blkdev.h 26 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886
  1. #ifndef _LINUX_BLKDEV_H
  2. #define _LINUX_BLKDEV_H
  3. #include <linux/sched.h>
  4. #include <linux/major.h>
  5. #include <linux/genhd.h>
  6. #include <linux/list.h>
  7. #include <linux/timer.h>
  8. #include <linux/workqueue.h>
  9. #include <linux/pagemap.h>
  10. #include <linux/backing-dev.h>
  11. #include <linux/wait.h>
  12. #include <linux/mempool.h>
  13. #include <linux/bio.h>
  14. #include <linux/module.h>
  15. #include <linux/stringify.h>
  16. #include <asm/scatterlist.h>
  17. #ifdef CONFIG_LBD
  18. # include <asm/div64.h>
  19. # define sector_div(a, b) do_div(a, b)
  20. #else
  21. # define sector_div(n, b)( \
  22. { \
  23. int _res; \
  24. _res = (n) % (b); \
  25. (n) /= (b); \
  26. _res; \
  27. } \
  28. )
  29. #endif
  30. #ifdef CONFIG_BLOCK
  31. struct scsi_ioctl_command;
  32. struct request_queue;
  33. typedef struct request_queue request_queue_t;
  34. struct elevator_queue;
  35. typedef struct elevator_queue elevator_t;
  36. struct request_pm_state;
  37. struct blk_trace;
  38. #define BLKDEV_MIN_RQ 4
  39. #define BLKDEV_MAX_RQ 128 /* Default maximum */
  40. /*
  41. * This is the per-process anticipatory I/O scheduler state.
  42. */
  43. struct as_io_context {
  44. spinlock_t lock;
  45. void (*dtor)(struct as_io_context *aic); /* destructor */
  46. void (*exit)(struct as_io_context *aic); /* called on task exit */
  47. unsigned long state;
  48. atomic_t nr_queued; /* queued reads & sync writes */
  49. atomic_t nr_dispatched; /* number of requests gone to the drivers */
  50. /* IO History tracking */
  51. /* Thinktime */
  52. unsigned long last_end_request;
  53. unsigned long ttime_total;
  54. unsigned long ttime_samples;
  55. unsigned long ttime_mean;
  56. /* Layout pattern */
  57. unsigned int seek_samples;
  58. sector_t last_request_pos;
  59. u64 seek_total;
  60. sector_t seek_mean;
  61. };
  62. struct cfq_queue;
  63. struct cfq_io_context {
  64. struct rb_node rb_node;
  65. void *key;
  66. struct cfq_queue *cfqq[2];
  67. struct io_context *ioc;
  68. unsigned long last_end_request;
  69. sector_t last_request_pos;
  70. unsigned long ttime_total;
  71. unsigned long ttime_samples;
  72. unsigned long ttime_mean;
  73. unsigned int seek_samples;
  74. u64 seek_total;
  75. sector_t seek_mean;
  76. struct list_head queue_list;
  77. void (*dtor)(struct io_context *); /* destructor */
  78. void (*exit)(struct io_context *); /* called on task exit */
  79. };
  80. /*
  81. * This is the per-process I/O subsystem state. It is refcounted and
  82. * kmalloc'ed. Currently all fields are modified in process io context
  83. * (apart from the atomic refcount), so require no locking.
