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" | |
16 | ||
6728cb0e | 17 | unsigned long blk_max_low_pfn; |
86db1e29 | 18 | EXPORT_SYMBOL(blk_max_low_pfn); |
6728cb0e JA |
19 | |
20 | unsigned long blk_max_pfn; | |
86db1e29 JA |
21 | |
22 | /** | |
23 | * blk_queue_prep_rq - set a prepare_request function for queue | |
24 | * @q: queue | |
25 | * @pfn: prepare_request function | |
26 | * | |
27 | * It's possible for a queue to register a prepare_request callback which | |
28 | * is invoked before the request is handed to the request_fn. The goal of | |
29 | * the function is to prepare a request for I/O, it can be used to build a | |
30 | * cdb from the request data for instance. | |
31 | * | |
32 | */ | |
33 | void blk_queue_prep_rq(struct request_queue *q, prep_rq_fn *pfn) | |
34 | { | |
35 | q->prep_rq_fn = pfn; | |
36 | } | |
86db1e29 JA |
37 | EXPORT_SYMBOL(blk_queue_prep_rq); |
38 | ||
28018c24 JB |
39 | /** |
40 | * blk_queue_unprep_rq - set an unprepare_request function for queue | |
41 | * @q: queue | |
42 | * @ufn: unprepare_request function | |
43 | * | |
44 | * It's possible for a queue to register an unprepare_request callback | |
45 | * which is invoked before the request is finally completed. The goal | |
46 | * of the function is to deallocate any data that was allocated in the | |
47 | * prepare_request callback. | |
48 | * | |
49 | */ | |
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 | ||
86db1e29 JA |
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 | */ | |
72 | void blk_queue_merge_bvec(struct request_queue *q, merge_bvec_fn *mbfn) | |
73 | { | |
74 | q->merge_bvec_fn = mbfn; | |
75 | } | |
86db1e29 JA |
76 | EXPORT_SYMBOL(blk_queue_merge_bvec); |
77 | ||
78 | void blk_queue_softirq_done(struct request_queue *q, softirq_done_fn *fn) | |
79 | { | |
80 | q->softirq_done_fn = fn; | |
81 | } | |
86db1e29 JA |
82 | EXPORT_SYMBOL(blk_queue_softirq_done); |
83 | ||
242f9dcb JA |
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 | ||
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 | ||
ef9e3fac KU |
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 | ||
e475bba2 MP |
102 | /** |
103 | * blk_set_default_limits - reset limits to default values | |
f740f5ca | 104 | * @lim: the queue_limits structure to reset |
e475bba2 MP |
105 | * |
106 | * Description: | |
b1bd055d | 107 | * Returns a queue_limit struct to its default state. |
e475bba2 MP |
108 | */ |
109 | void blk_set_default_limits(struct queue_limits *lim) | |
110 | { | |
8a78362c | 111 | lim->max_segments = BLK_MAX_SEGMENTS; |
13f05c8d | 112 | lim->max_integrity_segments = 0; |
e475bba2 | 113 | lim->seg_boundary_mask = BLK_SEG_BOUNDARY_MASK; |
eb28d31b | 114 | lim->max_segment_size = BLK_MAX_SEGMENT_SIZE; |
b1bd055d | 115 | lim->max_sectors = lim->max_hw_sectors = BLK_SAFE_MAX_SECTORS; |
86b37281 MP |
116 | lim->max_discard_sectors = 0; |
117 | lim->discard_granularity = 0; | |
118 | lim->discard_alignment = 0; | |
119 | lim->discard_misaligned = 0; | |
b1bd055d | 120 | lim->discard_zeroes_data = 0; |
e475bba2 | 121 | lim->logical_block_size = lim->physical_block_size = lim->io_min = 512; |
3a02c8e8 | 122 | lim->bounce_pfn = (unsigned long)(BLK_BOUNCE_ANY >> PAGE_SHIFT); |
e475bba2 MP |
123 | lim->alignment_offset = 0; |
124 | lim->io_opt = 0; | |
125 | lim->misaligned = 0; | |
e692cb66 | 126 | lim->cluster = 1; |
e475bba2 MP |
127 | } |
128 | EXPORT_SYMBOL(blk_set_default_limits); | |
129 | ||
b1bd055d MP |
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 | */ | |
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 | ||
147 | lim->max_sectors = BLK_DEF_MAX_SECTORS; | |
148 | } | |
149 | EXPORT_SYMBOL(blk_set_stacking_limits); | |
150 | ||
86db1e29 JA |
151 | /** |
152 | * blk_queue_make_request - define an alternate make_request function for a device | |
153 | * @q: the request queue for the device to be affected | |
154 | * @mfn: the alternate make_request function | |
155 | * | |
156 | * Description: | |
157 | * The normal way for &struct bios to be passed to a device | |
158 | * driver is for them to be collected into requests on a request | |
159 | * queue, and then to allow the device driver to select requests | |
