Commit | Line | Data |
---|---|---|
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" | |
87760e5e | 16 | #include "blk-wbt.h" |
86db1e29 | 17 | |
6728cb0e | 18 | unsigned long blk_max_low_pfn; |
86db1e29 | 19 | EXPORT_SYMBOL(blk_max_low_pfn); |
6728cb0e JA |
20 | |
21 | unsigned long blk_max_pfn; | |
86db1e29 JA |
22 | |
23 | /** | |
24 | * blk_queue_prep_rq - set a prepare_request function for queue | |
25 | * @q: queue | |
26 | * @pfn: prepare_request function | |
27 | * | |
28 | * It's possible for a queue to register a prepare_request callback which | |
29 | * is invoked before the request is handed to the request_fn. The goal of | |
30 | * the function is to prepare a request for I/O, it can be used to build a | |
31 | * cdb from the request data for instance. | |
32 | * | |
33 | */ | |
34 | void blk_queue_prep_rq(struct request_queue *q, prep_rq_fn *pfn) | |
35 | { | |
36 | q->prep_rq_fn = pfn; | |
37 | } | |
86db1e29 JA |
38 | EXPORT_SYMBOL(blk_queue_prep_rq); |
39 | ||
28018c24 JB |
40 | /** |
41 | * blk_queue_unprep_rq - set an unprepare_request function for queue | |
42 | * @q: queue | |
43 | * @ufn: unprepare_request function | |
44 | * | |
45 | * It's possible for a queue to register an unprepare_request callback | |
46 | * which is invoked before the request is finally completed. The goal | |
47 | * of the function is to deallocate any data that was allocated in the | |
48 | * prepare_request callback. | |
49 | * | |
50 | */ | |
51 | void blk_queue_unprep_rq(struct request_queue *q, unprep_rq_fn *ufn) | |
52 | { | |
53 | q->unprep_rq_fn = ufn; | |
54 | } | |
55 | EXPORT_SYMBOL(blk_queue_unprep_rq); | |
56 | ||
86db1e29 JA |
57 | void blk_queue_softirq_done(struct request_queue *q, softirq_done_fn *fn) |
58 | { | |
59 | q->softirq_done_fn = fn; | |
60 | } | |
86db1e29 JA |
61 | EXPORT_SYMBOL(blk_queue_softirq_done); |
62 | ||
242f9dcb JA |
63 | void blk_queue_rq_timeout(struct request_queue *q, unsigned int timeout) |
64 | { | |
65 | q->rq_timeout = timeout; | |
66 | } | |
67 | EXPORT_SYMBOL_GPL(blk_queue_rq_timeout); | |
68 | ||
69 | void blk_queue_rq_timed_out(struct request_queue *q, rq_timed_out_fn *fn) | |
70 | { | |
71 | q->rq_timed_out_fn = fn; | |
72 | } | |
73 | EXPORT_SYMBOL_GPL(blk_queue_rq_timed_out); | |
74 | ||
ef9e3fac KU |
75 | void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn) |
76 | { | |
77 | q->lld_busy_fn = fn; | |
78 | } | |
79 | EXPORT_SYMBOL_GPL(blk_queue_lld_busy); | |
80 | ||
e475bba2 MP |
81 | /** |
82 | * blk_set_default_limits - reset limits to default values | |
f740f5ca | 83 | * @lim: the queue_limits structure to reset |
e475bba2 MP |
84 | * |
85 | * Description: | |
b1bd055d | 86 | * Returns a queue_limit struct to its default state. |
e475bba2 MP |
87 | */ |
88 | void blk_set_default_limits(struct queue_limits *lim) | |
89 | { | |
8a78362c | 90 | lim->max_segments = BLK_MAX_SEGMENTS; |
13f05c8d | 91 | lim->max_integrity_segments = 0; |
e475bba2 | 92 | lim->seg_boundary_mask = BLK_SEG_BOUNDARY_MASK; |
03100aad | 93 | lim->virt_boundary_mask = 0; |
eb28d31b | 94 | lim->max_segment_size = BLK_MAX_SEGMENT_SIZE; |
5f009d3f KB |
95 | lim->max_sectors = lim->max_hw_sectors = BLK_SAFE_MAX_SECTORS; |
96 | lim->max_dev_sectors = 0; | |
762380ad | 97 | lim->chunk_sectors = 0; |
4363ac7c | 98 | lim->max_write_same_sectors = 0; |
a6f0788e | 99 | lim->max_write_zeroes_sectors = 0; |
86b37281 | 100 | lim->max_discard_sectors = 0; |
0034af03 | 101 | lim->max_hw_discard_sectors = 0; |
86b37281 MP |
102 | lim->discard_granularity = 0; |
103 | lim->discard_alignment = 0; | |
104 | lim->discard_misaligned = 0; | |
b1bd055d | 105 | lim->discard_zeroes_data = 0; |
e475bba2 | 106 | lim->logical_block_size = lim->physical_block_size = lim->io_min = 512; |
3a02c8e8 | 107 | lim->bounce_pfn = (unsigned long)(BLK_BOUNCE_ANY >> PAGE_SHIFT); |
e475bba2 MP |
108 | lim->alignment_offset = 0; |
109 | lim->io_opt = 0; | |
110 | lim->misaligned = 0; | |
e692cb66 | 111 | lim->cluster = 1; |
797476b8 | 112 | lim->zoned = BLK_ZONED_NONE; |
e475bba2 MP |
113 | } |
114 | EXPORT_SYMBOL(blk_set_default_limits); | |
115 | ||
b1bd055d MP |
116 | /** |
117 | * blk_set_stacking_limits - set default limits for stacking devices | |
118 | * @lim: the queue_limits structure to reset | |
119 | * | |
120 | * Description: | |
121 | * Returns a queue_limit struct to its default state. Should be used | |
122 | * by stacking drivers like DM that have no internal limits. | |
123 | */ | |
124 | void blk_set_stacking_limits(struct queue_limits *lim) | |
125 | { | |
126 | blk_set_default_limits(lim); | |
127 | ||
128 | /* Inherit limits from component devices */ | |
129 | lim->discard_zeroes_data = 1; | |
130 | lim->max_segments = USHRT_MAX; | |
131 | lim->max_hw_sectors = UINT_MAX; | |
d82ae52e | 132 | lim->max_segment_size = UINT_MAX; |
fe86cdce | 133 | lim->max_sectors = UINT_MAX; |
