blk-settings.c 26 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822
  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.
  97. */
  98. void blk_set_default_limits(struct queue_limits *lim)
  99. {
  100. lim->max_segments = BLK_MAX_SEGMENTS;
  101. lim->max_integrity_segments = 0;
  102. lim->seg_boundary_mask = BLK_SEG_BOUNDARY_MASK;
  103. lim->max_segment_size = BLK_MAX_SEGMENT_SIZE;
  104. lim->max_sectors = lim->max_hw_sectors = BLK_SAFE_MAX_SECTORS;
  105. lim->max_discard_sectors = 0;
  106. lim->discard_granularity = 0;
  107. lim->discard_alignment = 0;
  108. lim->discard_misaligned = 0;
  109. lim->discard_zeroes_data = 0;
  110. lim->logical_block_size = lim->physical_block_size = lim->io_min = 512;
  111. lim->bounce_pfn = (unsigned long)(BLK_BOUNCE_ANY >> PAGE_SHIFT);
  112. lim->alignment_offset = 0;
  113. lim->io_opt = 0;
  114. lim->misaligned = 0;
  115. lim->cluster = 1;
  116. }
  117. EXPORT_SYMBOL(blk_set_default_limits);
  118. /**
  119. * blk_set_stacking_limits - set default limits for stacking devices
  120. * @lim: the queue_limits structure to reset
  121. *
  122. * Description:
  123. * Returns a queue_limit struct to its default state. Should be used
  124. * by stacking drivers like DM that have no internal limits.
  125. */
  126. void blk_set_stacking_limits(struct queue_limits *lim)
  127. {
  128. blk_set_default_limits(lim);
  129. /* Inherit limits from component devices */
  130. lim->discard_zeroes_data = 1;
  131. lim->max_segments = USHRT_MAX;
  132. lim->max_hw_sectors = UINT_MAX;
  133. lim->max_sectors = BLK_DEF_MAX_SECTORS;
  134. }
  135. EXPORT_SYMBOL(blk_set_stacking_limits);
  136. /**
  137. * blk_queue_make_request - define an alternate make_request function for a device
  138. * @q: the request queue for the device to be affected
  139. * @mfn: the alternate make_request function
  140. *
  141. * Description:
  142. * The normal way for &struct bios to be passed to a device
  143. * driver is for them to be collected into requests on a request
  144. * queue, and then to allow the device driver to select requests
  145. * off that queue when it is ready. This works well for many block
  146. * devices. However some block devices (typically virtual devices
  147. * such as md or lvm) do not benefit from the processing on the
  148. * request queue, and are served best by having the requests passed
  149. * directly to them. This can be achieved by providing a function
  150. * to blk_queue_make_request().
  151. *
  152. * Caveat:
  153. * The driver that does this *must* be able to deal appropriately
  154. * with buffers in "highmemory". This can be accomplished by either calling
  155. * __bio_kmap_atomic() to get a temporary kernel mapping, or by calling
  156. * blk_queue_bounce() to create a buffer in normal memory.
  157. **/
  158. void blk_queue_make_request(struct request_queue *q, make_request_fn *mfn)
  159. {
  160. /*
  161. * set defaults
  162. */
  163. q->nr_requests = BLKDEV_MAX_RQ;
  164. q->make_request_fn = mfn;
  165. blk_queue_dma_alignment(q, 511);
  166. blk_queue_congestion_threshold(q);
  167. q->nr_batching = BLK_BATCH_REQ;
  168. blk_set_default_limits(&q->limits);
  169. /*
  170. * by default assume old behaviour and bounce for any highmem page
  171. */
  172. blk_queue_bounce_limit(q, BLK_BOUNCE_HIGH);
  173. }
  174. EXPORT_SYMBOL(blk_queue_make_request);
  175. /**
  176. * blk_queue_bounce_limit - set bounce buffer limit for queue
  177. * @q: the request queue for the device
  178. * @dma_mask: the maximum address the device can handle
  179. *
  180. * Description:
  181. * Different hardware can have different requirements as to what pages
  182. * it can do I/O directly to. A low level driver can call
  183. * blk_queue_bounce_limit to have lower memory pages allocated as bounce
  184. * buffers for doing I/O to pages residing above @dma_mask.
