328{
025146e1 329 q->limits.logical_block_size = size;
c72758f3
MP
330
331 if (q->limits.physical_block_size < size)
332 q->limits.physical_block_size = size;
333
334 if (q->limits.io_min < q->limits.physical_block_size)
335 q->limits.io_min = q->limits.physical_block_size;
86db1e29 336}
e1defc4f 337EXPORT_SYMBOL(blk_queue_logical_block_size);
86db1e29 338
c72758f3
MP
339/**
340 * blk_queue_physical_block_size - set physical block size for the queue
341 * @q: the request queue for the device
342 * @size: the physical block size, in bytes
343 *
344 * Description:
345 * This should be set to the lowest possible sector size that the
346 * hardware can operate on without reverting to read-modify-write
347 * operations.
348 */
892b6f90 349void blk_queue_physical_block_size(struct request_queue *q, unsigned int size)
c72758f3
MP
350{
351 q->limits.physical_block_size = size;
352
353 if (q->limits.physical_block_size < q->limits.logical_block_size)
354 q->limits.physical_block_size = q->limits.logical_block_size;
355
356 if (q->limits.io_min < q->limits.physical_block_size)
357 q->limits.io_min = q->limits.physical_block_size;
358}
359EXPORT_SYMBOL(blk_queue_physical_block_size);
360
361/**
362 * blk_queue_alignment_offset - set physical block alignment offset
363 * @q: the request queue for the device
8ebf9756 364 * @offset: alignment offset in bytes
c72758f3
MP
365 *
366 * Description:
367 * Some devices are naturally misaligned to compensate for things like
368 * the legacy DOS partition table 63-sector offset. Low-level drivers
369 * should call this function for devices whose first sector is not
370 * naturally aligned.
371 */
372void blk_queue_alignment_offset(struct request_queue *q, unsigned int offset)
373{
374 q->limits.alignment_offset =
375 offset & (q->limits.physical_block_size - 1);
376 q->limits.misaligned = 0;
377}
378EXPORT_SYMBOL(blk_queue_alignment_offset);
379
7c958e32
MP
380/**
381 * blk_limits_io_min - set minimum request size for a device
382 * @limits: the queue limits
383 * @min: smallest I/O size in bytes
384 *
385 * Description:
386 * Some devices have an internal block size bigger than the reported
387 * hardware sector size. This function can be used to signal the
388 * smallest I/O the device can perform without incurring a performance
389 * penalty.
390 */
391void blk_limits_io_min(struct queue_limits *limits, unsigned int min)
392{
393 limits->io_min = min;
394
395 if (limits->io_min < limits->logical_block_size)
396 limits->io_min = limits->logical_block_size;
397
398 if (limits->io_min < limits->physical_block_size)
399 limits->io_min = limits->physical_block_size;
400}
401EXPORT_SYMBOL(blk_limits_io_min);
402
c72758f3
MP
403/**
404 * blk_queue_io_min - set minimum request size for the queue
405 * @q: the request queue for the device
8ebf9756 406 * @min: smallest I/O size in bytes
c72758f3
MP
407 *
408 * Description:
7e5f5fb0
MP
409 * Storage devices may report a granularity or preferred minimum I/O
410 * size which is the smallest request the device can perform without
411 * incurring a performance penalty. For disk drives this is often the
412 * physical block size. For RAID arrays it is often the stripe chunk
413 * size. A properly aligned multiple of minimum_io_size is the
414 * preferred request size for workloads where a high number of I/O
415 * operations is desired.
c72758f3
MP
416 */
417void blk_queue_io_min(struct request_queue *q, unsigned int min)
418{
7c958e32 419 blk_limits_io_min(&q->limits, min);
c72758f3
MP
420}
421EXPORT_SYMBOL(blk_queue_io_min);
422
3c5820c7
MP
423/**
424 * blk_limits_io_opt - set optimal request size for a device
425 * @limits: the queue limits
426 * @opt: smallest I/O size in bytes
427 *
428 * Description:
429 * Storage devices may report an optimal I/O size, which is the
430 * device's preferred unit for sustained I/O. This is rarely reported
431 * for disk drives. For RAID arrays it is usually the stripe width or
432 * the internal track size. A properly aligned multiple of
433 * optimal_io_size is the preferred request size for workloads where
434 * sustained throughput is desired.
