1 #ifndef _LINUX_BLKDEV_H
2 #define _LINUX_BLKDEV_H
6 #include <linux/sched.h>
7 #include <linux/major.h>
8 #include <linux/genhd.h>
9 #include <linux/list.h>
10 #include <linux/timer.h>
11 #include <linux/workqueue.h>
12 #include <linux/pagemap.h>
13 #include <linux/backing-dev.h>
14 #include <linux/wait.h>
15 #include <linux/mempool.h>
16 #include <linux/bio.h>
17 #include <linux/module.h>
18 #include <linux/stringify.h>
19 #include <linux/gfp.h>
20 #include <linux/bsg.h>
21 #include <linux/smp.h>
23 #include <asm/scatterlist.h>
25 struct scsi_ioctl_command
;
28 struct elevator_queue
;
29 struct request_pm_state
;
34 #define BLKDEV_MIN_RQ 4
35 #define BLKDEV_MAX_RQ 128 /* Default maximum */
38 typedef void (rq_end_io_fn
)(struct request
*, int);
42 * count[], starved[], and wait[] are indexed by
43 * BLK_RW_SYNC/BLK_RW_ASYNC
49 wait_queue_head_t wait
[2];
53 * request command types
55 enum rq_cmd_type_bits
{
56 REQ_TYPE_FS
= 1, /* fs request */
57 REQ_TYPE_BLOCK_PC
, /* scsi command */
58 REQ_TYPE_SENSE
, /* sense request */
59 REQ_TYPE_PM_SUSPEND
, /* suspend request */
60 REQ_TYPE_PM_RESUME
, /* resume request */
61 REQ_TYPE_PM_SHUTDOWN
, /* shutdown request */
62 REQ_TYPE_SPECIAL
, /* driver defined type */
63 REQ_TYPE_LINUX_BLOCK
, /* generic block layer message */
65 * for ATA/ATAPI devices. this really doesn't belong here, ide should
66 * use REQ_TYPE_SPECIAL and use rq->cmd[0] with the range of driver
67 * private REQ_LB opcodes to differentiate what type of request this is
69 REQ_TYPE_ATA_TASKFILE
,
79 * For request of type REQ_TYPE_LINUX_BLOCK, rq->cmd[0] is the opcode being
80 * sent down (similar to how REQ_TYPE_BLOCK_PC means that ->cmd[] holds a
83 * 0x00 -> 0x3f are driver private, to be used for whatever purpose they need,
84 * typically to differentiate REQ_TYPE_SPECIAL requests.
88 REQ_LB_OP_EJECT
= 0x40, /* eject request */
89 REQ_LB_OP_FLUSH
= 0x41, /* flush request */
90 REQ_LB_OP_DISCARD
= 0x42, /* discard sectors */
94 * request type modified bits. first two bits match BIO_RW* bits, important
97 __REQ_RW
, /* not set, read. set, write */
98 __REQ_FAILFAST_DEV
, /* no driver retries of device errors */
99 __REQ_FAILFAST_TRANSPORT
, /* no driver retries of transport errors */
100 __REQ_FAILFAST_DRIVER
, /* no driver retries of driver errors */
101 __REQ_DISCARD
, /* request to discard sectors */
102 __REQ_SORTED
, /* elevator knows about this request */
103 __REQ_SOFTBARRIER
, /* may not be passed by ioscheduler */
104 __REQ_HARDBARRIER
, /* may not be passed by drive either */
105 __REQ_FUA
, /* forced unit access */
106 __REQ_NOMERGE
, /* don't touch this for merging */
107 __REQ_STARTED
, /* drive already may have started this one */
108 __REQ_DONTPREP
, /* don't call prep for this one */
109 __REQ_QUEUED
, /* uses queueing */
110 __REQ_ELVPRIV
, /* elevator private data attached */
111 __REQ_FAILED
, /* set if the request failed */
112 __REQ_QUIET
, /* don't worry about errors */
113 __REQ_PREEMPT
, /* set for "ide_preempt" requests */
114 __REQ_ORDERED_COLOR
, /* is before or after barrier */
115 __REQ_RW_SYNC
, /* request is sync (sync write or read) */
116 __REQ_ALLOCED
, /* request came from our alloc pool */
117 __REQ_RW_META
, /* metadata io request */
118 __REQ_COPY_USER
, /* contains copies of user pages */
119 __REQ_INTEGRITY
, /* integrity metadata has been remapped */
120 __REQ_NOIDLE
, /* Don't anticipate more IO after this one */
121 __REQ_IO_STAT
, /* account I/O stat */
122 __REQ_NR_BITS
, /* stops here */
125 #define REQ_RW (1 << __REQ_RW)
126 #define REQ_FAILFAST_DEV (1 << __REQ_FAILFAST_DEV)
127 #define REQ_FAILFAST_TRANSPORT (1 << __REQ_FAILFAST_TRANSPORT)
128 #define REQ_FAILFAST_DRIVER (1 << __REQ_FAILFAST_DRIVER)
129 #define REQ_DISCARD (1 << __REQ_DISCARD)
130 #define REQ_SORTED (1 << __REQ_SORTED)
131 #define REQ_SOFTBARRIER (1 << __REQ_SOFTBARRIER)
132 #define REQ_HARDBARRIER (1 << __REQ_HARDBARRIER)
133 #define REQ_FUA (1 << __REQ_FUA)
134 #define REQ_NOMERGE (1 << __REQ_NOMERGE)
135 #define REQ_STARTED (1 << __REQ_STARTED)
136 #define REQ_DONTPREP (1 << __REQ_DONTPREP)
137 #define REQ_QUEUED (1 << __REQ_QUEUED)
138 #define REQ_ELVPRIV (1 << __REQ_ELVPRIV)
139 #define REQ_FAILED (1 << __REQ_FAILED)
140 #define REQ_QUIET (1 << __REQ_QUIET)
141 #define REQ_PREEMPT (1 << __REQ_PREEMPT)
142 #define REQ_ORDERED_COLOR (1 << __REQ_ORDERED_COLOR)
143 #define REQ_RW_SYNC (1 << __REQ_RW_SYNC)
144 #define REQ_ALLOCED (1 << __REQ_ALLOCED)
145 #define REQ_RW_META (1 << __REQ_RW_META)
146 #define REQ_COPY_USER (1 << __REQ_COPY_USER)
147 #define REQ_INTEGRITY (1 << __REQ_INTEGRITY)
148 #define REQ_NOIDLE (1 << __REQ_NOIDLE)
149 #define REQ_IO_STAT (1 << __REQ_IO_STAT)
151 #define BLK_MAX_CDB 16
154 * try to put the fields that are referenced together in the same cacheline.