  84. */
  85. struct io_context {
  86. atomic_t refcount;
  87. struct task_struct *task;
  88. unsigned int ioprio_changed;
  89. /*
  90. * For request batching
  91. */
  92. unsigned long last_waited; /* Time last woken after wait for request */
  93. int nr_batch_requests; /* Number of requests left in the batch */
  94. struct as_io_context *aic;
  95. struct rb_root cic_root;
  96. };
  97. void put_io_context(struct io_context *ioc);
  98. void exit_io_context(void);
  99. struct io_context *get_io_context(gfp_t gfp_flags, int node);
  100. void copy_io_context(struct io_context **pdst, struct io_context **psrc);
  101. void swap_io_context(struct io_context **ioc1, struct io_context **ioc2);
  102. struct request;
  103. typedef void (rq_end_io_fn)(struct request *, int);
  104. struct request_list {
  105. int count[2];
  106. int starved[2];
  107. int elvpriv;
  108. mempool_t *rq_pool;
  109. wait_queue_head_t wait[2];
  110. };
  111. /*
  112. * request command types
  113. */
  114. enum rq_cmd_type_bits {
  115. REQ_TYPE_FS = 1, /* fs request */
  116. REQ_TYPE_BLOCK_PC, /* scsi command */
  117. REQ_TYPE_SENSE, /* sense request */
  118. REQ_TYPE_PM_SUSPEND, /* suspend request */
  119. REQ_TYPE_PM_RESUME, /* resume request */
  120. REQ_TYPE_PM_SHUTDOWN, /* shutdown request */
  121. REQ_TYPE_FLUSH, /* flush request */
  122. REQ_TYPE_SPECIAL, /* driver defined type */
  123. REQ_TYPE_LINUX_BLOCK, /* generic block layer message */
  124. /*
  125. * for ATA/ATAPI devices. this really doesn't belong here, ide should
  126. * use REQ_TYPE_SPECIAL and use rq->cmd[0] with the range of driver
  127. * private REQ_LB opcodes to differentiate what type of request this is
  128. */
  129. REQ_TYPE_ATA_CMD,
  130. REQ_TYPE_ATA_TASK,
  131. REQ_TYPE_ATA_TASKFILE,
  132. REQ_TYPE_ATA_PC,
  133. };
  134. /*
  135. * For request of type REQ_TYPE_LINUX_BLOCK, rq->cmd[0] is the opcode being
  136. * sent down (similar to how REQ_TYPE_BLOCK_PC means that ->cmd[] holds a
  137. * SCSI cdb.
  138. *
  139. * 0x00 -> 0x3f are driver private, to be used for whatever purpose they need,
  140. * typically to differentiate REQ_TYPE_SPECIAL requests.
  141. *
  142. */
  143. enum {
  144. /*
  145. * just examples for now
  146. */
  147. REQ_LB_OP_EJECT = 0x40, /* eject request */
  148. REQ_LB_OP_FLUSH = 0x41, /* flush device */
  149. };
  150. /*
  151. * request type modified bits. first three bits match BIO_RW* bits, important
  152. */
  153. enum rq_flag_bits {
  154. __REQ_RW, /* not set, read. set, write */
  155. __REQ_FAILFAST, /* no low level driver retries */
  156. __REQ_SORTED, /* elevator knows about this request */
  157. __REQ_SOFTBARRIER, /* may not be passed by ioscheduler */
  158. __REQ_HARDBARRIER, /* may not be passed by drive either */
  159. __REQ_FUA, /* forced unit access */
  160. __REQ_NOMERGE, /* don't touch this for merging */
  161. __REQ_STARTED, /* drive already may have started this one */
  162. __REQ_DONTPREP, /* don't call prep for this one */
  163. __REQ_QUEUED, /* uses queueing */
  164. __REQ_ELVPRIV, /* elevator private data attached */
  165. __REQ_FAILED, /* set if the request failed */
  166. __REQ_QUIET, /* don't worry about errors */
  167. __REQ_PREEMPT, /* set for "ide_preempt" requests */
  168. __REQ_ORDERED_COLOR, /* is before or after barrier */
  169. __REQ_RW_SYNC, /* request is sync (O_DIRECT) */
  170. __REQ_ALLOCED, /* request came from our alloc pool */
  171. __REQ_RW_META, /* metadata io request */
  172. __REQ_NR_BITS, /* stops here */
  173. };
  174. #define REQ_RW (1 << __REQ_RW)
  175. #define REQ_FAILFAST (1 << __REQ_FAILFAST)
  176. #define REQ_SORTED (1 << __REQ_SORTED)
  177. #define REQ_SOFTBARRIER (1 << __REQ_SOFTBARRIER)
  178. #define REQ_HARDBARRIER (1 << __REQ_HARDBARRIER)
  179. #define REQ_FUA (1 << __REQ_FUA)
  180. #define REQ_NOMERGE (1 << __REQ_NOMERGE)
  181. #define REQ_STARTED (1 << __REQ_STARTED)
  182. #define REQ_DONTPREP (1 << __REQ_DONTPREP)
  183. #define REQ_QUEUED (1 << __REQ_QUEUED)
  184. #define REQ_ELVPRIV (1 << __REQ_ELVPRIV)
  185. #define REQ_FAILED (1 << __REQ_FAILED)
  186. #define REQ_QUIET (1 << __REQ_QUIET)
  187. #define REQ_PREEMPT (1 << __REQ_PREEMPT)
  188. #define REQ_ORDERED_COLOR (1 << __REQ_ORDERED_COLOR)
  189. #define REQ_RW_SYNC (1 << __REQ_RW_SYNC)
  190. #define REQ_ALLOCED (1 << __REQ_ALLOCED)
  191. #define REQ_RW_META (1 << __REQ_RW_META)
  192. #define BLK_MAX_CDB 16
  193. /*
  194. * try to put the fields that are referenced together in the same cacheline
  195. */
  196. struct request {
  197. struct list_head queuelist;
  198. struct list_head donelist;
  199. request_queue_t *q;
  200. unsigned int cmd_flags;
  201. enum rq_cmd_type_bits cmd_type;
  202. /* Maintain bio traversal state for part by part I/O submission.