160 | * off that queue when it is ready. This works well for many block | |
161 | * devices. However some block devices (typically virtual devices | |
162 | * such as md or lvm) do not benefit from the processing on the | |
163 | * request queue, and are served best by having the requests passed | |
164 | * directly to them. This can be achieved by providing a function | |
165 | * to blk_queue_make_request(). | |
166 | * | |
167 | * Caveat: | |
168 | * The driver that does this *must* be able to deal appropriately | |
169 | * with buffers in "highmemory". This can be accomplished by either calling | |
170 | * __bio_kmap_atomic() to get a temporary kernel mapping, or by calling | |
171 | * blk_queue_bounce() to create a buffer in normal memory. | |
172 | **/ | |
6728cb0e | 173 | void blk_queue_make_request(struct request_queue *q, make_request_fn *mfn) |
86db1e29 JA |
174 | { |
175 | /* | |
176 | * set defaults | |
177 | */ | |
178 | q->nr_requests = BLKDEV_MAX_RQ; | |
0e435ac2 | 179 | |
86db1e29 | 180 | q->make_request_fn = mfn; |
86db1e29 JA |
181 | blk_queue_dma_alignment(q, 511); |
182 | blk_queue_congestion_threshold(q); | |
183 | q->nr_batching = BLK_BATCH_REQ; | |
184 | ||
e475bba2 MP |
185 | blk_set_default_limits(&q->limits); |
186 | ||
86db1e29 JA |
187 | /* |
188 | * by default assume old behaviour and bounce for any highmem page | |
189 | */ | |
190 | blk_queue_bounce_limit(q, BLK_BOUNCE_HIGH); | |
191 | } | |
86db1e29 JA |
192 | EXPORT_SYMBOL(blk_queue_make_request); |
193 | ||
194 | /** | |
195 | * blk_queue_bounce_limit - set bounce buffer limit for queue | |
cd0aca2d TH |
196 | * @q: the request queue for the device |
197 | * @dma_mask: the maximum address the device can handle | |
86db1e29 JA |
198 | * |
199 | * Description: | |
200 | * Different hardware can have different requirements as to what pages | |
201 | * it can do I/O directly to. A low level driver can call | |
202 | * blk_queue_bounce_limit to have lower memory pages allocated as bounce | |
cd0aca2d | 203 | * buffers for doing I/O to pages residing above @dma_mask. |
86db1e29 | 204 | **/ |
cd0aca2d | 205 | void blk_queue_bounce_limit(struct request_queue *q, u64 dma_mask) |
86db1e29 | 206 | { |
cd0aca2d | 207 | unsigned long b_pfn = dma_mask >> PAGE_SHIFT; |
86db1e29 JA |
208 | int dma = 0; |
209 | ||
210 | q->bounce_gfp = GFP_NOIO; | |
211 | #if BITS_PER_LONG == 64 | |
cd0aca2d TH |
212 | /* |
213 | * Assume anything <= 4GB can be handled by IOMMU. Actually | |
214 | * some IOMMUs can handle everything, but I don't know of a | |
215 | * way to test this here. | |
216 | */ | |
217 | if (b_pfn < (min_t(u64, 0xffffffffUL, BLK_BOUNCE_HIGH) >> PAGE_SHIFT)) | |
86db1e29 | 218 | dma = 1; |
efb012b3 | 219 | q->limits.bounce_pfn = max(max_low_pfn, b_pfn); |
86db1e29 | 220 | #else |
6728cb0e | 221 | if (b_pfn < blk_max_low_pfn) |
86db1e29 | 222 | dma = 1; |
c49825fa | 223 | q->limits.bounce_pfn = b_pfn; |
260a67a9 | 224 | #endif |
86db1e29 JA |
225 | if (dma) { |
226 | init_emergency_isa_pool(); | |
227 | q->bounce_gfp = GFP_NOIO | GFP_DMA; | |
260a67a9 | 228 | q->limits.bounce_pfn = b_pfn; |
86db1e29 JA |
229 | } |
230 | } | |
86db1e29 JA |
231 | EXPORT_SYMBOL(blk_queue_bounce_limit); |
232 | ||
233 | /** | |
72d4cd9f MS |
234 | * blk_limits_max_hw_sectors - set hard and soft limit of max sectors for request |
235 | * @limits: the queue limits | |
2800aac1 | 236 | * @max_hw_sectors: max hardware sectors in the usual 512b unit |
86db1e29 JA |
237 | * |
238 | * Description: | |
2800aac1 MP |
239 | * Enables a low level driver to set a hard upper limit, |
240 | * max_hw_sectors, on the size of requests. max_hw_sectors is set by | |
241 | * the device driver based upon the combined capabilities of I/O | |
242 | * controller and storage device. | |
243 | * | |
244 | * max_sectors is a soft limit imposed by the block layer for | |
245 | * filesystem type requests. This value can be overridden on a | |
246 | * per-device basis in /sys/block/<device>/queue/max_sectors_kb. | |
247 | * The soft limit can not exceed max_hw_sectors. | |
86db1e29 | 248 | **/ |
