]>
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; |
1e739730 | 91 | lim->max_discard_segments = 1; |
13f05c8d | 92 | lim->max_integrity_segments = 0; |
e475bba2 | 93 | lim->seg_boundary_mask = BLK_SEG_BOUNDARY_MASK; |
03100aad | 94 | lim->virt_boundary_mask = 0; |
eb28d31b | 95 | lim->max_segment_size = BLK_MAX_SEGMENT_SIZE; |
5f009d3f KB |
96 | lim->max_sectors = lim->max_hw_sectors = BLK_SAFE_MAX_SECTORS; |
97 | lim->max_dev_sectors = 0; | |
762380ad | 98 | lim->chunk_sectors = 0; |
4363ac7c | 99 | lim->max_write_same_sectors = 0; |
a6f0788e | 100 | lim->max_write_zeroes_sectors = 0; |
86b37281 | 101 | lim->max_discard_sectors = 0; |
0034af03 | 102 | lim->max_hw_discard_sectors = 0; |
86b37281 MP |
103 | lim->discard_granularity = 0; |
104 | lim->discard_alignment = 0; | |
105 | lim->discard_misaligned = 0; | |
b1bd055d | 106 | lim->discard_zeroes_data = 0; |
e475bba2 | 107 | lim->logical_block_size = lim->physical_block_size = lim->io_min = 512; |
3a02c8e8 | 108 | lim->bounce_pfn = (unsigned long)(BLK_BOUNCE_ANY >> PAGE_SHIFT); |
e475bba2 MP |
109 | lim->alignment_offset = 0; |
110 | lim->io_opt = 0; | |
111 | lim->misaligned = 0; | |
e692cb66 | 112 | lim->cluster = 1; |
797476b8 | 113 | lim->zoned = BLK_ZONED_NONE; |
e475bba2 MP |
114 | } |
115 | EXPORT_SYMBOL(blk_set_default_limits); | |
116 | ||
b1bd055d MP |
117 | /** |
118 | * blk_set_stacking_limits - set default limits for stacking devices | |
119 | * @lim: the queue_limits structure to reset | |
120 | * | |
121 | * Description: | |
122 | * Returns a queue_limit struct to its default state. Should be used | |
123 | * by stacking drivers like DM that have no internal limits. | |
124 | */ | |
125 | void blk_set_stacking_limits(struct queue_limits *lim) | |
126 | { | |
127 | blk_set_default_limits(lim); | |
128 | ||
129 | /* Inherit limits from component devices */ | |
130 | lim->discard_zeroes_data = 1; | |
131 | lim->max_segments = USHRT_MAX; | |
1e739730 | 132 | lim->max_discard_segments = 1; |
b1bd055d | 133 | lim->max_hw_sectors = UINT_MAX; |
d82ae52e | 134 | lim->max_segment_size = UINT_MAX; |
fe86cdce | 135 | lim->max_sectors = UINT_MAX; |
ca369d51 | 136 | lim->max_dev_sectors = UINT_MAX; |
4363ac7c | 137 | lim->max_write_same_sectors = UINT_MAX; |
a6f0788e | 138 | lim->max_write_zeroes_sectors = UINT_MAX; |
b1bd055d MP |
139 | } |
140 | EXPORT_SYMBOL(blk_set_stacking_limits); | |
141 | ||
86db1e29 JA |
142 | /** |
143 | * blk_queue_make_request - define an alternate make_request function for a device | |
144 | * @q: the request queue for the device to be affected | |
145 | * @mfn: the alternate make_request function | |
146 | * | |
147 | * Description: | |
148 | * The normal way for &struct bios to be passed to a device | |
149 | * driver is for them to be collected into requests on a request | |
150 | * queue, and then to allow the device driver to select requests | |
151 | * off that queue when it is ready. This works well for many block | |
152 | * devices. However some block devices (typically virtual devices | |
153 | * such as md or lvm) do not benefit from the processing on the | |
154 | * request queue, and are served best by having the requests passed | |
155 | * directly to them. This can be achieved by providing a function | |
156 | * to blk_queue_make_request(). | |
157 | * | |
158 | * Caveat: | |
159 | * The driver that does this *must* be able to deal appropriately | |
160 | * with buffers in "highmemory". This can be accomplished by either calling | |
161 | * __bio_kmap_atomic() to get a temporary kernel mapping, or by calling | |
162 | * blk_queue_bounce() to create a buffer in normal memory. | |
163 | **/ | |
6728cb0e | 164 | void blk_queue_make_request(struct request_queue *q, make_request_fn *mfn) |
86db1e29 JA |
165 | { |
166 | /* | |
167 | * set defaults | |
168 | */ | |
169 | q->nr_requests = BLKDEV_MAX_RQ; | |
0e435ac2 | 170 | |
86db1e29 | 171 | q->make_request_fn = mfn; |
86db1e29 JA |
172 | blk_queue_dma_alignment(q, 511); |
173 | blk_queue_congestion_threshold(q); | |
174 | q->nr_batching = BLK_BATCH_REQ; | |
175 | ||
e475bba2 MP |
176 | blk_set_default_limits(&q->limits); |
177 | ||
86db1e29 JA |
178 | /* |
179 | * by default assume old behaviour and bounce for any highmem page | |
180 | */ | |
181 | blk_queue_bounce_limit(q, BLK_BOUNCE_HIGH); | |
182 | } | |
86db1e29 JA |
183 | EXPORT_SYMBOL(blk_queue_make_request); |
184 | ||
185 | /** | |
186 | * blk_queue_bounce_limit - set bounce buffer limit for queue | |
cd0aca2d | 187 | * @q: the request queue for the device |
9f7e45d8 | 188 | * @max_addr: the maximum address the device can handle |
86db1e29 JA |
189 | * |