  203. * hard_* are block layer internals, no driver should touch them!
  204. */
  205. sector_t sector; /* next sector to submit */
  206. sector_t hard_sector; /* next sector to complete */
  207. unsigned long nr_sectors; /* no. of sectors left to submit */
  208. unsigned long hard_nr_sectors; /* no. of sectors left to complete */
  209. /* no. of sectors left to submit in the current segment */
  210. unsigned int current_nr_sectors;
  211. /* no. of sectors left to complete in the current segment */
  212. unsigned int hard_cur_sectors;
  213. struct bio *bio;
  214. struct bio *biotail;
  215. struct hlist_node hash; /* merge hash */
  216. /*
  217. * The rb_node is only used inside the io scheduler, requests
  218. * are pruned when moved to the dispatch queue. So let the
  219. * completion_data share space with the rb_node.
  220. */
  221. union {
  222. struct rb_node rb_node; /* sort/lookup */
  223. void *completion_data;
  224. };
  225. /*
  226. * two pointers are available for the IO schedulers, if they need
  227. * more they have to dynamically allocate it.
  228. */
  229. void *elevator_private;
  230. void *elevator_private2;
  231. struct gendisk *rq_disk;
  232. unsigned long start_time;
  233. /* Number of scatter-gather DMA addr+len pairs after
  234. * physical address coalescing is performed.
  235. */
  236. unsigned short nr_phys_segments;
  237. /* Number of scatter-gather addr+len pairs after
  238. * physical and DMA remapping hardware coalescing is performed.
  239. * This is the number of scatter-gather entries the driver
  240. * will actually have to deal with after DMA mapping is done.
  241. */
  242. unsigned short nr_hw_segments;
  243. unsigned short ioprio;
  244. void *special;
  245. char *buffer;
  246. int tag;
  247. int errors;
  248. int ref_count;
  249. /*
  250. * when request is used as a packet command carrier
  251. */
  252. unsigned int cmd_len;
  253. unsigned char cmd[BLK_MAX_CDB];
  254. unsigned int data_len;
  255. unsigned int sense_len;
  256. void *data;
  257. void *sense;
  258. unsigned int timeout;
  259. int retries;
  260. /*
  261. * completion callback.
  262. */
  263. rq_end_io_fn *end_io;
  264. void *end_io_data;
  265. };
  266. /*
  267. * State information carried for REQ_TYPE_PM_SUSPEND and REQ_TYPE_PM_RESUME
  268. * requests. Some step values could eventually be made generic.
  269. */
  270. struct request_pm_state
  271. {
  272. /* PM state machine step value, currently driver specific */
  273. int pm_step;
  274. /* requested PM state value (S1, S2, S3, S4, ...) */
  275. u32 pm_state;
  276. void* data; /* for driver use */
  277. };
  278. #include <linux/elevator.h>
  279. typedef void (request_fn_proc) (request_queue_t *q);
  280. typedef int (make_request_fn) (request_queue_t *q, struct bio *bio);
  281. typedef int (prep_rq_fn) (request_queue_t *, struct request *);
  282. typedef void (unplug_fn) (request_queue_t *);
  283. struct bio_vec;
  284. typedef int (merge_bvec_fn) (request_queue_t *, struct bio *, struct bio_vec *);
  285. typedef int (issue_flush_fn) (request_queue_t *, struct gendisk *, sector_t *);
  286. typedef void (prepare_flush_fn) (request_queue_t *, struct request *);
  287. typedef void (softirq_done_fn)(struct request *);
  288. enum blk_queue_state {
  289. Queue_down,
  290. Queue_up,
  291. };
  292. struct blk_queue_tag {
  293. struct request **tag_index; /* map of busy tags */
  294. unsigned long *tag_map; /* bit map of free/busy tags */
  295. struct list_head busy_list; /* fifo list of busy tags */
  296. int busy; /* current depth */
  297. int max_depth; /* what we will send to device */
  298. int real_max_depth; /* what the array can hold */
  299. atomic_t refcnt; /* map can be shared */
  300. };
  301. struct request_queue
  302. {
  303. /*
  304. * Together with queue_head for cacheline sharing
  305. */
  306. struct list_head queue_head;
  307. struct request *last_merge;
  308. elevator_t *elevator;
  309. /*
  310. * the queue request freelist, one for reads and one for writes
  311. */
  312. struct request_list rq;
  313. request_fn_proc *request_fn;
  314. make_request_fn *make_request_fn;
  315. prep_rq_fn *prep_rq_fn;
  316. unplug_fn *unplug_fn;
  317. merge_bvec_fn *merge_bvec_fn;
  318. issue_flush_fn *issue_flush_fn;
  319. prepare_flush_fn *prepare_flush_fn;
  320. softirq_done_fn *softirq_done_fn;
  321. /*
  322. * Dispatch queue sorting
  323. */
  324. sector_t end_sector;
  325. struct request *boundary_rq;
  326. /*
  327. * Auto-unplugging state
  328. */
  329. struct timer_list unplug_timer;
  330. int unplug_thresh; /* After this many requests */
  331. unsigned long unplug_delay; /* After this many jiffies */
  332. struct work_struct unplug_work;
  333. struct backing_dev_info backing_dev_info;
  334. /*
  335. * The queue owner gets to use this for whatever they like.
