1 #ifndef _LINUX_BLKDEV_H
2 #define _LINUX_BLKDEV_H
4 #include <linux/config.h>
5 #include <linux/major.h>
6 #include <linux/genhd.h>
7 #include <linux/list.h>
8 #include <linux/timer.h>
9 #include <linux/workqueue.h>
10 #include <linux/pagemap.h>
11 #include <linux/backing-dev.h>
12 #include <linux/wait.h>
13 #include <linux/mempool.h>
14 #include <linux/bio.h>
15 #include <linux/module.h>
16 #include <linux/stringify.h>
18 #include <asm/scatterlist.h>
21 typedef struct request_queue request_queue_t;
23 typedef struct elevator_s elevator_t;
24 struct request_pm_state;
26 #define BLKDEV_MIN_RQ 4
27 #define BLKDEV_MAX_RQ 128 /* Default maximum */
30 * This is the per-process anticipatory I/O scheduler state.
32 struct as_io_context {
35 void (*dtor)(struct as_io_context *aic); /* destructor */
36 void (*exit)(struct as_io_context *aic); /* called on task exit */
39 atomic_t nr_queued; /* queued reads & sync writes */
40 atomic_t nr_dispatched; /* number of requests gone to the drivers */
42 /* IO History tracking */
44 unsigned long last_end_request;
45 unsigned long ttime_total;
46 unsigned long ttime_samples;
47 unsigned long ttime_mean;
49 unsigned int seek_samples;
50 sector_t last_request_pos;
56 * This is the per-process I/O subsystem state. It is refcounted and
57 * kmalloc'ed. Currently all fields are modified in process io context
58 * (apart from the atomic refcount), so require no locking.
65 * For request batching
67 unsigned long last_waited; /* Time last woken after wait for request */
68 int nr_batch_requests; /* Number of requests left in the batch */
70 struct as_io_context *aic;
73 void put_io_context(struct io_context *ioc);
74 void exit_io_context(void);
75 struct io_context *get_io_context(int gfp_flags);
76 void copy_io_context(struct io_context **pdst, struct io_context **psrc);
77 void swap_io_context(struct io_context **ioc1, struct io_context **ioc2);
82 wait_queue_head_t wait[2];
85 #define BLK_MAX_CDB 16
88 * try to put the fields that are referenced together in the same cacheline
91 struct list_head queuelist; /* looking for ->queue? you must _not_
92 * access it directly, use
93 * blkdev_dequeue_request! */
94 unsigned long flags; /* see REQ_ bits below */
96 /* Maintain bio traversal state for part by part I/O submission.
97 * hard_* are block layer internals, no driver should touch them!
100 sector_t sector; /* next sector to submit */
101 unsigned long nr_sectors; /* no. of sectors left to submit */
102 /* no. of sectors left to submit in the current segment */
103 unsigned int current_nr_sectors;
105 sector_t hard_sector; /* next sector to complete */
106 unsigned long hard_nr_sectors; /* no. of sectors left to complete */
107 /* no. of sectors left to complete in the current segment */
108 unsigned int hard_cur_sectors;
110 /* no. of segments left to submit in the current bio */
111 unsigned short nr_cbio_segments;
112 /* no. of sectors left to submit in the current bio */
113 unsigned long nr_cbio_sectors;
115 struct bio *cbio; /* next bio to submit */
116 struct bio *bio; /* next unfinished bio to complete */
119 void *elevator_private;
121 int rq_status; /* should split this into a few status bits */
122 struct gendisk *rq_disk;
124 unsigned long start_time;
126 /* Number of scatter-gather DMA addr+len pairs after
127 * physical address coalescing is performed.
129 unsigned short nr_phys_segments;
131 /* Number of scatter-gather addr+len pairs after
132 * physical and DMA remapping hardware coalescing is performed.
133 * This is the number of scatter-gather entries the driver
134 * will actually have to deal with after DMA mapping is done.
