block: bump BLK_DEF_MAX_SECTORS to 2560
[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;
eb28d31b 92 lim->max_segment_size = BLK_MAX_SEGMENT_SIZE;
b1bd055d 93 lim->max_sectors = lim->max_hw_sectors = BLK_SAFE_MAX_SECTORS;
762380ad 94 lim->chunk_sectors = 0;
4363ac7c 95 lim->max_write_same_sectors = 0;
86b37281 96 lim->max_discard_sectors = 0;
0034af03 97 lim->max_hw_discard_sectors = 0;
86b37281
MP
98 lim->discard_granularity = 0;
99 lim->discard_alignment = 0;
100 lim->discard_misaligned = 0;
b1bd055d 101 lim->discard_zeroes_data = 0;
e475bba2 102 lim->logical_block_size = lim->physical_block_size = lim->io_min = 512;
3a02c8e8 103 lim->bounce_pfn = (unsigned long)(BLK_BOUNCE_ANY >> PAGE_SHIFT);
e475bba2
MP
104 lim->alignment_offset = 0;
105 lim->io_opt = 0;
106 lim->misaligned = 0;
e692cb66 107 lim->cluster = 1;
e475bba2
MP
108}
109EXPORT_SYMBOL(blk_set_default_limits);
110
b1bd055d
MP
111/**
112 * blk_set_stacking_limits - set default limits for stacking devices
113 * @lim: the queue_limits structure to reset
114 *
115 * Description:
116 * Returns a queue_limit struct to its default state. Should be used
117 * by stacking drivers like DM that have no internal limits.
118 */
119void blk_set_stacking_limits(struct queue_limits *lim)
120{
121 blk_set_default_limits(lim);
122
123 /* Inherit limits from component devices */
124 lim->discard_zeroes_data = 1;
125 lim->max_segments = USHRT_MAX;
126 lim->max_hw_sectors = UINT_MAX;
d82ae52e 127 lim->max_segment_size = UINT_MAX;
fe86cdce 128 lim->max_sectors = UINT_MAX;
4363ac7c 129 lim->max_write_same_sectors = UINT_MAX;
b1bd055d
MP
130}
131EXPORT_SYMBOL(blk_set_stacking_limits);
132
86db1e29
JA
133/**
134 * blk_queue_make_request - define an alternate make_request function for a device
135 * @q: the request queue for the device to be affected
136 * @mfn: the alternate make_request function
137 *
138 * Description:
139 * The normal way for &struct bios to be passed to a device
140 * driver is for them to be collected into requests on a request
141 * queue, and then to allow the device driver to select requests
142 * off that queue when it is ready. This works well for many block
143 * devices. However some block devices (typically virtual devices
144 * such as md or lvm) do not benefit from the processing on the
145 * request queue, and are served best by having the requests passed
146 * directly to them. This can be achieved by providing a function
147 * to blk_queue_make_request().
148 *
149 * Caveat:
150 * The driver that does this *must* be able to deal appropriately
151 * with buffers in "highmemory". This can be accomplished by either calling
152 * __bio_kmap_atomic() to get a temporary kernel mapping, or by calling
153 * blk_queue_bounce() to create a buffer in normal memory.
154 **/
6728cb0e 155void blk_queue_make_request(struct request_queue *q, make_request_fn *mfn)
86db1e29
JA
156{
157 /*
158 * set defaults
159 */
160 q->nr_requests = BLKDEV_MAX_RQ;
0e435ac2 161
86db1e29 162 q->make_request_fn = mfn;
86db1e29
JA
163 blk_queue_dma_alignment(q, 511);
164 blk_queue_congestion_threshold(q);
165 q->nr_batching = BLK_BATCH_REQ;
166
e475bba2
MP
167 blk_set_default_limits(&q->limits);
168
86db1e29
JA
169 /*
170 * by default assume old behaviour and bounce for any highmem page
171 */
172 blk_queue_bounce_limit(q, BLK_BOUNCE_HIGH);
173}
86db1e29
JA
174EXPORT_SYMBOL(blk_queue_make_request);
175
176/**
177 * blk_queue_bounce_limit - set bounce buffer limit for queue
cd0aca2d 178 * @q: the request queue for the device
9f7e45d8 179 * @max_addr: the maximum address the device can handle
86db1e29
JA
180 *
181 * Description:
182 * Different hardware can have different requirements as to what pages
183 * it can do I/O directly to. A low level driver can call
184 * blk_queue_bounce_limit to have lower memory pages allocated as bounce
9f7e45d8 185 * buffers for doing I/O to pages residing above @max_addr.