155 * if you modify this structure, be sure to check block/blk-core.c:rq_init()
159 struct list_head queuelist
;
160 struct call_single_data csd
;
163 struct request_queue
*q
;
165 unsigned int cmd_flags
;
166 enum rq_cmd_type_bits cmd_type
;
167 unsigned long atomic_flags
;
169 /* the following two fields are internal, NEVER access directly */
170 sector_t __sector
; /* sector cursor */
171 unsigned int __data_len
; /* total data len */
176 struct hlist_node hash
; /* merge hash */
178 * The rb_node is only used inside the io scheduler, requests
179 * are pruned when moved to the dispatch queue. So let the
180 * completion_data share space with the rb_node.
183 struct rb_node rb_node
; /* sort/lookup */
184 void *completion_data
;
188 * two pointers are available for the IO schedulers, if they need
189 * more they have to dynamically allocate it.
191 void *elevator_private
;
192 void *elevator_private2
;
194 struct gendisk
*rq_disk
;
195 unsigned long start_time
;
197 /* Number of scatter-gather DMA addr+len pairs after
198 * physical address coalescing is performed.
200 unsigned short nr_phys_segments
;
202 unsigned short ioprio
;
204 void *special
; /* opaque pointer available for LLD use */
205 char *buffer
; /* kaddr of the current segment if available */
213 * when request is used as a packet command carrier
215 unsigned short cmd_len
;
216 unsigned char __cmd
[BLK_MAX_CDB
];
219 unsigned int extra_len
; /* length of alignment and padding */
220 unsigned int sense_len
;
221 unsigned int resid_len
; /* residual count */
224 unsigned long deadline
;
225 struct list_head timeout_list
;
226 unsigned int timeout
;
230 * completion callback.
232 rq_end_io_fn
*end_io
;
236 struct request
*next_rq
;
239 static inline unsigned short req_get_ioprio(struct request
*req
)
245 * State information carried for REQ_TYPE_PM_SUSPEND and REQ_TYPE_PM_RESUME
246 * requests. Some step values could eventually be made generic.
248 struct request_pm_state
250 /* PM state machine step value, currently driver specific */
252 /* requested PM state value (S1, S2, S3, S4, ...) */
254 void* data
; /* for driver use */
257 #include <linux/elevator.h>
259 typedef void (request_fn_proc
) (struct request_queue
*q
);
260 typedef int (make_request_fn
) (struct request_queue
*q
, struct bio
*bio
);
261 typedef int (prep_rq_fn
) (struct request_queue
*, struct request
*);
262 typedef void (unplug_fn
) (struct request_queue
*);
263 typedef int (prepare_discard_fn
) (struct request_queue
*, struct request
*);
266 struct bvec_merge_data
{
267 struct block_device
*bi_bdev
;
272 typedef int (merge_bvec_fn
) (struct request_queue
*, struct bvec_merge_data
*,
274 typedef void (prepare_flush_fn
) (struct request_queue
*, struct request
*);
275 typedef void (softirq_done_fn
)(struct request
*);
276 typedef int (dma_drain_needed_fn
)(struct request
*);
277 typedef int (lld_busy_fn
) (struct request_queue
*q
);
279 enum blk_eh_timer_return
{
285 typedef enum blk_eh_timer_return (rq_timed_out_fn
)(struct request
*);
287 enum blk_queue_state
{
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 int busy
; /* current depth */
296 int max_depth
; /* what we will send to device */
297 int real_max_depth
; /* what the array can hold */
298 atomic_t refcnt
; /* map can be shared */
301 #define BLK_SCSI_MAX_CMDS (256)
302 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
304 struct blk_cmd_filter
{
305 unsigned long read_ok
[BLK_SCSI_CMD_PER_LONG
];
306 unsigned long write_ok
[BLK_SCSI_CMD_PER_LONG
];
313 * Together with queue_head for cacheline sharing
315 struct list_head queue_head
;
316 struct request
*last_merge
;
317 struct elevator_queue
*elevator
;
320 * the queue request freelist, one for reads and one for writes
322 struct request_list rq
;
324 request_fn_proc
*request_fn
;
325 make_request_fn
*make_request_fn
;
326 prep_rq_fn
*prep_rq_fn
;
327 unplug_fn
*unplug_fn
;
328 prepare_discard_fn
*prepare_discard_fn
;
329 merge_bvec_fn
*merge_bvec_fn
;
330 prepare_flush_fn
*prepare_flush_fn
;
331 softirq_done_fn
*softirq_done_fn
;
332 rq_timed_out_fn
*rq_timed_out_fn
;
333 dma_drain_needed_fn
*dma_drain_needed
;
334 lld_busy_fn
*lld_busy_fn
;
337 * Dispatch queue sorting
340 struct request
*boundary_rq
;
343 * Auto-unplugging state
345 struct timer_list unplug_timer
;
346 int unplug_thresh
; /* After this many requests */
347 unsigned long unplug_delay
; /* After this many jiffies */
348 struct work_struct unplug_work
;
350 struct backing_dev_info backing_dev_info
;
353 * The queue owner gets to use this for whatever they like.