72d4cd9f | 249 | void blk_limits_max_hw_sectors(struct queue_limits *limits, unsigned int max_hw_sectors) |
86db1e29 | 250 | { |
2800aac1 MP |
251 | if ((max_hw_sectors << 9) < PAGE_CACHE_SIZE) { |
252 | max_hw_sectors = 1 << (PAGE_CACHE_SHIFT - 9); | |
24c03d47 | 253 | printk(KERN_INFO "%s: set to minimum %d\n", |
2800aac1 | 254 | __func__, max_hw_sectors); |
86db1e29 JA |
255 | } |
256 | ||
72d4cd9f MS |
257 | limits->max_hw_sectors = max_hw_sectors; |
258 | limits->max_sectors = min_t(unsigned int, max_hw_sectors, | |
259 | BLK_DEF_MAX_SECTORS); | |
260 | } | |
261 | EXPORT_SYMBOL(blk_limits_max_hw_sectors); | |
262 | ||
263 | /** | |
264 | * blk_queue_max_hw_sectors - set max sectors for a request for this queue | |
265 | * @q: the request queue for the device | |
266 | * @max_hw_sectors: max hardware sectors in the usual 512b unit | |
267 | * | |
268 | * Description: | |
269 | * See description for blk_limits_max_hw_sectors(). | |
270 | **/ | |
271 | void blk_queue_max_hw_sectors(struct request_queue *q, unsigned int max_hw_sectors) | |
272 | { | |
273 | blk_limits_max_hw_sectors(&q->limits, max_hw_sectors); | |
86db1e29 | 274 | } |
086fa5ff | 275 | EXPORT_SYMBOL(blk_queue_max_hw_sectors); |
86db1e29 | 276 | |
67efc925 CH |
277 | /** |
278 | * blk_queue_max_discard_sectors - set max sectors for a single discard | |
279 | * @q: the request queue for the device | |
c7ebf065 | 280 | * @max_discard_sectors: maximum number of sectors to discard |
67efc925 CH |
281 | **/ |
282 | void blk_queue_max_discard_sectors(struct request_queue *q, | |
283 | unsigned int max_discard_sectors) | |
284 | { | |
285 | q->limits.max_discard_sectors = max_discard_sectors; | |
286 | } | |
287 | EXPORT_SYMBOL(blk_queue_max_discard_sectors); | |
288 | ||
86db1e29 | 289 | /** |
8a78362c | 290 | * blk_queue_max_segments - set max hw segments for a request for this queue |
86db1e29 JA |
291 | * @q: the request queue for the device |
292 | * @max_segments: max number of segments | |
293 | * | |
294 | * Description: | |
295 | * Enables a low level driver to set an upper limit on the number of | |
8a78362c | 296 | * hw data segments in a request. |
86db1e29 | 297 | **/ |
8a78362c | 298 | void blk_queue_max_segments(struct request_queue *q, unsigned short max_segments) |
86db1e29 JA |
299 | { |
300 | if (!max_segments) { | |
301 | max_segments = 1; | |
24c03d47 HH |
302 | printk(KERN_INFO "%s: set to minimum %d\n", |
303 | __func__, max_segments); | |
86db1e29 JA |
304 | } |
305 | ||
8a78362c | 306 | q->limits.max_segments = max_segments; |
86db1e29 | 307 | } |
8a78362c | 308 | EXPORT_SYMBOL(blk_queue_max_segments); |
86db1e29 JA |
309 | |
310 | /** | |
311 | * blk_queue_max_segment_size - set max segment size for blk_rq_map_sg | |
312 | * @q: the request queue for the device | |
313 | * @max_size: max size of segment in bytes | |
314 | * | |
315 | * Description: | |
316 | * Enables a low level driver to set an upper limit on the size of a | |
317 | * coalesced segment | |
318 | **/ | |
319 | void blk_queue_max_segment_size(struct request_queue *q, unsigned int max_size) | |
320 | { | |
321 | if (max_size < PAGE_CACHE_SIZE) { | |
322 | max_size = PAGE_CACHE_SIZE; | |
24c03d47 HH |
323 | printk(KERN_INFO "%s: set to minimum %d\n", |
324 | __func__, max_size); | |
86db1e29 JA |
325 | } |
326 | ||
025146e1 | 327 | q->limits.max_segment_size = max_size; |
86db1e29 | 328 | } |
86db1e29 JA |
329 | EXPORT_SYMBOL(blk_queue_max_segment_size); |
330 | ||
331 | /** | |
e1defc4f | 332 | * blk_queue_logical_block_size - set logical block size for the queue |
86db1e29 | 333 | * @q: the request queue for the device |
e1defc4f | 334 | * @size: the logical block size, in bytes |
86db1e29 JA |
335 | * |
336 | * Description: | |
e1defc4f MP |
337 | * This should be set to the lowest possible block size that the |
338 | * storage device can address. The default of 512 covers most | |
339 | * hardware. | |
86db1e29 | 340 | **/ |
e1defc4f | 341 | void blk_queue_logical_block_size(struct request_queue *q, unsigned short size) |
86db1e29 | 342 | { |
025146e1 | 343 | q->limits.logical_block_size = size; |
c72758f3 MP |
344 | |
345 | if (q->limits.physical_block_size < size) | |
