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 */ | |
10 | ||
11 | #include "blk.h" | |
12 | ||
6728cb0e | 13 | unsigned long blk_max_low_pfn; |
86db1e29 | 14 | EXPORT_SYMBOL(blk_max_low_pfn); |
6728cb0e JA |
15 | |
16 | unsigned long blk_max_pfn; | |
86db1e29 JA |
17 | |
18 | /** | |
19 | * blk_queue_prep_rq - set a prepare_request function for queue | |
20 | * @q: queue | |
21 | * @pfn: prepare_request function | |
22 | * | |
23 | * It's possible for a queue to register a prepare_request callback which | |
24 | * is invoked before the request is handed to the request_fn. The goal of | |
25 | * the function is to prepare a request for I/O, it can be used to build a | |
26 | * cdb from the request data for instance. | |
27 | * | |
28 | */ | |
29 | void blk_queue_prep_rq(struct request_queue *q, prep_rq_fn *pfn) | |
30 | { | |
31 | q->prep_rq_fn = pfn; | |
32 | } | |
86db1e29 JA |
33 | EXPORT_SYMBOL(blk_queue_prep_rq); |
34 | ||
fb2dce86 DW |
35 | /** |
36 | * blk_queue_set_discard - set a discard_sectors function for queue | |
37 | * @q: queue | |
38 | * @dfn: prepare_discard function | |
39 | * | |
40 | * It's possible for a queue to register a discard callback which is used | |
41 | * to transform a discard request into the appropriate type for the | |
42 | * hardware. If none is registered, then discard requests are failed | |
43 | * with %EOPNOTSUPP. | |
44 | * | |
45 | */ | |
46 | void blk_queue_set_discard(struct request_queue *q, prepare_discard_fn *dfn) | |
47 | { | |
48 | q->prepare_discard_fn = dfn; | |
49 | } | |
50 | EXPORT_SYMBOL(blk_queue_set_discard); | |
51 | ||
86db1e29 JA |
52 | /** |
53 | * blk_queue_merge_bvec - set a merge_bvec function for queue | |
54 | * @q: queue | |
55 | * @mbfn: merge_bvec_fn | |
56 | * | |
57 | * Usually queues have static limitations on the max sectors or segments that | |
58 | * we can put in a request. Stacking drivers may have some settings that | |
59 | * are dynamic, and thus we have to query the queue whether it is ok to | |
60 | * add a new bio_vec to a bio at a given offset or not. If the block device | |
61 | * has such limitations, it needs to register a merge_bvec_fn to control | |
62 | * the size of bio's sent to it. Note that a block device *must* allow a | |
63 | * single page to be added to an empty bio. The block device driver may want | |
64 | * to use the bio_split() function to deal with these bio's. By default | |
65 | * no merge_bvec_fn is defined for a queue, and only the fixed limits are | |
66 | * honored. | |
67 | */ | |
68 | void blk_queue_merge_bvec(struct request_queue *q, merge_bvec_fn *mbfn) | |
69 | { | |
70 | q->merge_bvec_fn = mbfn; | |
71 | } | |
86db1e29 JA |
72 | EXPORT_SYMBOL(blk_queue_merge_bvec); |
73 | ||
74 | void blk_queue_softirq_done(struct request_queue *q, softirq_done_fn *fn) | |
75 | { | |
76 | q->softirq_done_fn = fn; | |
77 | } | |
86db1e29 JA |
78 | EXPORT_SYMBOL(blk_queue_softirq_done); |
79 | ||
242f9dcb JA |
80 | void blk_queue_rq_timeout(struct request_queue *q, unsigned int timeout) |
81 | { | |
82 | q->rq_timeout = timeout; | |
83 | } | |
84 | EXPORT_SYMBOL_GPL(blk_queue_rq_timeout); | |
85 | ||
86 | void blk_queue_rq_timed_out(struct request_queue *q, rq_timed_out_fn *fn) | |
87 | { | |
88 | q->rq_timed_out_fn = fn; | |