  336. * ll_rw_blk doesn't touch it.
  337. */
  338. void *queuedata;
  339. /*
  340. * queue needs bounce pages for pages above this limit
  341. */
  342. unsigned long bounce_pfn;
  343. gfp_t bounce_gfp;
  344. /*
  345. * various queue flags, see QUEUE_* below
  346. */
  347. unsigned long queue_flags;
  348. /*
  349. * protects queue structures from reentrancy. ->__queue_lock should
  350. * _never_ be used directly, it is queue private. always use
  351. * ->queue_lock.
  352. */
  353. spinlock_t __queue_lock;
  354. spinlock_t *queue_lock;
  355. /*
  356. * queue kobject
  357. */
  358. struct kobject kobj;
  359. /*
  360. * queue settings
  361. */
  362. unsigned long nr_requests; /* Max # of requests */
  363. unsigned int nr_congestion_on;
  364. unsigned int nr_congestion_off;
  365. unsigned int nr_batching;
  366. unsigned int max_sectors;
  367. unsigned int max_hw_sectors;
  368. unsigned short max_phys_segments;
  369. unsigned short max_hw_segments;
  370. unsigned short hardsect_size;
  371. unsigned int max_segment_size;
  372. unsigned long seg_boundary_mask;
  373. unsigned int dma_alignment;
  374. struct blk_queue_tag *queue_tags;
  375. unsigned int nr_sorted;
  376. unsigned int in_flight;
  377. /*
  378. * sg stuff
  379. */
  380. unsigned int sg_timeout;
  381. unsigned int sg_reserved_size;
  382. int node;
  383. #ifdef CONFIG_BLK_DEV_IO_TRACE
  384. struct blk_trace *blk_trace;
  385. #endif
  386. /*
  387. * reserved for flush operations
  388. */
  389. unsigned int ordered, next_ordered, ordseq;
  390. int orderr, ordcolor;
  391. struct request pre_flush_rq, bar_rq, post_flush_rq;
  392. struct request *orig_bar_rq;
  393. unsigned int bi_size;
  394. struct mutex sysfs_lock;
  395. };
  396. #define QUEUE_FLAG_CLUSTER 0 /* cluster several segments into 1 */
  397. #define QUEUE_FLAG_QUEUED 1 /* uses generic tag queueing */
  398. #define QUEUE_FLAG_STOPPED 2 /* queue is stopped */
  399. #define QUEUE_FLAG_READFULL 3 /* write queue has been filled */
  400. #define QUEUE_FLAG_WRITEFULL 4 /* read queue has been filled */
  401. #define QUEUE_FLAG_DEAD 5 /* queue being torn down */
  402. #define QUEUE_FLAG_REENTER 6 /* Re-entrancy avoidance */
  403. #define QUEUE_FLAG_PLUGGED 7 /* queue is plugged */
  404. #define QUEUE_FLAG_ELVSWITCH 8 /* don't use elevator, just do FIFO */
  405. enum {
  406. /*
  407. * Hardbarrier is supported with one of the following methods.