346 | q->limits.physical_block_size = size; | |
347 | ||
348 | if (q->limits.io_min < q->limits.physical_block_size) | |
349 | q->limits.io_min = q->limits.physical_block_size; | |
86db1e29 | 350 | } |
e1defc4f | 351 | EXPORT_SYMBOL(blk_queue_logical_block_size); |
86db1e29 | 352 | |
c72758f3 MP |
353 | /** |
354 | * blk_queue_physical_block_size - set physical block size for the queue | |
355 | * @q: the request queue for the device | |
356 | * @size: the physical block size, in bytes | |
357 | * | |
358 | * Description: | |
359 | * This should be set to the lowest possible sector size that the | |
360 | * hardware can operate on without reverting to read-modify-write | |
361 | * operations. | |
362 | */ | |
892b6f90 | 363 | void blk_queue_physical_block_size(struct request_queue *q, unsigned int size) |
c72758f3 MP |
364 | { |
365 | q->limits.physical_block_size = size; | |
366 | ||
367 | if (q->limits.physical_block_size < q->limits.logical_block_size) | |
368 | q->limits.physical_block_size = q->limits.logical_block_size; | |
369 | ||
370 | if (q->limits.io_min < q->limits.physical_block_size) | |
371 | q->limits.io_min = q->limits.physical_block_size; | |
372 | } | |
373 | EXPORT_SYMBOL(blk_queue_physical_block_size); | |
374 | ||
375 | /** | |
376 | * blk_queue_alignment_offset - set physical block alignment offset | |
377 | * @q: the request queue for the device | |
8ebf9756 | 378 | * @offset: alignment offset in bytes |
c72758f3 MP |
379 | * |
380 | * Description: | |
381 | * Some devices are naturally misaligned to compensate for things like | |
382 | * the legacy DOS partition table 63-sector offset. Low-level drivers | |
383 | * should call this function for devices whose first sector is not | |
384 | * naturally aligned. | |
385 | */ | |
386 | void blk_queue_alignment_offset(struct request_queue *q, unsigned int offset) | |
387 | { | |
388 | q->limits.alignment_offset = | |
389 | offset & (q->limits.physical_block_size - 1); | |
390 | q->limits.misaligned = 0; | |
391 | } | |
392 | EXPORT_SYMBOL(blk_queue_alignment_offset); | |
393 | ||
7c958e32 MP |
394 | /** |
395 | * blk_limits_io_min - set minimum request size for a device | |
396 | * @limits: the queue limits | |
397 | * @min: smallest I/O size in bytes | |
398 | * | |
399 | * Description: | |
400 | * Some devices have an internal block size bigger than the reported | |
401 | * hardware sector size. This function can be used to signal the | |
402 | * smallest I/O the device can perform without incurring a performance | |
403 | * penalty. | |
404 | */ | |
405 | void blk_limits_io_min(struct queue_limits *limits, unsigned int min) | |
406 | { | |
407 | limits->io_min = min; | |
408 | ||
409 | if (limits->io_min < limits->logical_block_size) | |
410 | limits->io_min = limits->logical_block_size; | |
411 | ||
412 | if (limits->io_min < limits->physical_block_size) | |
413 | limits->io_min = limits->physical_block_size; | |
414 | } | |
415 | EXPORT_SYMBOL(blk_limits_io_min); | |
416 | ||
c72758f3 MP |
417 | /** |
418 | * blk_queue_io_min - set minimum request size for the queue | |
419 | * @q: the request queue for the device | |
8ebf9756 | 420 | * @min: smallest I/O size in bytes |
c72758f3 MP |
421 | * |
422 | * Description: | |
7e5f5fb0 MP |
423 | * Storage devices may report a granularity or preferred minimum I/O |
424 | * size which is the smallest request the device can perform without | |
425 | * incurring a performance penalty. For disk drives this is often the | |
426 | * physical block size. For RAID arrays it is often the stripe chunk | |
427 | * size. A properly aligned multiple of minimum_io_size is the | |
428 | * preferred request size for workloads where a high number of I/O | |
429 | * operations is desired. | |
c72758f3 MP |
430 | */ |
431 | void blk_queue_io_min(struct request_queue *q, unsigned int min) | |
432 | { | |
7c958e32 | 433 | blk_limits_io_min(&q->limits, min); |
c72758f3 MP |
434 | } |
435 | EXPORT_SYMBOL(blk_queue_io_min); | |
436 | ||
3c5820c7 MP |
437 | /** |
438 | * blk_limits_io_opt - set optimal request size for a device | |
439 | * @limits: the queue limits | |
440 | * @opt: smallest I/O size in bytes | |