89 | } | |
90 | EXPORT_SYMBOL_GPL(blk_queue_rq_timed_out); | |
91 | ||
ef9e3fac KU |
92 | void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn) |
93 | { | |
94 | q->lld_busy_fn = fn; | |
95 | } | |
96 | EXPORT_SYMBOL_GPL(blk_queue_lld_busy); | |
97 | ||
86db1e29 JA |
98 | /** |
99 | * blk_queue_make_request - define an alternate make_request function for a device | |
100 | * @q: the request queue for the device to be affected | |
101 | * @mfn: the alternate make_request function | |
102 | * | |
103 | * Description: | |
104 | * The normal way for &struct bios to be passed to a device | |
105 | * driver is for them to be collected into requests on a request | |
106 | * queue, and then to allow the device driver to select requests | |
107 | * off that queue when it is ready. This works well for many block | |
108 | * devices. However some block devices (typically virtual devices | |
109 | * such as md or lvm) do not benefit from the processing on the | |
110 | * request queue, and are served best by having the requests passed | |
111 | * directly to them. This can be achieved by providing a function | |
112 | * to blk_queue_make_request(). | |
113 | * | |
114 | * Caveat: | |
115 | * The driver that does this *must* be able to deal appropriately | |
116 | * with buffers in "highmemory". This can be accomplished by either calling | |
117 | * __bio_kmap_atomic() to get a temporary kernel mapping, or by calling | |
118 | * blk_queue_bounce() to create a buffer in normal memory. | |
119 | **/ | |
6728cb0e | 120 | void blk_queue_make_request(struct request_queue *q, make_request_fn *mfn) |
86db1e29 JA |
121 | { |
122 | /* | |
123 | * set defaults | |
124 | */ | |
125 | q->nr_requests = BLKDEV_MAX_RQ; | |
126 | blk_queue_max_phys_segments(q, MAX_PHYS_SEGMENTS); | |
127 | blk_queue_max_hw_segments(q, MAX_HW_SEGMENTS); | |
0e435ac2 MB |
128 | blk_queue_segment_boundary(q, BLK_SEG_BOUNDARY_MASK); |
129 | blk_queue_max_segment_size(q, MAX_SEGMENT_SIZE); | |
130 | ||
86db1e29 | 131 | q->make_request_fn = mfn; |
6728cb0e JA |
132 | q->backing_dev_info.ra_pages = |
133 | (VM_MAX_READAHEAD * 1024) / PAGE_CACHE_SIZE; | |
86db1e29 JA |
134 | q->backing_dev_info.state = 0; |
135 | q->backing_dev_info.capabilities = BDI_CAP_MAP_COPY; | |
136 | blk_queue_max_sectors(q, SAFE_MAX_SECTORS); | |
137 | blk_queue_hardsect_size(q, 512); | |
138 | blk_queue_dma_alignment(q, 511); | |
139 | blk_queue_congestion_threshold(q); | |
140 | q->nr_batching = BLK_BATCH_REQ; | |
141 | ||
142 | q->unplug_thresh = 4; /* hmm */ | |
143 | q->unplug_delay = (3 * HZ) / 1000; /* 3 milliseconds */ | |
144 | if (q->unplug_delay == 0) | |
145 | q->unplug_delay = 1; | |
146 | ||
86db1e29 JA |
147 | q->unplug_timer.function = blk_unplug_timeout; |
148 | q->unplug_timer.data = (unsigned long)q; | |
149 | ||
150 | /* | |
151 | * by default assume old behaviour and bounce for any highmem page | |
152 | */ | |
153 | blk_queue_bounce_limit(q, BLK_BOUNCE_HIGH); | |
154 | } | |
86db1e29 JA |
155 | EXPORT_SYMBOL(blk_queue_make_request); |
156 | ||
157 | /** | |
158 | * blk_queue_bounce_limit - set bounce buffer limit for queue | |
cd0aca2d TH |
159 | * @q: the request queue for the device |
160 | * @dma_mask: the maximum address the device can handle | |
86db1e29 JA |
161 | * |