86db1e29 186 **/
9f7e45d8 187void blk_queue_bounce_limit(struct request_queue *q, u64 max_addr)
86db1e29 188{
9f7e45d8 189 unsigned long b_pfn = max_addr >> PAGE_SHIFT;
86db1e29
JA
190 int dma = 0;
191
192 q->bounce_gfp = GFP_NOIO;
193#if BITS_PER_LONG == 64
cd0aca2d
TH
194 /*
195 * Assume anything <= 4GB can be handled by IOMMU. Actually
196 * some IOMMUs can handle everything, but I don't know of a
197 * way to test this here.
198 */
199 if (b_pfn < (min_t(u64, 0xffffffffUL, BLK_BOUNCE_HIGH) >> PAGE_SHIFT))
86db1e29 200 dma = 1;
efb012b3 201 q->limits.bounce_pfn = max(max_low_pfn, b_pfn);
86db1e29 202#else
6728cb0e 203 if (b_pfn < blk_max_low_pfn)
86db1e29 204 dma = 1;
c49825fa 205 q->limits.bounce_pfn = b_pfn;
260a67a9 206#endif
86db1e29
JA
207 if (dma) {
208 init_emergency_isa_pool();
209 q->bounce_gfp = GFP_NOIO | GFP_DMA;
260a67a9 210 q->limits.bounce_pfn = b_pfn;
86db1e29
JA
211 }
212}
86db1e29
JA
213EXPORT_SYMBOL(blk_queue_bounce_limit);
214
215/**
72d4cd9f
MS
216 * blk_limits_max_hw_sectors - set hard and soft limit of max sectors for request
217 * @limits: the queue limits
2800aac1 218 * @max_hw_sectors: max hardware sectors in the usual 512b unit
86db1e29
JA
219 *
220 * Description:
2800aac1
MP
221 * Enables a low level driver to set a hard upper limit,
222 * max_hw_sectors, on the size of requests. max_hw_sectors is set by
223 * the device driver based upon the combined capabilities of I/O
224 * controller and storage device.
225 *
226 * max_sectors is a soft limit imposed by the block layer for
227 * filesystem type requests. This value can be overridden on a
228 * per-device basis in /sys/block/<device>/queue/max_sectors_kb.
229 * The soft limit can not exceed max_hw_sectors.
86db1e29 230 **/
72d4cd9f 231void blk_limits_max_hw_sectors(struct queue_limits *limits, unsigned int max_hw_sectors)
86db1e29 232{
2800aac1
MP
233 if ((max_hw_sectors << 9) < PAGE_CACHE_SIZE) {
234 max_hw_sectors = 1 << (PAGE_CACHE_SHIFT - 9);
24c03d47 235 printk(KERN_INFO "%s: set to minimum %d\n",
2800aac1 236 __func__, max_hw_sectors);
86db1e29
JA
237 }
238
30e2bc08
JM
239 limits->max_hw_sectors = max_hw_sectors;
240 limits->max_sectors = min_t(unsigned int, max_hw_sectors,
241 BLK_DEF_MAX_SECTORS);
72d4cd9f
MS
242}
243EXPORT_SYMBOL(blk_limits_max_hw_sectors);
244
245/**
246 * blk_queue_max_hw_sectors - set max sectors for a request for this queue
247 * @q: the request queue for the device
248 * @max_hw_sectors: max hardware sectors in the usual 512b unit
249 *
250 * Description:
251 * See description for blk_limits_max_hw_sectors().