354 * ll_rw_blk doesn't touch it.
359 * queue needs bounce pages for pages above this limit
361 unsigned long bounce_pfn
;
365 * various queue flags, see QUEUE_* below
367 unsigned long queue_flags
;
370 * protects queue structures from reentrancy. ->__queue_lock should
371 * _never_ be used directly, it is queue private. always use
374 spinlock_t __queue_lock
;
375 spinlock_t
*queue_lock
;
385 unsigned long nr_requests
; /* Max # of requests */
386 unsigned int nr_congestion_on
;
387 unsigned int nr_congestion_off
;
388 unsigned int nr_batching
;
390 unsigned int max_sectors
;
391 unsigned int max_hw_sectors
;
392 unsigned short max_phys_segments
;
393 unsigned short max_hw_segments
;
394 unsigned short hardsect_size
;
395 unsigned int max_segment_size
;
397 unsigned long seg_boundary_mask
;
398 void *dma_drain_buffer
;
399 unsigned int dma_drain_size
;
400 unsigned int dma_pad_mask
;
401 unsigned int dma_alignment
;
403 struct blk_queue_tag
*queue_tags
;
404 struct list_head tag_busy_list
;
406 unsigned int nr_sorted
;
407 unsigned int in_flight
;
409 unsigned int rq_timeout
;
410 struct timer_list timeout
;
411 struct list_head timeout_list
;
416 unsigned int sg_timeout
;
417 unsigned int sg_reserved_size
;
419 #ifdef CONFIG_BLK_DEV_IO_TRACE
420 struct blk_trace
*blk_trace
;
423 * reserved for flush operations
425 unsigned int ordered
, next_ordered
, ordseq
;
426 int orderr
, ordcolor
;
427 struct request pre_flush_rq
, bar_rq
, post_flush_rq
;
428 struct request
*orig_bar_rq
;
430 struct mutex sysfs_lock
;
432 #if defined(CONFIG_BLK_DEV_BSG)
433 struct bsg_class_device bsg_dev
;
435 struct blk_cmd_filter cmd_filter
;
438 #define QUEUE_FLAG_CLUSTER 0 /* cluster several segments into 1 */
439 #define QUEUE_FLAG_QUEUED 1 /* uses generic tag queueing */
440 #define QUEUE_FLAG_STOPPED 2 /* queue is stopped */
441 #define QUEUE_FLAG_SYNCFULL 3 /* read queue has been filled */
442 #define QUEUE_FLAG_ASYNCFULL 4 /* write queue has been filled */
443 #define QUEUE_FLAG_DEAD 5 /* queue being torn down */
444 #define QUEUE_FLAG_REENTER 6 /* Re-entrancy avoidance */
445 #define QUEUE_FLAG_PLUGGED 7 /* queue is plugged */
446 #define QUEUE_FLAG_ELVSWITCH 8 /* don't use elevator, just do FIFO */
447 #define QUEUE_FLAG_BIDI 9 /* queue supports bidi requests */
448 #define QUEUE_FLAG_NOMERGES 10 /* disable merge attempts */
449 #define QUEUE_FLAG_SAME_COMP 11 /* force complete on same CPU */
450 #define QUEUE_FLAG_FAIL_IO 12 /* fake timeout */
451 #define QUEUE_FLAG_STACKABLE 13 /* supports request stacking */
452 #define QUEUE_FLAG_NONROT 14 /* non-rotational device (SSD) */
453 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
454 #define QUEUE_FLAG_IO_STAT 15 /* do IO stats */
456 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
457 (1 << QUEUE_FLAG_CLUSTER) | \
458 (1 << QUEUE_FLAG_STACKABLE))
460 static inline int queue_is_locked(struct request_queue
*q
)
463 spinlock_t
*lock
= q
->queue_lock
;
464 return lock
&& spin_is_locked(lock
);
470 static inline void queue_flag_set_unlocked(unsigned int flag
,
471 struct request_queue
*q
)
473 __set_bit(flag
, &q
->queue_flags
);
476 static inline int queue_flag_test_and_clear(unsigned int flag
,
477 struct request_queue
*q
)
479 WARN_ON_ONCE(!queue_is_locked(q
));
481 if (test_bit(flag
, &q
->queue_flags
)) {
482 __clear_bit(flag
, &q
->queue_flags
);
489 static inline int queue_flag_test_and_set(unsigned int flag
,
490 struct request_queue
*q
)
492 WARN_ON_ONCE(!queue_is_locked(q
));
494 if (!test_bit(flag
, &q
->queue_flags
)) {
495 __set_bit(flag
, &q
->queue_flags
);
502 static inline void queue_flag_set(unsigned int flag
, struct request_queue
*q
)
504 WARN_ON_ONCE(!queue_is_locked(q
));
505 __set_bit(flag
, &q
->queue_flags
);
508 static inline void queue_flag_clear_unlocked(unsigned int flag
,
509 struct request_queue
*q
)
511 __clear_bit(flag
, &q
->queue_flags
);
514 static inline void queue_flag_clear(unsigned int flag
, struct request_queue
*q
)
516 WARN_ON_ONCE(!queue_is_locked(q
));
517 __clear_bit(flag
, &q
->queue_flags
);
522 * Hardbarrier is supported with one of the following methods.