ca369d51 | 134 | lim->max_dev_sectors = UINT_MAX; |
4363ac7c | 135 | lim->max_write_same_sectors = UINT_MAX; |
a6f0788e | 136 | lim->max_write_zeroes_sectors = UINT_MAX; |
b1bd055d MP |
137 | } |
138 | EXPORT_SYMBOL(blk_set_stacking_limits); | |
139 | ||
86db1e29 JA |
140 | /** |
141 | * blk_queue_make_request - define an alternate make_request function for a device | |
142 | * @q: the request queue for the device to be affected | |
143 | * @mfn: the alternate make_request function | |
144 | * | |
145 | * Description: | |
146 | * The normal way for &struct bios to be passed to a device | |
147 | * driver is for them to be collected into requests on a request | |
148 | * queue, and then to allow the device driver to select requests | |
149 | * off that queue when it is ready. This works well for many block | |
150 | * devices. However some block devices (typically virtual devices | |
151 | * such as md or lvm) do not benefit from the processing on the | |
152 | * request queue, and are served best by having the requests passed | |
153 | * directly to them. This can be achieved by providing a function | |
154 | * to blk_queue_make_request(). | |
155 | * | |
156 | * Caveat: | |
157 | * The driver that does this *must* be able to deal appropriately | |
158 | * with buffers in "highmemory". This can be accomplished by either calling | |
159 | * __bio_kmap_atomic() to get a temporary kernel mapping, or by calling | |
160 | * blk_queue_bounce() to create a buffer in normal memory. | |
161 | **/ | |
6728cb0e | 162 | void blk_queue_make_request(struct request_queue *q, make_request_fn *mfn) |
86db1e29 JA |
163 | { |
164 | /* | |
165 | * set defaults | |
166 | */ | |
167 | q->nr_requests = BLKDEV_MAX_RQ; | |
0e435ac2 | 168 | |
86db1e29 | 169 | q->make_request_fn = mfn; |
86db1e29 JA |
170 | blk_queue_dma_alignment(q, 511); |
171 | blk_queue_congestion_threshold(q); | |
172 | q->nr_batching = BLK_BATCH_REQ; | |
173 | ||
e475bba2 MP |
174 | blk_set_default_limits(&q->limits); |
175 | ||
86db1e29 JA |
176 | /* |
177 | * by default assume old behaviour and bounce for any highmem page | |
178 | */ | |
179 | blk_queue_bounce_limit(q, BLK_BOUNCE_HIGH); | |
180 | } | |
86db1e29 JA |
181 | EXPORT_SYMBOL(blk_queue_make_request); |
182 | ||
183 | /** | |
184 | * blk_queue_bounce_limit - set bounce buffer limit for queue | |
cd0aca2d | 185 | * @q: the request queue for the device |
9f7e45d8 | 186 | * @max_addr: the maximum address the device can handle |
86db1e29 JA |
187 | * |
188 | * Description: | |
189 | * Different hardware can have different requirements as to what pages | |
190 | * it can do I/O directly to. A low level driver can call | |
191 | * blk_queue_bounce_limit to have lower memory pages allocated as bounce | |
9f7e45d8 | 192 | * buffers for doing I/O to pages residing above @max_addr. |
86db1e29 | 193 | **/ |
9f7e45d8 | 194 | void blk_queue_bounce_limit(struct request_queue *q, u64 max_addr) |
86db1e29 | 195 | { |
9f7e45d8 | 196 | unsigned long b_pfn = max_addr >> PAGE_SHIFT; |
86db1e29 JA |
197 | int dma = 0; |
198 | ||
199 | q->bounce_gfp = GFP_NOIO; | |
200 | #if BITS_PER_LONG == 64 | |
cd0aca2d TH |
201 | /* |
202 | * Assume anything <= 4GB can be handled by IOMMU. Actually | |
203 | * some IOMMUs can handle everything, but I don't know of a | |
204 | * way to test this here. | |
205 | */ | |
206 | if (b_pfn < (min_t(u64, 0xffffffffUL, BLK_BOUNCE_HIGH) >> PAGE_SHIFT)) | |
86db1e29 | 207 | dma = 1; |
efb012b3 | 208 | q->limits.bounce_pfn = max(max_low_pfn, b_pfn); |
86db1e29 | 209 | #else |
6728cb0e | 210 | if (b_pfn < blk_max_low_pfn) |
86db1e29 | 211 | dma = 1; |
c49825fa | 212 | q->limits.bounce_pfn = b_pfn; |
260a67a9 | 213 | #endif |
86db1e29 JA |
214 | if (dma) { |
215 | init_emergency_isa_pool(); | |
216 | q->bounce_gfp = GFP_NOIO | GFP_DMA; | |
260a67a9 | 217 | q->limits.bounce_pfn = b_pfn; |
86db1e29 JA |
218 | } |
219 | } | |
86db1e29 JA |
220 | EXPORT_SYMBOL(blk_queue_bounce_limit); |
221 | ||
222 | /** | |
ca369d51 MP |
223 | * blk_queue_max_hw_sectors - set max sectors for a request for this queue |
224 | * @q: the request queue for the device | |
2800aac1 | 225 | * @max_hw_sectors: max hardware sectors in the usual 512b unit |
86db1e29 JA |
226 | * |
227 | * Description: | |
2800aac1 MP |
228 | * Enables a low level driver to set a hard upper limit, |
229 | * max_hw_sectors, on the size of requests. max_hw_sectors is set by | |
4f258a46 MP |
230 | * the device driver based upon the capabilities of the I/O |
231 | * controller. | |
2800aac1 | 232 | * |
ca369d51 MP |
233 | * max_dev_sectors is a hard limit imposed by the storage device for |
234 | * READ/WRITE requests. It is set by the disk driver. | |
235 | * | |
2800aac1 MP |
236 | * max_sectors is a soft limit imposed by the block layer for |
237 | * filesystem type requests. This value can be overridden on a | |
238 | * per-device basis in /sys/block/<device>/queue/max_sectors_kb. | |