162 | * Description: | |
163 | * Different hardware can have different requirements as to what pages | |
164 | * it can do I/O directly to. A low level driver can call | |
165 | * blk_queue_bounce_limit to have lower memory pages allocated as bounce | |
cd0aca2d | 166 | * buffers for doing I/O to pages residing above @dma_mask. |
86db1e29 | 167 | **/ |
cd0aca2d | 168 | void blk_queue_bounce_limit(struct request_queue *q, u64 dma_mask) |
86db1e29 | 169 | { |
cd0aca2d | 170 | unsigned long b_pfn = dma_mask >> PAGE_SHIFT; |
86db1e29 JA |
171 | int dma = 0; |
172 | ||
173 | q->bounce_gfp = GFP_NOIO; | |
174 | #if BITS_PER_LONG == 64 | |
cd0aca2d TH |
175 | /* |
176 | * Assume anything <= 4GB can be handled by IOMMU. Actually | |
177 | * some IOMMUs can handle everything, but I don't know of a | |
178 | * way to test this here. | |
179 | */ | |
180 | if (b_pfn < (min_t(u64, 0xffffffffUL, BLK_BOUNCE_HIGH) >> PAGE_SHIFT)) | |
86db1e29 JA |
181 | dma = 1; |
182 | q->bounce_pfn = max_low_pfn; | |
183 | #else | |
6728cb0e | 184 | if (b_pfn < blk_max_low_pfn) |
86db1e29 | 185 | dma = 1; |
6728cb0e | 186 | q->bounce_pfn = b_pfn; |
86db1e29 JA |
187 | #endif |
188 | if (dma) { | |
189 | init_emergency_isa_pool(); | |
190 | q->bounce_gfp = GFP_NOIO | GFP_DMA; | |
6728cb0e | 191 | q->bounce_pfn = b_pfn; |
86db1e29 JA |
192 | } |
193 | } | |
86db1e29 JA |
194 | EXPORT_SYMBOL(blk_queue_bounce_limit); |
195 | ||
196 | /** | |
197 | * blk_queue_max_sectors - set max sectors for a request for this queue | |
198 | * @q: the request queue for the device | |
199 | * @max_sectors: max sectors in the usual 512b unit | |
200 | * | |
201 | * Description: | |
202 | * Enables a low level driver to set an upper limit on the size of | |
203 | * received requests. | |
204 | **/ | |
205 | void blk_queue_max_sectors(struct request_queue *q, unsigned int max_sectors) | |
206 | { | |
207 | if ((max_sectors << 9) < PAGE_CACHE_SIZE) { | |
208 | max_sectors = 1 << (PAGE_CACHE_SHIFT - 9); | |
24c03d47 HH |
209 | printk(KERN_INFO "%s: set to minimum %d\n", |
210 | __func__, max_sectors); | |
86db1e29 JA |
211 | } |
212 | ||
213 | if (BLK_DEF_MAX_SECTORS > max_sectors) | |
214 | q->max_hw_sectors = q->max_sectors = max_sectors; | |
215 | else { | |
216 | q->max_sectors = BLK_DEF_MAX_SECTORS; | |
217 | q->max_hw_sectors = max_sectors; | |
218 | } | |
219 | } | |
86db1e29 JA |
220 | EXPORT_SYMBOL(blk_queue_max_sectors); |
221 | ||
222 | /** | |
223 | * blk_queue_max_phys_segments - set max phys segments for a request for this queue | |
224 | * @q: the request queue for the device | |
225 | * @max_segments: max number of segments | |
226 | * | |
227 | * Description: | |
228 | * Enables a low level driver to set an upper limit on the number of | |
229 | * physical data segments in a request. This would be the largest sized | |
230 | * scatter list the driver could handle. | |
231 | **/ | |
232 | void blk_queue_max_phys_segments(struct request_queue *q, | |
233 | unsigned short max_segments) | |
234 | { | |
235 | if (!max_segments) { | |
236 | max_segments = 1; | |
24c03d47 HH |
237 | printk(KERN_INFO "%s: set to minimum %d\n", |
238 | __func__, max_segments); | |
86db1e29 JA |
239 | } |
240 | ||
241 | q->max_phys_segments = max_segments; | |
242 | } | |
86db1e29 JA |