435 */
436void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt)
437{
438 limits->io_opt = opt;
439}
440EXPORT_SYMBOL(blk_limits_io_opt);
441
c72758f3
MP
442/**
443 * blk_queue_io_opt - set optimal request size for the queue
444 * @q: the request queue for the device
8ebf9756 445 * @opt: optimal request size in bytes
c72758f3
MP
446 *
447 * Description:
7e5f5fb0
MP
448 * Storage devices may report an optimal I/O size, which is the
449 * device's preferred unit for sustained I/O. This is rarely reported
450 * for disk drives. For RAID arrays it is usually the stripe width or
451 * the internal track size. A properly aligned multiple of
452 * optimal_io_size is the preferred request size for workloads where
453 * sustained throughput is desired.
c72758f3
MP
454 */
455void blk_queue_io_opt(struct request_queue *q, unsigned int opt)
456{
3c5820c7 457 blk_limits_io_opt(&q->limits, opt);
c72758f3
MP
458}
459EXPORT_SYMBOL(blk_queue_io_opt);
460
86db1e29
JA
461/**
462 * blk_queue_stack_limits - inherit underlying queue limits for stacked drivers
463 * @t: the stacking driver (top)
464 * @b: the underlying device (bottom)
465 **/
466void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b)
467{
fef24667 468 blk_stack_limits(&t->limits, &b->limits, 0);
86db1e29 469}
86db1e29
JA
470EXPORT_SYMBOL(blk_queue_stack_limits);
471
c72758f3
MP
472/**
473 * blk_stack_limits - adjust queue_limits for stacked devices
81744ee4
MP
474 * @t: the stacking driver limits (top device)
475 * @b: the underlying queue limits (bottom, component device)
e03a72e1 476 * @start: first data sector within component device
c72758f3
MP
477 *
478 * Description:
81744ee4
MP
479 * This function is used by stacking drivers like MD and DM to ensure
480 * that all component devices have compatible block sizes and
481 * alignments. The stacking driver must provide a queue_limits
482 * struct (top) and then iteratively call the stacking function for
483 * all component (bottom) devices. The stacking function will
484 * attempt to combine the values and ensure proper alignment.
485 *
486 * Returns 0 if the top and bottom queue_limits are compatible. The
487 * top device's block sizes and alignment offsets may be adjusted to
488 * ensure alignment with the bottom device. If no compatible sizes
489 * and alignments exist, -1 is returned and the resulting top
490 * queue_limits will have the misaligned flag set to indicate that
491 * the alignment_offset is undefined.