239 | * The soft limit can not exceed max_hw_sectors. | |
86db1e29 | 240 | **/ |
ca369d51 | 241 | void blk_queue_max_hw_sectors(struct request_queue *q, unsigned int max_hw_sectors) |
86db1e29 | 242 | { |
ca369d51 MP |
243 | struct queue_limits *limits = &q->limits; |
244 | unsigned int max_sectors; | |
245 | ||
09cbfeaf KS |
246 | if ((max_hw_sectors << 9) < PAGE_SIZE) { |
247 | max_hw_sectors = 1 << (PAGE_SHIFT - 9); | |
24c03d47 | 248 | printk(KERN_INFO "%s: set to minimum %d\n", |
2800aac1 | 249 | __func__, max_hw_sectors); |
86db1e29 JA |
250 | } |
251 | ||
30e2bc08 | 252 | limits->max_hw_sectors = max_hw_sectors; |
ca369d51 MP |
253 | max_sectors = min_not_zero(max_hw_sectors, limits->max_dev_sectors); |
254 | max_sectors = min_t(unsigned int, max_sectors, BLK_DEF_MAX_SECTORS); | |
255 | limits->max_sectors = max_sectors; | |
dc3b17cc | 256 | q->backing_dev_info->io_pages = max_sectors >> (PAGE_SHIFT - 9); |
86db1e29 | 257 | } |
086fa5ff | 258 | EXPORT_SYMBOL(blk_queue_max_hw_sectors); |
86db1e29 | 259 | |
762380ad JA |
260 | /** |
261 | * blk_queue_chunk_sectors - set size of the chunk for this queue | |
262 | * @q: the request queue for the device | |
263 | * @chunk_sectors: chunk sectors in the usual 512b unit | |
264 | * | |
265 | * Description: | |
266 | * If a driver doesn't want IOs to cross a given chunk size, it can set | |
267 | * this limit and prevent merging across chunks. Note that the chunk size | |
58a4915a JA |
268 | * must currently be a power-of-2 in sectors. Also note that the block |
269 | * layer must accept a page worth of data at any offset. So if the | |
270 | * crossing of chunks is a hard limitation in the driver, it must still be | |
271 | * prepared to split single page bios. | |
762380ad JA |
272 | **/ |
273 | void blk_queue_chunk_sectors(struct request_queue *q, unsigned int chunk_sectors) | |
274 | { | |
275 | BUG_ON(!is_power_of_2(chunk_sectors)); | |
276 | q->limits.chunk_sectors = chunk_sectors; | |
277 | } | |
278 | EXPORT_SYMBOL(blk_queue_chunk_sectors); | |
279 | ||
67efc925 CH |
280 | /** |
281 | * blk_queue_max_discard_sectors - set max sectors for a single discard | |
282 | * @q: the request queue for the device | |
c7ebf065 | 283 | * @max_discard_sectors: maximum number of sectors to discard |
67efc925 CH |
284 | **/ |
285 | void blk_queue_max_discard_sectors(struct request_queue *q, | |
286 | unsigned int max_discard_sectors) | |
287 | { | |
0034af03 | 288 | q->limits.max_hw_discard_sectors = max_discard_sectors; |
67efc925 CH |
289 | q->limits.max_discard_sectors = max_discard_sectors; |
290 | } | |
291 | EXPORT_SYMBOL(blk_queue_max_discard_sectors); | |
292 | ||
4363ac7c MP |
293 | /** |
294 | * blk_queue_max_write_same_sectors - set max sectors for a single write same | |
295 | * @q: the request queue for the device | |
296 | * @max_write_same_sectors: maximum number of sectors to write per command | |
297 | **/ | |
298 | void blk_queue_max_write_same_sectors(struct request_queue *q, | |
299 | unsigned int max_write_same_sectors) | |
300 | { | |
301 | q->limits.max_write_same_sectors = max_write_same_sectors; | |
302 | } | |
303 | EXPORT_SYMBOL(blk_queue_max_write_same_sectors); | |
304 | ||
a6f0788e CK |
305 | /** |
306 | * blk_queue_max_write_zeroes_sectors - set max sectors for a single | |
307 | * write zeroes | |
308 | * @q: the request queue for the device | |
309 | * @max_write_zeroes_sectors: maximum number of sectors to write per command | |
310 | **/ | |
311 | void blk_queue_max_write_zeroes_sectors(struct request_queue *q, | |
312 | unsigned int max_write_zeroes_sectors) | |
313 | { | |
314 | q->limits.max_write_zeroes_sectors = max_write_zeroes_sectors; | |
315 | } | |
316 | EXPORT_SYMBOL(blk_queue_max_write_zeroes_sectors); | |
317 | ||
86db1e29 | 318 | /** |
8a78362c | 319 | * blk_queue_max_segments - set max hw segments for a request for this queue |
86db1e29 JA |
320 | * @q: the request queue for the device |
321 | * @max_segments: max number of segments | |
322 | * | |
323 | * Description: | |
324 | * Enables a low level driver to set an upper limit on the number of | |
8a78362c | 325 | * hw data segments in a request. |
86db1e29 | 326 | **/ |
8a78362c | 327 | void blk_queue_max_segments(struct request_queue *q, unsigned short max_segments) |
86db1e29 JA |
328 | { |
329 | if (!max_segments) { | |
330 | max_segments = 1; | |
24c03d47 HH |
331 | printk(KERN_INFO "%s: set to minimum %d\n", |
332 | __func__, max_segments); | |
86db1e29 JA |
333 | } |
334 | ||
8a78362c | 335 | q->limits.max_segments = max_segments; |
86db1e29 | 336 | } |
8a78362c | 337 | EXPORT_SYMBOL(blk_queue_max_segments); |
86db1e29 JA |
338 | |
339 | /** | |
340 | * blk_queue_max_segment_size - set max segment size for blk_rq_map_sg | |
341 | * @q: the request queue for the device | |
342 | * @max_size: max size of segment in bytes | |
343 | * | |
