Merge tag 'metag-for-v4.9-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/jhoga...
[linux-2.6-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>
9#include <linux/bootmem.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"
16
6728cb0e 17unsigned long blk_max_low_pfn;
86db1e29 18EXPORT_SYMBOL(blk_max_low_pfn);
6728cb0e
JA
19
20unsigned long blk_max_pfn;
86db1e29
JA
21
22/**
23 * blk_queue_prep_rq - set a prepare_request function for queue
24 * @q: queue
25 * @pfn: prepare_request function
26 *
27 * It's possible for a queue to register a prepare_request callback which
28 * is invoked before the request is handed to the request_fn. The goal of
29 * the function is to prepare a request for I/O, it can be used to build a
30 * cdb from the request data for instance.
31 *
32 */
33void blk_queue_prep_rq(struct request_queue *q, prep_rq_fn *pfn)
34{
35 q->prep_rq_fn = pfn;
36}
86db1e29
JA
37EXPORT_SYMBOL(blk_queue_prep_rq);
38
28018c24
JB
39/**
40 * blk_queue_unprep_rq - set an unprepare_request function for queue
41 * @q: queue
42 * @ufn: unprepare_request function
43 *
44 * It's possible for a queue to register an unprepare_request callback
45 * which is invoked before the request is finally completed. The goal
46 * of the function is to deallocate any data that was allocated in the
47 * prepare_request callback.
48 *
49 */
50void blk_queue_unprep_rq(struct request_queue *q, unprep_rq_fn *ufn)
51{
52 q->unprep_rq_fn = ufn;
53}
54EXPORT_SYMBOL(blk_queue_unprep_rq);
55
86db1e29
JA
56void blk_queue_softirq_done(struct request_queue *q, softirq_done_fn *fn)
57{
58 q->softirq_done_fn = fn;
59}
86db1e29
JA
60EXPORT_SYMBOL(blk_queue_softirq_done);
61
242f9dcb
JA
62void blk_queue_rq_timeout(struct request_queue *q, unsigned int timeout)
63{
64 q->rq_timeout = timeout;
65}
66EXPORT_SYMBOL_GPL(blk_queue_rq_timeout);
67
68void blk_queue_rq_timed_out(struct request_queue *q, rq_timed_out_fn *fn)
69{
70 q->rq_timed_out_fn = fn;
71}
72EXPORT_SYMBOL_GPL(blk_queue_rq_timed_out);
73
ef9e3fac
KU
74void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn)
75{
76 q->lld_busy_fn = fn;
77}
78EXPORT_SYMBOL_GPL(blk_queue_lld_busy);
79
e475bba2
MP
80/**
81 * blk_set_default_limits - reset limits to default values
f740f5ca 82 * @lim: the queue_limits structure to reset
e475bba2
MP
83 *
84 * Description:
b1bd055d 85 * Returns a queue_limit struct to its default state.
e475bba2
MP
86 */
87void blk_set_default_limits(struct queue_limits *lim)
88{
8a78362c 89 lim->max_segments = BLK_MAX_SEGMENTS;
13f05c8d 90 lim->max_integrity_segments = 0;
e475bba2 91 lim->seg_boundary_mask = BLK_SEG_BOUNDARY_MASK;
03100aad 92 lim->virt_boundary_mask = 0;
eb28d31b 93 lim->max_segment_size = BLK_MAX_SEGMENT_SIZE;
5f009d3f
KB
94 lim->max_sectors = lim->max_hw_sectors = BLK_SAFE_MAX_SECTORS;
95 lim->max_dev_sectors = 0;
762380ad 96 lim->chunk_sectors = 0;
4363ac7c 97 lim->max_write_same_sectors = 0;
86b37281 98 lim->max_discard_sectors = 0;
0034af03 99 lim->max_hw_discard_sectors = 0;
86b37281
MP
100 lim->discard_granularity = 0;
101 lim->discard_alignment = 0;
102 lim->discard_misaligned = 0;
b1bd055d 103 lim->discard_zeroes_data = 0;
e475bba2 104 lim->logical_block_size = lim->physical_block_size = lim->io_min = 512;
3a02c8e8 105 lim->bounce_pfn = (unsigned long)(BLK_BOUNCE_ANY >> PAGE_SHIFT);
e475bba2
MP
106 lim->alignment_offset = 0;
107 lim->io_opt = 0;
108 lim->misaligned = 0;
e692cb66 109 lim->cluster = 1;
e475bba2
MP
110}
111EXPORT_SYMBOL(blk_set_default_limits);
112
b1bd055d
MP
113/**
114 * blk_set_stacking_limits - set default limits for stacking devices
115 * @lim: the queue_limits structure to reset
116 *
117 * Description:
118 * Returns a queue_limit struct to its default state. Should be used
119 * by stacking drivers like DM that have no internal limits.