441 | * | |
442 | * Description: | |
443 | * Storage devices may report an optimal I/O size, which is the | |
444 | * device's preferred unit for sustained I/O. This is rarely reported | |
445 | * for disk drives. For RAID arrays it is usually the stripe width or | |
446 | * the internal track size. A properly aligned multiple of | |
447 | * optimal_io_size is the preferred request size for workloads where | |
448 | * sustained throughput is desired. | |
449 | */ | |
450 | void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt) | |
451 | { | |
452 | limits->io_opt = opt; | |
453 | } | |
454 | EXPORT_SYMBOL(blk_limits_io_opt); | |
455 | ||
c72758f3 MP |
456 | /** |
457 | * blk_queue_io_opt - set optimal request size for the queue | |
458 | * @q: the request queue for the device | |
8ebf9756 | 459 | * @opt: optimal request size in bytes |
c72758f3 MP |
460 | * |
461 | * Description: | |
7e5f5fb0 MP |
462 | * Storage devices may report an optimal I/O size, which is the |
463 | * device's preferred unit for sustained I/O. This is rarely reported | |
464 | * for disk drives. For RAID arrays it is usually the stripe width or | |
465 | * the internal track size. A properly aligned multiple of | |
466 | * optimal_io_size is the preferred request size for workloads where | |
467 | * sustained throughput is desired. | |
c72758f3 MP |
468 | */ |
469 | void blk_queue_io_opt(struct request_queue *q, unsigned int opt) | |
470 | { | |
3c5820c7 | 471 | blk_limits_io_opt(&q->limits, opt); |
c72758f3 MP |
472 | } |
473 | EXPORT_SYMBOL(blk_queue_io_opt); | |
474 | ||
86db1e29 JA |
475 | /** |
476 | * blk_queue_stack_limits - inherit underlying queue limits for stacked drivers | |
477 | * @t: the stacking driver (top) | |
478 | * @b: the underlying device (bottom) | |
479 | **/ | |
480 | void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b) | |
481 | { | |
fef24667 | 482 | blk_stack_limits(&t->limits, &b->limits, 0); |
86db1e29 | 483 | } |
86db1e29 JA |
484 | EXPORT_SYMBOL(blk_queue_stack_limits); |
485 | ||
c72758f3 MP |
486 | /** |
487 | * blk_stack_limits - adjust queue_limits for stacked devices | |
81744ee4 MP |
488 | * @t: the stacking driver limits (top device) |
489 | * @b: the underlying queue limits (bottom, component device) | |
e03a72e1 | 490 | * @start: first data sector within component device |
c72758f3 MP |
491 | * |
492 | * Description: | |
81744ee4 MP |
493 | * This function is used by stacking drivers like MD and DM to ensure |
494 | * that all component devices have compatible block sizes and | |
495 | * alignments. The stacking driver must provide a queue_limits | |
496 | * struct (top) and then iteratively call the stacking function for | |
497 | * all component (bottom) devices. The stacking function will | |
498 | * attempt to combine the values and ensure proper alignment. | |
499 | * | |
500 | * Returns 0 if the top and bottom queue_limits are compatible. The | |
501 | * top device's block sizes and alignment offsets may be adjusted to | |
502 | * ensure alignment with the bottom device. If no compatible sizes | |
503 | * and alignments exist, -1 is returned and the resulting top | |
504 | * queue_limits will have the misaligned flag set to indicate that | |
505 | * the alignment_offset is undefined. | |
c72758f3 MP |
506 | */ |
507 | int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, | |
e03a72e1 | 508 | sector_t start) |
c72758f3 | 509 | { |
e03a72e1 | 510 | unsigned int top, bottom, alignment, ret = 0; |
86b37281 | 511 | |
c72758f3 MP |
512 | t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors); |
513 | t->max_hw_sectors = min_not_zero(t->max_hw_sectors, b->max_hw_sectors); | |
77634f33 | 514 | t->bounce_pfn = min_not_zero(t->bounce_pfn, b->bounce_pfn); |
c72758f3 MP |
515 | |
516 | t->seg_boundary_mask = min_not_zero(t->seg_boundary_mask, | |
517 | b->seg_boundary_mask); | |
518 | ||
8a78362c | 519 | t->max_segments = min_not_zero(t->max_segments, b->max_segments); |
13f05c8d MP |
520 | t->max_integrity_segments = min_not_zero(t->max_integrity_segments, |
521 | b->max_integrity_segments); | |
c72758f3 MP |
522 | |
523 | t->max_segment_size = min_not_zero(t->max_segment_size, | |