344 | * Description: | |
345 | * Enables a low level driver to set an upper limit on the size of a | |
346 | * coalesced segment | |
347 | **/ | |
348 | void blk_queue_max_segment_size(struct request_queue *q, unsigned int max_size) | |
349 | { | |
09cbfeaf KS |
350 | if (max_size < PAGE_SIZE) { |
351 | max_size = PAGE_SIZE; | |
24c03d47 HH |
352 | printk(KERN_INFO "%s: set to minimum %d\n", |
353 | __func__, max_size); | |
86db1e29 JA |
354 | } |
355 | ||
025146e1 | 356 | q->limits.max_segment_size = max_size; |
86db1e29 | 357 | } |
86db1e29 JA |
358 | EXPORT_SYMBOL(blk_queue_max_segment_size); |
359 | ||
360 | /** | |
e1defc4f | 361 | * blk_queue_logical_block_size - set logical block size for the queue |
86db1e29 | 362 | * @q: the request queue for the device |
e1defc4f | 363 | * @size: the logical block size, in bytes |
86db1e29 JA |
364 | * |
365 | * Description: | |
e1defc4f MP |
366 | * This should be set to the lowest possible block size that the |
367 | * storage device can address. The default of 512 covers most | |
368 | * hardware. | |
86db1e29 | 369 | **/ |
e1defc4f | 370 | void blk_queue_logical_block_size(struct request_queue *q, unsigned short size) |
86db1e29 | 371 | { |
025146e1 | 372 | q->limits.logical_block_size = size; |
c72758f3 MP |
373 | |
374 | if (q->limits.physical_block_size < size) | |
375 | q->limits.physical_block_size = size; | |
376 | ||
377 | if (q->limits.io_min < q->limits.physical_block_size) | |
378 | q->limits.io_min = q->limits.physical_block_size; | |
86db1e29 | 379 | } |
e1defc4f | 380 | EXPORT_SYMBOL(blk_queue_logical_block_size); |
86db1e29 | 381 | |
c72758f3 MP |
382 | /** |
383 | * blk_queue_physical_block_size - set physical block size for the queue | |
384 | * @q: the request queue for the device | |
385 | * @size: the physical block size, in bytes | |
386 | * | |
387 | * Description: | |
388 | * This should be set to the lowest possible sector size that the | |
389 | * hardware can operate on without reverting to read-modify-write | |
390 | * operations. | |
391 | */ | |
892b6f90 | 392 | void blk_queue_physical_block_size(struct request_queue *q, unsigned int size) |
c72758f3 MP |
393 | { |
394 | q->limits.physical_block_size = size; | |
395 | ||
396 | if (q->limits.physical_block_size < q->limits.logical_block_size) | |
397 | q->limits.physical_block_size = q->limits.logical_block_size; | |
398 | ||
399 | if (q->limits.io_min < q->limits.physical_block_size) | |
400 | q->limits.io_min = q->limits.physical_block_size; | |
401 | } | |
402 | EXPORT_SYMBOL(blk_queue_physical_block_size); | |
403 | ||
404 | /** | |
405 | * blk_queue_alignment_offset - set physical block alignment offset | |
406 | * @q: the request queue for the device | |
8ebf9756 | 407 | * @offset: alignment offset in bytes |
c72758f3 MP |
408 | * |
409 | * Description: | |
410 | * Some devices are naturally misaligned to compensate for things like | |
411 | * the legacy DOS partition table 63-sector offset. Low-level drivers | |
412 | * should call this function for devices whose first sector is not | |
413 | * naturally aligned. | |
414 | */ | |
415 | void blk_queue_alignment_offset(struct request_queue *q, unsigned int offset) | |
416 | { | |
417 | q->limits.alignment_offset = | |
418 | offset & (q->limits.physical_block_size - 1); | |
419 | q->limits.misaligned = 0; | |
420 | } | |
421 | EXPORT_SYMBOL(blk_queue_alignment_offset); | |
422 | ||
7c958e32 MP |
423 | /** |
424 | * blk_limits_io_min - set minimum request size for a device | |
425 | * @limits: the queue limits | |
426 | * @min: smallest I/O size in bytes | |
427 | * | |
428 | * Description: | |
429 | * Some devices have an internal block size bigger than the reported | |
430 | * hardware sector size. This function can be used to signal the | |
431 | * smallest I/O the device can perform without incurring a performance | |
432 | * penalty. | |
433 | */ | |
434 | void blk_limits_io_min(struct queue_limits *limits, unsigned int min) | |
435 | { | |
436 | limits->io_min = min; | |
437 | ||
438 | if (limits->io_min < limits->logical_block_size) | |
439 | limits->io_min = limits->logical_block_size; | |
440 | ||
441 | if (limits->io_min < limits->physical_block_size) | |
442 | limits->io_min = limits->physical_block_size; | |
443 | } | |
444 | EXPORT_SYMBOL(blk_limits_io_min); | |
445 | ||
c72758f3 MP |
446 | /** |
447 | * blk_queue_io_min - set minimum request size for the queue | |
448 | * @q: the request queue for the device | |
8ebf9756 | 449 | * @min: smallest I/O size in bytes |
c72758f3 MP |
450 | * |
451 | * Description: | |
7e5f5fb0 MP |
452 | * Storage devices may report a granularity or preferred minimum I/O |
453 | * size which is the smallest request the device can perform without | |
454 | * incurring a performance penalty. For disk drives this is often the | |
455 | * physical block size. For RAID arrays it is often the stripe chunk | |