190 | * Description: | |
191 | * Different hardware can have different requirements as to what pages | |
192 | * it can do I/O directly to. A low level driver can call | |
193 | * blk_queue_bounce_limit to have lower memory pages allocated as bounce | |
9f7e45d8 | 194 | * buffers for doing I/O to pages residing above @max_addr. |
86db1e29 | 195 | **/ |
9f7e45d8 | 196 | void blk_queue_bounce_limit(struct request_queue *q, u64 max_addr) |
86db1e29 | 197 | { |
9f7e45d8 | 198 | unsigned long b_pfn = max_addr >> PAGE_SHIFT; |
86db1e29 JA |
199 | int dma = 0; |
200 | ||
201 | q->bounce_gfp = GFP_NOIO; | |
202 | #if BITS_PER_LONG == 64 | |
cd0aca2d TH |
203 | /* |
204 | * Assume anything <= 4GB can be handled by IOMMU. Actually | |
205 | * some IOMMUs can handle everything, but I don't know of a | |
206 | * way to test this here. | |
207 | */ | |
208 | if (b_pfn < (min_t(u64, 0xffffffffUL, BLK_BOUNCE_HIGH) >> PAGE_SHIFT)) | |
86db1e29 | 209 | dma = 1; |
efb012b3 | 210 | q->limits.bounce_pfn = max(max_low_pfn, b_pfn); |
86db1e29 | 211 | #else |
6728cb0e | 212 | if (b_pfn < blk_max_low_pfn) |
86db1e29 | 213 | dma = 1; |
c49825fa | 214 | q->limits.bounce_pfn = b_pfn; |
260a67a9 | 215 | #endif |
86db1e29 JA |
216 | if (dma) { |
217 | init_emergency_isa_pool(); | |
218 | q->bounce_gfp = GFP_NOIO | GFP_DMA; | |
260a67a9 | 219 | q->limits.bounce_pfn = b_pfn; |
86db1e29 JA |
220 | } |
221 | } | |
86db1e29 JA |
222 | EXPORT_SYMBOL(blk_queue_bounce_limit); |
223 | ||
224 | /** | |
ca369d51 MP |
225 | * blk_queue_max_hw_sectors - set max sectors for a request for this queue |
226 | * @q: the request queue for the device | |
2800aac1 | 227 | * @max_hw_sectors: max hardware sectors in the usual 512b unit |
86db1e29 JA |
228 | * |
229 | * Description: | |
2800aac1 MP |
230 | * Enables a low level driver to set a hard upper limit, |
231 | * max_hw_sectors, on the size of requests. max_hw_sectors is set by | |
4f258a46 MP |
232 | * the device driver based upon the capabilities of the I/O |
233 | * controller. | |
2800aac1 | 234 | * |
ca369d51 MP |
235 | * max_dev_sectors is a hard limit imposed by the storage device for |
236 | * READ/WRITE requests. It is set by the disk driver. | |
237 | * | |
2800aac1 MP |
238 | * max_sectors is a soft limit imposed by the block layer for |
239 | * filesystem type requests. This value can be overridden on a | |
240 | * per-device basis in /sys/block/<device>/queue/max_sectors_kb. | |
241 | * The soft limit can not exceed max_hw_sectors. | |
86db1e29 | 242 | **/ |
ca369d51 | 243 | void blk_queue_max_hw_sectors(struct request_queue *q, unsigned int max_hw_sectors) |
86db1e29 | 244 | { |
ca369d51 MP |
245 | struct queue_limits *limits = &q->limits; |
246 | unsigned int max_sectors; | |
247 | ||
09cbfeaf KS |
248 | if ((max_hw_sectors << 9) < PAGE_SIZE) { |
249 | max_hw_sectors = 1 << (PAGE_SHIFT - 9); | |
24c03d47 | 250 | printk(KERN_INFO "%s: set to minimum %d\n", |
2800aac1 | 251 | __func__, max_hw_sectors); |
86db1e29 JA |
252 | } |
253 | ||
30e2bc08 | 254 | limits->max_hw_sectors = max_hw_sectors; |
ca369d51 MP |
255 | max_sectors = min_not_zero(max_hw_sectors, limits->max_dev_sectors); |
256 | max_sectors = min_t(unsigned int, max_sectors, BLK_DEF_MAX_SECTORS); | |
257 | limits->max_sectors = max_sectors; | |
dc3b17cc | 258 | q->backing_dev_info->io_pages = max_sectors >> (PAGE_SHIFT - 9); |
86db1e29 | 259 | } |
086fa5ff | 260 | EXPORT_SYMBOL(blk_queue_max_hw_sectors); |
86db1e29 | 261 | |
762380ad JA |
262 | /** |
263 | * blk_queue_chunk_sectors - set size of the chunk for this queue | |
264 | * @q: the request queue for the device | |
265 | * @chunk_sectors: chunk sectors in the usual 512b unit | |
266 | * | |
267 | * Description: | |
268 | * If a driver doesn't want IOs to cross a given chunk size, it can set | |
269 | * this limit and prevent merging across chunks. Note that the chunk size | |
58a4915a JA |
270 | * must currently be a power-of-2 in sectors. Also note that the block |
271 | * layer must accept a page worth of data at any offset. So if the | |
272 | * crossing of chunks is a hard limitation in the driver, it must still be | |
273 | * prepared to split single page bios. | |
762380ad JA |
274 | **/ |
275 | void blk_queue_chunk_sectors(struct request_queue *q, unsigned int chunk_sectors) | |
276 | { | |
277 | BUG_ON(!is_power_of_2(chunk_sectors)); | |
278 | q->limits.chunk_sectors = chunk_sectors; | |
279 | } | |
280 | EXPORT_SYMBOL(blk_queue_chunk_sectors); | |
281 | ||
67efc925 CH |
282 | /** |
283 | * blk_queue_max_discard_sectors - set max sectors for a single discard | |
284 | * @q: the request queue for the device | |