  408. *
  409. * NONE : hardbarrier unsupported
  410. * DRAIN : ordering by draining is enough
  411. * DRAIN_FLUSH : ordering by draining w/ pre and post flushes
  412. * DRAIN_FUA : ordering by draining w/ pre flush and FUA write
  413. * TAG : ordering by tag is enough
  414. * TAG_FLUSH : ordering by tag w/ pre and post flushes
  415. * TAG_FUA : ordering by tag w/ pre flush and FUA write
  416. */
  417. QUEUE_ORDERED_NONE = 0x00,
  418. QUEUE_ORDERED_DRAIN = 0x01,
  419. QUEUE_ORDERED_TAG = 0x02,
  420. QUEUE_ORDERED_PREFLUSH = 0x10,
  421. QUEUE_ORDERED_POSTFLUSH = 0x20,
  422. QUEUE_ORDERED_FUA = 0x40,
  423. QUEUE_ORDERED_DRAIN_FLUSH = QUEUE_ORDERED_DRAIN |
  424. QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_POSTFLUSH,
  425. QUEUE_ORDERED_DRAIN_FUA = QUEUE_ORDERED_DRAIN |
  426. QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_FUA,
  427. QUEUE_ORDERED_TAG_FLUSH = QUEUE_ORDERED_TAG |
  428. QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_POSTFLUSH,
  429. QUEUE_ORDERED_TAG_FUA = QUEUE_ORDERED_TAG |
  430. QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_FUA,
  431. /*
  432. * Ordered operation sequence
  433. */
  434. QUEUE_ORDSEQ_STARTED = 0x01, /* flushing in progress */
  435. QUEUE_ORDSEQ_DRAIN = 0x02, /* waiting for the queue to be drained */
  436. QUEUE_ORDSEQ_PREFLUSH = 0x04, /* pre-flushing in progress */
  437. QUEUE_ORDSEQ_BAR = 0x08, /* original barrier req in progress */
  438. QUEUE_ORDSEQ_POSTFLUSH = 0x10, /* post-flushing in progress */
  439. QUEUE_ORDSEQ_DONE = 0x20,
  440. };
  441. #define blk_queue_plugged(q) test_bit(QUEUE_FLAG_PLUGGED, &(q)->queue_flags)
  442. #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
  443. #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
  444. #define blk_queue_flushing(q) ((q)->ordseq)
  445. #define blk_fs_request(rq) ((rq)->cmd_type == REQ_TYPE_FS)
  446. #define blk_pc_request(rq) ((rq)->cmd_type == REQ_TYPE_BLOCK_PC)
  447. #define blk_special_request(rq) ((rq)->cmd_type == REQ_TYPE_SPECIAL)
  448. #define blk_sense_request(rq) ((rq)->cmd_type == REQ_TYPE_SENSE)
  449. #define blk_noretry_request(rq) ((rq)->cmd_flags & REQ_FAILFAST)
  450. #define blk_rq_started(rq) ((rq)->cmd_flags & REQ_STARTED)
  451. #define blk_account_rq(rq) (blk_rq_started(rq) && blk_fs_request(rq))
  452. #define blk_pm_suspend_request(rq) ((rq)->cmd_type == REQ_TYPE_PM_SUSPEND)
  453. #define blk_pm_resume_request(rq) ((rq)->cmd_type == REQ_TYPE_PM_RESUME)
  454. #define blk_pm_request(rq) \
  455. (blk_pm_suspend_request(rq) || blk_pm_resume_request(rq))
  456. #define blk_sorted_rq(rq) ((rq)->cmd_flags & REQ_SORTED)
  457. #define blk_barrier_rq(rq) ((rq)->cmd_flags & REQ_HARDBARRIER)
  458. #define blk_fua_rq(rq) ((rq)->cmd_flags & REQ_FUA)
  459. #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
  460. #define rq_data_dir(rq) ((rq)->cmd_flags & 1)
  461. /*
  462. * We regard a request as sync, if it's a READ or a SYNC write.
  463. */
  464. #define rq_is_sync(rq) (rq_data_dir((rq)) == READ || (rq)->cmd_flags & REQ_RW_SYNC)
  465. #define rq_is_meta(rq) ((rq)->cmd_flags & REQ_RW_META)
  466. static inline int blk_queue_full(struct request_queue *q, int rw)
  467. {
  468. if (rw == READ)
  469. return test_bit(QUEUE_FLAG_READFULL, &q->queue_flags);
  470. return test_bit(QUEUE_FLAG_WRITEFULL, &q->queue_flags);
  471. }
  472. static inline void blk_set_queue_full(struct request_queue *q, int rw)
  473. {
  474. if (rw == READ)
  475. set_bit(QUEUE_FLAG_READFULL, &q->queue_flags);
  476. else
  477. set_bit(QUEUE_FLAG_WRITEFULL, &q->queue_flags);
  478. }
  479. static inline void blk_clear_queue_full(struct request_queue *q, int rw)
  480. {
  481. if (rw == READ)
  482. clear_bit(QUEUE_FLAG_READFULL, &q->queue_flags);
  483. else
  484. clear_bit(QUEUE_FLAG_WRITEFULL, &q->queue_flags);
  485. }
  486. /*
  487. * mergeable request must not have _NOMERGE or _BARRIER bit set, nor may
  488. * it already be started by driver.