456 | * size. A properly aligned multiple of minimum_io_size is the | |
457 | * preferred request size for workloads where a high number of I/O | |
458 | * operations is desired. | |
c72758f3 MP |
459 | */ |
460 | void blk_queue_io_min(struct request_queue *q, unsigned int min) | |
461 | { | |
7c958e32 | 462 | blk_limits_io_min(&q->limits, min); |
c72758f3 MP |
463 | } |
464 | EXPORT_SYMBOL(blk_queue_io_min); | |
465 | ||
3c5820c7 MP |
466 | /** |
467 | * blk_limits_io_opt - set optimal request size for a device | |
468 | * @limits: the queue limits | |
469 | * @opt: smallest I/O size in bytes | |
470 | * | |
471 | * Description: | |
472 | * Storage devices may report an optimal I/O size, which is the | |
473 | * device's preferred unit for sustained I/O. This is rarely reported | |
474 | * for disk drives. For RAID arrays it is usually the stripe width or | |
475 | * the internal track size. A properly aligned multiple of | |
476 | * optimal_io_size is the preferred request size for workloads where | |
477 | * sustained throughput is desired. | |
478 | */ | |
479 | void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt) | |
480 | { | |
481 | limits->io_opt = opt; | |
482 | } | |
483 | EXPORT_SYMBOL(blk_limits_io_opt); | |
484 | ||
c72758f3 MP |
485 | /** |
486 | * blk_queue_io_opt - set optimal request size for the queue | |
487 | * @q: the request queue for the device | |
8ebf9756 | 488 | * @opt: optimal request size in bytes |
c72758f3 MP |
489 | * |
490 | * Description: | |
7e5f5fb0 MP |
491 | * Storage devices may report an optimal I/O size, which is the |
492 | * device's preferred unit for sustained I/O. This is rarely reported | |
493 | * for disk drives. For RAID arrays it is usually the stripe width or | |
494 | * the internal track size. A properly aligned multiple of | |
495 | * optimal_io_size is the preferred request size for workloads where | |
496 | * sustained throughput is desired. | |
c72758f3 MP |
497 | */ |
498 | void blk_queue_io_opt(struct request_queue *q, unsigned int opt) | |
499 | { | |
3c5820c7 | 500 | blk_limits_io_opt(&q->limits, opt); |
c72758f3 MP |
501 | } |
502 | EXPORT_SYMBOL(blk_queue_io_opt); | |
503 | ||
86db1e29 JA |
504 | /** |
505 | * blk_queue_stack_limits - inherit underlying queue limits for stacked drivers | |
506 | * @t: the stacking driver (top) | |
507 | * @b: the underlying device (bottom) | |
508 | **/ | |
509 | void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b) | |
510 | { | |
fef24667 | 511 | blk_stack_limits(&t->limits, &b->limits, 0); |
86db1e29 | 512 | } |
86db1e29 JA |
513 | EXPORT_SYMBOL(blk_queue_stack_limits); |
514 | ||
c72758f3 MP |
515 | /** |
516 | * blk_stack_limits - adjust queue_limits for stacked devices | |
81744ee4 MP |
517 | * @t: the stacking driver limits (top device) |
518 | * @b: the underlying queue limits (bottom, component device) | |
e03a72e1 | 519 | * @start: first data sector within component device |
c72758f3 MP |
520 | * |
521 | * Description: | |
81744ee4 MP |
522 | * This function is used by stacking drivers like MD and DM to ensure |
523 | * that all component devices have compatible block sizes and | |
524 | * alignments. The stacking driver must provide a queue_limits | |
525 | * struct (top) and then iteratively call the stacking function for | |
526 | * all component (bottom) devices. The stacking function will | |
527 | * attempt to combine the values and ensure proper alignment. | |
528 | * | |
529 | * Returns 0 if the top and bottom queue_limits are compatible. The | |
530 | * top device's block sizes and alignment offsets may be adjusted to | |
531 | * ensure alignment with the bottom device. If no compatible sizes | |
532 | * and alignments exist, -1 is returned and the resulting top | |
533 | * queue_limits will have the misaligned flag set to indicate that | |
534 | * the alignment_offset is undefined. | |
c72758f3 MP |
535 | */ |
536 | int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, | |
e03a72e1 | 537 | sector_t start) |
c72758f3 | 538 | { |
e03a72e1 | 539 | unsigned int top, bottom, alignment, ret = 0; |
86b37281 | 540 | |
c72758f3 MP |
541 | t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors); |
542 | t->max_hw_sectors = min_not_zero(t->max_hw_sectors, b->max_hw_sectors); | |
ca369d51 | 543 | t->max_dev_sectors = min_not_zero(t->max_dev_sectors, b->max_dev_sectors); |
4363ac7c MP |
544 | t->max_write_same_sectors = min(t->max_write_same_sectors, |
545 | b->max_write_same_sectors); | |
a6f0788e CK |
546 | t->max_write_zeroes_sectors = min(t->max_write_zeroes_sectors, |
547 | b->max_write_zeroes_sectors); | |
77634f33 | 548 | t->bounce_pfn = min_not_zero(t->bounce_pfn, b->bounce_pfn); |
c72758f3 MP |
549 | |
550 | t->seg_boundary_mask = min_not_zero(t->seg_boundary_mask, | |
551 | b->seg_boundary_mask); | |
03100aad KB |
552 | t->virt_boundary_mask = min_not_zero(t->virt_boundary_mask, |