120 */
121void blk_set_stacking_limits(struct queue_limits *lim)
122{
123 blk_set_default_limits(lim);
124
125 /* Inherit limits from component devices */
126 lim->discard_zeroes_data = 1;
127 lim->max_segments = USHRT_MAX;
128 lim->max_hw_sectors = UINT_MAX;
d82ae52e 129 lim->max_segment_size = UINT_MAX;
fe86cdce 130 lim->max_sectors = UINT_MAX;
ca369d51 131 lim->max_dev_sectors = UINT_MAX;
4363ac7c 132 lim->max_write_same_sectors = UINT_MAX;
b1bd055d
MP
133}
134EXPORT_SYMBOL(blk_set_stacking_limits);
135
86db1e29
JA
136/**
137 * blk_queue_make_request - define an alternate make_request function for a device
138 * @q: the request queue for the device to be affected
139 * @mfn: the alternate make_request function
140 *
141 * Description:
142 * The normal way for &struct bios to be passed to a device
143 * driver is for them to be collected into requests on a request
144 * queue, and then to allow the device driver to select requests
145 * off that queue when it is ready. This works well for many block
146 * devices. However some block devices (typically virtual devices
147 * such as md or lvm) do not benefit from the processing on the
148 * request queue, and are served best by having the requests passed
149 * directly to them. This can be achieved by providing a function
150 * to blk_queue_make_request().
151 *
152 * Caveat:
153 * The driver that does this *must* be able to deal appropriately
154 * with buffers in "highmemory". This can be accomplished by either calling
155 * __bio_kmap_atomic() to get a temporary kernel mapping, or by calling
156 * blk_queue_bounce() to create a buffer in normal memory.
157 **/
6728cb0e 158void blk_queue_make_request(struct request_queue *q, make_request_fn *mfn)
86db1e29
JA
159{
160 /*
161 * set defaults
162 */
163 q->nr_requests = BLKDEV_MAX_RQ;
0e435ac2 164
86db1e29 165 q->make_request_fn = mfn;
86db1e29
JA
166 blk_queue_dma_alignment(q, 511);
167 blk_queue_congestion_threshold(q);
168 q->nr_batching = BLK_BATCH_REQ;
169
e475bba2
MP
170 blk_set_default_limits(&q->limits);
171
86db1e29
JA
172 /*
173 * by default assume old behaviour and bounce for any highmem page
174 */
175 blk_queue_bounce_limit(q, BLK_BOUNCE_HIGH);
176}
86db1e29
JA
177EXPORT_SYMBOL(blk_queue_make_request);
178
179/**
180 * blk_queue_bounce_limit - set bounce buffer limit for queue
cd0aca2d 181 * @q: the request queue for the device
9f7e45d8 182 * @max_addr: the maximum address the device can handle
86db1e29
JA
183 *
184 * Description:
185 * Different hardware can have different requirements as to what pages
186 * it can do I/O directly to. A low level driver can call
187 * blk_queue_bounce_limit to have lower memory pages allocated as bounce
9f7e45d8 188 * buffers for doing I/O to pages residing above @max_addr.
86db1e29 189 **/
9f7e45d8 190void blk_queue_bounce_limit(struct request_queue *q, u64 max_addr)
86db1e29 191{
9f7e45d8 192 unsigned long b_pfn = max_addr >> PAGE_SHIFT;
86db1e29
JA
193 int dma = 0;
194
195 q->bounce_gfp = GFP_NOIO;
196#if BITS_PER_LONG == 64
cd0aca2d
TH
197 /*
198 * Assume anything <= 4GB can be handled by IOMMU. Actually
199 * some IOMMUs can handle everything, but I don't know of a
200 * way to test this here.
201 */
202 if (b_pfn < (min_t(u64, 0xffffffffUL, BLK_BOUNCE_HIGH) >> PAGE_SHIFT))
86db1e29 203 dma = 1;
efb012b3 204 q->limits.bounce_pfn = max(max_low_pfn, b_pfn);
86db1e29 205#else
6728cb0e 206 if (b_pfn < blk_max_low_pfn)
86db1e29 207 dma = 1;
c49825fa 208 q->limits.bounce_pfn = b_pfn;
260a67a9 209#endif
86db1e29
JA
210 if (dma) {
211 init_emergency_isa_pool();
212 q->bounce_gfp = GFP_NOIO | GFP_DMA;
260a67a9 213 q->limits.bounce_pfn = b_pfn;
86db1e29
JA
214 }
215}
86db1e29
JA
216EXPORT_SYMBOL(blk_queue_bounce_limit);
217
218/**
ca369d51
MP
219 * blk_queue_max_hw_sectors - set max sectors for a request for this queue
220 * @q: the request queue for the device
2800aac1 221 * @max_hw_sectors: max hardware sectors in the usual 512b unit
86db1e29
JA
222 *
223 * Description:
2800aac1
MP
224 * Enables a low level driver to set a hard upper limit,
225 * max_hw_sectors, on the size of requests. max_hw_sectors is set by
4f258a46
MP
226 * the device driver based upon the capabilities of the I/O
227 * controller.