524 * NONE : hardbarrier unsupported
525 * DRAIN : ordering by draining is enough
526 * DRAIN_FLUSH : ordering by draining w/ pre and post flushes
527 * DRAIN_FUA : ordering by draining w/ pre flush and FUA write
528 * TAG : ordering by tag is enough
529 * TAG_FLUSH : ordering by tag w/ pre and post flushes
530 * TAG_FUA : ordering by tag w/ pre flush and FUA write
532 QUEUE_ORDERED_BY_DRAIN
= 0x01,
533 QUEUE_ORDERED_BY_TAG
= 0x02,
534 QUEUE_ORDERED_DO_PREFLUSH
= 0x10,
535 QUEUE_ORDERED_DO_BAR
= 0x20,
536 QUEUE_ORDERED_DO_POSTFLUSH
= 0x40,
537 QUEUE_ORDERED_DO_FUA
= 0x80,
539 QUEUE_ORDERED_NONE
= 0x00,
541 QUEUE_ORDERED_DRAIN
= QUEUE_ORDERED_BY_DRAIN
|
542 QUEUE_ORDERED_DO_BAR
,
543 QUEUE_ORDERED_DRAIN_FLUSH
= QUEUE_ORDERED_DRAIN
|
544 QUEUE_ORDERED_DO_PREFLUSH
|
545 QUEUE_ORDERED_DO_POSTFLUSH
,
546 QUEUE_ORDERED_DRAIN_FUA
= QUEUE_ORDERED_DRAIN
|
547 QUEUE_ORDERED_DO_PREFLUSH
|
548 QUEUE_ORDERED_DO_FUA
,
550 QUEUE_ORDERED_TAG
= QUEUE_ORDERED_BY_TAG
|
551 QUEUE_ORDERED_DO_BAR
,
552 QUEUE_ORDERED_TAG_FLUSH
= QUEUE_ORDERED_TAG
|
553 QUEUE_ORDERED_DO_PREFLUSH
|
554 QUEUE_ORDERED_DO_POSTFLUSH
,
555 QUEUE_ORDERED_TAG_FUA
= QUEUE_ORDERED_TAG
|
556 QUEUE_ORDERED_DO_PREFLUSH
|
557 QUEUE_ORDERED_DO_FUA
,
560 * Ordered operation sequence
562 QUEUE_ORDSEQ_STARTED
= 0x01, /* flushing in progress */
563 QUEUE_ORDSEQ_DRAIN
= 0x02, /* waiting for the queue to be drained */
564 QUEUE_ORDSEQ_PREFLUSH
= 0x04, /* pre-flushing in progress */
565 QUEUE_ORDSEQ_BAR
= 0x08, /* original barrier req in progress */
566 QUEUE_ORDSEQ_POSTFLUSH
= 0x10, /* post-flushing in progress */
567 QUEUE_ORDSEQ_DONE
= 0x20,
570 #define blk_queue_plugged(q) test_bit(QUEUE_FLAG_PLUGGED, &(q)->queue_flags)
571 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
572 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
573 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
574 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
575 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
576 #define blk_queue_flushing(q) ((q)->ordseq)
577 #define blk_queue_stackable(q) \
578 test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
580 #define blk_fs_request(rq) ((rq)->cmd_type == REQ_TYPE_FS)
581 #define blk_pc_request(rq) ((rq)->cmd_type == REQ_TYPE_BLOCK_PC)
582 #define blk_special_request(rq) ((rq)->cmd_type == REQ_TYPE_SPECIAL)
583 #define blk_sense_request(rq) ((rq)->cmd_type == REQ_TYPE_SENSE)
585 #define blk_failfast_dev(rq) ((rq)->cmd_flags & REQ_FAILFAST_DEV)
586 #define blk_failfast_transport(rq) ((rq)->cmd_flags & REQ_FAILFAST_TRANSPORT)
587 #define blk_failfast_driver(rq) ((rq)->cmd_flags & REQ_FAILFAST_DRIVER)
588 #define blk_noretry_request(rq) (blk_failfast_dev(rq) || \
589 blk_failfast_transport(rq) || \
590 blk_failfast_driver(rq))
591 #define blk_rq_started(rq) ((rq)->cmd_flags & REQ_STARTED)
592 #define blk_rq_io_stat(rq) ((rq)->cmd_flags & REQ_IO_STAT)
594 #define blk_account_rq(rq) (blk_rq_started(rq) && (blk_fs_request(rq) || blk_discard_rq(rq)))
596 #define blk_pm_suspend_request(rq) ((rq)->cmd_type == REQ_TYPE_PM_SUSPEND)
597 #define blk_pm_resume_request(rq) ((rq)->cmd_type == REQ_TYPE_PM_RESUME)
598 #define blk_pm_request(rq) \
599 (blk_pm_suspend_request(rq) || blk_pm_resume_request(rq))
601 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
602 #define blk_sorted_rq(rq) ((rq)->cmd_flags & REQ_SORTED)
603 #define blk_barrier_rq(rq) ((rq)->cmd_flags & REQ_HARDBARRIER)
604 #define blk_fua_rq(rq) ((rq)->cmd_flags & REQ_FUA)
605 #define blk_discard_rq(rq) ((rq)->cmd_flags & REQ_DISCARD)
606 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
607 /* rq->queuelist of dequeued request must be list_empty() */
608 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
610 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
612 #define rq_data_dir(rq) ((rq)->cmd_flags & 1)
615 * We regard a request as sync, if either a read or a sync write
617 static inline bool rw_is_sync(unsigned int rw_flags
)
619 return !(rw_flags
& REQ_RW
) || (rw_flags
& REQ_RW_SYNC
);
622 static inline bool rq_is_sync(struct request
*rq
)
624 return rw_is_sync(rq
->cmd_flags
);
627 #define rq_is_meta(rq) ((rq)->cmd_flags & REQ_RW_META)
628 #define rq_noidle(rq) ((rq)->cmd_flags & REQ_NOIDLE)
630 static inline int blk_queue_full(struct request_queue
*q
, int sync
)
633 return test_bit(QUEUE_FLAG_SYNCFULL
, &q
->queue_flags
);
634 return test_bit(QUEUE_FLAG_ASYNCFULL
, &q
->queue_flags
);
637 static inline void blk_set_queue_full(struct request_queue
*q
, int sync
)
640 queue_flag_set(QUEUE_FLAG_SYNCFULL
, q
);
642 queue_flag_set(QUEUE_FLAG_ASYNCFULL
, q
);
645 static inline void blk_clear_queue_full(struct request_queue
*q
, int sync
)
648 queue_flag_clear(QUEUE_FLAG_SYNCFULL
, q
);
650 queue_flag_clear(QUEUE_FLAG_ASYNCFULL
, q
);
655 * mergeable request must not have _NOMERGE or _BARRIER bit set, nor may
656 * it already be started by driver.