  489. */
  490. #define RQ_NOMERGE_FLAGS \
  491. (REQ_NOMERGE | REQ_STARTED | REQ_HARDBARRIER | REQ_SOFTBARRIER)
  492. #define rq_mergeable(rq) \
  493. (!((rq)->cmd_flags & RQ_NOMERGE_FLAGS) && blk_fs_request((rq)))
  494. /*
  495. * q->prep_rq_fn return values
  496. */
  497. #define BLKPREP_OK 0 /* serve it */
  498. #define BLKPREP_KILL 1 /* fatal error, kill */
  499. #define BLKPREP_DEFER 2 /* leave on queue */
  500. extern unsigned long blk_max_low_pfn, blk_max_pfn;
  501. /*
  502. * standard bounce addresses:
  503. *
  504. * BLK_BOUNCE_HIGH : bounce all highmem pages
  505. * BLK_BOUNCE_ANY : don't bounce anything
  506. * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
  507. */
  508. #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
  509. #define BLK_BOUNCE_ANY ((u64)blk_max_pfn << PAGE_SHIFT)
  510. #define BLK_BOUNCE_ISA (ISA_DMA_THRESHOLD)
  511. #ifdef CONFIG_MMU
  512. extern int init_emergency_isa_pool(void);
  513. extern void blk_queue_bounce(request_queue_t *q, struct bio **bio);
  514. #else
  515. static inline int init_emergency_isa_pool(void)
  516. {
  517. return 0;
  518. }
  519. static inline void blk_queue_bounce(request_queue_t *q, struct bio **bio)
  520. {
  521. }
  522. #endif /* CONFIG_MMU */
  523. #define rq_for_each_bio(_bio, rq) \
  524. if ((rq->bio)) \
  525. for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
  526. extern int blk_register_queue(struct gendisk *disk);
  527. extern void blk_unregister_queue(struct gendisk *disk);
  528. extern void register_disk(struct gendisk *dev);
  529. extern void generic_make_request(struct bio *bio);
  530. extern void blk_put_request(struct request *);
  531. extern void __blk_put_request(request_queue_t *, struct request *);
  532. extern void blk_end_sync_rq(struct request *rq, int error);
  533. extern struct request *blk_get_request(request_queue_t *, int, gfp_t);
  534. extern void blk_insert_request(request_queue_t *, struct request *, int, void *);
  535. extern void blk_requeue_request(request_queue_t *, struct request *);
  536. extern void blk_plug_device(request_queue_t *);
  537. extern int blk_remove_plug(request_queue_t *);
  538. extern void blk_recount_segments(request_queue_t *, struct bio *);
  539. extern int scsi_cmd_ioctl(struct file *, struct gendisk *, unsigned int, void __user *);
  540. extern int sg_scsi_ioctl(struct file *, struct request_queue *,
  541. struct gendisk *, struct scsi_ioctl_command __user *);
  542. /*
  543. * Temporary export, until SCSI gets fixed up.
  544. */
  545. extern int ll_back_merge_fn(request_queue_t *, struct request *, struct bio *);
  546. /*
  547. * A queue has just exitted congestion. Note this in the global counter of
  548. * congested queues, and wake up anyone who was waiting for requests to be
  549. * put back.
  550. */
  551. static inline void blk_clear_queue_congested(request_queue_t *q, int rw)
  552. {
  553. clear_bdi_congested(&q->backing_dev_info, rw);
  554. }
  555. /*
  556. * A queue has just entered congestion. Flag that in the queue's VM-visible
  557. * state flags and increment the global gounter of congested queues.