2800aac1 228 *
ca369d51
MP
229 * max_dev_sectors is a hard limit imposed by the storage device for
230 * READ/WRITE requests. It is set by the disk driver.
231 *
2800aac1
MP
232 * max_sectors is a soft limit imposed by the block layer for
233 * filesystem type requests. This value can be overridden on a
234 * per-device basis in /sys/block/<device>/queue/max_sectors_kb.
235 * The soft limit can not exceed max_hw_sectors.
86db1e29 236 **/
ca369d51 237void blk_queue_max_hw_sectors(struct request_queue *q, unsigned int max_hw_sectors)
86db1e29 238{
ca369d51
MP
239 struct queue_limits *limits = &q->limits;
240 unsigned int max_sectors;
241
09cbfeaf
KS
242 if ((max_hw_sectors << 9) < PAGE_SIZE) {
243 max_hw_sectors = 1 << (PAGE_SHIFT - 9);
24c03d47 244 printk(KERN_INFO "%s: set to minimum %d\n",
2800aac1 245 __func__, max_hw_sectors);
86db1e29
JA
246 }
247
30e2bc08 248 limits->max_hw_sectors = max_hw_sectors;
ca369d51
MP
249 max_sectors = min_not_zero(max_hw_sectors, limits->max_dev_sectors);
250 max_sectors = min_t(unsigned int, max_sectors, BLK_DEF_MAX_SECTORS);
251 limits->max_sectors = max_sectors;
86db1e29 252}
086fa5ff 253EXPORT_SYMBOL(blk_queue_max_hw_sectors);
86db1e29 254
762380ad
JA
255/**
256 * blk_queue_chunk_sectors - set size of the chunk for this queue
257 * @q: the request queue for the device
258 * @chunk_sectors: chunk sectors in the usual 512b unit
259 *
260 * Description:
261 * If a driver doesn't want IOs to cross a given chunk size, it can set
262 * this limit and prevent merging across chunks. Note that the chunk size
58a4915a
JA
263 * must currently be a power-of-2 in sectors. Also note that the block
264 * layer must accept a page worth of data at any offset. So if the
265 * crossing of chunks is a hard limitation in the driver, it must still be
266 * prepared to split single page bios.
762380ad
JA
267 **/
268void blk_queue_chunk_sectors(struct request_queue *q, unsigned int chunk_sectors)
269{
270 BUG_ON(!is_power_of_2(chunk_sectors));
271 q->limits.chunk_sectors = chunk_sectors;
272}
273EXPORT_SYMBOL(blk_queue_chunk_sectors);
274
67efc925
CH
275/**
276 * blk_queue_max_discard_sectors - set max sectors for a single discard
277 * @q: the request queue for the device
c7ebf065 278 * @max_discard_sectors: maximum number of sectors to discard
67efc925
CH
279 **/
280void blk_queue_max_discard_sectors(struct request_queue *q,
281 unsigned int max_discard_sectors)
282{
0034af03 283 q->limits.max_hw_discard_sectors = max_discard_sectors;
67efc925
CH
284 q->limits.max_discard_sectors = max_discard_sectors;
285}
286EXPORT_SYMBOL(blk_queue_max_discard_sectors);
287
4363ac7c
MP
288/**
289 * blk_queue_max_write_same_sectors - set max sectors for a single write same
290 * @q: the request queue for the device
291 * @max_write_same_sectors: maximum number of sectors to write per command
292 **/
293void blk_queue_max_write_same_sectors(struct request_queue *q,
294 unsigned int max_write_same_sectors)
295{
296 q->limits.max_write_same_sectors = max_write_same_sectors;
297}
298EXPORT_SYMBOL(blk_queue_max_write_same_sectors);
299
86db1e29 300/**
8a78362c 301 * blk_queue_max_segments - set max hw segments for a request for this queue
86db1e29
JA
302 * @q: the request queue for the device
303 * @max_segments: max number of segments
304 *
305 * Description:
306 * Enables a low level driver to set an upper limit on the number of
8a78362c 307 * hw data segments in a request.