243 | EXPORT_SYMBOL(blk_queue_max_phys_segments); |
244 | ||
245 | /** | |
246 | * blk_queue_max_hw_segments - set max hw segments for a request for this queue | |
247 | * @q: the request queue for the device | |
248 | * @max_segments: max number of segments | |
249 | * | |
250 | * Description: | |
251 | * Enables a low level driver to set an upper limit on the number of | |
252 | * hw data segments in a request. This would be the largest number of | |
710027a4 | 253 | * address/length pairs the host adapter can actually give at once |
86db1e29 JA |
254 | * to the device. |
255 | **/ | |
256 | void blk_queue_max_hw_segments(struct request_queue *q, | |
257 | unsigned short max_segments) | |
258 | { | |
259 | if (!max_segments) { | |
260 | max_segments = 1; | |
24c03d47 HH |
261 | printk(KERN_INFO "%s: set to minimum %d\n", |
262 | __func__, max_segments); | |
86db1e29 JA |
263 | } |
264 | ||
265 | q->max_hw_segments = max_segments; | |
266 | } | |
86db1e29 JA |
267 | EXPORT_SYMBOL(blk_queue_max_hw_segments); |
268 | ||
269 | /** | |
270 | * blk_queue_max_segment_size - set max segment size for blk_rq_map_sg | |
271 | * @q: the request queue for the device | |
272 | * @max_size: max size of segment in bytes | |
273 | * | |
274 | * Description: | |
275 | * Enables a low level driver to set an upper limit on the size of a | |
276 | * coalesced segment | |
277 | **/ | |
278 | void blk_queue_max_segment_size(struct request_queue *q, unsigned int max_size) | |
279 | { | |
280 | if (max_size < PAGE_CACHE_SIZE) { | |
281 | max_size = PAGE_CACHE_SIZE; | |
24c03d47 HH |
282 | printk(KERN_INFO "%s: set to minimum %d\n", |
283 | __func__, max_size); | |
86db1e29 JA |
284 | } |
285 | ||
286 | q->max_segment_size = max_size; | |
287 | } | |
86db1e29 JA |
288 | EXPORT_SYMBOL(blk_queue_max_segment_size); |
289 | ||
290 | /** | |
291 | * blk_queue_hardsect_size - set hardware sector size for the queue | |
292 | * @q: the request queue for the device | |
293 | * @size: the hardware sector size, in bytes | |
294 | * | |
295 | * Description: | |
296 | * This should typically be set to the lowest possible sector size | |
297 | * that the hardware can operate on (possible without reverting to | |
298 | * even internal read-modify-write operations). Usually the default | |
299 | * of 512 covers most hardware. | |
300 | **/ | |
301 | void blk_queue_hardsect_size(struct request_queue *q, unsigned short size) | |
302 | { | |
303 | q->hardsect_size = size; | |
304 | } | |
86db1e29 JA |
305 | EXPORT_SYMBOL(blk_queue_hardsect_size); |
306 | ||
307 | /* | |
308 | * Returns the minimum that is _not_ zero, unless both are zero. | |
309 | */ | |
310 | #define min_not_zero(l, r) (l == 0) ? r : ((r == 0) ? l : min(l, r)) | |
311 | ||
312 | /** | |
313 | * blk_queue_stack_limits - inherit underlying queue limits for stacked drivers | |
314 | * @t: the stacking driver (top) | |
315 | * @b: the underlying device (bottom) | |
316 | **/ | |
317 | void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b) | |
318 | { | |
319 | /* zero is "infinity" */ | |
6728cb0e JA |
320 | t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors); |
321 | t->max_hw_sectors = min_not_zero(t->max_hw_sectors, b->max_hw_sectors); | |
0e435ac2 | 322 | t->seg_boundary_mask = min_not_zero(t->seg_boundary_mask, b->seg_boundary_mask); |