658 #define RQ_NOMERGE_FLAGS \
659 (REQ_NOMERGE | REQ_STARTED | REQ_HARDBARRIER | REQ_SOFTBARRIER)
660 #define rq_mergeable(rq) \
661 (!((rq)->cmd_flags & RQ_NOMERGE_FLAGS) && \
662 (blk_discard_rq(rq) || blk_fs_request((rq))))
665 * q->prep_rq_fn return values
667 #define BLKPREP_OK 0 /* serve it */
668 #define BLKPREP_KILL 1 /* fatal error, kill */
669 #define BLKPREP_DEFER 2 /* leave on queue */
671 extern unsigned long blk_max_low_pfn
, blk_max_pfn
;
674 * standard bounce addresses:
676 * BLK_BOUNCE_HIGH : bounce all highmem pages
677 * BLK_BOUNCE_ANY : don't bounce anything
678 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
681 #if BITS_PER_LONG == 32
682 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
684 #define BLK_BOUNCE_HIGH -1ULL
686 #define BLK_BOUNCE_ANY (-1ULL)
687 #define BLK_BOUNCE_ISA (ISA_DMA_THRESHOLD)
690 * default timeout for SG_IO if none specified
692 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
693 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
696 extern int init_emergency_isa_pool(void);
697 extern void blk_queue_bounce(struct request_queue
*q
, struct bio
**bio
);
699 static inline int init_emergency_isa_pool(void)
703 static inline void blk_queue_bounce(struct request_queue
*q
, struct bio
**bio
)
706 #endif /* CONFIG_MMU */
712 unsigned long offset
;
716 struct req_iterator
{
721 /* This should not be used directly - use rq_for_each_segment */
722 #define for_each_bio(_bio) \
723 for (; _bio; _bio = _bio->bi_next)
724 #define __rq_for_each_bio(_bio, rq) \
726 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
728 #define rq_for_each_segment(bvl, _rq, _iter) \
729 __rq_for_each_bio(_iter.bio, _rq) \
730 bio_for_each_segment(bvl, _iter.bio, _iter.i)
732 #define rq_iter_last(rq, _iter) \
733 (_iter.bio->bi_next == NULL && _iter.i == _iter.bio->bi_vcnt-1)
735 extern int blk_register_queue(struct gendisk
*disk
);
736 extern void blk_unregister_queue(struct gendisk
*disk
);
737 extern void register_disk(struct gendisk
*dev
);
738 extern void generic_make_request(struct bio
*bio
);
739 extern void blk_rq_init(struct request_queue
*q
, struct request
*rq
);
740 extern void blk_put_request(struct request
*);
741 extern void __blk_put_request(struct request_queue
*, struct request
*);
742 extern struct request
*blk_get_request(struct request_queue
*, int, gfp_t
);
743 extern void blk_insert_request(struct request_queue
*, struct request
*, int, void *);
744 extern void blk_requeue_request(struct request_queue
*, struct request
*);
745 extern int blk_rq_check_limits(struct request_queue
*q
, struct request
*rq
);
746 extern int blk_lld_busy(struct request_queue
*q
);
747 extern int blk_insert_cloned_request(struct request_queue
*q
,
749 extern void blk_plug_device(struct request_queue
*);
750 extern void blk_plug_device_unlocked(struct request_queue
*);
751 extern int blk_remove_plug(struct request_queue
*);
752 extern void blk_recount_segments(struct request_queue
*, struct bio
*);
753 extern int scsi_cmd_ioctl(struct request_queue
*, struct gendisk
*, fmode_t
,
754 unsigned int, void __user
*);
755 extern int sg_scsi_ioctl(struct request_queue
*, struct gendisk
*, fmode_t
,
756 struct scsi_ioctl_command __user
*);
759 * Temporary export, until SCSI gets fixed up.
761 extern int blk_rq_append_bio(struct request_queue
*q
, struct request
*rq
,
765 * A queue has just exitted congestion. Note this in the global counter of
766 * congested queues, and wake up anyone who was waiting for requests to be
769 static inline void blk_clear_queue_congested(struct request_queue
*q
, int rw
)
771 clear_bdi_congested(&q
->backing_dev_info
, rw
);
775 * A queue has just entered congestion. Flag that in the queue's VM-visible
776 * state flags and increment the global gounter of congested queues.