553 | b->virt_boundary_mask); | |
c72758f3 | 554 | |
8a78362c | 555 | t->max_segments = min_not_zero(t->max_segments, b->max_segments); |
13f05c8d MP |
556 | t->max_integrity_segments = min_not_zero(t->max_integrity_segments, |
557 | b->max_integrity_segments); | |
c72758f3 MP |
558 | |
559 | t->max_segment_size = min_not_zero(t->max_segment_size, | |
560 | b->max_segment_size); | |
561 | ||
fe0b393f MP |
562 | t->misaligned |= b->misaligned; |
563 | ||
e03a72e1 | 564 | alignment = queue_limit_alignment_offset(b, start); |
9504e086 | 565 | |
81744ee4 MP |
566 | /* Bottom device has different alignment. Check that it is |
567 | * compatible with the current top alignment. | |
568 | */ | |
9504e086 MP |
569 | if (t->alignment_offset != alignment) { |
570 | ||
571 | top = max(t->physical_block_size, t->io_min) | |
572 | + t->alignment_offset; | |
81744ee4 | 573 | bottom = max(b->physical_block_size, b->io_min) + alignment; |
9504e086 | 574 | |
81744ee4 | 575 | /* Verify that top and bottom intervals line up */ |
b8839b8c | 576 | if (max(top, bottom) % min(top, bottom)) { |
9504e086 | 577 | t->misaligned = 1; |
fe0b393f MP |
578 | ret = -1; |
579 | } | |
9504e086 MP |
580 | } |
581 | ||
c72758f3 MP |
582 | t->logical_block_size = max(t->logical_block_size, |
583 | b->logical_block_size); | |
584 | ||
585 | t->physical_block_size = max(t->physical_block_size, | |
586 | b->physical_block_size); | |
587 | ||
588 | t->io_min = max(t->io_min, b->io_min); | |
e9637415 | 589 | t->io_opt = lcm_not_zero(t->io_opt, b->io_opt); |
9504e086 | 590 | |
e692cb66 | 591 | t->cluster &= b->cluster; |
98262f27 | 592 | t->discard_zeroes_data &= b->discard_zeroes_data; |
c72758f3 | 593 | |
81744ee4 | 594 | /* Physical block size a multiple of the logical block size? */ |
9504e086 MP |
595 | if (t->physical_block_size & (t->logical_block_size - 1)) { |
596 | t->physical_block_size = t->logical_block_size; | |
c72758f3 | 597 | t->misaligned = 1; |
fe0b393f | 598 | ret = -1; |
86b37281 MP |
599 | } |
600 | ||
81744ee4 | 601 | /* Minimum I/O a multiple of the physical block size? */ |
9504e086 MP |
602 | if (t->io_min & (t->physical_block_size - 1)) { |
603 | t->io_min = t->physical_block_size; | |
604 | t->misaligned = 1; | |
fe0b393f | 605 | ret = -1; |
c72758f3 MP |
606 | } |
607 | ||
81744ee4 | 608 | /* Optimal I/O a multiple of the physical block size? */ |
9504e086 MP |
609 | if (t->io_opt & (t->physical_block_size - 1)) { |
610 | t->io_opt = 0; | |
611 | t->misaligned = 1; | |
fe0b393f | 612 | ret = -1; |
9504e086 | 613 | } |
c72758f3 | 614 | |
c78afc62 KO |
615 | t->raid_partial_stripes_expensive = |
616 | max(t->raid_partial_stripes_expensive, | |
617 | b->raid_partial_stripes_expensive); | |
618 | ||
81744ee4 | 619 | /* Find lowest common alignment_offset */ |
e9637415 | 620 | t->alignment_offset = lcm_not_zero(t->alignment_offset, alignment) |
b8839b8c | 621 | % max(t->physical_block_size, t->io_min); |
86b37281 | 622 | |
81744ee4 | 623 | /* Verify that new alignment_offset is on a logical block boundary */ |
fe0b393f | 624 | if (t->alignment_offset & (t->logical_block_size - 1)) { |
c72758f3 | 625 | t->misaligned = 1; |
fe0b393f MP |
626 | ret = -1; |
627 | } | |
c72758f3 | 628 | |
9504e086 MP |
629 | /* Discard alignment and granularity */ |
630 | if (b->discard_granularity) { | |
e03a72e1 | 631 | alignment = queue_limit_discard_alignment(b, start); |
9504e086 MP |
632 | |
633 | if (t->discard_granularity != 0 && | |
634 | t->discard_alignment != alignment) { | |
635 | top = t->discard_granularity + t->discard_alignment; | |
636 | bottom = b->discard_granularity + alignment; | |
70dd5bf3 | 637 | |
9504e086 | 638 | /* Verify that top and bottom intervals line up */ |
8dd2cb7e | 639 | if ((max(top, bottom) % min(top, bottom)) != 0) |
9504e086 MP |
640 | t->discard_misaligned = 1; |
641 | } | |
642 | ||
81744ee4 MP |
643 | t->max_discard_sectors = min_not_zero(t->max_discard_sectors, |
644 | b->max_discard_sectors); | |
0034af03 JA |
645 | t->max_hw_discard_sectors = min_not_zero(t->max_hw_discard_sectors, |
646 | b->max_hw_discard_sectors); | |
9504e086 MP |
647 | t->discard_granularity = max(t->discard_granularity, |
648 | b->discard_granularity); | |
e9637415 | 649 | t->discard_alignment = lcm_not_zero(t->discard_alignment, alignment) % |
8dd2cb7e | 650 | t->discard_granularity; |
9504e086 | 651 | } |
70dd5bf3 | 652 | |
987b3b26 HR |
653 | if (b->chunk_sectors) |
654 | t->chunk_sectors = min_not_zero(t->chunk_sectors, | |
655 | b->chunk_sectors); | |
656 | ||
fe0b393f | 657 | return ret; |
c72758f3 | 658 | } |
5d85d324 | 659 | EXPORT_SYMBOL(blk_stack_limits); |
c72758f3 | 660 | |
17be8c24 MP |
661 | /** |
662 | * bdev_stack_limits - adjust queue limits for stacked drivers | |
663 | * @t: the stacking driver limits (top device) | |
664 | * @bdev: the component block_device (bottom) | |