c72758f3
MP
492 */
493int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
e03a72e1 494 sector_t start)
c72758f3 495{
e03a72e1 496 unsigned int top, bottom, alignment, ret = 0;
86b37281 497
c72758f3
MP
498 t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors);
499 t->max_hw_sectors = min_not_zero(t->max_hw_sectors, b->max_hw_sectors);
ca369d51 500 t->max_dev_sectors = min_not_zero(t->max_dev_sectors, b->max_dev_sectors);
4363ac7c
MP
501 t->max_write_same_sectors = min(t->max_write_same_sectors,
502 b->max_write_same_sectors);
a6f0788e
CK
503 t->max_write_zeroes_sectors = min(t->max_write_zeroes_sectors,
504 b->max_write_zeroes_sectors);
77634f33 505 t->bounce_pfn = min_not_zero(t->bounce_pfn, b->bounce_pfn);
c72758f3
MP
506
507 t->seg_boundary_mask = min_not_zero(t->seg_boundary_mask,
508 b->seg_boundary_mask);
03100aad
KB
509 t->virt_boundary_mask = min_not_zero(t->virt_boundary_mask,
510 b->virt_boundary_mask);
c72758f3 511
8a78362c 512 t->max_segments = min_not_zero(t->max_segments, b->max_segments);
1e739730
CH
513 t->max_discard_segments = min_not_zero(t->max_discard_segments,
514 b->max_discard_segments);
13f05c8d
MP
515 t->max_integrity_segments = min_not_zero(t->max_integrity_segments,
516 b->max_integrity_segments);
c72758f3
MP
517
518 t->max_segment_size = min_not_zero(t->max_segment_size,
519 b->max_segment_size);
520
fe0b393f
MP
521 t->misaligned |= b->misaligned;
522
e03a72e1 523 alignment = queue_limit_alignment_offset(b, start);
9504e086 524
81744ee4
MP
525 /* Bottom device has different alignment. Check that it is
526 * compatible with the current top alignment.
527 */
9504e086
MP
528 if (t->alignment_offset != alignment) {
529
530 top = max(t->physical_block_size, t->io_min)
531 + t->alignment_offset;
81744ee4 532 bottom = max(b->physical_block_size, b->io_min) + alignment;
9504e086 533
81744ee4 534 /* Verify that top and bottom intervals line up */
b8839b8c 535 if (max(top, bottom) % min(top, bottom)) {
9504e086 536 t->misaligned = 1;
fe0b393f
MP
537 ret = -1;
538 }
9504e086
MP
539 }
540
c72758f3
MP
541 t->logical_block_size = max(t->logical_block_size,
542 b->logical_block_size);
543
544 t->physical_block_size = max(t->physical_block_size,
545 b->physical_block_size);
546
547 t->io_min = max(t->io_min, b->io_min);
e9637415 548 t->io_opt = lcm_not_zero(t->io_opt, b->io_opt);
9504e086 549
e692cb66 550 t->cluster &= b->cluster;
c72758f3 551
81744ee4 552 /* Physical block size a multiple of the logical block size? */
9504e086
MP
553 if (t->physical_block_size & (t->logical_block_size - 1)) {
554 t->physical_block_size = t->logical_block_size;
c72758f3 555 t->misaligned = 1;
fe0b393f 556 ret = -1;
86b37281
MP
557 }
558
81744ee4 559 /* Minimum I/O a multiple of the physical block size? */
9504e086
MP
560 if (t->io_min & (t->physical_block_size - 1)) {
561 t->io_min = t->physical_block_size;
562 t->misaligned = 1;
fe0b393f 563 ret = -1;
c72758f3
MP
564 }
565
81744ee4 566 /* Optimal I/O a multiple of the physical block size? */
9504e086
MP
567 if (t->io_opt & (t->physical_block_size - 1)) {
568 t->io_opt = 0;
569 t->misaligned = 1;
fe0b393f 570 ret = -1;
9504e086 571 }
c72758f3 572
c78afc62
KO
573 t->raid_partial_stripes_expensive =
574 max(t->raid_partial_stripes_expensive,
575 b->raid_partial_stripes_expensive);
576
81744ee4 577 /* Find lowest common alignment_offset */
e9637415 578 t->alignment_offset = lcm_not_zero(t->alignment_offset, alignment)
b8839b8c 579 % max(t->physical_block_size, t->io_min);
86b37281 580
81744ee4 581 /* Verify that new alignment_offset is on a logical block boundary */
fe0b393f 582 if (t->alignment_offset & (t->logical_block_size - 1)) {