524 | b->max_segment_size); | |
525 | ||
fe0b393f MP |
526 | t->misaligned |= b->misaligned; |
527 | ||
e03a72e1 | 528 | alignment = queue_limit_alignment_offset(b, start); |
9504e086 | 529 | |
81744ee4 MP |
530 | /* Bottom device has different alignment. Check that it is |
531 | * compatible with the current top alignment. | |
532 | */ | |
9504e086 MP |
533 | if (t->alignment_offset != alignment) { |
534 | ||
535 | top = max(t->physical_block_size, t->io_min) | |
536 | + t->alignment_offset; | |
81744ee4 | 537 | bottom = max(b->physical_block_size, b->io_min) + alignment; |
9504e086 | 538 | |
81744ee4 | 539 | /* Verify that top and bottom intervals line up */ |
fe0b393f | 540 | if (max(top, bottom) & (min(top, bottom) - 1)) { |
9504e086 | 541 | t->misaligned = 1; |
fe0b393f MP |
542 | ret = -1; |
543 | } | |
9504e086 MP |
544 | } |
545 | ||
c72758f3 MP |
546 | t->logical_block_size = max(t->logical_block_size, |
547 | b->logical_block_size); | |
548 | ||
549 | t->physical_block_size = max(t->physical_block_size, | |
550 | b->physical_block_size); | |
551 | ||
552 | t->io_min = max(t->io_min, b->io_min); | |
9504e086 MP |
553 | t->io_opt = lcm(t->io_opt, b->io_opt); |
554 | ||
e692cb66 | 555 | t->cluster &= b->cluster; |
98262f27 | 556 | t->discard_zeroes_data &= b->discard_zeroes_data; |
c72758f3 | 557 | |
81744ee4 | 558 | /* Physical block size a multiple of the logical block size? */ |
9504e086 MP |
559 | if (t->physical_block_size & (t->logical_block_size - 1)) { |
560 | t->physical_block_size = t->logical_block_size; | |
c72758f3 | 561 | t->misaligned = 1; |
fe0b393f | 562 | ret = -1; |
86b37281 MP |
563 | } |
564 | ||
81744ee4 | 565 | /* Minimum I/O a multiple of the physical block size? */ |
9504e086 MP |
566 | if (t->io_min & (t->physical_block_size - 1)) { |
567 | t->io_min = t->physical_block_size; | |
568 | t->misaligned = 1; | |
fe0b393f | 569 | ret = -1; |
c72758f3 MP |
570 | } |
571 | ||
81744ee4 | 572 | /* Optimal I/O a multiple of the physical block size? */ |
9504e086 MP |
573 | if (t->io_opt & (t->physical_block_size - 1)) { |
574 | t->io_opt = 0; | |
575 | t->misaligned = 1; | |
fe0b393f | 576 | ret = -1; |
9504e086 | 577 | } |
c72758f3 | 578 | |
81744ee4 | 579 | /* Find lowest common alignment_offset */ |
9504e086 MP |
580 | t->alignment_offset = lcm(t->alignment_offset, alignment) |
581 | & (max(t->physical_block_size, t->io_min) - 1); | |
86b37281 | 582 | |
81744ee4 | 583 | /* Verify that new alignment_offset is on a logical block boundary */ |
fe0b393f | 584 | if (t->alignment_offset & (t->logical_block_size - 1)) { |
c72758f3 | 585 | t->misaligned = 1; |
fe0b393f MP |
586 | ret = -1; |
587 | } | |
c72758f3 | 588 | |
9504e086 MP |
589 | /* Discard alignment and granularity */ |
590 | if (b->discard_granularity) { | |
e03a72e1 | 591 | alignment = queue_limit_discard_alignment(b, start); |
9504e086 MP |
592 | |
593 | if (t->discard_granularity != 0 && | |
594 | t->discard_alignment != alignment) { | |
595 | top = t->discard_granularity + t->discard_alignment; | |
596 | bottom = b->discard_granularity + alignment; | |
70dd5bf3 | 597 | |
9504e086 MP |
598 | /* Verify that top and bottom intervals line up */ |
599 | if (max(top, bottom) & (min(top, bottom) - 1)) | |
600 | t->discard_misaligned = 1; | |
601 | } | |
602 | ||
81744ee4 MP |
603 | t->max_discard_sectors = min_not_zero(t->max_discard_sectors, |
604 | b->max_discard_sectors); | |
9504e086 MP |
605 | t->discard_granularity = max(t->discard_granularity, |
606 | b->discard_granularity); | |
607 | t->discard_alignment = lcm(t->discard_alignment, alignment) & | |
608 | (t->discard_granularity - 1); | |
609 | } | |
70dd5bf3 | 610 | |
fe0b393f | 611 | return ret; |
c72758f3 | 612 | } |
5d85d324 | 613 | EXPORT_SYMBOL(blk_stack_limits); |
c72758f3 | 614 | |
17be8c24 MP |
615 | /** |
616 | * bdev_stack_limits - adjust queue limits for stacked drivers | |
617 | * @t: the stacking driver limits (top device) | |
618 | * @bdev: the component block_device (bottom) | |
619 | * @start: first data sector within component device | |
620 | * | |
621 | * Description: | |
622 | * Merges queue limits for a top device and a block_device. Returns | |
