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