blk-settings.c 25 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807
  1. /*
  2. * Functions related to setting various queue properties from drivers
  3. */
  4. #include <linux/kernel.h>
  5. #include <linux/module.h>
  6. #include <linux/init.h>
  7. #include <linux/bio.h>
  8. #include <linux/blkdev.h>
  9. #include <linux/bootmem.h> /* for max_pfn/max_low_pfn */
  10. #include <linux/gcd.h>
  11. #include <linux/lcm.h>
  12. #include <linux/jiffies.h>
  13. #include <linux/gfp.h>
  14. #include "blk.h"
  15. unsigned long blk_max_low_pfn;
  16. EXPORT_SYMBOL(blk_max_low_pfn);
  17. unsigned long blk_max_pfn;
  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. }
  33. EXPORT_SYMBOL(blk_queue_prep_rq);
  34. /**
  35. * blk_queue_unprep_rq - set an unprepare_request function for queue
  36. * @q: queue
  37. * @ufn: unprepare_request function
  38. *
  39. * It's possible for a queue to register an unprepare_request callback
  40. * which is invoked before the request is finally completed. The goal
  41. * of the function is to deallocate any data that was allocated in the
  42. * prepare_request callback.
  43. *
  44. */
  45. void blk_queue_unprep_rq(struct request_queue *q, unprep_rq_fn *ufn)
  46. {
  47. q->unprep_rq_fn = ufn;
  48. }
  49. EXPORT_SYMBOL(blk_queue_unprep_rq);
  50. /**
  51. * blk_queue_merge_bvec - set a merge_bvec function for queue
  52. * @q: queue
  53. * @mbfn: merge_bvec_fn
  54. *
  55. * Usually queues have static limitations on the max sectors or segments that
  56. * we can put in a request. Stacking drivers may have some settings that
  57. * are dynamic, and thus we have to query the queue whether it is ok to
  58. * add a new bio_vec to a bio at a given offset or not. If the block device
  59. * has such limitations, it needs to register a merge_bvec_fn to control
  60. * the size of bio's sent to it. Note that a block device *must* allow a
  61. * single page to be added to an empty bio. The block device driver may want
  62. * to use the bio_split() function to deal with these bio's. By default
  63. * no merge_bvec_fn is defined for a queue, and only the fixed limits are
  64. * honored.
  65. */
  66. void blk_queue_merge_bvec(struct request_queue *q, merge_bvec_fn *mbfn)
  67. {
  68. q->merge_bvec_fn = mbfn;
  69. }
  70. EXPORT_SYMBOL(blk_queue_merge_bvec);
  71. void blk_queue_softirq_done(struct request_queue *q, softirq_done_fn *fn)
  72. {
  73. q->softirq_done_fn = fn;
  74. }
  75. EXPORT_SYMBOL(blk_queue_softirq_done);
  76. void blk_queue_rq_timeout(struct request_queue *q, unsigned int timeout)
  77. {
  78. q->rq_timeout = timeout;
  79. }
  80. EXPORT_SYMBOL_GPL(blk_queue_rq_timeout);
  81. void blk_queue_rq_timed_out(struct request_queue *q, rq_timed_out_fn *fn)
  82. {
  83. q->rq_timed_out_fn = fn;
  84. }
  85. EXPORT_SYMBOL_GPL(blk_queue_rq_timed_out);
  86. void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn)
  87. {
  88. q->lld_busy_fn = fn;
  89. }
  90. EXPORT_SYMBOL_GPL(blk_queue_lld_busy);
  91. /**
  92. * blk_set_default_limits - reset limits to default values
  93. * @lim: the queue_limits structure to reset
  94. *
  95. * Description:
  96. * Returns a queue_limit struct to its default state. Can be used by
  97. * stacking drivers like DM that stage table swaps and reuse an
  98. * existing device queue.