  185. **/
  186. void blk_queue_bounce_limit(struct request_queue *q, u64 dma_mask)
  187. {
  188. unsigned long b_pfn = dma_mask >> PAGE_SHIFT;
  189. int dma = 0;
  190. q->bounce_gfp = GFP_NOIO;
  191. #if BITS_PER_LONG == 64
  192. /*
  193. * Assume anything <= 4GB can be handled by IOMMU. Actually
  194. * some IOMMUs can handle everything, but I don't know of a
  195. * way to test this here.
  196. */
  197. if (b_pfn < (min_t(u64, 0xffffffffUL, BLK_BOUNCE_HIGH) >> PAGE_SHIFT))
  198. dma = 1;
  199. q->limits.bounce_pfn = max(max_low_pfn, b_pfn);
  200. #else
  201. if (b_pfn < blk_max_low_pfn)
  202. dma = 1;
  203. q->limits.bounce_pfn = b_pfn;
  204. #endif
  205. if (dma) {
  206. init_emergency_isa_pool();
  207. q->bounce_gfp = GFP_NOIO | GFP_DMA;
  208. q->limits.bounce_pfn = b_pfn;
  209. }
  210. }
  211. EXPORT_SYMBOL(blk_queue_bounce_limit);
  212. /**
  213. * blk_limits_max_hw_sectors - set hard and soft limit of max sectors for request
  214. * @limits: the queue limits
  215. * @max_hw_sectors: max hardware sectors in the usual 512b unit
  216. *
  217. * Description:
  218. * Enables a low level driver to set a hard upper limit,
  219. * max_hw_sectors, on the size of requests. max_hw_sectors is set by
  220. * the device driver based upon the combined capabilities of I/O
  221. * controller and storage device.
  222. *
  223. * max_sectors is a soft limit imposed by the block layer for
  224. * filesystem type requests. This value can be overridden on a
  225. * per-device basis in /sys/block/<device>/queue/max_sectors_kb.
  226. * The soft limit can not exceed max_hw_sectors.
  227. **/
  228. void blk_limits_max_hw_sectors(struct queue_limits *limits, unsigned int max_hw_sectors)
  229. {
  230. if ((max_hw_sectors << 9) < PAGE_CACHE_SIZE) {
  231. max_hw_sectors = 1 << (PAGE_CACHE_SHIFT - 9);
  232. printk(KERN_INFO "%s: set to minimum %d\n",
  233. __func__, max_hw_sectors);
  234. }
  235. limits->max_hw_sectors = max_hw_sectors;
  236. limits->max_sectors = min_t(unsigned int, max_hw_sectors,
  237. BLK_DEF_MAX_SECTORS);
  238. }
  239. EXPORT_SYMBOL(blk_limits_max_hw_sectors);
  240. /**
  241. * blk_queue_max_hw_sectors - set max sectors for a request for this queue
  242. * @q: the request queue for the device
  243. * @max_hw_sectors: max hardware sectors in the usual 512b unit
  244. *
  245. * Description:
  246. * See description for blk_limits_max_hw_sectors().
  247. **/
  248. void blk_queue_max_hw_sectors(struct request_queue *q, unsigned int max_hw_sectors)
  249. {
  250. blk_limits_max_hw_sectors(&q->limits, max_hw_sectors);
  251. }
  252. EXPORT_SYMBOL(blk_queue_max_hw_sectors);
  253. /**
  254. * blk_queue_max_discard_sectors - set max sectors for a single discard
  255. * @q: the request queue for the device
  256. * @max_discard_sectors: maximum number of sectors to discard
  257. **/
  258. void blk_queue_max_discard_sectors(struct request_queue *q,
  259. unsigned int max_discard_sectors)
  260. {
  261. q->limits.max_discard_sectors = max_discard_sectors;
  262. }
  263. EXPORT_SYMBOL(blk_queue_max_discard_sectors);
  264. /**
  265. * blk_queue_max_segments - set max hw segments for a request for this queue
  266. * @q: the request queue for the device
  267. * @max_segments: max number of segments
  268. *
  269. * Description:
  270. * Enables a low level driver to set an upper limit on the number of
  271. * hw data segments in a request.