252 **/
253void blk_queue_max_hw_sectors(struct request_queue *q, unsigned int max_hw_sectors)
254{
255 blk_limits_max_hw_sectors(&q->limits, max_hw_sectors);
86db1e29 256}
086fa5ff 257EXPORT_SYMBOL(blk_queue_max_hw_sectors);
86db1e29 258
762380ad
JA
259/**
260 * blk_queue_chunk_sectors - set size of the chunk for this queue
261 * @q: the request queue for the device
262 * @chunk_sectors: chunk sectors in the usual 512b unit
263 *
264 * Description:
265 * If a driver doesn't want IOs to cross a given chunk size, it can set
266 * this limit and prevent merging across chunks. Note that the chunk size
58a4915a
JA
267 * must currently be a power-of-2 in sectors. Also note that the block
268 * layer must accept a page worth of data at any offset. So if the
269 * crossing of chunks is a hard limitation in the driver, it must still be
270 * prepared to split single page bios.
762380ad
JA
271 **/
272void blk_queue_chunk_sectors(struct request_queue *q, unsigned int chunk_sectors)
273{
274 BUG_ON(!is_power_of_2(chunk_sectors));
275 q->limits.chunk_sectors = chunk_sectors;
276}
277EXPORT_SYMBOL(blk_queue_chunk_sectors);
278
67efc925
CH
279/**
280 * blk_queue_max_discard_sectors - set max sectors for a single discard
281 * @q: the request queue for the device
c7ebf065 282 * @max_discard_sectors: maximum number of sectors to discard
67efc925
CH
283 **/
284void blk_queue_max_discard_sectors(struct request_queue *q,
285 unsigned int max_discard_sectors)
286{
0034af03 287 q->limits.max_hw_discard_sectors = max_discard_sectors;
67efc925
CH
288 q->limits.max_discard_sectors = max_discard_sectors;
289}
290EXPORT_SYMBOL(blk_queue_max_discard_sectors);
291
4363ac7c
MP
292/**
293 * blk_queue_max_write_same_sectors - set max sectors for a single write same
294 * @q: the request queue for the device
295 * @max_write_same_sectors: maximum number of sectors to write per command
296 **/
297void blk_queue_max_write_same_sectors(struct request_queue *q,
298 unsigned int max_write_same_sectors)
299{
300 q->limits.max_write_same_sectors = max_write_same_sectors;
301}
302EXPORT_SYMBOL(blk_queue_max_write_same_sectors);
303
86db1e29 304/**
8a78362c 305 * blk_queue_max_segments - set max hw segments for a request for this queue
86db1e29
JA
306 * @q: the request queue for the device
307 * @max_segments: max number of segments
308 *
309 * Description:
310 * Enables a low level driver to set an upper limit on the number of
8a78362c 311 * hw data segments in a request.
86db1e29 312 **/
8a78362c 313void blk_queue_max_segments(struct request_queue *q, unsigned short max_segments)
86db1e29
JA
314{
315 if (!max_segments) {
316 max_segments = 1;
24c03d47
HH
317 printk(KERN_INFO "%s: set to minimum %d\n",
318 __func__, max_segments);
86db1e29
JA
319 }
320
8a78362c 321 q->limits.max_segments = max_segments;
86db1e29 322}
8a78362c 323EXPORT_SYMBOL(blk_queue_max_segments);
86db1e29
JA
324
325/**
326 * blk_queue_max_segment_size - set max segment size for blk_rq_map_sg
327 * @q: the request queue for the device
328 * @max_size: max size of segment in bytes
329 *
330 * Description:
331 * Enables a low level driver to set an upper limit on the size of a
332 * coalesced segment
333 **/
334void blk_queue_max_segment_size(struct request_queue *q, unsigned int max_size)
335{
336 if (max_size < PAGE_CACHE_SIZE) {
337 max_size = PAGE_CACHE_SIZE;
24c03d47
HH
338 printk(KERN_INFO "%s: set to minimum %d\n",
339 __func__, max_size);
86db1e29
JA
340 }
341
025146e1 342 q->limits.max_segment_size = max_size;
86db1e29 343}
86db1e29
JA
344EXPORT_SYMBOL(blk_queue_max_segment_size);
345
346/**
e1defc4f 347 * blk_queue_logical_block_size - set logical block size for the queue
86db1e29 348 * @q: the request queue for the device
e1defc4f 349 * @size: the logical block size, in bytes
86db1e29
JA
350 *
351 * Description:
e1defc4f
MP
352 * This should be set to the lowest possible block size that the
353 * storage device can address. The default of 512 covers most
354 * hardware.