c7ebf065 | 285 | * @max_discard_sectors: maximum number of sectors to discard |
67efc925 CH |
286 | **/ |
287 | void blk_queue_max_discard_sectors(struct request_queue *q, | |
288 | unsigned int max_discard_sectors) | |
289 | { | |
0034af03 | 290 | q->limits.max_hw_discard_sectors = max_discard_sectors; |
67efc925 CH |
291 | q->limits.max_discard_sectors = max_discard_sectors; |
292 | } | |
293 | EXPORT_SYMBOL(blk_queue_max_discard_sectors); | |
294 | ||
4363ac7c MP |
295 | /** |
296 | * blk_queue_max_write_same_sectors - set max sectors for a single write same | |
297 | * @q: the request queue for the device | |
298 | * @max_write_same_sectors: maximum number of sectors to write per command | |
299 | **/ | |
300 | void blk_queue_max_write_same_sectors(struct request_queue *q, | |
301 | unsigned int max_write_same_sectors) | |
302 | { | |
303 | q->limits.max_write_same_sectors = max_write_same_sectors; | |
304 | } | |
305 | EXPORT_SYMBOL(blk_queue_max_write_same_sectors); | |
306 | ||
a6f0788e CK |
307 | /** |
308 | * blk_queue_max_write_zeroes_sectors - set max sectors for a single | |
309 | * write zeroes | |
310 | * @q: the request queue for the device | |
311 | * @max_write_zeroes_sectors: maximum number of sectors to write per command | |
312 | **/ | |
313 | void blk_queue_max_write_zeroes_sectors(struct request_queue *q, | |
314 | unsigned int max_write_zeroes_sectors) | |
315 | { | |
316 | q->limits.max_write_zeroes_sectors = max_write_zeroes_sectors; | |
317 | } | |
318 | EXPORT_SYMBOL(blk_queue_max_write_zeroes_sectors); | |
319 | ||
86db1e29 | 320 | /** |
8a78362c | 321 | * blk_queue_max_segments - set max hw segments for a request for this queue |
86db1e29 JA |
322 | * @q: the request queue for the device |
323 | * @max_segments: max number of segments | |
324 | * | |
325 | * Description: | |
326 | * Enables a low level driver to set an upper limit on the number of | |
8a78362c | 327 | * hw data segments in a request. |
86db1e29 | 328 | **/ |
8a78362c | 329 | void blk_queue_max_segments(struct request_queue *q, unsigned short max_segments) |
86db1e29 JA |
330 | { |
331 | if (!max_segments) { | |
332 | max_segments = 1; | |
24c03d47 HH |
333 | printk(KERN_INFO "%s: set to minimum %d\n", |
334 | __func__, max_segments); | |
86db1e29 JA |
335 | } |
336 | ||
8a78362c | 337 | q->limits.max_segments = max_segments; |
86db1e29 | 338 | } |
8a78362c | 339 | EXPORT_SYMBOL(blk_queue_max_segments); |
86db1e29 | 340 | |
1e739730 CH |
341 | /** |
342 | * blk_queue_max_discard_segments - set max segments for discard requests | |
343 | * @q: the request queue for the device | |
344 | * @max_segments: max number of segments | |
345 | * | |
346 | * Description: | |
347 | * Enables a low level driver to set an upper limit on the number of | |
348 | * segments in a discard request. | |
349 | **/ | |
350 | void blk_queue_max_discard_segments(struct request_queue *q, | |
351 | unsigned short max_segments) | |
352 | { | |
353 | q->limits.max_discard_segments = max_segments; | |
354 | } | |
355 | EXPORT_SYMBOL_GPL(blk_queue_max_discard_segments); | |
356 | ||
86db1e29 JA |
357 | /** |
358 | * blk_queue_max_segment_size - set max segment size for blk_rq_map_sg | |
359 | * @q: the request queue for the device | |
360 | * @max_size: max size of segment in bytes | |
361 | * | |
362 | * Description: | |
363 | * Enables a low level driver to set an upper limit on the size of a | |
364 | * coalesced segment | |
365 | **/ | |
366 | void blk_queue_max_segment_size(struct request_queue *q, unsigned int max_size) | |
367 | { | |
09cbfeaf KS |
368 | if (max_size < PAGE_SIZE) { |
369 | max_size = PAGE_SIZE; | |
24c03d47 HH |
370 | printk(KERN_INFO "%s: set to minimum %d\n", |
371 | __func__, max_size); | |
86db1e29 JA |
372 | } |
373 | ||
025146e1 | 374 | q->limits.max_segment_size = max_size; |
86db1e29 | 375 | } |
86db1e29 JA |
376 | EXPORT_SYMBOL(blk_queue_max_segment_size); |
377 | ||
378 | /** | |
e1defc4f | 379 | * blk_queue_logical_block_size - set logical block size for the queue |
86db1e29 | 380 | * @q: the request queue for the device |
e1defc4f | 381 | * @size: the logical block size, in bytes |
86db1e29 JA |
382 | * |
383 | * Description: | |
e1defc4f MP |
384 | * This should be set to the lowest possible block size that the |
385 | * storage device can address. The default of 512 covers most | |
386 | * hardware. | |
86db1e29 | 387 | **/ |
e1defc4f | 388 | void blk_queue_logical_block_size(struct request_queue *q, unsigned short size) |
86db1e29 | 389 | { |
025146e1 | 390 | q->limits.logical_block_size = size; |
c72758f3 MP |
391 | |
392 | if (q->limits.physical_block_size < size) | |