623 | * 0 if alignment didn't change. Returns -1 if adding the bottom | |
624 | * device caused misalignment. | |
625 | */ | |
626 | int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev, | |
627 | sector_t start) | |
628 | { | |
629 | struct request_queue *bq = bdev_get_queue(bdev); | |
630 | ||
631 | start += get_start_sect(bdev); | |
632 | ||
e03a72e1 | 633 | return blk_stack_limits(t, &bq->limits, start); |
17be8c24 MP |
634 | } |
635 | EXPORT_SYMBOL(bdev_stack_limits); | |
636 | ||
c72758f3 MP |
637 | /** |
638 | * disk_stack_limits - adjust queue limits for stacked drivers | |
77634f33 | 639 | * @disk: MD/DM gendisk (top) |
c72758f3 MP |
640 | * @bdev: the underlying block device (bottom) |
641 | * @offset: offset to beginning of data within component device | |
642 | * | |
643 | * Description: | |
e03a72e1 MP |
644 | * Merges the limits for a top level gendisk and a bottom level |
645 | * block_device. | |
c72758f3 MP |
646 | */ |
647 | void disk_stack_limits(struct gendisk *disk, struct block_device *bdev, | |
648 | sector_t offset) | |
649 | { | |
650 | struct request_queue *t = disk->queue; | |
c72758f3 | 651 | |
e03a72e1 | 652 | if (bdev_stack_limits(&t->limits, bdev, offset >> 9) < 0) { |
c72758f3 MP |
653 | char top[BDEVNAME_SIZE], bottom[BDEVNAME_SIZE]; |
654 | ||
655 | disk_name(disk, 0, top); | |
656 | bdevname(bdev, bottom); | |
657 | ||
658 | printk(KERN_NOTICE "%s: Warning: Device %s is misaligned\n", | |
659 | top, bottom); | |
660 | } | |
c72758f3 MP |
661 | } |
662 | EXPORT_SYMBOL(disk_stack_limits); | |
663 | ||
e3790c7d TH |
664 | /** |
665 | * blk_queue_dma_pad - set pad mask | |
666 | * @q: the request queue for the device | |
667 | * @mask: pad mask | |
668 | * | |
27f8221a | 669 | * Set dma pad mask. |
e3790c7d | 670 | * |
27f8221a FT |
671 | * Appending pad buffer to a request modifies the last entry of a |
672 | * scatter list such that it includes the pad buffer. | |
e3790c7d TH |
673 | **/ |
674 | void blk_queue_dma_pad(struct request_queue *q, unsigned int mask) | |
675 | { | |
676 | q->dma_pad_mask = mask; | |
677 | } | |
678 | EXPORT_SYMBOL(blk_queue_dma_pad); | |
679 | ||
27f8221a FT |
680 | /** |
681 | * blk_queue_update_dma_pad - update pad mask | |
682 | * @q: the request queue for the device | |
683 | * @mask: pad mask | |
684 | * | |
685 | * Update dma pad mask. | |
686 | * | |
687 | * Appending pad buffer to a request modifies the last entry of a | |
688 | * scatter list such that it includes the pad buffer. | |
689 | **/ | |
690 | void blk_queue_update_dma_pad(struct request_queue *q, unsigned int mask) | |
691 | { | |
692 | if (mask > q->dma_pad_mask) | |
693 | q->dma_pad_mask = mask; | |
694 | } | |
695 | EXPORT_SYMBOL(blk_queue_update_dma_pad); | |
696 | ||
86db1e29 JA |
697 | /** |
698 | * blk_queue_dma_drain - Set up a drain buffer for excess dma. | |
86db1e29 | 699 | * @q: the request queue for the device |
2fb98e84 | 700 | * @dma_drain_needed: fn which returns non-zero if drain is necessary |
86db1e29 JA |
701 | * @buf: physically contiguous buffer |
702 | * @size: size of the buffer in bytes | |
703 | * | |
704 | * Some devices have excess DMA problems and can't simply discard (or | |
705 | * zero fill) the unwanted piece of the transfer. They have to have a | |
706 | * real area of memory to transfer it into. The use case for this is | |
707 | * ATAPI devices in DMA mode. If the packet command causes a transfer | |
708 | * bigger than the transfer size some HBAs will lock up if there | |
709 | * aren't DMA elements to contain the excess transfer. What this API | |
710 | * does is adjust the queue so that the buf is always appended | |
711 | * silently to the scatterlist. | |
712 | * | |
8a78362c MP |
713 | * Note: This routine adjusts max_hw_segments to make room for appending |
714 | * the drain buffer. If you call blk_queue_max_segments() after calling | |
715 | * this routine, you must set the limit to one fewer than your device | |
716 | * can support otherwise there won't be room for the drain buffer. | |
86db1e29 | 717 | */ |
448da4d2 | 718 | int blk_queue_dma_drain(struct request_queue *q, |
2fb98e84 TH |