  272. **/
  273. void blk_queue_max_segments(struct request_queue *q, unsigned short max_segments)
  274. {
  275. if (!max_segments) {
  276. max_segments = 1;
  277. printk(KERN_INFO "%s: set to minimum %d\n",
  278. __func__, max_segments);
  279. }
  280. q->limits.max_segments = max_segments;
  281. }
  282. EXPORT_SYMBOL(blk_queue_max_segments);
  283. /**
  284. * blk_queue_max_segment_size - set max segment size for blk_rq_map_sg
  285. * @q: the request queue for the device
  286. * @max_size: max size of segment in bytes
  287. *
  288. * Description:
  289. * Enables a low level driver to set an upper limit on the size of a
  290. * coalesced segment
  291. **/
  292. void blk_queue_max_segment_size(struct request_queue *q, unsigned int max_size)
  293. {
  294. if (max_size < PAGE_CACHE_SIZE) {
  295. max_size = PAGE_CACHE_SIZE;
  296. printk(KERN_INFO "%s: set to minimum %d\n",
  297. __func__, max_size);
  298. }
  299. q->limits.max_segment_size = max_size;
  300. }
  301. EXPORT_SYMBOL(blk_queue_max_segment_size);
  302. /**
  303. * blk_queue_logical_block_size - set logical block size for the queue
  304. * @q: the request queue for the device
  305. * @size: the logical block size, in bytes
  306. *
  307. * Description:
  308. * This should be set to the lowest possible block size that the
  309. * storage device can address. The default of 512 covers most
  310. * hardware.
  311. **/
  312. void blk_queue_logical_block_size(struct request_queue *q, unsigned short size)
  313. {
  314. q->limits.logical_block_size = size;
  315. if (q->limits.physical_block_size < size)
  316. q->limits.physical_block_size = size;
  317. if (q->limits.io_min < q->limits.physical_block_size)
  318. q->limits.io_min = q->limits.physical_block_size;
  319. }
  320. EXPORT_SYMBOL(blk_queue_logical_block_size);
  321. /**
  322. * blk_queue_physical_block_size - set physical block size for the queue
  323. * @q: the request queue for the device
  324. * @size: the physical block size, in bytes
  325. *
  326. * Description:
  327. * This should be set to the lowest possible sector size that the
  328. * hardware can operate on without reverting to read-modify-write
  329. * operations.
  330. */
  331. void blk_queue_physical_block_size(struct request_queue *q, unsigned int size)
  332. {
  333. q->limits.physical_block_size = size;
  334. if (q->limits.physical_block_size < q->limits.logical_block_size)
  335. q->limits.physical_block_size = q->limits.logical_block_size;
  336. if (q->limits.io_min < q->limits.physical_block_size)
  337. q->limits.io_min = q->limits.physical_block_size;
  338. }
  339. EXPORT_SYMBOL(blk_queue_physical_block_size);
  340. /**
  341. * blk_queue_alignment_offset - set physical block alignment offset
  342. * @q: the request queue for the device
  343. * @offset: alignment offset in bytes
  344. *
  345. * Description:
  346. * Some devices are naturally misaligned to compensate for things like
  347. * the legacy DOS partition table 63-sector offset. Low-level drivers
  348. * should call this function for devices whose first sector is not
  349. * naturally aligned.
  350. */
  351. void blk_queue_alignment_offset(struct request_queue *q, unsigned int offset)
  352. {
  353. q->limits.alignment_offset =
  354. offset & (q->limits.physical_block_size - 1);
  355. q->limits.misaligned = 0;
  356. }
  357. EXPORT_SYMBOL(blk_queue_alignment_offset);
  358. /**
  359. * blk_limits_io_min - set minimum request size for a device
  360. * @limits: the queue limits
  361. * @min: smallest I/O size in bytes
  362. *
  363. * Description:
  364. * Some devices have an internal block size bigger than the reported
  365. * hardware sector size. This function can be used to signal the
  366. * smallest I/O the device can perform without incurring a performance
  367. * penalty.