393 | q->limits.physical_block_size = size; | |
394 | ||
395 | if (q->limits.io_min < q->limits.physical_block_size) | |
396 | q->limits.io_min = q->limits.physical_block_size; | |
86db1e29 | 397 | } |
e1defc4f | 398 | EXPORT_SYMBOL(blk_queue_logical_block_size); |
86db1e29 | 399 | |
c72758f3 MP |
400 | /** |
401 | * blk_queue_physical_block_size - set physical block size for the queue | |
402 | * @q: the request queue for the device | |
403 | * @size: the physical block size, in bytes | |
404 | * | |
405 | * Description: | |
406 | * This should be set to the lowest possible sector size that the | |
407 | * hardware can operate on without reverting to read-modify-write | |
408 | * operations. | |
409 | */ | |
892b6f90 | 410 | void blk_queue_physical_block_size(struct request_queue *q, unsigned int size) |
c72758f3 MP |
411 | { |
412 | q->limits.physical_block_size = size; | |
413 | ||
414 | if (q->limits.physical_block_size < q->limits.logical_block_size) | |
415 | q->limits.physical_block_size = q->limits.logical_block_size; | |
416 | ||
417 | if (q->limits.io_min < q->limits.physical_block_size) | |
418 | q->limits.io_min = q->limits.physical_block_size; | |
419 | } | |
420 | EXPORT_SYMBOL(blk_queue_physical_block_size); | |
421 | ||
422 | /** | |
423 | * blk_queue_alignment_offset - set physical block alignment offset | |
424 | * @q: the request queue for the device | |
8ebf9756 | 425 | * @offset: alignment offset in bytes |
c72758f3 MP |
426 | * |
427 | * Description: | |
428 | * Some devices are naturally misaligned to compensate for things like | |
429 | * the legacy DOS partition table 63-sector offset. Low-level drivers | |
430 | * should call this function for devices whose first sector is not | |
431 | * naturally aligned. | |
432 | */ | |
433 | void blk_queue_alignment_offset(struct request_queue *q, unsigned int offset) | |
434 | { | |
435 | q->limits.alignment_offset = | |
436 | offset & (q->limits.physical_block_size - 1); | |
437 | q->limits.misaligned = 0; | |
438 | } | |
439 | EXPORT_SYMBOL(blk_queue_alignment_offset); | |
440 | ||
7c958e32 MP |
441 | /** |
442 | * blk_limits_io_min - set minimum request size for a device | |
443 | * @limits: the queue limits | |
444 | * @min: smallest I/O size in bytes | |
445 | * | |
446 | * Description: | |
447 | * Some devices have an internal block size bigger than the reported | |
448 | * hardware sector size. This function can be used to signal the | |
449 | * smallest I/O the device can perform without incurring a performance | |
450 | * penalty. | |
451 | */ | |
452 | void blk_limits_io_min(struct queue_limits *limits, unsigned int min) | |
453 | { | |
454 | limits->io_min = min; | |
455 | ||
456 | if (limits->io_min < limits->logical_block_size) | |
457 | limits->io_min = limits->logical_block_size; | |
458 | ||
459 | if (limits->io_min < limits->physical_block_size) | |
460 | limits->io_min = limits->physical_block_size; | |
461 | } | |
462 | EXPORT_SYMBOL(blk_limits_io_min); | |
463 | ||
c72758f3 MP |
464 | /** |
465 | * blk_queue_io_min - set minimum request size for the queue | |
466 | * @q: the request queue for the device | |
8ebf9756 | 467 | * @min: smallest I/O size in bytes |
c72758f3 MP |
468 | * |
469 | * Description: | |
7e5f5fb0 MP |
470 | * Storage devices may report a granularity or preferred minimum I/O |
471 | * size which is the smallest request the device can perform without | |
472 | * incurring a performance penalty. For disk drives this is often the | |
473 | * physical block size. For RAID arrays it is often the stripe chunk | |
474 | * size. A properly aligned multiple of minimum_io_size is the | |
475 | * preferred request size for workloads where a high number of I/O | |
476 | * operations is desired. | |
c72758f3 MP |
477 | */ |
478 | void blk_queue_io_min(struct request_queue *q, unsigned int min) | |
479 | { | |
7c958e32 | 480 | blk_limits_io_min(&q->limits, min); |
c72758f3 MP |
481 | } |
482 | EXPORT_SYMBOL(blk_queue_io_min); | |
483 | ||
3c5820c7 MP |
484 | /** |
485 | * blk_limits_io_opt - set optimal request size for a device | |
486 | * @limits: the queue limits | |
487 | * @opt: smallest I/O size in bytes | |
488 | * | |
489 | * Description: | |
490 | * Storage devices may report an optimal I/O size, which is the | |
491 | * device's preferred unit for sustained I/O. This is rarely reported | |
492 | * for disk drives. For RAID arrays it is usually the stripe width or | |
493 | * the internal track size. A properly aligned multiple of | |
494 | * optimal_io_size is the preferred request size for workloads where | |
495 | * sustained throughput is desired. | |
496 | */ | |