  99. */
  100. void blk_set_default_limits(struct queue_limits *lim)
  101. {
  102. lim->max_segments = BLK_MAX_SEGMENTS;
  103. lim->max_integrity_segments = 0;
  104. lim->seg_boundary_mask = BLK_SEG_BOUNDARY_MASK;
  105. lim->max_segment_size = BLK_MAX_SEGMENT_SIZE;
  106. lim->max_sectors = BLK_DEF_MAX_SECTORS;
  107. lim->max_hw_sectors = INT_MAX;
  108. lim->max_discard_sectors = 0;
  109. lim->discard_granularity = 0;
  110. lim->discard_alignment = 0;
  111. lim->discard_misaligned = 0;
  112. lim->discard_zeroes_data = 1;
  113. lim->logical_block_size = lim->physical_block_size = lim->io_min = 512;
  114. lim->bounce_pfn = (unsigned long)(BLK_BOUNCE_ANY >> PAGE_SHIFT);
  115. lim->alignment_offset = 0;
  116. lim->io_opt = 0;
  117. lim->misaligned = 0;
  118. lim->cluster = 1;
  119. }
  120. EXPORT_SYMBOL(blk_set_default_limits);
  121. /**
  122. * blk_queue_make_request - define an alternate make_request function for a device
  123. * @q: the request queue for the device to be affected
  124. * @mfn: the alternate make_request function
  125. *
  126. * Description:
  127. * The normal way for &struct bios to be passed to a device
  128. * driver is for them to be collected into requests on a request
  129. * queue, and then to allow the device driver to select requests
  130. * off that queue when it is ready. This works well for many block
  131. * devices. However some block devices (typically virtual devices
  132. * such as md or lvm) do not benefit from the processing on the
  133. * request queue, and are served best by having the requests passed
  134. * directly to them. This can be achieved by providing a function
  135. * to blk_queue_make_request().
  136. *
  137. * Caveat:
  138. * The driver that does this *must* be able to deal appropriately
  139. * with buffers in "highmemory". This can be accomplished by either calling
  140. * __bio_kmap_atomic() to get a temporary kernel mapping, or by calling
  141. * blk_queue_bounce() to create a buffer in normal memory.
  142. **/
  143. void blk_queue_make_request(struct request_queue *q, make_request_fn *mfn)
  144. {
  145. /*
  146. * set defaults
  147. */
  148. q->nr_requests = BLKDEV_MAX_RQ;
  149. q->make_request_fn = mfn;
  150. blk_queue_dma_alignment(q, 511);
  151. blk_queue_congestion_threshold(q);
  152. q->nr_batching = BLK_BATCH_REQ;
  153. blk_set_default_limits(&q->limits);
  154. blk_queue_max_hw_sectors(q, BLK_SAFE_MAX_SECTORS);
  155. q->limits.discard_zeroes_data = 0;
  156. /*
  157. * by default assume old behaviour and bounce for any highmem page
  158. */
  159. blk_queue_bounce_limit(q, BLK_BOUNCE_HIGH);
  160. }
  161. EXPORT_SYMBOL(blk_queue_make_request);
  162. /**
  163. * blk_queue_bounce_limit - set bounce buffer limit for queue
  164. * @q: the request queue for the device
  165. * @dma_mask: the maximum address the device can handle
  166. *
  167. * Description:
  168. * Different hardware can have different requirements as to what pages
  169. * it can do I/O directly to. A low level driver can call
  170. * blk_queue_bounce_limit to have lower memory pages allocated as bounce
  171. * buffers for doing I/O to pages residing above @dma_mask.
  172. **/
  173. void blk_queue_bounce_limit(struct request_queue *q, u64 dma_mask)
  174. {
  175. unsigned long b_pfn = dma_mask >> PAGE_SHIFT;
  176. int dma = 0;
  177. q->bounce_gfp = GFP_NOIO;
  178. #if BITS_PER_LONG == 64
  179. /*
  180. * Assume anything <= 4GB can be handled by IOMMU. Actually
  181. * some IOMMUs can handle everything, but I don't know of a
  182. * way to test this here.