  558. */
  559. static inline void blk_set_queue_congested(request_queue_t *q, int rw)
  560. {
  561. set_bdi_congested(&q->backing_dev_info, rw);
  562. }
  563. extern void blk_start_queue(request_queue_t *q);
  564. extern void blk_stop_queue(request_queue_t *q);
  565. extern void blk_sync_queue(struct request_queue *q);
  566. extern void __blk_stop_queue(request_queue_t *q);
  567. extern void blk_run_queue(request_queue_t *);
  568. extern void blk_start_queueing(request_queue_t *);
  569. extern int blk_rq_map_user(request_queue_t *, struct request *, void __user *, unsigned long);
  570. extern int blk_rq_unmap_user(struct bio *);
  571. extern int blk_rq_map_kern(request_queue_t *, struct request *, void *, unsigned int, gfp_t);
  572. extern int blk_rq_map_user_iov(request_queue_t *, struct request *,
  573. struct sg_iovec *, int, unsigned int);
  574. extern int blk_execute_rq(request_queue_t *, struct gendisk *,
  575. struct request *, int);
  576. extern void blk_execute_rq_nowait(request_queue_t *, struct gendisk *,
  577. struct request *, int, rq_end_io_fn *);
  578. static inline request_queue_t *bdev_get_queue(struct block_device *bdev)
  579. {
  580. return bdev->bd_disk->queue;
  581. }
  582. static inline void blk_run_backing_dev(struct backing_dev_info *bdi,
  583. struct page *page)
  584. {
  585. if (bdi && bdi->unplug_io_fn)
  586. bdi->unplug_io_fn(bdi, page);
  587. }
  588. static inline void blk_run_address_space(struct address_space *mapping)
  589. {
  590. if (mapping)
  591. blk_run_backing_dev(mapping->backing_dev_info, NULL);
  592. }
  593. /*
  594. * end_request() and friends. Must be called with the request queue spinlock
  595. * acquired. All functions called within end_request() _must_be_ atomic.
  596. *
  597. * Several drivers define their own end_request and call
  598. * end_that_request_first() and end_that_request_last()
  599. * for parts of the original function. This prevents
  600. * code duplication in drivers.
  601. */
  602. extern int end_that_request_first(struct request *, int, int);
  603. extern int end_that_request_chunk(struct request *, int, int);
  604. extern void end_that_request_last(struct request *, int);
  605. extern void end_request(struct request *req, int uptodate);
  606. extern void blk_complete_request(struct request *);
  607. /*
  608. * end_that_request_first/chunk() takes an uptodate argument. we account
  609. * any value <= as an io error. 0 means -EIO for compatability reasons,
  610. * any other < 0 value is the direct error type. An uptodate value of
  611. * 1 indicates successful io completion
  612. */
  613. #define end_io_error(uptodate) (unlikely((uptodate) <= 0))
  614. static inline void blkdev_dequeue_request(struct request *req)
  615. {
  616. elv_dequeue_request(req->q, req);
  617. }
  618. /*
  619. * Access functions for manipulating queue properties
  620. */
  621. extern request_queue_t *blk_init_queue_node(request_fn_proc *rfn,
  622. spinlock_t *lock, int node_id);
  623. extern request_queue_t *blk_init_queue(request_fn_proc *, spinlock_t *);
  624. extern void blk_cleanup_queue(request_queue_t *);
  625. extern void blk_queue_make_request(request_queue_t *, make_request_fn *);
  626. extern void blk_queue_bounce_limit(request_queue_t *, u64);
  627. extern void blk_queue_max_sectors(request_queue_t *, unsigned int);
  628. extern void blk_queue_max_phys_segments(request_queue_t *, unsigned short);
  629. extern void blk_queue_max_hw_segments(request_queue_t *, unsigned short);
  630. extern void blk_queue_max_segment_size(request_queue_t *, unsigned int);
  631. extern void blk_queue_hardsect_size(request_queue_t *, unsigned short);
  632. extern void blk_queue_stack_limits(request_queue_t *t, request_queue_t *b);
  633. extern void blk_queue_segment_boundary(request_queue_t *, unsigned long);
  634. extern void blk_queue_prep_rq(request_queue_t *, prep_rq_fn *pfn);
  635. extern void blk_queue_merge_bvec(request_queue_t *, merge_bvec_fn *);
  636. extern void blk_queue_dma_alignment(request_queue_t *, int);
  637. extern void blk_queue_softirq_done(request_queue_t *, softirq_done_fn *);
  638. extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
  639. extern int blk_queue_ordered(request_queue_t *, unsigned, prepare_flush_fn *);