665 | * @start: first data sector within component device | |
666 | * | |
667 | * Description: | |
668 | * Merges queue limits for a top device and a block_device. Returns | |
669 | * 0 if alignment didn't change. Returns -1 if adding the bottom | |
670 | * device caused misalignment. | |
671 | */ | |
672 | int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev, | |
673 | sector_t start) | |
674 | { | |
675 | struct request_queue *bq = bdev_get_queue(bdev); | |
676 | ||
677 | start += get_start_sect(bdev); | |
678 | ||
e03a72e1 | 679 | return blk_stack_limits(t, &bq->limits, start); |
17be8c24 MP |
680 | } |
681 | EXPORT_SYMBOL(bdev_stack_limits); | |
682 | ||
c72758f3 MP |
683 | /** |
684 | * disk_stack_limits - adjust queue limits for stacked drivers | |
77634f33 | 685 | * @disk: MD/DM gendisk (top) |
c72758f3 MP |
686 | * @bdev: the underlying block device (bottom) |
687 | * @offset: offset to beginning of data within component device | |
688 | * | |
689 | * Description: | |
e03a72e1 MP |
690 | * Merges the limits for a top level gendisk and a bottom level |
691 | * block_device. | |
c72758f3 MP |
692 | */ |
693 | void disk_stack_limits(struct gendisk *disk, struct block_device *bdev, | |
694 | sector_t offset) | |
695 | { | |
696 | struct request_queue *t = disk->queue; | |
c72758f3 | 697 | |
e03a72e1 | 698 | if (bdev_stack_limits(&t->limits, bdev, offset >> 9) < 0) { |
c72758f3 MP |
699 | char top[BDEVNAME_SIZE], bottom[BDEVNAME_SIZE]; |
700 | ||
701 | disk_name(disk, 0, top); | |
702 | bdevname(bdev, bottom); | |
703 | ||
704 | printk(KERN_NOTICE "%s: Warning: Device %s is misaligned\n", | |
705 | top, bottom); | |
706 | } | |
c72758f3 MP |
707 | } |
708 | EXPORT_SYMBOL(disk_stack_limits); | |
709 | ||
e3790c7d TH |
710 | /** |
711 | * blk_queue_dma_pad - set pad mask | |
712 | * @q: the request queue for the device | |
713 | * @mask: pad mask | |
714 | * | |
27f8221a | 715 | * Set dma pad mask. |
e3790c7d | 716 | * |
27f8221a FT |
717 | * Appending pad buffer to a request modifies the last entry of a |
718 | * scatter list such that it includes the pad buffer. | |
e3790c7d TH |
719 | **/ |
720 | void blk_queue_dma_pad(struct request_queue *q, unsigned int mask) | |
721 | { | |
722 | q->dma_pad_mask = mask; | |
723 | } | |
724 | EXPORT_SYMBOL(blk_queue_dma_pad); | |
725 | ||
27f8221a FT |
726 | /** |
727 | * blk_queue_update_dma_pad - update pad mask | |
728 | * @q: the request queue for the device | |
729 | * @mask: pad mask | |
730 | * | |
731 | * Update dma pad mask. | |
732 | * | |
733 | * Appending pad buffer to a request modifies the last entry of a | |
734 | * scatter list such that it includes the pad buffer. | |
735 | **/ | |
736 | void blk_queue_update_dma_pad(struct request_queue *q, unsigned int mask) | |
737 | { | |
738 | if (mask > q->dma_pad_mask) | |
739 | q->dma_pad_mask = mask; | |
740 | } | |
741 | EXPORT_SYMBOL(blk_queue_update_dma_pad); | |
742 | ||
86db1e29 JA |
743 | /** |
744 | * blk_queue_dma_drain - Set up a drain buffer for excess dma. | |
86db1e29 | 745 | * @q: the request queue for the device |
2fb98e84 | 746 | * @dma_drain_needed: fn which returns non-zero if drain is necessary |
86db1e29 JA |
747 | * @buf: physically contiguous buffer |
748 | * @size: size of the buffer in bytes | |
749 | * | |
750 | * Some devices have excess DMA problems and can't simply discard (or | |
751 | * zero fill) the unwanted piece of the transfer. They have to have a | |
752 | * real area of memory to transfer it into. The use case for this is | |
753 | * ATAPI devices in DMA mode. If the packet command causes a transfer | |
754 | * bigger than the transfer size some HBAs will lock up if there | |
755 | * aren't DMA elements to contain the excess transfer. What this API | |
756 | * does is adjust the queue so that the buf is always appended | |
757 | * silently to the scatterlist. | |
758 | * | |
8a78362c MP |
759 | * Note: This routine adjusts max_hw_segments to make room for appending |
760 | * the drain buffer. If you call blk_queue_max_segments() after calling | |
761 | * this routine, you must set the limit to one fewer than your device | |
762 | * can support otherwise there won't be room for the drain buffer. | |
86db1e29 | 763 | */ |
448da4d2 | 764 | int blk_queue_dma_drain(struct request_queue *q, |
2fb98e84 TH |
765 | dma_drain_needed_fn *dma_drain_needed, |
766 | void *buf, unsigned int size) | |
86db1e29 | 767 | { |
8a78362c | 768 | if (queue_max_segments(q) < 2) |
86db1e29 JA |
769 | return -EINVAL; |
770 | /* make room for appending the drain */ | |
8a78362c | 771 | blk_queue_max_segments(q, queue_max_segments(q) - 1); |
2fb98e84 | 772 | q->dma_drain_needed = dma_drain_needed; |
86db1e29 JA |
773 | q->dma_drain_buffer = buf; |
774 | q->dma_drain_size = size; | |
775 | ||
776 | return 0; | |
777 | } | |
86db1e29 JA |
778 | EXPORT_SYMBOL_GPL(blk_queue_dma_drain); |
779 | ||
780 | /** | |