86db1e29 355 **/
e1defc4f 356void blk_queue_logical_block_size(struct request_queue *q, unsigned short size)
86db1e29 357{
025146e1 358 q->limits.logical_block_size = size;
c72758f3
MP
359
360 if (q->limits.physical_block_size < size)
361 q->limits.physical_block_size = size;
362
363 if (q->limits.io_min < q->limits.physical_block_size)
364 q->limits.io_min = q->limits.physical_block_size;
86db1e29 365}
e1defc4f 366EXPORT_SYMBOL(blk_queue_logical_block_size);
86db1e29 367
c72758f3
MP
368/**
369 * blk_queue_physical_block_size - set physical block size for the queue
370 * @q: the request queue for the device
371 * @size: the physical block size, in bytes
372 *
373 * Description:
374 * This should be set to the lowest possible sector size that the
375 * hardware can operate on without reverting to read-modify-write
376 * operations.
377 */
892b6f90 378void blk_queue_physical_block_size(struct request_queue *q, unsigned int size)
c72758f3
MP
379{
380 q->limits.physical_block_size = size;
381
382 if (q->limits.physical_block_size < q->limits.logical_block_size)
383 q->limits.physical_block_size = q->limits.logical_block_size;
384
385 if (q->limits.io_min < q->limits.physical_block_size)
386 q->limits.io_min = q->limits.physical_block_size;
387}
388EXPORT_SYMBOL(blk_queue_physical_block_size);
389
390/**
391 * blk_queue_alignment_offset - set physical block alignment offset
392 * @q: the request queue for the device
8ebf9756 393 * @offset: alignment offset in bytes
c72758f3
MP
394 *
395 * Description:
396 * Some devices are naturally misaligned to compensate for things like
397 * the legacy DOS partition table 63-sector offset. Low-level drivers
398 * should call this function for devices whose first sector is not
399 * naturally aligned.
400 */
401void blk_queue_alignment_offset(struct request_queue *q, unsigned int offset)
402{
403 q->limits.alignment_offset =
404 offset & (q->limits.physical_block_size - 1);
405 q->limits.misaligned = 0;
406}
407EXPORT_SYMBOL(blk_queue_alignment_offset);
408
7c958e32
MP
409/**
410 * blk_limits_io_min - set minimum request size for a device
411 * @limits: the queue limits
412 * @min: smallest I/O size in bytes
413 *
414 * Description:
415 * Some devices have an internal block size bigger than the reported
416 * hardware sector size. This function can be used to signal the
417 * smallest I/O the device can perform without incurring a performance
418 * penalty.
419 */
420void blk_limits_io_min(struct queue_limits *limits, unsigned int min)
421{
422 limits->io_min = min;
423
424 if (limits->io_min < limits->logical_block_size)
425 limits->io_min = limits->logical_block_size;
426
427 if (limits->io_min < limits->physical_block_size)
428 limits->io_min = limits->physical_block_size;
429}
430EXPORT_SYMBOL(blk_limits_io_min);
431
c72758f3
MP
432/**
433 * blk_queue_io_min - set minimum request size for the queue
434 * @q: the request queue for the device
8ebf9756 435 * @min: smallest I/O size in bytes
c72758f3
MP
436 *
437 * Description:
7e5f5fb0
MP
438 * Storage devices may report a granularity or preferred minimum I/O
439 * size which is the smallest request the device can perform without
440 * incurring a performance penalty. For disk drives this is often the
441 * physical block size. For RAID arrays it is often the stripe chunk
442 * size. A properly aligned multiple of minimum_io_size is the
443 * preferred request size for workloads where a high number of I/O
444 * operations is desired.
c72758f3
MP
445 */
446void blk_queue_io_min(struct request_queue *q, unsigned int min)
447{
7c958e32 448 blk_limits_io_min(&q->limits, min);
c72758f3
MP
449}
450EXPORT_SYMBOL(blk_queue_io_min);
451
3c5820c7
MP
452/**
453 * blk_limits_io_opt - set optimal request size for a device
454 * @limits: the queue limits
455 * @opt: smallest I/O size in bytes
456 *
457 * Description:
458 * Storage devices may report an optimal I/O size, which is the
459 * device's preferred unit for sustained I/O. This is rarely reported
460 * for disk drives. For RAID arrays it is usually the stripe width or
461 * the internal track size. A properly aligned multiple of
462 * optimal_io_size is the preferred request size for workloads where
463 * sustained throughput is desired.