86db1e29 308 **/
8a78362c 309void blk_queue_max_segments(struct request_queue *q, unsigned short max_segments)
86db1e29
JA
310{
311 if (!max_segments) {
312 max_segments = 1;
24c03d47
HH
313 printk(KERN_INFO "%s: set to minimum %d\n",
314 __func__, max_segments);
86db1e29
JA
315 }
316
8a78362c 317 q->limits.max_segments = max_segments;
86db1e29 318}
8a78362c 319EXPORT_SYMBOL(blk_queue_max_segments);
86db1e29
JA
320
321/**
322 * blk_queue_max_segment_size - set max segment size for blk_rq_map_sg
323 * @q: the request queue for the device
324 * @max_size: max size of segment in bytes
325 *
326 * Description:
327 * Enables a low level driver to set an upper limit on the size of a
328 * coalesced segment
329 **/
330void blk_queue_max_segment_size(struct request_queue *q, unsigned int max_size)
331{
09cbfeaf
KS
332 if (max_size < PAGE_SIZE) {
333 max_size = PAGE_SIZE;
24c03d47
HH
334 printk(KERN_INFO "%s: set to minimum %d\n",
335 __func__, max_size);
86db1e29
JA
336 }
337
025146e1 338 q->limits.max_segment_size = max_size;
86db1e29 339}
86db1e29
JA
340EXPORT_SYMBOL(blk_queue_max_segment_size);
341
342/**
e1defc4f 343 * blk_queue_logical_block_size - set logical block size for the queue
86db1e29 344 * @q: the request queue for the device
e1defc4f 345 * @size: the logical block size, in bytes
86db1e29
JA
346 *
347 * Description:
e1defc4f
MP
348 * This should be set to the lowest possible block size that the
349 * storage device can address. The default of 512 covers most
350 * hardware.
86db1e29 351 **/
e1defc4f 352void blk_queue_logical_block_size(struct request_queue *q, unsigned short size)
86db1e29 353{
025146e1 354 q->limits.logical_block_size = size;
c72758f3
MP
355
356 if (q->limits.physical_block_size < size)
357 q->limits.physical_block_size = size;
358
359 if (q->limits.io_min < q->limits.physical_block_size)
360 q->limits.io_min = q->limits.physical_block_size;
86db1e29 361}
e1defc4f 362EXPORT_SYMBOL(blk_queue_logical_block_size);
86db1e29 363
c72758f3
MP
364/**
365 * blk_queue_physical_block_size - set physical block size for the queue
366 * @q: the request queue for the device
367 * @size: the physical block size, in bytes
368 *
369 * Description:
370 * This should be set to the lowest possible sector size that the
371 * hardware can operate on without reverting to read-modify-write
372 * operations.
373 */
892b6f90 374void blk_queue_physical_block_size(struct request_queue *q, unsigned int size)
c72758f3
MP
375{
376 q->limits.physical_block_size = size;
377
378 if (q->limits.physical_block_size < q->limits.logical_block_size)
379 q->limits.physical_block_size = q->limits.logical_block_size;
380
381 if (q->limits.io_min < q->limits.physical_block_size)
382 q->limits.io_min = q->limits.physical_block_size;
383}
384EXPORT_SYMBOL(blk_queue_physical_block_size);
385
386/**
387 * blk_queue_alignment_offset - set physical block alignment offset
388 * @q: the request queue for the device
8ebf9756 389 * @offset: alignment offset in bytes
c72758f3
MP
390 *
391 * Description:
392 * Some devices are naturally misaligned to compensate for things like
393 * the legacy DOS partition table 63-sector offset. Low-level drivers
394 * should call this function for devices whose first sector is not
395 * naturally aligned.
396 */
397void blk_queue_alignment_offset(struct request_queue *q, unsigned int offset)
398{
399 q->limits.alignment_offset =
400 offset & (q->limits.physical_block_size - 1);
401 q->limits.misaligned = 0;
402}
403EXPORT_SYMBOL(blk_queue_alignment_offset);
404
7c958e32
MP
405/**
406 * blk_limits_io_min - set minimum request size for a device
407 * @limits: the queue limits
408 * @min: smallest I/O size in bytes
409 *
410 * Description:
411 * Some devices have an internal block size bigger than the reported
412 * hardware sector size. This function can be used to signal the
413 * smallest I/O the device can perform without incurring a performance
414 * penalty.