  183. */
  184. if (b_pfn < (min_t(u64, 0xffffffffUL, BLK_BOUNCE_HIGH) >> PAGE_SHIFT))
  185. dma = 1;
  186. q->limits.bounce_pfn = max(max_low_pfn, b_pfn);
  187. #else
  188. if (b_pfn < blk_max_low_pfn)
  189. dma = 1;
  190. q->limits.bounce_pfn = b_pfn;
  191. #endif
  192. if (dma) {
  193. init_emergency_isa_pool();
  194. q->bounce_gfp = GFP_NOIO | GFP_DMA;
  195. q->limits.bounce_pfn = b_pfn;
  196. }
  197. }
  198. EXPORT_SYMBOL(blk_queue_bounce_limit);
  199. /**
  200. * blk_limits_max_hw_sectors - set hard and soft limit of max sectors for request
  201. * @limits: the queue limits
  202. * @max_hw_sectors: max hardware sectors in the usual 512b unit
  203. *
  204. * Description:
  205. * Enables a low level driver to set a hard upper limit,
  206. * max_hw_sectors, on the size of requests. max_hw_sectors is set by
  207. * the device driver based upon the combined capabilities of I/O
  208. * controller and storage device.
  209. *
  210. * max_sectors is a soft limit imposed by the block layer for
  211. * filesystem type requests. This value can be overridden on a
  212. * per-device basis in /sys/block/<device>/queue/max_sectors_kb.
  213. * The soft limit can not exceed max_hw_sectors.
  214. **/
  215. void blk_limits_max_hw_sectors(struct queue_limits *limits, unsigned int max_hw_sectors)
  216. {
  217. if ((max_hw_sectors << 9) < PAGE_CACHE_SIZE) {
  218. max_hw_sectors = 1 << (PAGE_CACHE_SHIFT - 9);
  219. printk(KERN_INFO "%s: set to minimum %d\n",
  220. __func__, max_hw_sectors);
  221. }
  222. limits->max_hw_sectors = max_hw_sectors;
  223. limits->max_sectors = min_t(unsigned int, max_hw_sectors,
  224. BLK_DEF_MAX_SECTORS);
  225. }
  226. EXPORT_SYMBOL(blk_limits_max_hw_sectors);
  227. /**
  228. * blk_queue_max_hw_sectors - set max sectors for a request for this queue
  229. * @q: the request queue for the device
  230. * @max_hw_sectors: max hardware sectors in the usual 512b unit
  231. *
  232. * Description:
  233. * See description for blk_limits_max_hw_sectors().
  234. **/
  235. void blk_queue_max_hw_sectors(struct request_queue *q, unsigned int max_hw_sectors)
  236. {
  237. blk_limits_max_hw_sectors(&q->limits, max_hw_sectors);
  238. }
  239. EXPORT_SYMBOL(blk_queue_max_hw_sectors);
  240. /**
  241. * blk_queue_max_discard_sectors - set max sectors for a single discard
  242. * @q: the request queue for the device
  243. * @max_discard_sectors: maximum number of sectors to discard
  244. **/
  245. void blk_queue_max_discard_sectors(struct request_queue *q,
  246. unsigned int max_discard_sectors)
  247. {
  248. q->limits.max_discard_sectors = max_discard_sectors;
  249. }
  250. EXPORT_SYMBOL(blk_queue_max_discard_sectors);
  251. /**
  252. * blk_queue_max_segments - set max hw segments for a request for this queue
  253. * @q: the request queue for the device
  254. * @max_segments: max number of segments
  255. *
  256. * Description:
  257. * Enables a low level driver to set an upper limit on the number of
  258. * hw data segments in a request.