  368. */
  369. void blk_limits_io_min(struct queue_limits *limits, unsigned int min)
  370. {
  371. limits->io_min = min;
  372. if (limits->io_min < limits->logical_block_size)
  373. limits->io_min = limits->logical_block_size;
  374. if (limits->io_min < limits->physical_block_size)
  375. limits->io_min = limits->physical_block_size;
  376. }
  377. EXPORT_SYMBOL(blk_limits_io_min);
  378. /**
  379. * blk_queue_io_min - set minimum request size for the queue
  380. * @q: the request queue for the device
  381. * @min: smallest I/O size in bytes
  382. *
  383. * Description:
  384. * Storage devices may report a granularity or preferred minimum I/O
  385. * size which is the smallest request the device can perform without
  386. * incurring a performance penalty. For disk drives this is often the
  387. * physical block size. For RAID arrays it is often the stripe chunk
  388. * size. A properly aligned multiple of minimum_io_size is the
  389. * preferred request size for workloads where a high number of I/O
  390. * operations is desired.
  391. */
  392. void blk_queue_io_min(struct request_queue *q, unsigned int min)
  393. {
  394. blk_limits_io_min(&q->limits, min);
  395. }
  396. EXPORT_SYMBOL(blk_queue_io_min);
  397. /**
  398. * blk_limits_io_opt - set optimal request size for a device
  399. * @limits: the queue limits
  400. * @opt: smallest I/O size in bytes
  401. *
  402. * Description:
  403. * Storage devices may report an optimal I/O size, which is the
  404. * device's preferred unit for sustained I/O. This is rarely reported
  405. * for disk drives. For RAID arrays it is usually the stripe width or
  406. * the internal track size. A properly aligned multiple of
  407. * optimal_io_size is the preferred request size for workloads where
  408. * sustained throughput is desired.
  409. */
  410. void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt)
  411. {
  412. limits->io_opt = opt;
  413. }
  414. EXPORT_SYMBOL(blk_limits_io_opt);
  415. /**
  416. * blk_queue_io_opt - set optimal request size for the queue
  417. * @q: the request queue for the device
  418. * @opt: optimal request size in bytes
  419. *
  420. * Description:
  421. * Storage devices may report an optimal I/O size, which is the
  422. * device's preferred unit for sustained I/O. This is rarely reported
  423. * for disk drives. For RAID arrays it is usually the stripe width or
  424. * the internal track size. A properly aligned multiple of
  425. * optimal_io_size is the preferred request size for workloads where
  426. * sustained throughput is desired.
  427. */
  428. void blk_queue_io_opt(struct request_queue *q, unsigned int opt)
  429. {
  430. blk_limits_io_opt(&q->limits, opt);
  431. }
  432. EXPORT_SYMBOL(blk_queue_io_opt);
  433. /**
  434. * blk_queue_stack_limits - inherit underlying queue limits for stacked drivers
  435. * @t: the stacking driver (top)
  436. * @b: the underlying device (bottom)
  437. **/
  438. void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b)
  439. {
  440. blk_stack_limits(&t->limits, &b->limits, 0);
  441. }
  442. EXPORT_SYMBOL(blk_queue_stack_limits);
  443. /**
  444. * blk_stack_limits - adjust queue_limits for stacked devices
  445. * @t: the stacking driver limits (top device)
  446. * @b: the underlying queue limits (bottom, component device)
  447. * @start: first data sector within component device
  448. *
  449. * Description:
  450. * This function is used by stacking drivers like MD and DM to ensure
  451. * that all component devices have compatible block sizes and
  452. * alignments. The stacking driver must provide a queue_limits
  453. * struct (top) and then iteratively call the stacking function for
  454. * all component (bottom) devices. The stacking function will
  455. * attempt to combine the values and ensure proper alignment.
  456. *
  457. * Returns 0 if the top and bottom queue_limits are compatible. The
  458. * top device's block sizes and alignment offsets may be adjusted to
  459. * ensure alignment with the bottom device. If no compatible sizes
  460. * and alignments exist, -1 is returned and the resulting top
  461. * queue_limits will have the misaligned flag set to indicate that
  462. * the alignment_offset is undefined.