c72758f3 583 t->misaligned = 1;
fe0b393f
MP
584 ret = -1;
585 }
c72758f3 586
9504e086
MP
587 /* Discard alignment and granularity */
588 if (b->discard_granularity) {
e03a72e1 589 alignment = queue_limit_discard_alignment(b, start);
9504e086
MP
590
591 if (t->discard_granularity != 0 &&
592 t->discard_alignment != alignment) {
593 top = t->discard_granularity + t->discard_alignment;
594 bottom = b->discard_granularity + alignment;
70dd5bf3 595
9504e086 596 /* Verify that top and bottom intervals line up */
8dd2cb7e 597 if ((max(top, bottom) % min(top, bottom)) != 0)
9504e086
MP
598 t->discard_misaligned = 1;
599 }
600
81744ee4
MP
601 t->max_discard_sectors = min_not_zero(t->max_discard_sectors,
602 b->max_discard_sectors);
0034af03
JA
603 t->max_hw_discard_sectors = min_not_zero(t->max_hw_discard_sectors,
604 b->max_hw_discard_sectors);
9504e086
MP
605 t->discard_granularity = max(t->discard_granularity,
606 b->discard_granularity);
e9637415 607 t->discard_alignment = lcm_not_zero(t->discard_alignment, alignment) %
8dd2cb7e 608 t->discard_granularity;
9504e086 609 }
70dd5bf3 610
987b3b26
HR
611 if (b->chunk_sectors)
612 t->chunk_sectors = min_not_zero(t->chunk_sectors,
613 b->chunk_sectors);
614
fe0b393f 615 return ret;
c72758f3 616}
5d85d324 617EXPORT_SYMBOL(blk_stack_limits);
c72758f3 618
17be8c24
MP
619/**
620 * bdev_stack_limits - adjust queue limits for stacked drivers
621 * @t: the stacking driver limits (top device)
622 * @bdev: the component block_device (bottom)
623 * @start: first data sector within component device
624 *
625 * Description:
626 * Merges queue limits for a top device and a block_device. Returns
627 * 0 if alignment didn't change. Returns -1 if adding the bottom
628 * device caused misalignment.
629 */
630int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
631 sector_t start)
632{
633 struct request_queue *bq = bdev_get_queue(bdev);
634
635 start += get_start_sect(bdev);
636
e03a72e1 637 return blk_stack_limits(t, &bq->limits, start);
17be8c24
MP
638}
639EXPORT_SYMBOL(bdev_stack_limits);
640
c72758f3
MP
641/**
642 * disk_stack_limits - adjust queue limits for stacked drivers
77634f33 643 * @disk: MD/DM gendisk (top)
c72758f3
MP
644 * @bdev: the underlying block device (bottom)
645 * @offset: offset to beginning of data within component device
646 *
647 * Description:
e03a72e1
MP
648 * Merges the limits for a top level gendisk and a bottom level
649 * block_device.
c72758f3
MP
650 */
651void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
652 sector_t offset)
653{
654 struct request_queue *t = disk->queue;
c72758f3 655
e03a72e1 656 if (bdev_stack_limits(&t->limits, bdev, offset >> 9) < 0) {
c72758f3
MP
657 char top[BDEVNAME_SIZE], bottom[BDEVNAME_SIZE];
658
659 disk_name(disk, 0, top);
660 bdevname(bdev, bottom);
661
662 printk(KERN_NOTICE "%s: Warning: Device %s is misaligned\n",
663 top, bottom);
664 }
c72758f3
MP
665}
666EXPORT_SYMBOL(disk_stack_limits);
667
e3790c7d
TH
668/**
669 * blk_queue_dma_pad - set pad mask
670 * @q: the request queue for the device
671 * @mask: pad mask
672 *
27f8221a 673 * Set dma pad mask.
e3790c7d 674 *
27f8221a
FT
675 * Appending pad buffer to a request modifies the last entry of a
676 * scatter list such that it includes the pad buffer.
e3790c7d
TH
677 **/
678void blk_queue_dma_pad(struct request_queue *q, unsigned int mask)
679{
680 q->dma_pad_mask = mask;
681}
682EXPORT_SYMBOL(blk_queue_dma_pad);
683
27f8221a
FT
684/**
685 * blk_queue_update_dma_pad - update pad mask
686 * @q: the request queue for the device
687 * @mask: pad mask
688 *
689 * Update dma pad mask.