464 */
465void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt)
466{
467 limits->io_opt = opt;
468}
469EXPORT_SYMBOL(blk_limits_io_opt);
470
c72758f3
MP
471/**
472 * blk_queue_io_opt - set optimal request size for the queue
473 * @q: the request queue for the device
8ebf9756 474 * @opt: optimal request size in bytes
c72758f3
MP
475 *
476 * Description:
7e5f5fb0
MP
477 * Storage devices may report an optimal I/O size, which is the
478 * device's preferred unit for sustained I/O. This is rarely reported
479 * for disk drives. For RAID arrays it is usually the stripe width or
480 * the internal track size. A properly aligned multiple of
481 * optimal_io_size is the preferred request size for workloads where
482 * sustained throughput is desired.
c72758f3
MP
483 */
484void blk_queue_io_opt(struct request_queue *q, unsigned int opt)
485{
3c5820c7 486 blk_limits_io_opt(&q->limits, opt);
c72758f3
MP
487}
488EXPORT_SYMBOL(blk_queue_io_opt);
489
86db1e29
JA
490/**
491 * blk_queue_stack_limits - inherit underlying queue limits for stacked drivers
492 * @t: the stacking driver (top)
493 * @b: the underlying device (bottom)
494 **/
495void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b)
496{
fef24667 497 blk_stack_limits(&t->limits, &b->limits, 0);
86db1e29 498}
86db1e29
JA
499EXPORT_SYMBOL(blk_queue_stack_limits);
500
c72758f3
MP
501/**
502 * blk_stack_limits - adjust queue_limits for stacked devices
81744ee4
MP
503 * @t: the stacking driver limits (top device)
504 * @b: the underlying queue limits (bottom, component device)
e03a72e1 505 * @start: first data sector within component device
c72758f3
MP
506 *
507 * Description:
81744ee4
MP
508 * This function is used by stacking drivers like MD and DM to ensure
509 * that all component devices have compatible block sizes and
510 * alignments. The stacking driver must provide a queue_limits
511 * struct (top) and then iteratively call the stacking function for
512 * all component (bottom) devices. The stacking function will
513 * attempt to combine the values and ensure proper alignment.
514 *
515 * Returns 0 if the top and bottom queue_limits are compatible. The
516 * top device's block sizes and alignment offsets may be adjusted to
517 * ensure alignment with the bottom device. If no compatible sizes
518 * and alignments exist, -1 is returned and the resulting top
519 * queue_limits will have the misaligned flag set to indicate that
520 * the alignment_offset is undefined.
c72758f3
MP
521 */
522int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
e03a72e1 523 sector_t start)
c72758f3 524{
e03a72e1 525 unsigned int top, bottom, alignment, ret = 0;
86b37281 526
c72758f3
MP
527 t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors);
528 t->max_hw_sectors = min_not_zero(t->max_hw_sectors, b->max_hw_sectors);
4363ac7c
MP
529 t->max_write_same_sectors = min(t->max_write_same_sectors,
530 b->max_write_same_sectors);
77634f33 531 t->bounce_pfn = min_not_zero(t->bounce_pfn, b->bounce_pfn);
c72758f3
MP
532
533 t->seg_boundary_mask = min_not_zero(t->seg_boundary_mask,
534 b->seg_boundary_mask);
535
8a78362c 536 t->max_segments = min_not_zero(t->max_segments, b->max_segments);
13f05c8d
MP
537 t->max_integrity_segments = min_not_zero(t->max_integrity_segments,
538 b->max_integrity_segments);
c72758f3
MP
539
540 t->max_segment_size = min_not_zero(t->max_segment_size,
541 b->max_segment_size);
542
fe0b393f
MP
543 t->misaligned |= b->misaligned;
544
e03a72e1 545 alignment = queue_limit_alignment_offset(b, start);
9504e086 546
81744ee4
MP
547 /* Bottom device has different alignment. Check that it is
548 * compatible with the current top alignment.