497 | void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt) | |
498 | { | |
499 | limits->io_opt = opt; | |
500 | } | |
501 | EXPORT_SYMBOL(blk_limits_io_opt); | |
502 | ||
c72758f3 MP |
503 | /** |
504 | * blk_queue_io_opt - set optimal request size for the queue | |
505 | * @q: the request queue for the device | |
8ebf9756 | 506 | * @opt: optimal request size in bytes |
c72758f3 MP |
507 | * |
508 | * Description: | |
7e5f5fb0 MP |
509 | * Storage devices may report an optimal I/O size, which is the |
510 | * device's preferred unit for sustained I/O. This is rarely reported | |
511 | * for disk drives. For RAID arrays it is usually the stripe width or | |
512 | * the internal track size. A properly aligned multiple of | |
513 | * optimal_io_size is the preferred request size for workloads where | |
514 | * sustained throughput is desired. | |
c72758f3 MP |
515 | */ |
516 | void blk_queue_io_opt(struct request_queue *q, unsigned int opt) | |
517 | { | |
3c5820c7 | 518 | blk_limits_io_opt(&q->limits, opt); |
c72758f3 MP |
519 | } |
520 | EXPORT_SYMBOL(blk_queue_io_opt); | |
521 | ||
86db1e29 JA |
522 | /** |
523 | * blk_queue_stack_limits - inherit underlying queue limits for stacked drivers | |
524 | * @t: the stacking driver (top) | |
525 | * @b: the underlying device (bottom) | |
526 | **/ | |
527 | void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b) | |
528 | { | |
fef24667 | 529 | blk_stack_limits(&t->limits, &b->limits, 0); |
86db1e29 | 530 | } |
86db1e29 JA |
531 | EXPORT_SYMBOL(blk_queue_stack_limits); |
532 | ||
c72758f3 MP |
533 | /** |
534 | * blk_stack_limits - adjust queue_limits for stacked devices | |
81744ee4 MP |
535 | * @t: the stacking driver limits (top device) |
536 | * @b: the underlying queue limits (bottom, component device) | |
e03a72e1 | 537 | * @start: first data sector within component device |
c72758f3 MP |
538 | * |
539 | * Description: | |
81744ee4 MP |
540 | * This function is used by stacking drivers like MD and DM to ensure |
541 | * that all component devices have compatible block sizes and | |
542 | * alignments. The stacking driver must provide a queue_limits | |
543 | * struct (top) and then iteratively call the stacking function for | |
544 | * all component (bottom) devices. The stacking function will | |
545 | * attempt to combine the values and ensure proper alignment. | |
546 | * | |
547 | * Returns 0 if the top and bottom queue_limits are compatible. The | |
548 | * top device's block sizes and alignment offsets may be adjusted to | |
549 | * ensure alignment with the bottom device. If no compatible sizes | |
550 | * and alignments exist, -1 is returned and the resulting top | |
551 | * queue_limits will have the misaligned flag set to indicate that | |
552 | * the alignment_offset is undefined. | |
c72758f3 MP |
553 | */ |
554 | int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, | |
e03a72e1 | 555 | sector_t start) |
c72758f3 | 556 | { |
e03a72e1 | 557 | unsigned int top, bottom, alignment, ret = 0; |
86b37281 | 558 | |
c72758f3 MP |
559 | t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors); |
560 | t->max_hw_sectors = min_not_zero(t->max_hw_sectors, b->max_hw_sectors); | |
ca369d51 | 561 | t->max_dev_sectors = min_not_zero(t->max_dev_sectors, b->max_dev_sectors); |
4363ac7c MP |
562 | t->max_write_same_sectors = min(t->max_write_same_sectors, |
563 | b->max_write_same_sectors); | |
a6f0788e CK |
564 | t->max_write_zeroes_sectors = min(t->max_write_zeroes_sectors, |
565 | b->max_write_zeroes_sectors); | |
77634f33 | 566 | t->bounce_pfn = min_not_zero(t->bounce_pfn, b->bounce_pfn); |
c72758f3 MP |
567 | |
568 | t->seg_boundary_mask = min_not_zero(t->seg_boundary_mask, | |
569 | b->seg_boundary_mask); | |
03100aad KB |
570 | t->virt_boundary_mask = min_not_zero(t->virt_boundary_mask, |
571 | b->virt_boundary_mask); | |
c72758f3 | 572 | |
8a78362c | 573 | t->max_segments = min_not_zero(t->max_segments, b->max_segments); |
1e739730 CH |
574 | t->max_discard_segments = min_not_zero(t->max_discard_segments, |
575 | b->max_discard_segments); | |
13f05c8d MP |
576 | t->max_integrity_segments = min_not_zero(t->max_integrity_segments, |
577 | b->max_integrity_segments); | |
c72758f3 MP |
578 | |
579 | t->max_segment_size = min_not_zero(t->max_segment_size, | |
580 | b->max_segment_size); | |
581 | ||
fe0b393f MP |
582 | t->misaligned |= b->misaligned; |
583 | ||
e03a72e1 | 584 | alignment = queue_limit_alignment_offset(b, start); |
9504e086 | 585 | |
81744ee4 MP |
586 | /* Bottom device has different alignment. Check that it is |
587 | * compatible with the current top alignment. | |
588 | */ | |
9504e086 MP |