781 | * blk_queue_segment_boundary - set boundary rules for segment merging | |
782 | * @q: the request queue for the device | |
783 | * @mask: the memory boundary mask | |
784 | **/ | |
785 | void blk_queue_segment_boundary(struct request_queue *q, unsigned long mask) | |
786 | { | |
09cbfeaf KS |
787 | if (mask < PAGE_SIZE - 1) { |
788 | mask = PAGE_SIZE - 1; | |
24c03d47 HH |
789 | printk(KERN_INFO "%s: set to minimum %lx\n", |
790 | __func__, mask); | |
86db1e29 JA |
791 | } |
792 | ||
025146e1 | 793 | q->limits.seg_boundary_mask = mask; |
86db1e29 | 794 | } |
86db1e29 JA |
795 | EXPORT_SYMBOL(blk_queue_segment_boundary); |
796 | ||
03100aad KB |
797 | /** |
798 | * blk_queue_virt_boundary - set boundary rules for bio merging | |
799 | * @q: the request queue for the device | |
800 | * @mask: the memory boundary mask | |
801 | **/ | |
802 | void blk_queue_virt_boundary(struct request_queue *q, unsigned long mask) | |
803 | { | |
804 | q->limits.virt_boundary_mask = mask; | |
805 | } | |
806 | EXPORT_SYMBOL(blk_queue_virt_boundary); | |
807 | ||
86db1e29 JA |
808 | /** |
809 | * blk_queue_dma_alignment - set dma length and memory alignment | |
810 | * @q: the request queue for the device | |
811 | * @mask: alignment mask | |
812 | * | |
813 | * description: | |
710027a4 | 814 | * set required memory and length alignment for direct dma transactions. |
8feb4d20 | 815 | * this is used when building direct io requests for the queue. |
86db1e29 JA |
816 | * |
817 | **/ | |
818 | void blk_queue_dma_alignment(struct request_queue *q, int mask) | |
819 | { | |
820 | q->dma_alignment = mask; | |
821 | } | |
86db1e29 JA |
822 | EXPORT_SYMBOL(blk_queue_dma_alignment); |
823 | ||
824 | /** | |
825 | * blk_queue_update_dma_alignment - update dma length and memory alignment | |
826 | * @q: the request queue for the device | |
827 | * @mask: alignment mask | |
828 | * | |
829 | * description: | |
710027a4 | 830 | * update required memory and length alignment for direct dma transactions. |
86db1e29 JA |
831 | * If the requested alignment is larger than the current alignment, then |
832 | * the current queue alignment is updated to the new value, otherwise it | |
833 | * is left alone. The design of this is to allow multiple objects | |
834 | * (driver, device, transport etc) to set their respective | |
835 | * alignments without having them interfere. | |
836 | * | |
837 | **/ | |
838 | void blk_queue_update_dma_alignment(struct request_queue *q, int mask) | |
839 | { | |
840 | BUG_ON(mask > PAGE_SIZE); | |
841 | ||
842 | if (mask > q->dma_alignment) | |
843 | q->dma_alignment = mask; | |
844 | } | |
86db1e29 JA |
845 | EXPORT_SYMBOL(blk_queue_update_dma_alignment); |
846 | ||
f3876930 | 847 | void blk_queue_flush_queueable(struct request_queue *q, bool queueable) |
848 | { | |
c888a8f9 JA |
849 | spin_lock_irq(q->queue_lock); |
850 | if (queueable) | |
851 | clear_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags); | |
852 | else | |
853 | set_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags); | |
854 | spin_unlock_irq(q->queue_lock); | |
f3876930 | 855 | } |
856 | EXPORT_SYMBOL_GPL(blk_queue_flush_queueable); | |
857 | ||
d278d4a8 JA |
858 | /** |
859 | * blk_set_queue_depth - tell the block layer about the device queue depth | |
860 | * @q: the request queue for the device | |
861 | * @depth: queue depth | |
862 | * | |
863 | */ | |
864 | void blk_set_queue_depth(struct request_queue *q, unsigned int depth) | |
865 | { | |
866 | q->queue_depth = depth; | |
87760e5e | 867 | wbt_set_queue_depth(q->rq_wb, depth); |
d278d4a8 JA |
868 | } |
869 | EXPORT_SYMBOL(blk_set_queue_depth); | |
870 | ||
93e9d8e8 JA |
871 | /** |
872 | * blk_queue_write_cache - configure queue's write cache | |
873 | * @q: the request queue for the device | |
874 | * @wc: write back cache on or off | |
875 | * @fua: device supports FUA writes, if true | |
876 | * | |
877 | * Tell the block layer about the write cache of @q. | |
878 | */ | |
879 | void blk_queue_write_cache(struct request_queue *q, bool wc, bool fua) | |
880 | { | |
881 | spin_lock_irq(q->queue_lock); | |
c888a8f9 | 882 | if (wc) |
93e9d8e8 | 883 | queue_flag_set(QUEUE_FLAG_WC, q); |
c888a8f9 | 884 | else |
93e9d8e8 | 885 | queue_flag_clear(QUEUE_FLAG_WC, q); |
c888a8f9 | 886 | if (fua) |
93e9d8e8 | 887 | queue_flag_set(QUEUE_FLAG_FUA, q); |
c888a8f9 | 888 | else |
93e9d8e8 JA |
889 | queue_flag_clear(QUEUE_FLAG_FUA, q); |
890 | spin_unlock_irq(q->queue_lock); | |
87760e5e JA |
891 | |
892 | wbt_set_write_cache(q->rq_wb, test_bit(QUEUE_FLAG_WC, &q->queue_flags)); | |
93e9d8e8 JA |
893 | } |
894 | EXPORT_SYMBOL_GPL(blk_queue_write_cache); | |
895 | ||
aeb3d3a8 | 896 | static int __init blk_settings_init(void) |
86db1e29 JA |
897 | { |
898 | blk_max_low_pfn = max_low_pfn - 1; | |
899 | blk_max_pfn = max_pfn - 1; | |
900 | return 0; | |
901 | } | |
902 | subsys_initcall(blk_settings_init); |