136 unsigned short nr_hw_segments;
143 struct request_list *rl;
145 struct completion *waiting;
149 * when request is used as a packet command carrier
151 unsigned int cmd_len;
152 unsigned char cmd[BLK_MAX_CDB];
154 unsigned int data_len;
157 unsigned int sense_len;
160 unsigned int timeout;
163 * For Power Management requests
165 struct request_pm_state *pm;
169 * first three bits match BIO_RW* bits, important
172 __REQ_RW, /* not set, read. set, write */
173 __REQ_FAILFAST, /* no low level driver retries */
174 __REQ_SOFTBARRIER, /* may not be passed by ioscheduler */
175 __REQ_HARDBARRIER, /* may not be passed by drive either */
176 __REQ_CMD, /* is a regular fs rw request */
177 __REQ_NOMERGE, /* don't touch this for merging */
178 __REQ_STARTED, /* drive already may have started this one */
179 __REQ_DONTPREP, /* don't call prep for this one */
180 __REQ_QUEUED, /* uses queueing */
182 * for ATA/ATAPI devices
184 __REQ_PC, /* packet command (special) */
185 __REQ_BLOCK_PC, /* queued down pc from block layer */
186 __REQ_SENSE, /* sense retrival */
188 __REQ_FAILED, /* set if the request failed */
189 __REQ_QUIET, /* don't worry about errors */
190 __REQ_SPECIAL, /* driver suplied command */
193 __REQ_DRIVE_TASKFILE,
194 __REQ_PREEMPT, /* set for "ide_preempt" requests */
195 __REQ_PM_SUSPEND, /* suspend request */
196 __REQ_PM_RESUME, /* resume request */
197 __REQ_PM_SHUTDOWN, /* shutdown request */
198 __REQ_NR_BITS, /* stops here */
201 #define REQ_RW (1 << __REQ_RW)
202 #define REQ_FAILFAST (1 << __REQ_FAILFAST)
203 #define REQ_SOFTBARRIER (1 << __REQ_SOFTBARRIER)
204 #define REQ_HARDBARRIER (1 << __REQ_HARDBARRIER)
205 #define REQ_CMD (1 << __REQ_CMD)
206 #define REQ_NOMERGE (1 << __REQ_NOMERGE)
207 #define REQ_STARTED (1 << __REQ_STARTED)
208 #define REQ_DONTPREP (1 << __REQ_DONTPREP)
209 #define REQ_QUEUED (1 << __REQ_QUEUED)
210 #define REQ_PC (1 << __REQ_PC)
211 #define REQ_BLOCK_PC (1 << __REQ_BLOCK_PC)
212 #define REQ_SENSE (1 << __REQ_SENSE)
213 #define REQ_FAILED (1 << __REQ_FAILED)
214 #define REQ_QUIET (1 << __REQ_QUIET)
215 #define REQ_SPECIAL (1 << __REQ_SPECIAL)
216 #define REQ_DRIVE_CMD (1 << __REQ_DRIVE_CMD)
217 #define REQ_DRIVE_TASK (1 << __REQ_DRIVE_TASK)
218 #define REQ_DRIVE_TASKFILE (1 << __REQ_DRIVE_TASKFILE)
219 #define REQ_PREEMPT (1 << __REQ_PREEMPT)
220 #define REQ_PM_SUSPEND (1 << __REQ_PM_SUSPEND)
221 #define REQ_PM_RESUME (1 << __REQ_PM_RESUME)
222 #define REQ_PM_SHUTDOWN (1 << __REQ_PM_SHUTDOWN)
225 * State information carried for REQ_PM_SUSPEND and REQ_PM_RESUME
226 * requests. Some step values could eventually be made generic.