778 static inline void blk_set_queue_congested(struct request_queue
*q
, int rw
)
780 set_bdi_congested(&q
->backing_dev_info
, rw
);
783 extern void blk_start_queue(struct request_queue
*q
);
784 extern void blk_stop_queue(struct request_queue
*q
);
785 extern void blk_sync_queue(struct request_queue
*q
);
786 extern void __blk_stop_queue(struct request_queue
*q
);
787 extern void __blk_run_queue(struct request_queue
*);
788 extern void blk_run_queue(struct request_queue
*);
789 extern int blk_rq_map_user(struct request_queue
*, struct request
*,
790 struct rq_map_data
*, void __user
*, unsigned long,
792 extern int blk_rq_unmap_user(struct bio
*);
793 extern int blk_rq_map_kern(struct request_queue
*, struct request
*, void *, unsigned int, gfp_t
);
794 extern int blk_rq_map_user_iov(struct request_queue
*, struct request
*,
795 struct rq_map_data
*, struct sg_iovec
*, int,
796 unsigned int, gfp_t
);
797 extern int blk_execute_rq(struct request_queue
*, struct gendisk
*,
798 struct request
*, int);
799 extern void blk_execute_rq_nowait(struct request_queue
*, struct gendisk
*,
800 struct request
*, int, rq_end_io_fn
*);
801 extern void blk_unplug(struct request_queue
*q
);
803 static inline struct request_queue
*bdev_get_queue(struct block_device
*bdev
)
805 return bdev
->bd_disk
->queue
;
808 static inline void blk_run_backing_dev(struct backing_dev_info
*bdi
,
811 if (bdi
&& bdi
->unplug_io_fn
)
812 bdi
->unplug_io_fn(bdi
, page
);
815 static inline void blk_run_address_space(struct address_space
*mapping
)
818 blk_run_backing_dev(mapping
->backing_dev_info
, NULL
);
821 extern void blkdev_dequeue_request(struct request
*req
);
824 * blk_rq_pos() : the current sector
825 * blk_rq_bytes() : bytes left in the entire request
826 * blk_rq_cur_bytes() : bytes left in the current segment
827 * blk_rq_sectors() : sectors left in the entire request
828 * blk_rq_cur_sectors() : sectors left in the current segment
830 static inline sector_t
blk_rq_pos(const struct request
*rq
)
835 static inline unsigned int blk_rq_bytes(const struct request
*rq
)
837 return rq
->__data_len
;
840 static inline int blk_rq_cur_bytes(const struct request
*rq
)
842 return rq
->bio
? bio_cur_bytes(rq
->bio
) : 0;
845 static inline unsigned int blk_rq_sectors(const struct request
*rq
)
847 return blk_rq_bytes(rq
) >> 9;
850 static inline unsigned int blk_rq_cur_sectors(const struct request
*rq
)
852 return blk_rq_cur_bytes(rq
) >> 9;
856 * Request completion related functions.
858 * blk_update_request() completes given number of bytes and updates
859 * the request without completing it.
861 * blk_end_request() and friends. __blk_end_request() must be called
862 * with the request queue spinlock acquired.
864 * Several drivers define their own end_request and call
865 * blk_end_request() for parts of the original function.
866 * This prevents code duplication in drivers.
868 extern bool blk_update_request(struct request
*rq
, int error
,
869 unsigned int nr_bytes
);
870 extern bool blk_end_bidi_request(struct request
*rq
, int error
,
871 unsigned int nr_bytes
,
872 unsigned int bidi_bytes
);
873 extern bool __blk_end_bidi_request(struct request
*rq
, int error
,
874 unsigned int nr_bytes
,
875 unsigned int bidi_bytes
);
878 * blk_end_request - Helper function for drivers to complete the request.
879 * @rq: the request being processed
880 * @error: %0 for success, < %0 for error
881 * @nr_bytes: number of bytes to complete
884 * Ends I/O on a number of bytes attached to @rq.
885 * If @rq has leftover, sets it up for the next range of segments.
888 * %false - we are done with this request
889 * %true - still buffers pending for this request
891 static inline bool blk_end_request(struct request
*rq
, int error
,
892 unsigned int nr_bytes
)
894 return blk_end_bidi_request(rq
, error
, nr_bytes
, 0);
898 * blk_end_request_all - Helper function for drives to finish the request.
899 * @rq: the request to finish
900 * @err: %0 for success, < %0 for error
903 * Completely finish @rq.
905 static inline void blk_end_request_all(struct request
*rq
, int error
)
909 pending
= blk_end_request(rq
, error
, blk_rq_bytes(rq
));
914 * blk_end_request_cur - Helper function to finish the current request chunk.
915 * @rq: the request to finish the current chunk for
916 * @err: %0 for success, < %0 for error
919 * Complete the current consecutively mapped chunk from @rq.
922 * %false - we are done with this request
923 * %true - still buffers pending for this request
925 static inline bool blk_end_request_cur(struct request
*rq
, int error
)
927 return blk_end_request(rq
, error
, blk_rq_cur_bytes(rq
));
931 * __blk_end_request - Helper function for drivers to complete the request.
932 * @rq: the request being processed
933 * @error: %0 for success, < %0 for error
934 * @nr_bytes: number of bytes to complete
937 * Must be called with queue lock held unlike blk_end_request().
940 * %false - we are done with this request
941 * %true - still buffers pending for this request
943 static inline bool __blk_end_request(struct request
*rq
, int error
,
944 unsigned int nr_bytes
)
946 return __blk_end_bidi_request(rq
, error
, nr_bytes
, 0);
950 * __blk_end_request_all - Helper function for drives to finish the request.
951 * @rq: the request to finish
952 * @err: %0 for success, < %0 for error
955 * Completely finish @rq. Must be called with queue lock held.
957 static inline void __blk_end_request_all(struct request
*rq
, int error
)
961 pending
= __blk_end_request(rq
, error
, blk_rq_bytes(rq
));
966 * __blk_end_request_cur - Helper function to finish the current request chunk.