719 | dma_drain_needed_fn *dma_drain_needed, |
720 | void *buf, unsigned int size) | |
86db1e29 | 721 | { |
8a78362c | 722 | if (queue_max_segments(q) < 2) |
86db1e29 JA |
723 | return -EINVAL; |
724 | /* make room for appending the drain */ | |
8a78362c | 725 | blk_queue_max_segments(q, queue_max_segments(q) - 1); |
2fb98e84 | 726 | q->dma_drain_needed = dma_drain_needed; |
86db1e29 JA |
727 | q->dma_drain_buffer = buf; |
728 | q->dma_drain_size = size; | |
729 | ||
730 | return 0; | |
731 | } | |
86db1e29 JA |
732 | EXPORT_SYMBOL_GPL(blk_queue_dma_drain); |
733 | ||
734 | /** | |
735 | * blk_queue_segment_boundary - set boundary rules for segment merging | |
736 | * @q: the request queue for the device | |
737 | * @mask: the memory boundary mask | |
738 | **/ | |
739 | void blk_queue_segment_boundary(struct request_queue *q, unsigned long mask) | |
740 | { | |
741 | if (mask < PAGE_CACHE_SIZE - 1) { | |
742 | mask = PAGE_CACHE_SIZE - 1; | |
24c03d47 HH |
743 | printk(KERN_INFO "%s: set to minimum %lx\n", |
744 | __func__, mask); | |
86db1e29 JA |
745 | } |
746 | ||
025146e1 | 747 | q->limits.seg_boundary_mask = mask; |
86db1e29 | 748 | } |
86db1e29 JA |
749 | EXPORT_SYMBOL(blk_queue_segment_boundary); |
750 | ||
751 | /** | |
752 | * blk_queue_dma_alignment - set dma length and memory alignment | |
753 | * @q: the request queue for the device | |
754 | * @mask: alignment mask | |
755 | * | |
756 | * description: | |
710027a4 | 757 | * set required memory and length alignment for direct dma transactions. |
8feb4d20 | 758 | * this is used when building direct io requests for the queue. |
86db1e29 JA |
759 | * |
760 | **/ | |
761 | void blk_queue_dma_alignment(struct request_queue *q, int mask) | |
762 | { | |
763 | q->dma_alignment = mask; | |
764 | } | |
86db1e29 JA |
765 | EXPORT_SYMBOL(blk_queue_dma_alignment); |
766 | ||
767 | /** | |
768 | * blk_queue_update_dma_alignment - update dma length and memory alignment | |
769 | * @q: the request queue for the device | |
770 | * @mask: alignment mask | |
771 | * | |
772 | * description: | |
710027a4 | 773 | * update required memory and length alignment for direct dma transactions. |
86db1e29 JA |
774 | * If the requested alignment is larger than the current alignment, then |
775 | * the current queue alignment is updated to the new value, otherwise it | |
776 | * is left alone. The design of this is to allow multiple objects | |
777 | * (driver, device, transport etc) to set their respective | |
778 | * alignments without having them interfere. | |
779 | * | |
780 | **/ | |
781 | void blk_queue_update_dma_alignment(struct request_queue *q, int mask) | |
782 | { | |
783 | BUG_ON(mask > PAGE_SIZE); | |
784 | ||
785 | if (mask > q->dma_alignment) | |
786 | q->dma_alignment = mask; | |
787 | } | |
86db1e29 JA |
788 | EXPORT_SYMBOL(blk_queue_update_dma_alignment); |
789 | ||
4913efe4 TH |
790 | /** |
791 | * blk_queue_flush - configure queue's cache flush capability | |
792 | * @q: the request queue for the device | |
793 | * @flush: 0, REQ_FLUSH or REQ_FLUSH | REQ_FUA | |
794 | * | |
795 | * Tell block layer cache flush capability of @q. If it supports | |
796 | * flushing, REQ_FLUSH should be set. If it supports bypassing | |
797 | * write cache for individual writes, REQ_FUA should be set. | |
798 | */ | |
799 | void blk_queue_flush(struct request_queue *q, unsigned int flush) | |
800 | { | |
801 | WARN_ON_ONCE(flush & ~(REQ_FLUSH | REQ_FUA)); | |
802 | ||
803 | if (WARN_ON_ONCE(!(flush & REQ_FLUSH) && (flush & REQ_FUA))) | |
804 | flush &= ~REQ_FUA; | |
805 | ||
806 | q->flush_flags = flush & (REQ_FLUSH | REQ_FUA); | |
807 | } | |
808 | EXPORT_SYMBOL_GPL(blk_queue_flush); | |
809 | ||
f3876930 | 810 | void blk_queue_flush_queueable(struct request_queue *q, bool queueable) |
811 | { | |
812 | q->flush_not_queueable = !queueable; | |
813 | } | |
814 | EXPORT_SYMBOL_GPL(blk_queue_flush_queueable); | |
815 | ||
aeb3d3a8 | 816 | static int __init blk_settings_init(void) |
86db1e29 JA |
817 | { |
818 | blk_max_low_pfn = max_low_pfn - 1; | |
819 | blk_max_pfn = max_pfn - 1; | |
820 | return 0; | |
821 | } | |
822 | subsys_initcall(blk_settings_init); |