228 struct request_pm_state
230 /* PM state machine step value, currently driver specific */
232 /* requested PM state value (S1, S2, S3, S4, ...) */
234 void* data; /* for driver use */
237 #include <linux/elevator.h>
239 typedef int (merge_request_fn) (request_queue_t *, struct request *,
241 typedef int (merge_requests_fn) (request_queue_t *, struct request *,
243 typedef void (request_fn_proc) (request_queue_t *q);
244 typedef int (make_request_fn) (request_queue_t *q, struct bio *bio);
245 typedef int (prep_rq_fn) (request_queue_t *, struct request *);
246 typedef void (unplug_fn) (request_queue_t *);
249 typedef int (merge_bvec_fn) (request_queue_t *, struct bio *, struct bio_vec *);
250 typedef void (activity_fn) (void *data, int rw);
252 enum blk_queue_state {
257 #define BLK_TAGS_PER_LONG (sizeof(unsigned long) * 8)
258 #define BLK_TAGS_MASK (BLK_TAGS_PER_LONG - 1)
260 struct blk_queue_tag {
261 struct request **tag_index; /* map of busy tags */
262 unsigned long *tag_map; /* bit map of free/busy tags */
263 struct list_head busy_list; /* fifo list of busy tags */
264 int busy; /* current depth */
265 int max_depth; /* what we will send to device */
266 int real_max_depth; /* what the array can hold */
267 atomic_t refcnt; /* map can be shared */
273 * Together with queue_head for cacheline sharing
275 struct list_head queue_head;
276 struct request *last_merge;
280 * the queue request freelist, one for reads and one for writes
282 struct request_list rq;
284 request_fn_proc *request_fn;
285 merge_request_fn *back_merge_fn;
286 merge_request_fn *front_merge_fn;
287 merge_requests_fn *merge_requests_fn;
288 make_request_fn *make_request_fn;
289 prep_rq_fn *prep_rq_fn;
290 unplug_fn *unplug_fn;
291 merge_bvec_fn *merge_bvec_fn;
292 activity_fn *activity_fn;
295 * Auto-unplugging state
297 struct timer_list unplug_timer;
298 int unplug_thresh; /* After this many requests */
299 unsigned long unplug_delay; /* After this many jiffies */
300 struct work_struct unplug_work;
302 struct backing_dev_info backing_dev_info;
305 * The queue owner gets to use this for whatever they like.
306 * ll_rw_blk doesn't touch it.
313 * queue needs bounce pages for pages above this limit
315 unsigned long bounce_pfn;
319 * various queue flags, see QUEUE_* below
321 unsigned long queue_flags;
324 * protects queue structures from reentrancy
326 spinlock_t *queue_lock;
336 unsigned long nr_requests; /* Max # of requests */
337 unsigned int nr_congestion_on;
338 unsigned int nr_congestion_off;
340 unsigned short max_sectors;
341 unsigned short max_phys_segments;
342 unsigned short max_hw_segments;
343 unsigned short hardsect_size;
344 unsigned int max_segment_size;
346 unsigned long seg_boundary_mask;
347 unsigned int dma_alignment;
349 struct blk_queue_tag *queue_tags;
353 unsigned int in_flight;
358 unsigned int sg_timeout;
359 unsigned int sg_reserved_size;
362 #define RQ_INACTIVE (-1)
364 #define RQ_SCSI_BUSY 0xffff
365 #define RQ_SCSI_DONE 0xfffe
366 #define RQ_SCSI_DISCONNECTING 0xffe0
368 #define QUEUE_FLAG_CLUSTER 0 /* cluster several segments into 1 */
369 #define QUEUE_FLAG_QUEUED 1 /* uses generic tag queueing */
370 #define QUEUE_FLAG_STOPPED 2 /* queue is stopped */