  259. **/
  260. void blk_queue_max_segments(struct request_queue *q, unsigned short max_segments)
  261. {
  262. if (!max_segments) {
  263. max_segments = 1;
  264. printk(KERN_INFO "%s: set to minimum %d\n",
  265. __func__, max_segments);
  266. }
  267. q->limits.max_segments = max_segments;
  268. }
  269. EXPORT_SYMBOL(blk_queue_max_segments);
  270. /**
  271. * blk_queue_max_segment_size - set max segment size for blk_rq_map_sg
  272. * @q: the request queue for the device
  273. * @max_size: max size of segment in bytes
  274. *
  275. * Description:
  276. * Enables a low level driver to set an upper limit on the size of a
  277. * coalesced segment
  278. **/
  279. void blk_queue_max_segment_size(struct request_queue *q, unsigned int max_size)
  280. {
  281. if (max_size < PAGE_CACHE_SIZE) {
  282. max_size = PAGE_CACHE_SIZE;
  283. printk(KERN_INFO "%s: set to minimum %d\n",
  284. __func__, max_size);
  285. }
  286. q->limits.max_segment_size = max_size;
  287. }
  288. EXPORT_SYMBOL(blk_queue_max_segment_size);
  289. /**
  290. * blk_queue_logical_block_size - set logical block size for the queue
  291. * @q: the request queue for the device
  292. * @size: the logical block size, in bytes
  293. *
  294. * Description:
  295. * This should be set to the lowest possible block size that the
  296. * storage device can address. The default of 512 covers most
  297. * hardware.
  298. **/
  299. void blk_queue_logical_block_size(struct request_queue *q, unsigned short size)
  300. {
  301. q->limits.logical_block_size = size;
  302. if (q->limits.physical_block_size < size)
  303. q->limits.physical_block_size = size;
  304. if (q->limits.io_min < q->limits.physical_block_size)
  305. q->limits.io_min = q->limits.physical_block_size;
  306. }
  307. EXPORT_SYMBOL(blk_queue_logical_block_size);
  308. /**
  309. * blk_queue_physical_block_size - set physical block size for the queue
  310. * @q: the request queue for the device
  311. * @size: the physical block size, in bytes
  312. *
  313. * Description:
  314. * This should be set to the lowest possible sector size that the
  315. * hardware can operate on without reverting to read-modify-write
  316. * operations.
  317. */
  318. void blk_queue_physical_block_size(struct request_queue *q, unsigned int size)
  319. {
  320. q->limits.physical_block_size = size;
  321. if (q->limits.physical_block_size < q->limits.logical_block_size)
  322. q->limits.physical_block_size = q->limits.logical_block_size;
  323. if (q->limits.io_min < q->limits.physical_block_size)
  324. q->limits.io_min = q->limits.physical_block_size;
  325. }
  326. EXPORT_SYMBOL(blk_queue_physical_block_size);
  327. /**
  328. * blk_queue_alignment_offset - set physical block alignment offset
  329. * @q: the request queue for the device
  330. * @offset: alignment offset in bytes
  331. *
  332. * Description:
  333. * Some devices are naturally misaligned to compensate for things like
  334. * the legacy DOS partition table 63-sector offset. Low-level drivers
  335. * should call this function for devices whose first sector is not
  336. * naturally aligned.
  337. */
  338. void blk_queue_alignment_offset(struct request_queue *q, unsigned int offset)
  339. {
  340. q->limits.alignment_offset =
  341. offset & (q->limits.physical_block_size - 1);
  342. q->limits.misaligned = 0;
  343. }
  344. EXPORT_SYMBOL(blk_queue_alignment_offset);
  345. /**
  346. * blk_limits_io_min - set minimum request size for a device
  347. * @limits: the queue limits
  348. * @min: smallest I/O size in bytes
  349. *
  350. * Description:
  351. * Some devices have an internal block size bigger than the reported
  352. * hardware sector size. This function can be used to signal the
  353. * smallest I/O the device can perform without incurring a performance
  354. * penalty.