415 */
416void blk_limits_io_min(struct queue_limits *limits, unsigned int min)
417{
418 limits->io_min = min;
419
420 if (limits->io_min < limits->logical_block_size)
421 limits->io_min = limits->logical_block_size;
422
423 if (limits->io_min < limits->physical_block_size)
424 limits->io_min = limits->physical_block_size;
425}
426EXPORT_SYMBOL(blk_limits_io_min);
427
c72758f3
MP
428/**
429 * blk_queue_io_min - set minimum request size for the queue
430 * @q: the request queue for the device
8ebf9756 431 * @min: smallest I/O size in bytes
c72758f3
MP
432 *
433 * Description:
7e5f5fb0
MP
434 * Storage devices may report a granularity or preferred minimum I/O
435 * size which is the smallest request the device can perform without
436 * incurring a performance penalty. For disk drives this is often the
437 * physical block size. For RAID arrays it is often the stripe chunk
438 * size. A properly aligned multiple of minimum_io_size is the
439 * preferred request size for workloads where a high number of I/O
440 * operations is desired.
c72758f3
MP
441 */
442void blk_queue_io_min(struct request_queue *q, unsigned int min)
443{
7c958e32 444 blk_limits_io_min(&q->limits, min);
c72758f3
MP
445}
446EXPORT_SYMBOL(blk_queue_io_min);
447
3c5820c7
MP
448/**
449 * blk_limits_io_opt - set optimal request size for a device
450 * @limits: the queue limits
451 * @opt: smallest I/O size in bytes
452 *
453 * Description:
454 * Storage devices may report an optimal I/O size, which is the
455 * device's preferred unit for sustained I/O. This is rarely reported
456 * for disk drives. For RAID arrays it is usually the stripe width or
457 * the internal track size. A properly aligned multiple of
458 * optimal_io_size is the preferred request size for workloads where
459 * sustained throughput is desired.
460 */
461void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt)
462{
463 limits->io_opt = opt;
464}
465EXPORT_SYMBOL(blk_limits_io_opt);
466
c72758f3
MP
467/**
468 * blk_queue_io_opt - set optimal request size for the queue
469 * @q: the request queue for the device
8ebf9756 470 * @opt: optimal request size in bytes
c72758f3
MP
471 *
472 * Description:
7e5f5fb0
MP
473 * Storage devices may report an optimal I/O size, which is the
474 * device's preferred unit for sustained I/O. This is rarely reported
475 * for disk drives. For RAID arrays it is usually the stripe width or
476 * the internal track size. A properly aligned multiple of
477 * optimal_io_size is the preferred request size for workloads where
478 * sustained throughput is desired.
c72758f3
MP
479 */
480void blk_queue_io_opt(struct request_queue *q, unsigned int opt)
481{
3c5820c7 482 blk_limits_io_opt(&q->limits, opt);
c72758f3
MP
483}
484EXPORT_SYMBOL(blk_queue_io_opt);
485
86db1e29
JA
486/**
487 * blk_queue_stack_limits - inherit underlying queue limits for stacked drivers
488 * @t: the stacking driver (top)
489 * @b: the underlying device (bottom)
490 **/
491void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b)
492{
fef24667 493 blk_stack_limits(&t->limits, &b->limits, 0);
86db1e29 494}
86db1e29
JA
495EXPORT_SYMBOL(blk_queue_stack_limits);
496
c72758f3
MP
497/**
498 * blk_stack_limits - adjust queue_limits for stacked devices
81744ee4
MP
499 * @t: the stacking driver limits (top device)
500 * @b: the underlying queue limits (bottom, component device)
e03a72e1 501 * @start: first data sector within component device
c72758f3
MP
502 *
503 * Description:
81744ee4
MP
504 * This function is used by stacking drivers like MD and DM to ensure
505 * that all component devices have compatible block sizes and
506 * alignments. The stacking driver must provide a queue_limits
507 * struct (top) and then iteratively call the stacking function for
508 * all component (bottom) devices. The stacking function will
509 * attempt to combine the values and ensure proper alignment.
510 *
511 * Returns 0 if the top and bottom queue_limits are compatible. The
512 * top device's block sizes and alignment offsets may be adjusted to
513 * ensure alignment with the bottom device. If no compatible sizes
514 * and alignments exist, -1 is returned and the resulting top
515 * queue_limits will have the misaligned flag set to indicate that
516 * the alignment_offset is undefined.
c72758f3
MP
517 */
518int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
e03a72e1 519 sector_t start)
c72758f3 520{
e03a72e1 521 unsigned int top, bottom, alignment, ret = 0;
86b37281 522
c72758f3
MP
523 t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors);
524 t->max_hw_sectors = min_not_zero(t->max_hw_sectors, b->max_hw_sectors);
ca369d51 525 t->max_dev_sectors = min_not_zero(t->max_dev_sectors, b->max_dev_sectors);
4363ac7c
MP
526 t->max_write_same_sectors = min(t->max_write_same_sectors,
527 b->max_write_same_sectors);
77634f33 528 t->bounce_pfn = min_not_zero(t->bounce_pfn, b->bounce_pfn);
c72758f3
MP
529
530 t->seg_boundary_mask = min_not_zero(t->seg_boundary_mask,
531 b->seg_boundary_mask);
03100aad
KB
532 t->virt_boundary_mask = min_not_zero(t->virt_boundary_mask,
533 b->virt_boundary_mask);
c72758f3 534
8a78362c 535 t->max_segments = min_not_zero(t->max_segments, b->max_segments);
13f05c8d
MP
536 t->max_integrity_segments = min_not_zero(t->max_integrity_segments,
537 b->max_integrity_segments);
c72758f3
MP
538
539 t->max_segment_size = min_not_zero(t->max_segment_size,
540 b->max_segment_size);
541
fe0b393f
MP
542 t->misaligned |= b->misaligned;
543
e03a72e1 544 alignment = queue_limit_alignment_offset(b, start);
9504e086 545
81744ee4
MP
546 /* Bottom device has different alignment. Check that it is
547 * compatible with the current top alignment.