589 | if (t->alignment_offset != alignment) { |
590 | ||
591 | top = max(t->physical_block_size, t->io_min) | |
592 | + t->alignment_offset; | |
81744ee4 | 593 | bottom = max(b->physical_block_size, b->io_min) + alignment; |
9504e086 | 594 | |
81744ee4 | 595 | /* Verify that top and bottom intervals line up */ |
b8839b8c | 596 | if (max(top, bottom) % min(top, bottom)) { |
9504e086 | 597 | t->misaligned = 1; |
fe0b393f MP |
598 | ret = -1; |
599 | } | |
9504e086 MP |
600 | } |
601 | ||
c72758f3 MP |
602 | t->logical_block_size = max(t->logical_block_size, |
603 | b->logical_block_size); | |
604 | ||
605 | t->physical_block_size = max(t->physical_block_size, | |
606 | b->physical_block_size); | |
607 | ||
608 | t->io_min = max(t->io_min, b->io_min); | |
e9637415 | 609 | t->io_opt = lcm_not_zero(t->io_opt, b->io_opt); |
9504e086 | 610 | |
e692cb66 | 611 | t->cluster &= b->cluster; |
98262f27 | 612 | t->discard_zeroes_data &= b->discard_zeroes_data; |
c72758f3 | 613 | |
81744ee4 | 614 | /* Physical block size a multiple of the logical block size? */ |
9504e086 MP |
615 | if (t->physical_block_size & (t->logical_block_size - 1)) { |
616 | t->physical_block_size = t->logical_block_size; | |
c72758f3 | 617 | t->misaligned = 1; |
fe0b393f | 618 | ret = -1; |
86b37281 MP |
619 | } |
620 | ||
81744ee4 | 621 | /* Minimum I/O a multiple of the physical block size? */ |
9504e086 MP |
622 | if (t->io_min & (t->physical_block_size - 1)) { |
623 | t->io_min = t->physical_block_size; | |
624 | t->misaligned = 1; | |
fe0b393f | 625 | ret = -1; |
c72758f3 MP |
626 | } |
627 | ||
81744ee4 | 628 | /* Optimal I/O a multiple of the physical block size? */ |
9504e086 MP |
629 | if (t->io_opt & (t->physical_block_size - 1)) { |
630 | t->io_opt = 0; | |
631 | t->misaligned = 1; | |
fe0b393f | 632 | ret = -1; |
9504e086 | 633 | } |
c72758f3 | 634 | |
c78afc62 KO |
635 | t->raid_partial_stripes_expensive = |
636 | max(t->raid_partial_stripes_expensive, | |
637 | b->raid_partial_stripes_expensive); | |
638 | ||
81744ee4 | 639 | /* Find lowest common alignment_offset */ |
e9637415 | 640 | t->alignment_offset = lcm_not_zero(t->alignment_offset, alignment) |
b8839b8c | 641 | % max(t->physical_block_size, t->io_min); |
86b37281 | 642 | |
81744ee4 | 643 | /* Verify that new alignment_offset is on a logical block boundary */ |
fe0b393f | 644 | if (t->alignment_offset & (t->logical_block_size - 1)) { |
c72758f3 | 645 | t->misaligned = 1; |
fe0b393f MP |
646 | ret = -1; |
647 | } | |
c72758f3 | 648 | |
9504e086 MP |
649 | /* Discard alignment and granularity */ |
650 | if (b->discard_granularity) { | |
e03a72e1 | 651 | alignment = queue_limit_discard_alignment(b, start); |
9504e086 MP |
652 | |
653 | if (t->discard_granularity != 0 && | |
654 | t->discard_alignment != alignment) { | |
655 | top = t->discard_granularity + t->discard_alignment; | |
656 | bottom = b->discard_granularity + alignment; | |
70dd5bf3 | 657 | |
9504e086 | 658 | /* Verify that top and bottom intervals line up */ |
8dd2cb7e | 659 | if ((max(top, bottom) % min(top, bottom)) != 0) |
9504e086 MP |
660 | t->discard_misaligned = 1; |
661 | } | |
662 | ||
81744ee4 MP |
663 | t->max_discard_sectors = min_not_zero(t->max_discard_sectors, |
664 | b->max_discard_sectors); | |
0034af03 JA |
665 | t->max_hw_discard_sectors = min_not_zero(t->max_hw_discard_sectors, |
666 | b->max_hw_discard_sectors); | |
9504e086 MP |
667 | t->discard_granularity = max(t->discard_granularity, |
668 | b->discard_granularity); | |
e9637415 | 669 | t->discard_alignment = lcm_not_zero(t->discard_alignment, alignment) % |
8dd2cb7e | 670 | t->discard_granularity; |
9504e086 | 671 | } |
70dd5bf3 | 672 | |
987b3b26 HR |
673 | if (b->chunk_sectors) |
674 | t->chunk_sectors = min_not_zero(t->chunk_sectors, | |
675 | b->chunk_sectors); | |
676 | ||
fe0b393f | 677 | return ret; |
c72758f3 | 678 | } |
5d85d324 | 679 | EXPORT_SYMBOL(blk_stack_limits); |
c72758f3 | 680 | |
17be8c24 MP |
681 | /** |
682 | * bdev_stack_limits - adjust queue limits for stacked drivers | |
683 | * @t: the stacking driver limits (top device) | |
684 | * @bdev: the component block_device (bottom) | |
685 | * @start: first data sector within component device | |
686 | * | |
687 | * Description: | |
688 | * Merges queue limits for a top device and a block_device. Returns | |
689 | * 0 if alignment didn't change. Returns -1 if adding the bottom | |
690 | * device caused misalignment. | |
691 | */ | |
692 | int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev, | |
693 | sector_t start) | |
694 | { | |
695 | struct request_queue *bq = bdev_get_queue(bdev); | |
696 | ||
697 | start += get_start_sect(bdev); | |