  355. */
  356. void blk_limits_io_min(struct queue_limits *limits, unsigned int min)
  357. {
  358. limits->io_min = min;
  359. if (limits->io_min < limits->logical_block_size)
  360. limits->io_min = limits->logical_block_size;
  361. if (limits->io_min < limits->physical_block_size)
  362. limits->io_min = limits->physical_block_size;
  363. }
  364. EXPORT_SYMBOL(blk_limits_io_min);
  365. /**
  366. * blk_queue_io_min - set minimum request size for the queue
  367. * @q: the request queue for the device
  368. * @min: smallest I/O size in bytes
  369. *
  370. * Description:
  371. * Storage devices may report a granularity or preferred minimum I/O
  372. * size which is the smallest request the device can perform without
  373. * incurring a performance penalty. For disk drives this is often the
  374. * physical block size. For RAID arrays it is often the stripe chunk
  375. * size. A properly aligned multiple of minimum_io_size is the
  376. * preferred request size for workloads where a high number of I/O
  377. * operations is desired.
  378. */
  379. void blk_queue_io_min(struct request_queue *q, unsigned int min)
  380. {
  381. blk_limits_io_min(&q->limits, min);
  382. }
  383. EXPORT_SYMBOL(blk_queue_io_min);
  384. /**
  385. * blk_limits_io_opt - set optimal request size for a device
  386. * @limits: the queue limits
  387. * @opt: smallest I/O size in bytes
  388. *
  389. * Description:
  390. * Storage devices may report an optimal I/O size, which is the
  391. * device's preferred unit for sustained I/O. This is rarely reported
  392. * for disk drives. For RAID arrays it is usually the stripe width or
  393. * the internal track size. A properly aligned multiple of
  394. * optimal_io_size is the preferred request size for workloads where
  395. * sustained throughput is desired.
  396. */
  397. void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt)
  398. {
  399. limits->io_opt = opt;
  400. }
  401. EXPORT_SYMBOL(blk_limits_io_opt);
  402. /**
  403. * blk_queue_io_opt - set optimal request size for the queue
  404. * @q: the request queue for the device
  405. * @opt: optimal request size in bytes
  406. *
  407. * Description:
  408. * Storage devices may report an optimal I/O size, which is the
  409. * device's preferred unit for sustained I/O. This is rarely reported
  410. * for disk drives. For RAID arrays it is usually the stripe width or
  411. * the internal track size. A properly aligned multiple of
  412. * optimal_io_size is the preferred request size for workloads where
  413. * sustained throughput is desired.
  414. */
  415. void blk_queue_io_opt(struct request_queue *q, unsigned int opt)
  416. {
  417. blk_limits_io_opt(&q->limits, opt);
  418. }
  419. EXPORT_SYMBOL(blk_queue_io_opt);
  420. /**
  421. * blk_queue_stack_limits - inherit underlying queue limits for stacked drivers
  422. * @t: the stacking driver (top)
  423. * @b: the underlying device (bottom)
  424. **/
  425. void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b)
  426. {
  427. blk_stack_limits(&t->limits, &b->limits, 0);
  428. }
  429. EXPORT_SYMBOL(blk_queue_stack_limits);
  430. /**
  431. * blk_stack_limits - adjust queue_limits for stacked devices
  432. * @t: the stacking driver limits (top device)
  433. * @b: the underlying queue limits (bottom, component device)
  434. * @start: first data sector within component device
  435. *
  436. * Description:
  437. * This function is used by stacking drivers like MD and DM to ensure
  438. * that all component devices have compatible block sizes and
  439. * alignments. The stacking driver must provide a queue_limits
  440. * struct (top) and then iteratively call the stacking function for
  441. * all component (bottom) devices. The stacking function will
  442. * attempt to combine the values and ensure proper alignment.
  443. *
  444. * Returns 0 if the top and bottom queue_limits are compatible. The
  445. * top device's block sizes and alignment offsets may be adjusted to
  446. * ensure alignment with the bottom device. If no compatible sizes
  447. * and alignments exist, -1 is returned and the resulting top
  448. * queue_limits will have the misaligned flag set to indicate that
  449. * the alignment_offset is undefined.