548 */
9504e086
MP
549 if (t->alignment_offset != alignment) {
550
551 top = max(t->physical_block_size, t->io_min)
552 + t->alignment_offset;
81744ee4 553 bottom = max(b->physical_block_size, b->io_min) + alignment;
9504e086 554
81744ee4 555 /* Verify that top and bottom intervals line up */
b8839b8c 556 if (max(top, bottom) % min(top, bottom)) {
9504e086 557 t->misaligned = 1;
fe0b393f
MP
558 ret = -1;
559 }
9504e086
MP
560 }
561
c72758f3
MP
562 t->logical_block_size = max(t->logical_block_size,
563 b->logical_block_size);
564
565 t->physical_block_size = max(t->physical_block_size,
566 b->physical_block_size);
567
568 t->io_min = max(t->io_min, b->io_min);
e9637415 569 t->io_opt = lcm_not_zero(t->io_opt, b->io_opt);
9504e086 570
e692cb66 571 t->cluster &= b->cluster;
98262f27 572 t->discard_zeroes_data &= b->discard_zeroes_data;
c72758f3 573
81744ee4 574 /* Physical block size a multiple of the logical block size? */
9504e086
MP
575 if (t->physical_block_size & (t->logical_block_size - 1)) {
576 t->physical_block_size = t->logical_block_size;
c72758f3 577 t->misaligned = 1;
fe0b393f 578 ret = -1;
86b37281
MP
579 }
580
81744ee4 581 /* Minimum I/O a multiple of the physical block size? */
9504e086
MP
582 if (t->io_min & (t->physical_block_size - 1)) {
583 t->io_min = t->physical_block_size;
584 t->misaligned = 1;
fe0b393f 585 ret = -1;
c72758f3
MP
586 }
587
81744ee4 588 /* Optimal I/O a multiple of the physical block size? */
9504e086
MP
589 if (t->io_opt & (t->physical_block_size - 1)) {
590 t->io_opt = 0;
591 t->misaligned = 1;
fe0b393f 592 ret = -1;
9504e086 593 }
c72758f3 594
c78afc62
KO
595 t->raid_partial_stripes_expensive =
596 max(t->raid_partial_stripes_expensive,
597 b->raid_partial_stripes_expensive);
598
81744ee4 599 /* Find lowest common alignment_offset */
e9637415 600 t->alignment_offset = lcm_not_zero(t->alignment_offset, alignment)
b8839b8c 601 % max(t->physical_block_size, t->io_min);
86b37281 602
81744ee4 603 /* Verify that new alignment_offset is on a logical block boundary */
fe0b393f 604 if (t->alignment_offset & (t->logical_block_size - 1)) {
c72758f3 605 t->misaligned = 1;
fe0b393f
MP
606 ret = -1;
607 }
c72758f3 608
9504e086
MP
609 /* Discard alignment and granularity */
610 if (b->discard_granularity) {
e03a72e1 611 alignment = queue_limit_discard_alignment(b, start);
9504e086
MP
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;
70dd5bf3 617
9504e086 618 /* Verify that top and bottom intervals line up */
8dd2cb7e 619 if ((max(top, bottom) % min(top, bottom)) != 0)
9504e086
MP
620 t->discard_misaligned = 1;
621 }
622
81744ee4
MP
623 t->max_discard_sectors = min_not_zero(t->max_discard_sectors,
624 b->max_discard_sectors);
0034af03
JA
625 t->max_hw_discard_sectors = min_not_zero(t->max_hw_discard_sectors,
626 b->max_hw_discard_sectors);
9504e086
MP
627 t->discard_granularity = max(t->discard_granularity,
628 b->discard_granularity);
e9637415 629 t->discard_alignment = lcm_not_zero(t->discard_alignment, alignment) %
8dd2cb7e 630 t->discard_granularity;
9504e086 631 }
70dd5bf3 632
fe0b393f 633 return ret;
c72758f3 634}
5d85d324 635EXPORT_SYMBOL(blk_stack_limits);
c72758f3 636
17be8c24
MP
637/**
638 * bdev_stack_limits - adjust queue limits for stacked drivers
639 * @t: the stacking driver limits (top device)
640 * @bdev: the component block_device (bottom)
641 * @start: first data sector within component device
642 *
643 * Description:
644 * Merges queue limits for a top device and a block_device. Returns
645 * 0 if alignment didn't change. Returns -1 if adding the bottom
646 * device caused misalignment.