  640. extern void blk_queue_issue_flush_fn(request_queue_t *, issue_flush_fn *);
  641. extern int blk_do_ordered(request_queue_t *, struct request **);
  642. extern unsigned blk_ordered_cur_seq(request_queue_t *);
  643. extern unsigned blk_ordered_req_seq(struct request *);
  644. extern void blk_ordered_complete_seq(request_queue_t *, unsigned, int);
  645. extern int blk_rq_map_sg(request_queue_t *, struct request *, struct scatterlist *);
  646. extern void blk_dump_rq_flags(struct request *, char *);
  647. extern void generic_unplug_device(request_queue_t *);
  648. extern void __generic_unplug_device(request_queue_t *);
  649. extern long nr_blockdev_pages(void);
  650. int blk_get_queue(request_queue_t *);
  651. request_queue_t *blk_alloc_queue(gfp_t);
  652. request_queue_t *blk_alloc_queue_node(gfp_t, int);
  653. extern void blk_put_queue(request_queue_t *);
  654. /*
  655. * tag stuff
  656. */
  657. #define blk_queue_tag_depth(q) ((q)->queue_tags->busy)
  658. #define blk_queue_tag_queue(q) ((q)->queue_tags->busy < (q)->queue_tags->max_depth)
  659. #define blk_rq_tagged(rq) ((rq)->cmd_flags & REQ_QUEUED)
  660. extern int blk_queue_start_tag(request_queue_t *, struct request *);
  661. extern struct request *blk_queue_find_tag(request_queue_t *, int);
  662. extern void blk_queue_end_tag(request_queue_t *, struct request *);
  663. extern int blk_queue_init_tags(request_queue_t *, int, struct blk_queue_tag *);
  664. extern void blk_queue_free_tags(request_queue_t *);
  665. extern int blk_queue_resize_tags(request_queue_t *, int);
  666. extern void blk_queue_invalidate_tags(request_queue_t *);
  667. extern struct blk_queue_tag *blk_init_tags(int);
  668. extern void blk_free_tags(struct blk_queue_tag *);
  669. static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
  670. int tag)
  671. {
  672. if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
  673. return NULL;
  674. return bqt->tag_index[tag];
  675. }
  676. extern void blk_rq_bio_prep(request_queue_t *, struct request *, struct bio *);
  677. extern int blkdev_issue_flush(struct block_device *, sector_t *);
  678. #define MAX_PHYS_SEGMENTS 128
  679. #define MAX_HW_SEGMENTS 128
  680. #define SAFE_MAX_SECTORS 255
  681. #define BLK_DEF_MAX_SECTORS 1024
  682. #define MAX_SEGMENT_SIZE 65536
  683. #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
  684. static inline int queue_hardsect_size(request_queue_t *q)
  685. {
  686. int retval = 512;
  687. if (q && q->hardsect_size)
  688. retval = q->hardsect_size;
  689. return retval;
  690. }
  691. static inline int bdev_hardsect_size(struct block_device *bdev)
  692. {
  693. return queue_hardsect_size(bdev_get_queue(bdev));
  694. }
  695. static inline int queue_dma_alignment(request_queue_t *q)
  696. {
  697. int retval = 511;
  698. if (q && q->dma_alignment)
  699. retval = q->dma_alignment;
  700. return retval;
  701. }
  702. /* assumes size > 256 */
  703. static inline unsigned int blksize_bits(unsigned int size)
  704. {
  705. unsigned int bits = 8;
  706. do {
  707. bits++;
  708. size >>= 1;
  709. } while (size > 256);
  710. return bits;
  711. }
  712. static inline unsigned int block_size(struct block_device *bdev)
  713. {
  714. return bdev->bd_block_size;
  715. }
  716. typedef struct {struct page *v;} Sector;
  717. unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
  718. static inline void put_dev_sector(Sector p)
  719. {
  720. page_cache_release(p.v);
  721. }
  722. struct work_struct;
  723. int kblockd_schedule_work(struct work_struct *work);
  724. void kblockd_flush(void);
  725. #define MODULE_ALIAS_BLOCKDEV(major,minor) \
  726. MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
  727. #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
  728. MODULE_ALIAS("block-major-" __stringify(major) "-*")
  729. #else /* CONFIG_BLOCK */
  730. /*
  731. * stubs for when the block layer is configured out
  732. */
  733. #define buffer_heads_over_limit 0
  734. static inline long blk_congestion_wait(int rw, long timeout)
  735. {
  736. return io_schedule_timeout(timeout);
  737. }
  738. static inline long nr_blockdev_pages(void)
  739. {
  740. return 0;
  741. }
  742. static inline void exit_io_context(void)
  743. {
  744. }
  745. #endif /* CONFIG_BLOCK */
  746. #endif