371 #define QUEUE_FLAG_READFULL 3 /* write queue has been filled */
372 #define QUEUE_FLAG_WRITEFULL 4 /* read queue has been filled */
373 #define QUEUE_FLAG_DEAD 5 /* queue being torn down */
374 #define QUEUE_FLAG_REENTER 6 /* Re-entrancy avoidance */
375 #define QUEUE_FLAG_PLUGGED 7 /* queue is plugged */
377 #define blk_queue_plugged(q) test_bit(QUEUE_FLAG_PLUGGED, &(q)->queue_flags)
378 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
379 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
381 #define blk_fs_request(rq) ((rq)->flags & REQ_CMD)
382 #define blk_pc_request(rq) ((rq)->flags & REQ_BLOCK_PC)
383 #define blk_noretry_request(rq) ((rq)->flags & REQ_FAILFAST)
384 #define blk_rq_started(rq) ((rq)->flags & REQ_STARTED)
386 #define blk_account_rq(rq) (blk_rq_started(rq) && blk_fs_request(rq))
388 #define blk_pm_suspend_request(rq) ((rq)->flags & REQ_PM_SUSPEND)
389 #define blk_pm_resume_request(rq) ((rq)->flags & REQ_PM_RESUME)
390 #define blk_pm_request(rq) \
391 ((rq)->flags & (REQ_PM_SUSPEND | REQ_PM_RESUME))
393 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
395 #define rq_data_dir(rq) ((rq)->flags & 1)
397 static inline int blk_queue_full(struct request_queue *q, int rw)
400 return test_bit(QUEUE_FLAG_READFULL, &q->queue_flags);
401 return test_bit(QUEUE_FLAG_WRITEFULL, &q->queue_flags);
404 static inline void blk_set_queue_full(struct request_queue *q, int rw)
407 set_bit(QUEUE_FLAG_READFULL, &q->queue_flags);
409 set_bit(QUEUE_FLAG_WRITEFULL, &q->queue_flags);
412 static inline void blk_clear_queue_full(struct request_queue *q, int rw)
415 clear_bit(QUEUE_FLAG_READFULL, &q->queue_flags);
417 clear_bit(QUEUE_FLAG_WRITEFULL, &q->queue_flags);
422 * mergeable request must not have _NOMERGE or _BARRIER bit set, nor may
423 * it already be started by driver.
425 #define RQ_NOMERGE_FLAGS \
426 (REQ_NOMERGE | REQ_STARTED | REQ_HARDBARRIER | REQ_SOFTBARRIER)
427 #define rq_mergeable(rq) \
428 (!((rq)->flags & RQ_NOMERGE_FLAGS) && blk_fs_request((rq)))
431 * noop, requests are automagically marked as active/inactive by I/O
432 * scheduler -- see elv_next_request
434 #define blk_queue_headactive(q, head_active)
436 /* current index into bio being processed for submission */
437 #define blk_rq_idx(rq) ((rq)->cbio->bi_vcnt - (rq)->nr_cbio_segments)
439 /* current bio vector being processed */
440 #define blk_rq_vec(rq) (bio_iovec_idx((rq)->cbio, blk_rq_idx(rq)))
442 /* current offset with respect to start of the segment being submitted */
443 #define blk_rq_offset(rq) \
444 (((rq)->hard_cur_sectors - (rq)->current_nr_sectors) << 9)
447 * temporarily mapping a (possible) highmem bio (typically for PIO transfer)
450 /* Assumes rq->cbio != NULL */
451 static inline char * rq_map_buffer(struct request *rq, unsigned long *flags)
453 return (__bio_kmap_irq(rq->cbio, blk_rq_idx(rq), flags)
454 + blk_rq_offset(rq));
457 static inline void rq_unmap_buffer(char *buffer, unsigned long *flags)
459 __bio_kunmap_irq(buffer, flags);
463 * q->prep_rq_fn return values
465 #define BLKPREP_OK 0 /* serve it */
466 #define BLKPREP_KILL 1 /* fatal error, kill */
467 #define BLKPREP_DEFER 2 /* leave on queue */
469 extern unsigned long blk_max_low_pfn, blk_max_pfn;