647 */
648int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
649 sector_t start)
650{
651 struct request_queue *bq = bdev_get_queue(bdev);
652
653 start += get_start_sect(bdev);
654
e03a72e1 655 return blk_stack_limits(t, &bq->limits, start);
17be8c24
MP
656}
657EXPORT_SYMBOL(bdev_stack_limits);
658
c72758f3
MP
659/**
660 * disk_stack_limits - adjust queue limits for stacked drivers
77634f33 661 * @disk: MD/DM gendisk (top)
c72758f3
MP
662 * @bdev: the underlying block device (bottom)
663 * @offset: offset to beginning of data within component device
664 *
665 * Description:
e03a72e1
MP
666 * Merges the limits for a top level gendisk and a bottom level
667 * block_device.
c72758f3
MP
668 */
669void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
670 sector_t offset)
671{
672 struct request_queue *t = disk->queue;
c72758f3 673
e03a72e1 674 if (bdev_stack_limits(&t->limits, bdev, offset >> 9) < 0) {
c72758f3
MP
675 char top[BDEVNAME_SIZE], bottom[BDEVNAME_SIZE];
676
677 disk_name(disk, 0, top);
678 bdevname(bdev, bottom);
679
680 printk(KERN_NOTICE "%s: Warning: Device %s is misaligned\n",
681 top, bottom);
682 }
c72758f3
MP
683}
684EXPORT_SYMBOL(disk_stack_limits);
685
e3790c7d
TH
686/**
687 * blk_queue_dma_pad - set pad mask
688 * @q: the request queue for the device
689 * @mask: pad mask
690 *
27f8221a 691 * Set dma pad mask.
e3790c7d 692 *
27f8221a
FT
693 * Appending pad buffer to a request modifies the last entry of a
694 * scatter list such that it includes the pad buffer.
e3790c7d
TH
695 **/
696void blk_queue_dma_pad(struct request_queue *q, unsigned int mask)
697{
698 q->dma_pad_mask = mask;
699}
700EXPORT_SYMBOL(blk_queue_dma_pad);
701
27f8221a
FT
702/**
703 * blk_queue_update_dma_pad - update pad mask
704 * @q: the request queue for the device
705 * @mask: pad mask
706 *
707 * Update dma pad mask.
708 *
709 * Appending pad buffer to a request modifies the last entry of a
710 * scatter list such that it includes the pad buffer.
711 **/
712void blk_queue_update_dma_pad(struct request_queue *q, unsigned int mask)
713{
714 if (mask > q->dma_pad_mask)
715 q->dma_pad_mask = mask;
716}
717EXPORT_SYMBOL(blk_queue_update_dma_pad);
718
86db1e29
JA
719/**
720 * blk_queue_dma_drain - Set up a drain buffer for excess dma.
86db1e29 721 * @q: the request queue for the device
2fb98e84 722 * @dma_drain_needed: fn which returns non-zero if drain is necessary
86db1e29
JA
723 * @buf: physically contiguous buffer
724 * @size: size of the buffer in bytes
725 *
726 * Some devices have excess DMA problems and can't simply discard (or
727 * zero fill) the unwanted piece of the transfer. They have to have a
728 * real area of memory to transfer it into. The use case for this is
729 * ATAPI devices in DMA mode. If the packet command causes a transfer
730 * bigger than the transfer size some HBAs will lock up if there
731 * aren't DMA elements to contain the excess transfer. What this API
732 * does is adjust the queue so that the buf is always appended
733 * silently to the scatterlist.
734 *
8a78362c
MP
735 * Note: This routine adjusts max_hw_segments to make room for appending
736 * the drain buffer. If you call blk_queue_max_segments() after calling
737 * this routine, you must set the limit to one fewer than your device
738 * can support otherwise there won't be room for the drain buffer.