549 */
9504e086
MP
550 if (t->alignment_offset != alignment) {
551
552 top = max(t->physical_block_size, t->io_min)
553 + t->alignment_offset;
81744ee4 554 bottom = max(b->physical_block_size, b->io_min) + alignment;
9504e086 555
81744ee4 556 /* Verify that top and bottom intervals line up */
b8839b8c 557 if (max(top, bottom) % min(top, bottom)) {
9504e086 558 t->misaligned = 1;
fe0b393f
MP
559 ret = -1;
560 }
9504e086
MP
561 }
562
c72758f3
MP
563 t->logical_block_size = max(t->logical_block_size,
564 b->logical_block_size);
565
566 t->physical_block_size = max(t->physical_block_size,
567 b->physical_block_size);
568
569 t->io_min = max(t->io_min, b->io_min);
e9637415 570 t->io_opt = lcm_not_zero(t->io_opt, b->io_opt);
9504e086 571
e692cb66 572 t->cluster &= b->cluster;
98262f27 573 t->discard_zeroes_data &= b->discard_zeroes_data;
c72758f3 574
81744ee4 575 /* Physical block size a multiple of the logical block size? */
9504e086
MP
576 if (t->physical_block_size & (t->logical_block_size - 1)) {
577 t->physical_block_size = t->logical_block_size;
c72758f3 578 t->misaligned = 1;
fe0b393f 579 ret = -1;
86b37281
MP
580 }
581
81744ee4 582 /* Minimum I/O a multiple of the physical block size? */
9504e086
MP
583 if (t->io_min & (t->physical_block_size - 1)) {
584 t->io_min = t->physical_block_size;
585 t->misaligned = 1;
fe0b393f 586 ret = -1;
c72758f3
MP
587 }
588
81744ee4 589 /* Optimal I/O a multiple of the physical block size? */
9504e086
MP
590 if (t->io_opt & (t->physical_block_size - 1)) {
591 t->io_opt = 0;
592 t->misaligned = 1;
fe0b393f 593 ret = -1;
9504e086 594 }
c72758f3 595
c78afc62
KO
596 t->raid_partial_stripes_expensive =
597 max(t->raid_partial_stripes_expensive,
598 b->raid_partial_stripes_expensive);
599
81744ee4 600 /* Find lowest common alignment_offset */
e9637415 601 t->alignment_offset = lcm_not_zero(t->alignment_offset, alignment)
b8839b8c 602 % max(t->physical_block_size, t->io_min);
86b37281 603
81744ee4 604 /* Verify that new alignment_offset is on a logical block boundary */
fe0b393f 605 if (t->alignment_offset & (t->logical_block_size - 1)) {
c72758f3 606 t->misaligned = 1;
fe0b393f
MP
607 ret = -1;
608 }
c72758f3 609
9504e086
MP
610 /* Discard alignment and granularity */
611 if (b->discard_granularity) {
e03a72e1 612 alignment = queue_limit_discard_alignment(b, start);
9504e086
MP
613
614 if (t->discard_granularity != 0 &&
615 t->discard_alignment != alignment) {
616 top = t->discard_granularity + t->discard_alignment;
617 bottom = b->discard_granularity + alignment;
70dd5bf3 618
9504e086 619 /* Verify that top and bottom intervals line up */
8dd2cb7e 620 if ((max(top, bottom) % min(top, bottom)) != 0)
9504e086
MP
621 t->discard_misaligned = 1;
622 }
623
81744ee4
MP
624 t->max_discard_sectors = min_not_zero(t->max_discard_sectors,
625 b->max_discard_sectors);
0034af03
JA
626 t->max_hw_discard_sectors = min_not_zero(t->max_hw_discard_sectors,
627 b->max_hw_discard_sectors);
9504e086
MP
628 t->discard_granularity = max(t->discard_granularity,
629 b->discard_granularity);
e9637415 630 t->discard_alignment = lcm_not_zero(t->discard_alignment, alignment) %
8dd2cb7e 631 t->discard_granularity;
9504e086 632 }
70dd5bf3 633
fe0b393f 634 return ret;
c72758f3 635}
5d85d324 636EXPORT_SYMBOL(blk_stack_limits);
c72758f3 637
17be8c24
MP
638/**
639 * bdev_stack_limits - adjust queue limits for stacked drivers
640 * @t: the stacking driver limits (top device)
641 * @bdev: the component block_device (bottom)
642 * @start: first data sector within component device
643 *
644 * Description:
645 * Merges queue limits for a top device and a block_device. Returns
646 * 0 if alignment didn't change. Returns -1 if adding the bottom
647 * device caused misalignment.