  463. */
  464. int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
  465. sector_t start)
  466. {
  467. unsigned int top, bottom, alignment, ret = 0;
  468. t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors);
  469. t->max_hw_sectors = min_not_zero(t->max_hw_sectors, b->max_hw_sectors);
  470. t->bounce_pfn = min_not_zero(t->bounce_pfn, b->bounce_pfn);
  471. t->seg_boundary_mask = min_not_zero(t->seg_boundary_mask,
  472. b->seg_boundary_mask);
  473. t->max_segments = min_not_zero(t->max_segments, b->max_segments);
  474. t->max_integrity_segments = min_not_zero(t->max_integrity_segments,
  475. b->max_integrity_segments);
  476. t->max_segment_size = min_not_zero(t->max_segment_size,
  477. b->max_segment_size);
  478. t->misaligned |= b->misaligned;
  479. alignment = queue_limit_alignment_offset(b, start);
  480. /* Bottom device has different alignment. Check that it is
  481. * compatible with the current top alignment.
  482. */
  483. if (t->alignment_offset != alignment) {
  484. top = max(t->physical_block_size, t->io_min)
  485. + t->alignment_offset;
  486. bottom = max(b->physical_block_size, b->io_min) + alignment;
  487. /* Verify that top and bottom intervals line up */
  488. if (max(top, bottom) & (min(top, bottom) - 1)) {
  489. t->misaligned = 1;
  490. ret = -1;
  491. }
  492. }
  493. t->logical_block_size = max(t->logical_block_size,
  494. b->logical_block_size);
  495. t->physical_block_size = max(t->physical_block_size,
  496. b->physical_block_size);
  497. t->io_min = max(t->io_min, b->io_min);
  498. t->io_opt = lcm(t->io_opt, b->io_opt);
  499. t->cluster &= b->cluster;
  500. t->discard_zeroes_data &= b->discard_zeroes_data;
  501. /* Physical block size a multiple of the logical block size? */
  502. if (t->physical_block_size & (t->logical_block_size - 1)) {
  503. t->physical_block_size = t->logical_block_size;
  504. t->misaligned = 1;
  505. ret = -1;
  506. }
  507. /* Minimum I/O a multiple of the physical block size? */
  508. if (t->io_min & (t->physical_block_size - 1)) {
  509. t->io_min = t->physical_block_size;
  510. t->misaligned = 1;
  511. ret = -1;
  512. }
  513. /* Optimal I/O a multiple of the physical block size? */
  514. if (t->io_opt & (t->physical_block_size - 1)) {
  515. t->io_opt = 0;
  516. t->misaligned = 1;
  517. ret = -1;
  518. }
  519. /* Find lowest common alignment_offset */
  520. t->alignment_offset = lcm(t->alignment_offset, alignment)
  521. & (max(t->physical_block_size, t->io_min) - 1);
  522. /* Verify that new alignment_offset is on a logical block boundary */
  523. if (t->alignment_offset & (t->logical_block_size - 1)) {
  524. t->misaligned = 1;
  525. ret = -1;
  526. }
  527. /* Discard alignment and granularity */
  528. if (b->discard_granularity) {
  529. alignment = queue_limit_discard_alignment(b, start);
  530. if (t->discard_granularity != 0 &&
  531. t->discard_alignment != alignment) {
  532. top = t->discard_granularity + t->discard_alignment;
  533. bottom = b->discard_granularity + alignment;
  534. /* Verify that top and bottom intervals line up */
  535. if (max(top, bottom) & (min(top, bottom) - 1))
  536. t->discard_misaligned = 1;
  537. }
  538. t->max_discard_sectors = min_not_zero(t->max_discard_sectors,
  539. b->max_discard_sectors);
  540. t->discard_granularity = max(t->discard_granularity,
  541. b->discard_granularity);
  542. t->discard_alignment = lcm(t->discard_alignment, alignment) &
  543. (t->discard_granularity - 1);
  544. }
  545. return ret;
  546. }
  547. EXPORT_SYMBOL(blk_stack_limits);
  548. /**
  549. * bdev_stack_limits - adjust queue limits for stacked drivers
  550. * @t: the stacking driver limits (top device)
  551. * @bdev: the component block_device (bottom)
  552. * @start: first data sector within component device
  553. *
  554. * Description:
  555. * Merges queue limits for a top device and a block_device. Returns
  556. * 0 if alignment didn't change. Returns -1 if adding the bottom
  557. * device caused misalignment.