86db1e29 739 */
448da4d2 740int blk_queue_dma_drain(struct request_queue *q,
2fb98e84
TH
741 dma_drain_needed_fn *dma_drain_needed,
742 void *buf, unsigned int size)
86db1e29 743{
8a78362c 744 if (queue_max_segments(q) < 2)
86db1e29
JA
745 return -EINVAL;
746 /* make room for appending the drain */
8a78362c 747 blk_queue_max_segments(q, queue_max_segments(q) - 1);
2fb98e84 748 q->dma_drain_needed = dma_drain_needed;
86db1e29
JA
749 q->dma_drain_buffer = buf;
750 q->dma_drain_size = size;
751
752 return 0;
753}
86db1e29
JA
754EXPORT_SYMBOL_GPL(blk_queue_dma_drain);
755
756/**
757 * blk_queue_segment_boundary - set boundary rules for segment merging
758 * @q: the request queue for the device
759 * @mask: the memory boundary mask
760 **/
761void blk_queue_segment_boundary(struct request_queue *q, unsigned long mask)
762{
09cbfeaf
KS
763 if (mask < PAGE_SIZE - 1) {
764 mask = PAGE_SIZE - 1;
24c03d47
HH
765 printk(KERN_INFO "%s: set to minimum %lx\n",
766 __func__, mask);
86db1e29
JA
767 }
768
025146e1 769 q->limits.seg_boundary_mask = mask;
86db1e29 770}
86db1e29
JA
771EXPORT_SYMBOL(blk_queue_segment_boundary);
772
03100aad
KB
773/**
774 * blk_queue_virt_boundary - set boundary rules for bio merging
775 * @q: the request queue for the device
776 * @mask: the memory boundary mask
777 **/
778void blk_queue_virt_boundary(struct request_queue *q, unsigned long mask)
779{
780 q->limits.virt_boundary_mask = mask;
781}
782EXPORT_SYMBOL(blk_queue_virt_boundary);
783
86db1e29
JA
784/**
785 * blk_queue_dma_alignment - set dma length and memory alignment
786 * @q: the request queue for the device
787 * @mask: alignment mask
788 *
789 * description:
710027a4 790 * set required memory and length alignment for direct dma transactions.
8feb4d20 791 * this is used when building direct io requests for the queue.
86db1e29
JA
792 *
793 **/
794void blk_queue_dma_alignment(struct request_queue *q, int mask)
795{
796 q->dma_alignment = mask;
797}
86db1e29
JA
798EXPORT_SYMBOL(blk_queue_dma_alignment);
799
800/**
801 * blk_queue_update_dma_alignment - update dma length and memory alignment
802 * @q: the request queue for the device
803 * @mask: alignment mask
804 *
805 * description:
710027a4 806 * update required memory and length alignment for direct dma transactions.
86db1e29
JA
807 * If the requested alignment is larger than the current alignment, then
808 * the current queue alignment is updated to the new value, otherwise it
809 * is left alone. The design of this is to allow multiple objects
810 * (driver, device, transport etc) to set their respective
811 * alignments without having them interfere.
812 *
813 **/
814void blk_queue_update_dma_alignment(struct request_queue *q, int mask)
815{
816 BUG_ON(mask > PAGE_SIZE);
817
818 if (mask > q->dma_alignment)
819 q->dma_alignment = mask;
820}
86db1e29
JA
821EXPORT_SYMBOL(blk_queue_update_dma_alignment);
822
f3876930 823void blk_queue_flush_queueable(struct request_queue *q, bool queueable)
824{
c888a8f9
JA
825 spin_lock_irq(q->queue_lock);
826 if (queueable)
827 clear_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags);
828 else
829 set_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags);
830 spin_unlock_irq(q->queue_lock);
f3876930 831}
832EXPORT_SYMBOL_GPL(blk_queue_flush_queueable);
833
93e9d8e8
JA
834/**
835 * blk_queue_write_cache - configure queue's write cache
836 * @q: the request queue for the device
837 * @wc: write back cache on or off
838 * @fua: device supports FUA writes, if true
839 *
840 * Tell the block layer about the write cache of @q.
841 */
842void blk_queue_write_cache(struct request_queue *q, bool wc, bool fua)
843{
844 spin_lock_irq(q->queue_lock);
c888a8f9 845 if (wc)
93e9d8e8 846 queue_flag_set(QUEUE_FLAG_WC, q);
c888a8f9 847 else
93e9d8e8 848 queue_flag_clear(QUEUE_FLAG_WC, q);
c888a8f9 849 if (fua)
93e9d8e8 850 queue_flag_set(QUEUE_FLAG_FUA, q);
c888a8f9 851 else
93e9d8e8
JA
852 queue_flag_clear(QUEUE_FLAG_FUA, q);
853 spin_unlock_irq(q->queue_lock);
854}
855EXPORT_SYMBOL_GPL(blk_queue_write_cache);
856
aeb3d3a8 857static int __init blk_settings_init(void)
86db1e29
JA
858{
859 blk_max_low_pfn = max_low_pfn - 1;
860 blk_max_pfn = max_pfn - 1;
861 return 0;
862}
863subsys_initcall(blk_settings_init);