698 | ||
e03a72e1 | 699 | return blk_stack_limits(t, &bq->limits, start); |
17be8c24 MP |
700 | } |
701 | EXPORT_SYMBOL(bdev_stack_limits); | |
702 | ||
c72758f3 MP |
703 | /** |
704 | * disk_stack_limits - adjust queue limits for stacked drivers | |
77634f33 | 705 | * @disk: MD/DM gendisk (top) |
c72758f3 MP |
706 | * @bdev: the underlying block device (bottom) |
707 | * @offset: offset to beginning of data within component device | |
708 | * | |
709 | * Description: | |
e03a72e1 MP |
710 | * Merges the limits for a top level gendisk and a bottom level |
711 | * block_device. | |
c72758f3 MP |
712 | */ |
713 | void disk_stack_limits(struct gendisk *disk, struct block_device *bdev, | |
714 | sector_t offset) | |
715 | { | |
716 | struct request_queue *t = disk->queue; | |
c72758f3 | 717 | |
e03a72e1 | 718 | if (bdev_stack_limits(&t->limits, bdev, offset >> 9) < 0) { |
c72758f3 MP |
719 | char top[BDEVNAME_SIZE], bottom[BDEVNAME_SIZE]; |
720 | ||
721 | disk_name(disk, 0, top); | |
722 | bdevname(bdev, bottom); | |
723 | ||
724 | printk(KERN_NOTICE "%s: Warning: Device %s is misaligned\n", | |
725 | top, bottom); | |
726 | } | |
c72758f3 MP |
727 | } |
728 | EXPORT_SYMBOL(disk_stack_limits); | |
729 | ||
e3790c7d TH |
730 | /** |
731 | * blk_queue_dma_pad - set pad mask | |
732 | * @q: the request queue for the device | |
733 | * @mask: pad mask | |
734 | * | |
27f8221a | 735 | * Set dma pad mask. |
e3790c7d | 736 | * |
27f8221a FT |
737 | * Appending pad buffer to a request modifies the last entry of a |
738 | * scatter list such that it includes the pad buffer. | |
e3790c7d TH |
739 | **/ |
740 | void blk_queue_dma_pad(struct request_queue *q, unsigned int mask) | |
741 | { | |
742 | q->dma_pad_mask = mask; | |
743 | } | |
744 | EXPORT_SYMBOL(blk_queue_dma_pad); | |
745 | ||
27f8221a FT |
746 | /** |
747 | * blk_queue_update_dma_pad - update pad mask | |
748 | * @q: the request queue for the device | |
749 | * @mask: pad mask | |
750 | * | |
751 | * Update dma pad mask. | |
752 | * | |
753 | * Appending pad buffer to a request modifies the last entry of a | |
754 | * scatter list such that it includes the pad buffer. | |
755 | **/ | |
756 | void blk_queue_update_dma_pad(struct request_queue *q, unsigned int mask) | |
757 | { | |
758 | if (mask > q->dma_pad_mask) | |
759 | q->dma_pad_mask = mask; | |
760 | } | |
761 | EXPORT_SYMBOL(blk_queue_update_dma_pad); | |
762 | ||
86db1e29 JA |
763 | /** |
764 | * blk_queue_dma_drain - Set up a drain buffer for excess dma. | |
86db1e29 | 765 | * @q: the request queue for the device |
2fb98e84 | 766 | * @dma_drain_needed: fn which returns non-zero if drain is necessary |
86db1e29 JA |
767 | * @buf: physically contiguous buffer |
768 | * @size: size of the buffer in bytes | |
769 | * | |
770 | * Some devices have excess DMA problems and can't simply discard (or | |
771 | * zero fill) the unwanted piece of the transfer. They have to have a | |
772 | * real area of memory to transfer it into. The use case for this is | |
773 | * ATAPI devices in DMA mode. If the packet command causes a transfer | |
774 | * bigger than the transfer size some HBAs will lock up if there | |
775 | * aren't DMA elements to contain the excess transfer. What this API | |
776 | * does is adjust the queue so that the buf is always appended | |
777 | * silently to the scatterlist. | |
778 | * | |
8a78362c MP |
779 | * Note: This routine adjusts max_hw_segments to make room for appending |
780 | * the drain buffer. If you call blk_queue_max_segments() after calling | |
781 | * this routine, you must set the limit to one fewer than your device | |
782 | * can support otherwise there won't be room for the drain buffer. | |
86db1e29 | 783 | */ |
448da4d2 | 784 | int blk_queue_dma_drain(struct request_queue *q, |
2fb98e84 TH |
785 | dma_drain_needed_fn *dma_drain_needed, |
786 | void *buf, unsigned int size) | |
86db1e29 | 787 | { |
8a78362c | 788 | if (queue_max_segments(q) < 2) |
86db1e29 JA |
789 | return -EINVAL; |
790 | /* make room for appending the drain */ | |
8a78362c | 791 | blk_queue_max_segments(q, queue_max_segments(q) - 1); |
2fb98e84 | 792 | q->dma_drain_needed = dma_drain_needed; |
86db1e29 JA |
793 | q->dma_drain_buffer = buf; |
794 | q->dma_drain_size = size; | |
795 | ||
796 | return 0; | |
797 | } | |
86db1e29 JA |
798 | EXPORT_SYMBOL_GPL(blk_queue_dma_drain); |
799 | ||
800 | /** | |
801 | * blk_queue_segment_boundary - set boundary rules for segment merging | |
802 | * @q: the request queue for the device | |
803 | * @mask: the memory boundary mask | |
804 | **/ | |
805 | void blk_queue_segment_boundary(struct request_queue *q, unsigned long mask) | |