86db1e29 | 323 | |
18af8b2c FT |
324 | t->max_phys_segments = min_not_zero(t->max_phys_segments, b->max_phys_segments); |
325 | t->max_hw_segments = min_not_zero(t->max_hw_segments, b->max_hw_segments); | |
326 | t->max_segment_size = min_not_zero(t->max_segment_size, b->max_segment_size); | |
6728cb0e | 327 | t->hardsect_size = max(t->hardsect_size, b->hardsect_size); |
e7e72bf6 NB |
328 | if (!t->queue_lock) |
329 | WARN_ON_ONCE(1); | |
330 | else if (!test_bit(QUEUE_FLAG_CLUSTER, &b->queue_flags)) { | |
331 | unsigned long flags; | |
332 | spin_lock_irqsave(t->queue_lock, flags); | |
75ad23bc | 333 | queue_flag_clear(QUEUE_FLAG_CLUSTER, t); |
e7e72bf6 NB |
334 | spin_unlock_irqrestore(t->queue_lock, flags); |
335 | } | |
86db1e29 | 336 | } |
86db1e29 JA |
337 | EXPORT_SYMBOL(blk_queue_stack_limits); |
338 | ||
e3790c7d TH |
339 | /** |
340 | * blk_queue_dma_pad - set pad mask | |
341 | * @q: the request queue for the device | |
342 | * @mask: pad mask | |
343 | * | |
27f8221a | 344 | * Set dma pad mask. |
e3790c7d | 345 | * |
27f8221a FT |
346 | * Appending pad buffer to a request modifies the last entry of a |
347 | * scatter list such that it includes the pad buffer. | |
e3790c7d TH |
348 | **/ |
349 | void blk_queue_dma_pad(struct request_queue *q, unsigned int mask) | |
350 | { | |
351 | q->dma_pad_mask = mask; | |
352 | } | |
353 | EXPORT_SYMBOL(blk_queue_dma_pad); | |
354 | ||
27f8221a FT |
355 | /** |
356 | * blk_queue_update_dma_pad - update pad mask | |
357 | * @q: the request queue for the device | |
358 | * @mask: pad mask | |
359 | * | |
360 | * Update dma pad mask. | |
361 | * | |
362 | * Appending pad buffer to a request modifies the last entry of a | |
363 | * scatter list such that it includes the pad buffer. | |
364 | **/ | |
365 | void blk_queue_update_dma_pad(struct request_queue *q, unsigned int mask) | |
366 | { | |
367 | if (mask > q->dma_pad_mask) | |
368 | q->dma_pad_mask = mask; | |
369 | } | |
370 | EXPORT_SYMBOL(blk_queue_update_dma_pad); | |
371 | ||
86db1e29 JA |
372 | /** |
373 | * blk_queue_dma_drain - Set up a drain buffer for excess dma. | |
86db1e29 | 374 | * @q: the request queue for the device |
2fb98e84 | 375 | * @dma_drain_needed: fn which returns non-zero if drain is necessary |
86db1e29 JA |
376 | * @buf: physically contiguous buffer |
377 | * @size: size of the buffer in bytes | |
378 | * | |
379 | * Some devices have excess DMA problems and can't simply discard (or | |
380 | * zero fill) the unwanted piece of the transfer. They have to have a | |
381 | * real area of memory to transfer it into. The use case for this is | |
382 | * ATAPI devices in DMA mode. If the packet command causes a transfer | |
383 | * bigger than the transfer size some HBAs will lock up if there | |
384 | * aren't DMA elements to contain the excess transfer. What this API | |
385 | * does is adjust the queue so that the buf is always appended | |
386 | * silently to the scatterlist. | |
387 | * | |
388 | * Note: This routine adjusts max_hw_segments to make room for | |
389 | * appending the drain buffer. If you call | |
390 | * blk_queue_max_hw_segments() or blk_queue_max_phys_segments() after | |
391 | * calling this routine, you must set the limit to one fewer than your | |