  450. */
  451. int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
  452. sector_t start)
  453. {
  454. unsigned int top, bottom, alignment, ret = 0;
  455. t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors);
  456. t->max_hw_sectors = min_not_zero(t->max_hw_sectors, b->max_hw_sectors);
  457. t->bounce_pfn = min_not_zero(t->bounce_pfn, b->bounce_pfn);
  458. t->seg_boundary_mask = min_not_zero(t->seg_boundary_mask,
  459. b->seg_boundary_mask);
  460. t->max_segments = min_not_zero(t->max_segments, b->max_segments);
  461. t->max_integrity_segments = min_not_zero(t->max_integrity_segments,
  462. b->max_integrity_segments);
  463. t->max_segment_size = min_not_zero(t->max_segment_size,
  464. b->max_segment_size);
  465. t->misaligned |= b->misaligned;
  466. alignment = queue_limit_alignment_offset(b, start);
  467. /* Bottom device has different alignment. Check that it is
  468. * compatible with the current top alignment.
  469. */
  470. if (t->alignment_offset != alignment) {
  471. top = max(t->physical_block_size, t->io_min)
  472. + t->alignment_offset;
  473. bottom = max(b->physical_block_size, b->io_min) + alignment;
  474. /* Verify that top and bottom intervals line up */
  475. if (max(top, bottom) & (min(top, bottom) - 1)) {
  476. t->misaligned = 1;
  477. ret = -1;
  478. }
  479. }
  480. t->logical_block_size = max(t->logical_block_size,
  481. b->logical_block_size);
  482. t->physical_block_size = max(t->physical_block_size,
  483. b->physical_block_size);
  484. t->io_min = max(t->io_min, b->io_min);
  485. t->io_opt = lcm(t->io_opt, b->io_opt);
  486. t->cluster &= b->cluster;
  487. t->discard_zeroes_data &= b->discard_zeroes_data;
  488. /* Physical block size a multiple of the logical block size? */
  489. if (t->physical_block_size & (t->logical_block_size - 1)) {
  490. t->physical_block_size = t->logical_block_size;
  491. t->misaligned = 1;
  492. ret = -1;
  493. }
  494. /* Minimum I/O a multiple of the physical block size? */
  495. if (t->io_min & (t->physical_block_size - 1)) {
  496. t->io_min = t->physical_block_size;
  497. t->misaligned = 1;
  498. ret = -1;
  499. }
  500. /* Optimal I/O a multiple of the physical block size? */
  501. if (t->io_opt & (t->physical_block_size - 1)) {
  502. t->io_opt = 0;
  503. t->misaligned = 1;
  504. ret = -1;
  505. }
  506. /* Find lowest common alignment_offset */
  507. t->alignment_offset = lcm(t->alignment_offset, alignment)
  508. & (max(t->physical_block_size, t->io_min) - 1);
  509. /* Verify that new alignment_offset is on a logical block boundary */
  510. if (t->alignment_offset & (t->logical_block_size - 1)) {
  511. t->misaligned = 1;
  512. ret = -1;
  513. }
  514. /* Discard alignment and granularity */
  515. if (b->discard_granularity) {
  516. alignment = queue_limit_discard_alignment(b, start);
  517. if (t->discard_granularity != 0 &&
  518. t->discard_alignment != alignment) {
  519. top = t->discard_granularity + t->discard_alignment;
  520. bottom = b->discard_granularity + alignment;
  521. /* Verify that top and bottom intervals line up */
  522. if (max(top, bottom) & (min(top, bottom) - 1))
  523. t->discard_misaligned = 1;
  524. }
  525. t->max_discard_sectors = min_not_zero(t->max_discard_sectors,
  526. b->max_discard_sectors);
  527. t->discard_granularity = max(t->discard_granularity,
  528. b->discard_granularity);
  529. t->discard_alignment = lcm(t->discard_alignment, alignment) &
  530. (t->discard_granularity - 1);
  531. }
  532. return ret;
  533. }
  534. EXPORT_SYMBOL(blk_stack_limits);
  535. /**
  536. * bdev_stack_limits - adjust queue limits for stacked drivers
  537. * @t: the stacking driver limits (top device)
  538. * @bdev: the component block_device (bottom)
  539. * @start: first data sector within component device
  540. *
  541. * Description:
  542. * Merges queue limits for a top device and a block_device. Returns
  543. * 0 if alignment didn't change. Returns -1 if adding the bottom
  544. * device caused misalignment.