648 */
649int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
650 sector_t start)
651{
652 struct request_queue *bq = bdev_get_queue(bdev);
653
654 start += get_start_sect(bdev);
655
e03a72e1 656 return blk_stack_limits(t, &bq->limits, start);
17be8c24
MP
657}
658EXPORT_SYMBOL(bdev_stack_limits);
659
c72758f3
MP
660/**
661 * disk_stack_limits - adjust queue limits for stacked drivers
77634f33 662 * @disk: MD/DM gendisk (top)
c72758f3
MP
663 * @bdev: the underlying block device (bottom)
664 * @offset: offset to beginning of data within component device
665 *
666 * Description:
e03a72e1
MP
667 * Merges the limits for a top level gendisk and a bottom level
668 * block_device.
c72758f3
MP
669 */
670void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
671 sector_t offset)
672{
673 struct request_queue *t = disk->queue;
c72758f3 674
e03a72e1 675 if (bdev_stack_limits(&t->limits, bdev, offset >> 9) < 0) {
c72758f3
MP
676 char top[BDEVNAME_SIZE], bottom[BDEVNAME_SIZE];
677
678 disk_name(disk, 0, top);
679 bdevname(bdev, bottom);
680
681 printk(KERN_NOTICE "%s: Warning: Device %s is misaligned\n",
682 top, bottom);
683 }
c72758f3
MP
684}
685EXPORT_SYMBOL(disk_stack_limits);
686
e3790c7d
TH
687/**
688 * blk_queue_dma_pad - set pad mask
689 * @q: the request queue for the device
690 * @mask: pad mask
691 *
27f8221a 692 * Set dma pad mask.
e3790c7d 693 *
27f8221a
FT
694 * Appending pad buffer to a request modifies the last entry of a
695 * scatter list such that it includes the pad buffer.
e3790c7d
TH
696 **/
697void blk_queue_dma_pad(struct request_queue *q, unsigned int mask)
698{
699 q->dma_pad_mask = mask;
700}
701EXPORT_SYMBOL(blk_queue_dma_pad);
702
27f8221a
FT
703/**
704 * blk_queue_update_dma_pad - update pad mask
705 * @q: the request queue for the device
706 * @mask: pad mask
707 *
708 * Update dma pad mask.
709 *
710 * Appending pad buffer to a request modifies the last entry of a
711 * scatter list such that it includes the pad buffer.
712 **/
713void blk_queue_update_dma_pad(struct request_queue *q, unsigned int mask)
714{
715 if (mask > q->dma_pad_mask)
716 q->dma_pad_mask = mask;
717}
718EXPORT_SYMBOL(blk_queue_update_dma_pad);
719
86db1e29
JA
720/**
721 * blk_queue_dma_drain - Set up a drain buffer for excess dma.
86db1e29 722 * @q: the request queue for the device
2fb98e84 723 * @dma_drain_needed: fn which returns non-zero if drain is necessary
86db1e29
JA
724 * @buf: physically contiguous buffer
725 * @size: size of the buffer in bytes
726 *
727 * Some devices have excess DMA problems and can't simply discard (or
728 * zero fill) the unwanted piece of the transfer. They have to have a
729 * real area of memory to transfer it into. The use case for this is
730 * ATAPI devices in DMA mode. If the packet command causes a transfer
731 * bigger than the transfer size some HBAs will lock up if there
732 * aren't DMA elements to contain the excess transfer. What this API
733 * does is adjust the queue so that the buf is always appended
734 * silently to the scatterlist.
735 *
8a78362c
MP
736 * Note: This routine adjusts max_hw_segments to make room for appending
737 * the drain buffer. If you call blk_queue_max_segments() after calling
738 * this routine, you must set the limit to one fewer than your device
739 * can support otherwise there won't be room for the drain buffer.