472 * standard bounce addresses:
474 * BLK_BOUNCE_HIGH : bounce all highmem pages
475 * BLK_BOUNCE_ANY : don't bounce anything
476 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
478 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
479 #define BLK_BOUNCE_ANY ((u64)blk_max_pfn << PAGE_SHIFT)
480 #define BLK_BOUNCE_ISA (ISA_DMA_THRESHOLD)
483 extern int init_emergency_isa_pool(void);
484 extern void blk_queue_bounce(request_queue_t *q, struct bio **bio);
486 static inline int init_emergency_isa_pool(void)
490 static inline void blk_queue_bounce(request_queue_t *q, struct bio **bio)
493 #endif /* CONFIG_MMU */
495 #define rq_for_each_bio(_bio, rq) \
497 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
501 unsigned block_size_bits;
504 extern int blk_register_queue(struct gendisk *disk);
505 extern void blk_unregister_queue(struct gendisk *disk);
506 extern void register_disk(struct gendisk *dev);
507 extern void generic_make_request(struct bio *bio);
508 extern void blk_put_request(struct request *);
509 extern void blk_attempt_remerge(request_queue_t *, struct request *);
510 extern void __blk_attempt_remerge(request_queue_t *, struct request *);
511 extern struct request *blk_get_request(request_queue_t *, int, int);
512 extern void blk_put_request(struct request *);
513 extern void blk_insert_request(request_queue_t *, struct request *, int, void *, int);
514 extern void blk_requeue_request(request_queue_t *, struct request *);
515 extern void blk_plug_device(request_queue_t *);
516 extern int blk_remove_plug(request_queue_t *);
517 extern void blk_recount_segments(request_queue_t *, struct bio *);
518 extern int blk_phys_contig_segment(request_queue_t *q, struct bio *, struct bio *);
519 extern int blk_hw_contig_segment(request_queue_t *q, struct bio *, struct bio *);
520 extern int scsi_cmd_ioctl(struct file *, struct gendisk *, unsigned int, void __user *);
521 extern void blk_start_queue(request_queue_t *q);
522 extern void blk_stop_queue(request_queue_t *q);
523 extern void __blk_stop_queue(request_queue_t *q);
524 extern void blk_run_queue(request_queue_t *);
525 extern void blk_queue_activity_fn(request_queue_t *, activity_fn *, void *);
526 extern struct request *blk_rq_map_user(request_queue_t *, int, void __user *, unsigned int);
527 extern int blk_rq_unmap_user(struct request *, struct bio *, unsigned int);
528 extern int blk_execute_rq(request_queue_t *, struct gendisk *, struct request *);
530 static inline request_queue_t *bdev_get_queue(struct block_device *bdev)
532 return bdev->bd_disk->queue;
535 static inline void blk_run_backing_dev(struct backing_dev_info *bdi,
538 if (bdi && bdi->unplug_io_fn)
539 bdi->unplug_io_fn(bdi, page);
542 static inline void blk_run_address_space(struct address_space *mapping)
545 blk_run_backing_dev(mapping->backing_dev_info, NULL);
549 * end_request() and friends. Must be called with the request queue spinlock
550 * acquired. All functions called within end_request() _must_be_ atomic.
552 * Several drivers define their own end_request and call
553 * end_that_request_first() and end_that_request_last()
554 * for parts of the original function. This prevents
555 * code duplication in drivers.