  545. */
  546. int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
  547. sector_t start)
  548. {
  549. struct request_queue *bq = bdev_get_queue(bdev);
  550. start += get_start_sect(bdev);
  551. return blk_stack_limits(t, &bq->limits, start);
  552. }
  553. EXPORT_SYMBOL(bdev_stack_limits);
  554. /**
  555. * disk_stack_limits - adjust queue limits for stacked drivers
  556. * @disk: MD/DM gendisk (top)
  557. * @bdev: the underlying block device (bottom)
  558. * @offset: offset to beginning of data within component device
  559. *
  560. * Description:
  561. * Merges the limits for a top level gendisk and a bottom level
  562. * block_device.
  563. */
  564. void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
  565. sector_t offset)
  566. {
  567. struct request_queue *t = disk->queue;
  568. if (bdev_stack_limits(&t->limits, bdev, offset >> 9) < 0) {
  569. char top[BDEVNAME_SIZE], bottom[BDEVNAME_SIZE];
  570. disk_name(disk, 0, top);
  571. bdevname(bdev, bottom);
  572. printk(KERN_NOTICE "%s: Warning: Device %s is misaligned\n",
  573. top, bottom);
  574. }
  575. }
  576. EXPORT_SYMBOL(disk_stack_limits);
  577. /**
  578. * blk_queue_dma_pad - set pad mask
  579. * @q: the request queue for the device
  580. * @mask: pad mask
  581. *
  582. * Set dma pad mask.
  583. *
  584. * Appending pad buffer to a request modifies the last entry of a
  585. * scatter list such that it includes the pad buffer.
  586. **/
  587. void blk_queue_dma_pad(struct request_queue *q, unsigned int mask)
  588. {
  589. q->dma_pad_mask = mask;
  590. }
  591. EXPORT_SYMBOL(blk_queue_dma_pad);
  592. /**
  593. * blk_queue_update_dma_pad - update pad mask
  594. * @q: the request queue for the device
  595. * @mask: pad mask
  596. *
  597. * Update dma pad mask.
  598. *
  599. * Appending pad buffer to a request modifies the last entry of a
  600. * scatter list such that it includes the pad buffer.
  601. **/
  602. void blk_queue_update_dma_pad(struct request_queue *q, unsigned int mask)
  603. {
  604. if (mask > q->dma_pad_mask)
  605. q->dma_pad_mask = mask;
  606. }
  607. EXPORT_SYMBOL(blk_queue_update_dma_pad);
  608. /**
  609. * blk_queue_dma_drain - Set up a drain buffer for excess dma.
  610. * @q: the request queue for the device
  611. * @dma_drain_needed: fn which returns non-zero if drain is necessary
  612. * @buf: physically contiguous buffer
  613. * @size: size of the buffer in bytes
  614. *
  615. * Some devices have excess DMA problems and can't simply discard (or
  616. * zero fill) the unwanted piece of the transfer. They have to have a
  617. * real area of memory to transfer it into. The use case for this is
  618. * ATAPI devices in DMA mode. If the packet command causes a transfer
  619. * bigger than the transfer size some HBAs will lock up if there
  620. * aren't DMA elements to contain the excess transfer. What this API
  621. * does is adjust the queue so that the buf is always appended
  622. * silently to the scatterlist.
  623. *
  624. * Note: This routine adjusts max_hw_segments to make room for appending
  625. * the drain buffer. If you call blk_queue_max_segments() after calling
  626. * this routine, you must set the limit to one fewer than your device
  627. * can support otherwise there won't be room for the drain buffer.