  558. */
  559. int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
  560. sector_t start)
  561. {
  562. struct request_queue *bq = bdev_get_queue(bdev);
  563. start += get_start_sect(bdev);
  564. return blk_stack_limits(t, &bq->limits, start);
  565. }
  566. EXPORT_SYMBOL(bdev_stack_limits);
  567. /**
  568. * disk_stack_limits - adjust queue limits for stacked drivers
  569. * @disk: MD/DM gendisk (top)
  570. * @bdev: the underlying block device (bottom)
  571. * @offset: offset to beginning of data within component device
  572. *
  573. * Description:
  574. * Merges the limits for a top level gendisk and a bottom level
  575. * block_device.
  576. */
  577. void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
  578. sector_t offset)
  579. {
  580. struct request_queue *t = disk->queue;
  581. if (bdev_stack_limits(&t->limits, bdev, offset >> 9) < 0) {
  582. char top[BDEVNAME_SIZE], bottom[BDEVNAME_SIZE];
  583. disk_name(disk, 0, top);
  584. bdevname(bdev, bottom);
  585. printk(KERN_NOTICE "%s: Warning: Device %s is misaligned\n",
  586. top, bottom);
  587. }
  588. }
  589. EXPORT_SYMBOL(disk_stack_limits);
  590. /**
  591. * blk_queue_dma_pad - set pad mask
  592. * @q: the request queue for the device
  593. * @mask: pad mask
  594. *
  595. * Set dma pad mask.
  596. *
  597. * Appending pad buffer to a request modifies the last entry of a
  598. * scatter list such that it includes the pad buffer.
  599. **/
  600. void blk_queue_dma_pad(struct request_queue *q, unsigned int mask)
  601. {
  602. q->dma_pad_mask = mask;
  603. }
  604. EXPORT_SYMBOL(blk_queue_dma_pad);
  605. /**
  606. * blk_queue_update_dma_pad - update pad mask
  607. * @q: the request queue for the device
  608. * @mask: pad mask
  609. *
  610. * Update dma pad mask.
  611. *
  612. * Appending pad buffer to a request modifies the last entry of a
  613. * scatter list such that it includes the pad buffer.
  614. **/
  615. void blk_queue_update_dma_pad(struct request_queue *q, unsigned int mask)
  616. {
  617. if (mask > q->dma_pad_mask)
  618. q->dma_pad_mask = mask;
  619. }
  620. EXPORT_SYMBOL(blk_queue_update_dma_pad);
  621. /**
  622. * blk_queue_dma_drain - Set up a drain buffer for excess dma.
  623. * @q: the request queue for the device
  624. * @dma_drain_needed: fn which returns non-zero if drain is necessary
  625. * @buf: physically contiguous buffer
  626. * @size: size of the buffer in bytes
  627. *
  628. * Some devices have excess DMA problems and can't simply discard (or
  629. * zero fill) the unwanted piece of the transfer. They have to have a
  630. * real area of memory to transfer it into. The use case for this is
  631. * ATAPI devices in DMA mode. If the packet command causes a transfer
  632. * bigger than the transfer size some HBAs will lock up if there
  633. * aren't DMA elements to contain the excess transfer. What this API
  634. * does is adjust the queue so that the buf is always appended
  635. * silently to the scatterlist.
  636. *
  637. * Note: This routine adjusts max_hw_segments to make room for appending
  638. * the drain buffer. If you call blk_queue_max_segments() after calling
  639. * this routine, you must set the limit to one fewer than your device
  640. * can support otherwise there won't be room for the drain buffer.