967 * @rq: the request to finish the current chunk for
968 * @err: %0 for success, < %0 for error
971 * Complete the current consecutively mapped chunk from @rq. Must
972 * be called with queue lock held.
975 * %false - we are done with this request
976 * %true - still buffers pending for this request
978 static inline bool __blk_end_request_cur(struct request
*rq
, int error
)
980 return __blk_end_request(rq
, error
, blk_rq_cur_bytes(rq
));
983 extern void blk_complete_request(struct request
*);
984 extern void __blk_complete_request(struct request
*);
985 extern void blk_abort_request(struct request
*);
986 extern void blk_abort_queue(struct request_queue
*);
989 * Access functions for manipulating queue properties
991 extern struct request_queue
*blk_init_queue_node(request_fn_proc
*rfn
,
992 spinlock_t
*lock
, int node_id
);
993 extern struct request_queue
*blk_init_queue(request_fn_proc
*, spinlock_t
*);
994 extern void blk_cleanup_queue(struct request_queue
*);
995 extern void blk_queue_make_request(struct request_queue
*, make_request_fn
*);
996 extern void blk_queue_bounce_limit(struct request_queue
*, u64
);
997 extern void blk_queue_max_sectors(struct request_queue
*, unsigned int);
998 extern void blk_queue_max_phys_segments(struct request_queue
*, unsigned short);
999 extern void blk_queue_max_hw_segments(struct request_queue
*, unsigned short);
1000 extern void blk_queue_max_segment_size(struct request_queue
*, unsigned int);
1001 extern void blk_queue_hardsect_size(struct request_queue
*, unsigned short);
1002 extern void blk_queue_stack_limits(struct request_queue
*t
, struct request_queue
*b
);
1003 extern void blk_queue_dma_pad(struct request_queue
*, unsigned int);
1004 extern void blk_queue_update_dma_pad(struct request_queue
*, unsigned int);
1005 extern int blk_queue_dma_drain(struct request_queue
*q
,
1006 dma_drain_needed_fn
*dma_drain_needed
,
1007 void *buf
, unsigned int size
);
1008 extern void blk_queue_lld_busy(struct request_queue
*q
, lld_busy_fn
*fn
);
1009 extern void blk_queue_segment_boundary(struct request_queue
*, unsigned long);
1010 extern void blk_queue_prep_rq(struct request_queue
*, prep_rq_fn
*pfn
);
1011 extern void blk_queue_merge_bvec(struct request_queue
*, merge_bvec_fn
*);
1012 extern void blk_queue_dma_alignment(struct request_queue
*, int);
1013 extern void blk_queue_update_dma_alignment(struct request_queue
*, int);
1014 extern void blk_queue_softirq_done(struct request_queue
*, softirq_done_fn
*);
1015 extern void blk_queue_set_discard(struct request_queue
*, prepare_discard_fn
*);
1016 extern void blk_queue_rq_timed_out(struct request_queue
*, rq_timed_out_fn
*);
1017 extern void blk_queue_rq_timeout(struct request_queue
*, unsigned int);
1018 extern struct backing_dev_info
*blk_get_backing_dev_info(struct block_device
*bdev
);
1019 extern int blk_queue_ordered(struct request_queue
*, unsigned, prepare_flush_fn
*);
1020 extern bool blk_do_ordered(struct request_queue
*, struct request
**);
1021 extern unsigned blk_ordered_cur_seq(struct request_queue
*);
1022 extern unsigned blk_ordered_req_seq(struct request
*);
1023 extern bool blk_ordered_complete_seq(struct request_queue
*, unsigned, int);
1025 extern int blk_rq_map_sg(struct request_queue
*, struct request
*, struct scatterlist
*);
1026 extern void blk_dump_rq_flags(struct request
*, char *);
1027 extern void generic_unplug_device(struct request_queue
*);
1028 extern long nr_blockdev_pages(void);
1030 int blk_get_queue(struct request_queue
*);
1031 struct request_queue
*blk_alloc_queue(gfp_t
);
1032 struct request_queue
*blk_alloc_queue_node(gfp_t
, int);
1033 extern void blk_put_queue(struct request_queue
*);
1038 #define blk_rq_tagged(rq) ((rq)->cmd_flags & REQ_QUEUED)
1039 extern int blk_queue_start_tag(struct request_queue
*, struct request
*);
1040 extern struct request
*blk_queue_find_tag(struct request_queue
*, int);
1041 extern void blk_queue_end_tag(struct request_queue
*, struct request
*);
1042 extern int blk_queue_init_tags(struct request_queue
*, int, struct blk_queue_tag
*);
1043 extern void blk_queue_free_tags(struct request_queue
*);
1044 extern int blk_queue_resize_tags(struct request_queue
*, int);
1045 extern void blk_queue_invalidate_tags(struct request_queue
*);
1046 extern struct blk_queue_tag
*blk_init_tags(int);
1047 extern void blk_free_tags(struct blk_queue_tag
*);
1049 static inline struct request
*blk_map_queue_find_tag(struct blk_queue_tag
*bqt
,
1052 if (unlikely(bqt
== NULL
|| tag
>= bqt
->real_max_depth
))
1054 return bqt
->tag_index
[tag
];
1057 extern int blkdev_issue_flush(struct block_device
*, sector_t
*);
1058 extern int blkdev_issue_discard(struct block_device
*,
1059 sector_t sector
, sector_t nr_sects
, gfp_t
);
1061 static inline int sb_issue_discard(struct super_block
*sb
,
1062 sector_t block
, sector_t nr_blocks
)
1064 block
<<= (sb
->s_blocksize_bits