  628. */
  629. int blk_queue_dma_drain(struct request_queue *q,
  630. dma_drain_needed_fn *dma_drain_needed,
  631. void *buf, unsigned int size)
  632. {
  633. if (queue_max_segments(q) < 2)
  634. return -EINVAL;
  635. /* make room for appending the drain */
  636. blk_queue_max_segments(q, queue_max_segments(q) - 1);
  637. q->dma_drain_needed = dma_drain_needed;
  638. q->dma_drain_buffer = buf;
  639. q->dma_drain_size = size;
  640. return 0;
  641. }
  642. EXPORT_SYMBOL_GPL(blk_queue_dma_drain);
  643. /**
  644. * blk_queue_segment_boundary - set boundary rules for segment merging
  645. * @q: the request queue for the device
  646. * @mask: the memory boundary mask
  647. **/
  648. void blk_queue_segment_boundary(struct request_queue *q, unsigned long mask)
  649. {
  650. if (mask < PAGE_CACHE_SIZE - 1) {
  651. mask = PAGE_CACHE_SIZE - 1;
  652. printk(KERN_INFO "%s: set to minimum %lx\n",
  653. __func__, mask);
  654. }
  655. q->limits.seg_boundary_mask = mask;
  656. }
  657. EXPORT_SYMBOL(blk_queue_segment_boundary);
  658. /**
  659. * blk_queue_dma_alignment - set dma length and memory alignment
  660. * @q: the request queue for the device
  661. * @mask: alignment mask
  662. *
  663. * description:
  664. * set required memory and length alignment for direct dma transactions.
  665. * this is used when building direct io requests for the queue.
  666. *
  667. **/
  668. void blk_queue_dma_alignment(struct request_queue *q, int mask)
  669. {
  670. q->dma_alignment = mask;
  671. }
  672. EXPORT_SYMBOL(blk_queue_dma_alignment);
  673. /**
  674. * blk_queue_update_dma_alignment - update dma length and memory alignment
  675. * @q: the request queue for the device
  676. * @mask: alignment mask
  677. *
  678. * description:
  679. * update required memory and length alignment for direct dma transactions.
  680. * If the requested alignment is larger than the current alignment, then
  681. * the current queue alignment is updated to the new value, otherwise it
  682. * is left alone. The design of this is to allow multiple objects
  683. * (driver, device, transport etc) to set their respective
  684. * alignments without having them interfere.
  685. *
  686. **/
  687. void blk_queue_update_dma_alignment(struct request_queue *q, int mask)
  688. {
  689. BUG_ON(mask > PAGE_SIZE);
  690. if (mask > q->dma_alignment)
  691. q->dma_alignment = mask;
  692. }
  693. EXPORT_SYMBOL(blk_queue_update_dma_alignment);
  694. /**
  695. * blk_queue_flush - configure queue's cache flush capability
  696. * @q: the request queue for the device
  697. * @flush: 0, REQ_FLUSH or REQ_FLUSH | REQ_FUA
  698. *
  699. * Tell block layer cache flush capability of @q. If it supports
  700. * flushing, REQ_FLUSH should be set. If it supports bypassing
  701. * write cache for individual writes, REQ_FUA should be set.
  702. */
  703. void blk_queue_flush(struct request_queue *q, unsigned int flush)
  704. {
  705. WARN_ON_ONCE(flush & ~(REQ_FLUSH | REQ_FUA));
  706. if (WARN_ON_ONCE(!(flush & REQ_FLUSH) && (flush & REQ_FUA)))
  707. flush &= ~REQ_FUA;
  708. q->flush_flags = flush & (REQ_FLUSH | REQ_FUA);
  709. }
  710. EXPORT_SYMBOL_GPL(blk_queue_flush);
  711. void blk_queue_flush_queueable(struct request_queue *q, bool queueable)
  712. {
  713. q->flush_not_queueable = !queueable;
  714. }
  715. EXPORT_SYMBOL_GPL(blk_queue_flush_queueable);
  716. static int __init blk_settings_init(void)
  717. {
  718. blk_max_low_pfn = max_low_pfn - 1;
  719. blk_max_pfn = max_pfn - 1;
  720. return 0;
  721. }
  722. subsys_initcall(blk_settings_init);