392 | * device can support otherwise there won't be room for the drain | |
393 | * buffer. | |
394 | */ | |
448da4d2 | 395 | int blk_queue_dma_drain(struct request_queue *q, |
2fb98e84 TH |
396 | dma_drain_needed_fn *dma_drain_needed, |
397 | void *buf, unsigned int size) | |
86db1e29 JA |
398 | { |
399 | if (q->max_hw_segments < 2 || q->max_phys_segments < 2) | |
400 | return -EINVAL; | |
401 | /* make room for appending the drain */ | |
402 | --q->max_hw_segments; | |
403 | --q->max_phys_segments; | |
2fb98e84 | 404 | q->dma_drain_needed = dma_drain_needed; |
86db1e29 JA |
405 | q->dma_drain_buffer = buf; |
406 | q->dma_drain_size = size; | |
407 | ||
408 | return 0; | |
409 | } | |
86db1e29 JA |
410 | EXPORT_SYMBOL_GPL(blk_queue_dma_drain); |
411 | ||
412 | /** | |
413 | * blk_queue_segment_boundary - set boundary rules for segment merging | |
414 | * @q: the request queue for the device | |
415 | * @mask: the memory boundary mask | |
416 | **/ | |
417 | void blk_queue_segment_boundary(struct request_queue *q, unsigned long mask) | |
418 | { | |
419 | if (mask < PAGE_CACHE_SIZE - 1) { | |
420 | mask = PAGE_CACHE_SIZE - 1; | |
24c03d47 HH |
421 | printk(KERN_INFO "%s: set to minimum %lx\n", |
422 | __func__, mask); | |
86db1e29 JA |
423 | } |
424 | ||
425 | q->seg_boundary_mask = mask; | |
426 | } | |
86db1e29 JA |
427 | EXPORT_SYMBOL(blk_queue_segment_boundary); |
428 | ||
429 | /** | |
430 | * blk_queue_dma_alignment - set dma length and memory alignment | |
431 | * @q: the request queue for the device | |
432 | * @mask: alignment mask | |
433 | * | |
434 | * description: | |
710027a4 | 435 | * set required memory and length alignment for direct dma transactions. |
8feb4d20 | 436 | * this is used when building direct io requests for the queue. |
86db1e29 JA |
437 | * |
438 | **/ | |
439 | void blk_queue_dma_alignment(struct request_queue *q, int mask) | |
440 | { | |
441 | q->dma_alignment = mask; | |
442 | } | |
86db1e29 JA |
443 | EXPORT_SYMBOL(blk_queue_dma_alignment); |
444 | ||
445 | /** | |
446 | * blk_queue_update_dma_alignment - update dma length and memory alignment | |
447 | * @q: the request queue for the device | |
448 | * @mask: alignment mask | |
449 | * | |
450 | * description: | |
710027a4 | 451 | * update required memory and length alignment for direct dma transactions. |
86db1e29 JA |
452 | * If the requested alignment is larger than the current alignment, then |
453 | * the current queue alignment is updated to the new value, otherwise it | |
454 | * is left alone. The design of this is to allow multiple objects | |
455 | * (driver, device, transport etc) to set their respective | |
456 | * alignments without having them interfere. | |
457 | * | |
458 | **/ | |
459 | void blk_queue_update_dma_alignment(struct request_queue *q, int mask) | |
460 | { | |
461 | BUG_ON(mask > PAGE_SIZE); | |
462 | ||
463 | if (mask > q->dma_alignment) | |
464 | q->dma_alignment = mask; | |
465 | } | |
86db1e29 JA |
466 | EXPORT_SYMBOL(blk_queue_update_dma_alignment); |
467 | ||
aeb3d3a8 | 468 | static int __init blk_settings_init(void) |
86db1e29 JA |
469 | { |
470 | blk_max_low_pfn = max_low_pfn - 1; | |
471 | blk_max_pfn = max_pfn - 1; | |
472 | return 0; | |
473 | } | |
474 | subsys_initcall(blk_settings_init); |