  641. */
  642. int blk_queue_dma_drain(struct request_queue *q,
  643. dma_drain_needed_fn *dma_drain_needed,
  644. void *buf, unsigned int size)
  645. {
  646. if (queue_max_segments(q) < 2)
  647. return -EINVAL;
  648. /* make room for appending the drain */
  649. blk_queue_max_segments(q, queue_max_segments(q) - 1);
  650. q->dma_drain_needed = dma_drain_needed;
  651. q->dma_drain_buffer = buf;
  652. q->dma_drain_size = size;
  653. return 0;
  654. }
  655. EXPORT_SYMBOL_GPL(blk_queue_dma_drain);
  656. /**
  657. * blk_queue_segment_boundary - set boundary rules for segment merging
  658. * @q: the request queue for the device
  659. * @mask: the memory boundary mask
  660. **/
  661. void blk_queue_segment_boundary(struct request_queue *q, unsigned long mask)
  662. {
  663. if (mask < PAGE_CACHE_SIZE - 1) {
  664. mask = PAGE_CACHE_SIZE - 1;
  665. printk(KERN_INFO "%s: set to minimum %lx\n",
  666. __func__, mask);
  667. }
  668. q->limits.seg_boundary_mask = mask;
  669. }
  670. EXPORT_SYMBOL(blk_queue_segment_boundary);
  671. /**
  672. * blk_queue_dma_alignment - set dma length and memory alignment
  673. * @q: the request queue for the device
  674. * @mask: alignment mask
  675. *
  676. * description:
  677. * set required memory and length alignment for direct dma transactions.
  678. * this is used when building direct io requests for the queue.
  679. *
  680. **/
  681. void blk_queue_dma_alignment(struct request_queue *q, int mask)
  682. {
  683. q->dma_alignment = mask;
  684. }
  685. EXPORT_SYMBOL(blk_queue_dma_alignment);
  686. /**
  687. * blk_queue_update_dma_alignment - update dma length and memory alignment
  688. * @q: the request queue for the device
  689. * @mask: alignment mask
  690. *
  691. * description:
  692. * update required memory and length alignment for direct dma transactions.
  693. * If the requested alignment is larger than the current alignment, then
  694. * the current queue alignment is updated to the new value, otherwise it
  695. * is left alone. The design of this is to allow multiple objects
  696. * (driver, device, transport etc) to set their respective
  697. * alignments without having them interfere.
  698. *
  699. **/
  700. void blk_queue_update_dma_alignment(struct request_queue *q, int mask)
  701. {
  702. BUG_ON(mask > PAGE_SIZE);
  703. if (mask > q->dma_alignment)
  704. q->dma_alignment = mask;
  705. }
  706. EXPORT_SYMBOL(blk_queue_update_dma_alignment);
  707. /**
  708. * blk_queue_flush - configure queue's cache flush capability
  709. * @q: the request queue for the device
  710. * @flush: 0, REQ_FLUSH or REQ_FLUSH | REQ_FUA
  711. *
  712. * Tell block layer cache flush capability of @q. If it supports
  713. * flushing, REQ_FLUSH should be set. If it supports bypassing
  714. * write cache for individual writes, REQ_FUA should be set.
  715. */
  716. void blk_queue_flush(struct request_queue *q, unsigned int flush)
  717. {
  718. WARN_ON_ONCE(flush & ~(REQ_FLUSH | REQ_FUA));
  719. if (WARN_ON_ONCE(!(flush & REQ_FLUSH) && (flush & REQ_FUA)))
  720. flush &= ~REQ_FUA;
  721. q->flush_flags = flush & (REQ_FLUSH | REQ_FUA);
  722. }
  723. EXPORT_SYMBOL_GPL(blk_queue_flush);
  724. void blk_queue_flush_queueable(struct request_queue *q, bool queueable)
  725. {
  726. q->flush_not_queueable = !queueable;
  727. }
  728. EXPORT_SYMBOL_GPL(blk_queue_flush_queueable);
  729. static int __init blk_settings_init(void)
  730. {
  731. blk_max_low_pfn = max_low_pfn - 1;
  732. blk_max_pfn = max_pfn - 1;
  733. return 0;
  734. }
  735. subsys_initcall(blk_settings_init);