557 extern int end_that_request_first(struct request *, int, int);
558 extern int end_that_request_chunk(struct request *, int, int);
559 extern void end_that_request_last(struct request *);
560 extern int process_that_request_first(struct request *, unsigned int);
561 extern void end_request(struct request *req, int uptodate);
563 static inline void blkdev_dequeue_request(struct request *req)
565 BUG_ON(list_empty(&req->queuelist));
567 list_del_init(&req->queuelist);
570 elv_remove_request(req->q, req);
574 * Access functions for manipulating queue properties
576 extern request_queue_t *blk_init_queue(request_fn_proc *, spinlock_t *);
577 extern void blk_cleanup_queue(request_queue_t *);
578 extern void blk_queue_make_request(request_queue_t *, make_request_fn *);
579 extern void blk_queue_bounce_limit(request_queue_t *, u64);
580 extern void blk_queue_max_sectors(request_queue_t *, unsigned short);
581 extern void blk_queue_max_phys_segments(request_queue_t *, unsigned short);
582 extern void blk_queue_max_hw_segments(request_queue_t *, unsigned short);
583 extern void blk_queue_max_segment_size(request_queue_t *, unsigned int);
584 extern void blk_queue_hardsect_size(request_queue_t *, unsigned short);
585 extern void blk_queue_stack_limits(request_queue_t *t, request_queue_t *b);
586 extern void blk_queue_segment_boundary(request_queue_t *, unsigned long);
587 extern void blk_queue_prep_rq(request_queue_t *, prep_rq_fn *pfn);
588 extern void blk_queue_merge_bvec(request_queue_t *, merge_bvec_fn *);
589 extern void blk_queue_dma_alignment(request_queue_t *, int);
590 extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
592 extern int blk_rq_map_sg(request_queue_t *, struct request *, struct scatterlist *);
593 extern void blk_dump_rq_flags(struct request *, char *);
594 extern void generic_unplug_device(request_queue_t *);
595 extern void __generic_unplug_device(request_queue_t *);
596 extern long nr_blockdev_pages(void);
598 int blk_get_queue(request_queue_t *);
599 request_queue_t *blk_alloc_queue(int);
600 #define blk_put_queue(q) blk_cleanup_queue((q))
605 #define blk_queue_tag_depth(q) ((q)->queue_tags->busy)
606 #define blk_queue_tag_queue(q) ((q)->queue_tags->busy < (q)->queue_tags->max_depth)
607 #define blk_rq_tagged(rq) ((rq)->flags & REQ_QUEUED)
608 extern int blk_queue_start_tag(request_queue_t *, struct request *);
609 extern struct request *blk_queue_find_tag(request_queue_t *, int);
610 extern void blk_queue_end_tag(request_queue_t *, struct request *);
611 extern int blk_queue_init_tags(request_queue_t *, int, struct blk_queue_tag *);
612 extern void blk_queue_free_tags(request_queue_t *);
613 extern int blk_queue_resize_tags(request_queue_t *, int);
614 extern void blk_queue_invalidate_tags(request_queue_t *);
615 extern long blk_congestion_wait(int rw, long timeout);
617 extern void blk_rq_bio_prep(request_queue_t *, struct request *, struct bio *);
618 extern void blk_rq_prep_restart(struct request *);
620 #define MAX_PHYS_SEGMENTS 128
621 #define MAX_HW_SEGMENTS 128
622 #define MAX_SECTORS 255
624 #define MAX_SEGMENT_SIZE 65536
626 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
628 extern void drive_stat_acct(struct request *, int, int);
630 static inline int queue_hardsect_size(request_queue_t *q)
634 if (q && q->hardsect_size)
635 retval = q->hardsect_size;
640 static inline int bdev_hardsect_size(struct block_device *bdev)
642 return queue_hardsect_size(bdev_get_queue(bdev));
645 static inline int queue_dma_alignment(request_queue_t *q)
649 if (q && q->dma_alignment)
650 retval = q->dma_alignment;
655 static inline int bdev_dma_aligment(struct block_device *bdev)
657 return queue_dma_alignment(bdev_get_queue(bdev));
660 #define blk_finished_io(nsects) do { } while (0)
661 #define blk_started_io(nsects) do { } while (0)
663 /* assumes size > 256 */
664 static inline unsigned int blksize_bits(unsigned int size)
666 unsigned int bits = 8;
670 } while (size > 256);
674 extern inline unsigned int block_size(struct block_device *bdev)
676 return bdev->bd_block_size;
679 typedef struct {struct page *v;} Sector;
681 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
683 static inline void put_dev_sector(Sector p)
685 page_cache_release(p.v);
689 int kblockd_schedule_work(struct work_struct *work);
690 void kblockd_flush(void);
693 # include <asm/div64.h>
694 # define sector_div(a, b) do_div(a, b)
696 # define sector_div(n, b)( \
706 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
707 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
708 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
709 MODULE_ALIAS("block-major-" __stringify(major) "-*")