690 *
691 * Appending pad buffer to a request modifies the last entry of a
692 * scatter list such that it includes the pad buffer.
693 **/
694void blk_queue_update_dma_pad(struct request_queue *q, unsigned int mask)
695{
696 if (mask > q->dma_pad_mask)
697 q->dma_pad_mask = mask;
698}
699EXPORT_SYMBOL(blk_queue_update_dma_pad);
700
86db1e29
JA
701/**
702 * blk_queue_dma_drain - Set up a drain buffer for excess dma.
86db1e29 703 * @q: the request queue for the device
2fb98e84 704 * @dma_drain_needed: fn which returns non-zero if drain is necessary
86db1e29
JA
705 * @buf: physically contiguous buffer
706 * @size: size of the buffer in bytes
707 *
708 * Some devices have excess DMA problems and can't simply discard (or
709 * zero fill) the unwanted piece of the transfer. They have to have a
710 * real area of memory to transfer it into. The use case for this is
711 * ATAPI devices in DMA mode. If the packet command causes a transfer
712 * bigger than the transfer size some HBAs will lock up if there
713 * aren't DMA elements to contain the excess transfer. What this API
714 * does is adjust the queue so that the buf is always appended
715 * silently to the scatterlist.
716 *
8a78362c
MP
717 * Note: This routine adjusts max_hw_segments to make room for appending
718 * the drain buffer. If you call blk_queue_max_segments() after calling
719 * this routine, you must set the limit to one fewer than your device
720 * can support otherwise there won't be room for the drain buffer.
86db1e29 721 */
448da4d2 722int blk_queue_dma_drain(struct request_queue *q,
2fb98e84
TH
723 dma_drain_needed_fn *dma_drain_needed,
724 void *buf, unsigned int size)
86db1e29 725{
8a78362c 726 if (queue_max_segments(q) < 2)
86db1e29
JA
727 return -EINVAL;
728 /* make room for appending the drain */
8a78362c 729 blk_queue_max_segments(q, queue_max_segments(q) - 1);
2fb98e84 730 q->dma_drain_needed = dma_drain_needed;
86db1e29
JA
731 q->dma_drain_buffer = buf;
732 q->dma_drain_size = size;
733
734 return 0;
735}
86db1e29
JA
736EXPORT_SYMBOL_GPL(blk_queue_dma_drain);
737
738/**
739 * blk_queue_segment_boundary - set boundary rules for segment merging
740 * @q: the request queue for the device
741 * @mask: the memory boundary mask
742 **/
743void blk_queue_segment_boundary(struct request_queue *q, unsigned long mask)
744{
09cbfeaf
KS
745 if (mask < PAGE_SIZE - 1) {
746 mask = PAGE_SIZE - 1;
24c03d47
HH
747 printk(KERN_INFO "%s: set to minimum %lx\n",
748 __func__, mask);
86db1e29
JA
749 }
750
025146e1 751 q->limits.seg_boundary_mask = mask;
86db1e29 752}
86db1e29
JA
753EXPORT_SYMBOL(blk_queue_segment_boundary);
754
03100aad
KB
755/**
756 * blk_queue_virt_boundary - set boundary rules for bio merging
757 * @q: the request queue for the device
758 * @mask: the memory boundary mask
759 **/
760void blk_queue_virt_boundary(struct request_queue *q, unsigned long mask)
761{
762 q->limits.virt_boundary_mask = mask;
763}
764EXPORT_SYMBOL(blk_queue_virt_boundary);
765
86db1e29
JA
766/**
767 * blk_queue_dma_alignment - set dma length and memory alignment
768 * @q: the request queue for the device
769 * @mask: alignment mask
770 *
771 * description:
710027a4 772 * set required memory and length alignment for direct dma transactions.