806 | { | |
09cbfeaf KS |
807 | if (mask < PAGE_SIZE - 1) { |
808 | mask = PAGE_SIZE - 1; | |
24c03d47 HH |
809 | printk(KERN_INFO "%s: set to minimum %lx\n", |
810 | __func__, mask); | |
86db1e29 JA |
811 | } |
812 | ||
025146e1 | 813 | q->limits.seg_boundary_mask = mask; |
86db1e29 | 814 | } |
86db1e29 JA |
815 | EXPORT_SYMBOL(blk_queue_segment_boundary); |
816 | ||
03100aad KB |
817 | /** |
818 | * blk_queue_virt_boundary - set boundary rules for bio merging | |
819 | * @q: the request queue for the device | |
820 | * @mask: the memory boundary mask | |
821 | **/ | |
822 | void blk_queue_virt_boundary(struct request_queue *q, unsigned long mask) | |
823 | { | |
824 | q->limits.virt_boundary_mask = mask; | |
825 | } | |
826 | EXPORT_SYMBOL(blk_queue_virt_boundary); | |
827 | ||
86db1e29 JA |
828 | /** |
829 | * blk_queue_dma_alignment - set dma length and memory alignment | |
830 | * @q: the request queue for the device | |
831 | * @mask: alignment mask | |
832 | * | |
833 | * description: | |
710027a4 | 834 | * set required memory and length alignment for direct dma transactions. |
8feb4d20 | 835 | * this is used when building direct io requests for the queue. |
86db1e29 JA |
836 | * |
837 | **/ | |
838 | void blk_queue_dma_alignment(struct request_queue *q, int mask) | |
839 | { | |
840 | q->dma_alignment = mask; | |
841 | } | |
86db1e29 JA |
842 | EXPORT_SYMBOL(blk_queue_dma_alignment); |
843 | ||
844 | /** | |
845 | * blk_queue_update_dma_alignment - update dma length and memory alignment | |
846 | * @q: the request queue for the device | |
847 | * @mask: alignment mask | |
848 | * | |
849 | * description: | |
710027a4 | 850 | * update required memory and length alignment for direct dma transactions. |
86db1e29 JA |
851 | * If the requested alignment is larger than the current alignment, then |
852 | * the current queue alignment is updated to the new value, otherwise it | |
853 | * is left alone. The design of this is to allow multiple objects | |
854 | * (driver, device, transport etc) to set their respective | |
855 | * alignments without having them interfere. | |
856 | * | |
857 | **/ | |
858 | void blk_queue_update_dma_alignment(struct request_queue *q, int mask) | |
859 | { | |
860 | BUG_ON(mask > PAGE_SIZE); | |
861 | ||
862 | if (mask > q->dma_alignment) | |
863 | q->dma_alignment = mask; | |
864 | } | |
86db1e29 JA |
865 | EXPORT_SYMBOL(blk_queue_update_dma_alignment); |
866 | ||
f3876930 | 867 | void blk_queue_flush_queueable(struct request_queue *q, bool queueable) |
868 | { | |
c888a8f9 JA |
869 | spin_lock_irq(q->queue_lock); |
870 | if (queueable) | |
871 | clear_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags); | |
872 | else | |
873 | set_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags); | |
874 | spin_unlock_irq(q->queue_lock); | |
f3876930 | 875 | } |
876 | EXPORT_SYMBOL_GPL(blk_queue_flush_queueable); | |
877 | ||
d278d4a8 JA |
878 | /** |
879 | * blk_set_queue_depth - tell the block layer about the device queue depth | |
880 | * @q: the request queue for the device | |
881 | * @depth: queue depth | |
882 | * | |
883 | */ | |
884 | void blk_set_queue_depth(struct request_queue *q, unsigned int depth) | |
885 | { | |
886 | q->queue_depth = depth; | |
87760e5e | 887 | wbt_set_queue_depth(q->rq_wb, depth); |
d278d4a8 JA |
888 | } |
889 | EXPORT_SYMBOL(blk_set_queue_depth); | |
890 | ||
93e9d8e8 JA |
891 | /** |
892 | * blk_queue_write_cache - configure queue's write cache | |
893 | * @q: the request queue for the device | |
894 | * @wc: write back cache on or off | |
895 | * @fua: device supports FUA writes, if true | |
896 | * | |
897 | * Tell the block layer about the write cache of @q. | |
898 | */ | |
899 | void blk_queue_write_cache(struct request_queue *q, bool wc, bool fua) | |
900 | { | |
901 | spin_lock_irq(q->queue_lock); | |
c888a8f9 | 902 | if (wc) |
93e9d8e8 | 903 | queue_flag_set(QUEUE_FLAG_WC, q); |
c888a8f9 | 904 | else |
93e9d8e8 | 905 | queue_flag_clear(QUEUE_FLAG_WC, q); |
c888a8f9 | 906 | if (fua) |
93e9d8e8 | 907 | queue_flag_set(QUEUE_FLAG_FUA, q); |
c888a8f9 | 908 | else |
93e9d8e8 JA |
909 | queue_flag_clear(QUEUE_FLAG_FUA, q); |
910 | spin_unlock_irq(q->queue_lock); | |
87760e5e JA |
911 | |
912 | wbt_set_write_cache(q->rq_wb, test_bit(QUEUE_FLAG_WC, &q->queue_flags)); | |
93e9d8e8 JA |
913 | } |
914 | EXPORT_SYMBOL_GPL(blk_queue_write_cache); | |
915 | ||
aeb3d3a8 | 916 | static int __init blk_settings_init(void) |
86db1e29 JA |
917 | { |
918 | blk_max_low_pfn = max_low_pfn - 1; | |
919 | blk_max_pfn = max_pfn - 1; | |
920 | return 0; | |
921 | } | |
922 | subsys_initcall(blk_settings_init); |