86db1e29 740 */
448da4d2 741int blk_queue_dma_drain(struct request_queue *q,
2fb98e84
TH
742 dma_drain_needed_fn *dma_drain_needed,
743 void *buf, unsigned int size)
86db1e29 744{
8a78362c 745 if (queue_max_segments(q) < 2)
86db1e29
JA
746 return -EINVAL;
747 /* make room for appending the drain */
8a78362c 748 blk_queue_max_segments(q, queue_max_segments(q) - 1);
2fb98e84 749 q->dma_drain_needed = dma_drain_needed;
86db1e29
JA
750 q->dma_drain_buffer = buf;
751 q->dma_drain_size = size;
752
753 return 0;
754}
86db1e29
JA
755EXPORT_SYMBOL_GPL(blk_queue_dma_drain);
756
757/**
758 * blk_queue_segment_boundary - set boundary rules for segment merging
759 * @q: the request queue for the device
760 * @mask: the memory boundary mask
761 **/
762void blk_queue_segment_boundary(struct request_queue *q, unsigned long mask)
763{
764 if (mask < PAGE_CACHE_SIZE - 1) {
765 mask = PAGE_CACHE_SIZE - 1;
24c03d47
HH
766 printk(KERN_INFO "%s: set to minimum %lx\n",
767 __func__, mask);
86db1e29
JA
768 }
769
025146e1 770 q->limits.seg_boundary_mask = mask;
86db1e29 771}
86db1e29
JA
772EXPORT_SYMBOL(blk_queue_segment_boundary);
773
774/**
775 * blk_queue_dma_alignment - set dma length and memory alignment
776 * @q: the request queue for the device
777 * @mask: alignment mask
778 *
779 * description:
710027a4 780 * set required memory and length alignment for direct dma transactions.
8feb4d20 781 * this is used when building direct io requests for the queue.
86db1e29
JA
782 *
783 **/
784void blk_queue_dma_alignment(struct request_queue *q, int mask)
785{
786 q->dma_alignment = mask;
787}
86db1e29
JA
788EXPORT_SYMBOL(blk_queue_dma_alignment);
789
790/**
791 * blk_queue_update_dma_alignment - update dma length and memory alignment
792 * @q: the request queue for the device
793 * @mask: alignment mask
794 *
795 * description:
710027a4 796 * update required memory and length alignment for direct dma transactions.
86db1e29
JA
797 * If the requested alignment is larger than the current alignment, then
798 * the current queue alignment is updated to the new value, otherwise it
799 * is left alone. The design of this is to allow multiple objects
800 * (driver, device, transport etc) to set their respective
801 * alignments without having them interfere.
802 *
803 **/
804void blk_queue_update_dma_alignment(struct request_queue *q, int mask)
805{
806 BUG_ON(mask > PAGE_SIZE);
807
808 if (mask > q->dma_alignment)
809 q->dma_alignment = mask;
810}
86db1e29
JA
811EXPORT_SYMBOL(blk_queue_update_dma_alignment);
812
4913efe4
TH
813/**
814 * blk_queue_flush - configure queue's cache flush capability
815 * @q: the request queue for the device
816 * @flush: 0, REQ_FLUSH or REQ_FLUSH | REQ_FUA
817 *
818 * Tell block layer cache flush capability of @q. If it supports
819 * flushing, REQ_FLUSH should be set. If it supports bypassing
820 * write cache for individual writes, REQ_FUA should be set.
821 */
822void blk_queue_flush(struct request_queue *q, unsigned int flush)
823{
824 WARN_ON_ONCE(flush & ~(REQ_FLUSH | REQ_FUA));
825
826 if (WARN_ON_ONCE(!(flush & REQ_FLUSH) && (flush & REQ_FUA)))
827 flush &= ~REQ_FUA;
828
829 q->flush_flags = flush & (REQ_FLUSH | REQ_FUA);
830}
831EXPORT_SYMBOL_GPL(blk_queue_flush);
832
f3876930 833void blk_queue_flush_queueable(struct request_queue *q, bool queueable)
834{
835 q->flush_not_queueable = !queueable;
836}
837EXPORT_SYMBOL_GPL(blk_queue_flush_queueable);
838
aeb3d3a8 839static int __init blk_settings_init(void)
86db1e29
JA
840{
841 blk_max_low_pfn = max_low_pfn - 1;
842 blk_max_pfn = max_pfn - 1;
843 return 0;
844}
845subsys_initcall(blk_settings_init);