8feb4d20 773 * this is used when building direct io requests for the queue.
86db1e29
JA
774 *
775 **/
776void blk_queue_dma_alignment(struct request_queue *q, int mask)
777{
778 q->dma_alignment = mask;
779}
86db1e29
JA
780EXPORT_SYMBOL(blk_queue_dma_alignment);
781
782/**
783 * blk_queue_update_dma_alignment - update dma length and memory alignment
784 * @q: the request queue for the device
785 * @mask: alignment mask
786 *
787 * description:
710027a4 788 * update required memory and length alignment for direct dma transactions.
86db1e29
JA
789 * If the requested alignment is larger than the current alignment, then
790 * the current queue alignment is updated to the new value, otherwise it
791 * is left alone. The design of this is to allow multiple objects
792 * (driver, device, transport etc) to set their respective
793 * alignments without having them interfere.
794 *
795 **/
796void blk_queue_update_dma_alignment(struct request_queue *q, int mask)
797{
798 BUG_ON(mask > PAGE_SIZE);
799
800 if (mask > q->dma_alignment)
801 q->dma_alignment = mask;
802}
86db1e29
JA
803EXPORT_SYMBOL(blk_queue_update_dma_alignment);
804
f3876930 805void blk_queue_flush_queueable(struct request_queue *q, bool queueable)
806{
c888a8f9 807 if (queueable)
8814ce8a 808 blk_queue_flag_clear(QUEUE_FLAG_FLUSH_NQ, q);
c888a8f9 809 else
8814ce8a 810 blk_queue_flag_set(QUEUE_FLAG_FLUSH_NQ, q);
f3876930 811}
812EXPORT_SYMBOL_GPL(blk_queue_flush_queueable);
813
d278d4a8
JA
814/**
815 * blk_set_queue_depth - tell the block layer about the device queue depth
816 * @q: the request queue for the device
817 * @depth: queue depth
818 *
819 */
820void blk_set_queue_depth(struct request_queue *q, unsigned int depth)
821{
822 q->queue_depth = depth;
a7905043 823 wbt_set_queue_depth(q, depth);
d278d4a8
JA
824}
825EXPORT_SYMBOL(blk_set_queue_depth);
826
93e9d8e8
JA
827/**
828 * blk_queue_write_cache - configure queue's write cache
829 * @q: the request queue for the device
830 * @wc: write back cache on or off
831 * @fua: device supports FUA writes, if true
832 *
833 * Tell the block layer about the write cache of @q.
834 */
835void blk_queue_write_cache(struct request_queue *q, bool wc, bool fua)
836{
837 spin_lock_irq(q->queue_lock);
c888a8f9 838 if (wc)
93e9d8e8 839 queue_flag_set(QUEUE_FLAG_WC, q);
c888a8f9 840 else
93e9d8e8 841 queue_flag_clear(QUEUE_FLAG_WC, q);
c888a8f9 842 if (fua)
93e9d8e8 843 queue_flag_set(QUEUE_FLAG_FUA, q);
c888a8f9 844 else
93e9d8e8
JA
845 queue_flag_clear(QUEUE_FLAG_FUA, q);
846 spin_unlock_irq(q->queue_lock);
87760e5e 847
a7905043 848 wbt_set_write_cache(q, test_bit(QUEUE_FLAG_WC, &q->queue_flags));
93e9d8e8
JA
849}
850EXPORT_SYMBOL_GPL(blk_queue_write_cache);
851
aeb3d3a8 852static int __init blk_settings_init(void)
86db1e29
JA
853{
854 blk_max_low_pfn = max_low_pfn - 1;
855 blk_max_pfn = max_pfn - 1;
856 return 0;
857}
858subsys_initcall(blk_settings_init);