- 9);
1065 nr_blocks
<<= (sb
->s_blocksize_bits
- 9);
1066 return blkdev_issue_discard(sb
->s_bdev
, block
, nr_blocks
, GFP_KERNEL
);
1070 * command filter functions
1072 extern int blk_verify_command(struct blk_cmd_filter
*filter
,
1073 unsigned char *cmd
, fmode_t has_write_perm
);
1074 extern void blk_unregister_filter(struct gendisk
*disk
);
1075 extern void blk_set_cmd_filter_defaults(struct blk_cmd_filter
*filter
);
1077 #define MAX_PHYS_SEGMENTS 128
1078 #define MAX_HW_SEGMENTS 128
1079 #define SAFE_MAX_SECTORS 255
1080 #define BLK_DEF_MAX_SECTORS 1024
1082 #define MAX_SEGMENT_SIZE 65536
1084 #define BLK_SEG_BOUNDARY_MASK 0xFFFFFFFFUL
1086 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1088 static inline int queue_hardsect_size(struct request_queue
*q
)
1092 if (q
&& q
->hardsect_size
)
1093 retval
= q
->hardsect_size
;
1098 static inline int bdev_hardsect_size(struct block_device
*bdev
)
1100 return queue_hardsect_size(bdev_get_queue(bdev
));
1103 static inline int queue_dma_alignment(struct request_queue
*q
)
1105 return q
? q
->dma_alignment
: 511;
1108 static inline int blk_rq_aligned(struct request_queue
*q
, void *addr
,
1111 unsigned int alignment
= queue_dma_alignment(q
) | q
->dma_pad_mask
;
1112 return !((unsigned long)addr
& alignment
) && !(len
& alignment
);
1115 /* assumes size > 256 */
1116 static inline unsigned int blksize_bits(unsigned int size
)
1118 unsigned int bits
= 8;
1122 } while (size
> 256);
1126 static inline unsigned int block_size(struct block_device
*bdev
)
1128 return bdev
->bd_block_size
;
1131 typedef struct {struct page
*v
;} Sector
;
1133 unsigned char *read_dev_sector(struct block_device
*, sector_t
, Sector
*);
1135 static inline void put_dev_sector(Sector p
)
1137 page_cache_release(p
.v
);
1141 int kblockd_schedule_work(struct request_queue
*q
, struct work_struct
*work
);
1143 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1144 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1145 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1146 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1148 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1150 #define INTEGRITY_FLAG_READ 2 /* verify data integrity on read */
1151 #define INTEGRITY_FLAG_WRITE 4 /* generate data integrity on write */
1153 struct blk_integrity_exchg
{
1157 unsigned int data_size
;
1158 unsigned short sector_size
;
1159 const char *disk_name
;
1162 typedef void (integrity_gen_fn
) (struct blk_integrity_exchg
*);
1163 typedef int (integrity_vrfy_fn
) (struct blk_integrity_exchg
*);
1164 typedef void (integrity_set_tag_fn
) (void *, void *, unsigned int);
1165 typedef void (integrity_get_tag_fn
) (void *, void *, unsigned int);
1167 struct blk_integrity
{
1168 integrity_gen_fn
*generate_fn
;
1169 integrity_vrfy_fn
*verify_fn
;
1170 integrity_set_tag_fn
*set_tag_fn
;
1171 integrity_get_tag_fn
*get_tag_fn
;
1173 unsigned short flags
;
1174 unsigned short tuple_size
;
1175 unsigned short sector_size
;
1176 unsigned short tag_size
;
1180 struct kobject kobj
;
1183 extern int blk_integrity_register(struct gendisk
*, struct blk_integrity
*);
1184 extern void blk_integrity_unregister(struct gendisk
*);
1185 extern int blk_integrity_compare(struct gendisk
*, struct gendisk
*);
1186 extern int blk_rq_map_integrity_sg(struct request
*, struct scatterlist
*);
1187 extern int blk_rq_count_integrity_sg(struct request
*);
1190 struct blk_integrity
*bdev_get_integrity(struct block_device
*bdev
)
1192 return bdev
->bd_disk
->integrity
;
1195 static inline struct blk_integrity
*blk_get_integrity(struct gendisk
*disk
)
1197 return disk
->integrity
;
1200 static inline int blk_integrity_rq(struct request
*rq
)
1202 if (rq
->bio
== NULL
)
1205 return bio_integrity(rq
->bio
);
1208 #else /* CONFIG_BLK_DEV_INTEGRITY */
1210 #define blk_integrity_rq(rq) (0)
1211 #define blk_rq_count_integrity_sg(a) (0)
1212 #define blk_rq_map_integrity_sg(a, b) (0)
1213 #define bdev_get_integrity(a) (0)
1214 #define blk_get_integrity(a) (0)
1215 #define blk_integrity_compare(a, b) (0)
1216 #define blk_integrity_register(a, b) (0)
1217 #define blk_integrity_unregister(a) do { } while (0);
1219 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1221 struct block_device_operations
{
1222 int (*open
) (struct block_device
*, fmode_t
);
1223 int (*release
) (struct gendisk
*, fmode_t
);
1224 int (*locked_ioctl
) (struct block_device
*, fmode_t
, unsigned, unsigned long);
1225 int (*ioctl
) (struct block_device
*, fmode_t
, unsigned, unsigned long);
1226 int (*compat_ioctl
) (struct block_device
*, fmode_t
, unsigned, unsigned long);
1227 int (*direct_access
) (struct block_device
*, sector_t
,
1228 void **, unsigned long *);
1229 int (*media_changed
) (struct gendisk
*);
1230 int (*revalidate_disk
) (struct gendisk
*);
1231 int (*getgeo
)(struct block_device
*, struct hd_geometry
*);
1232 struct module
*owner
;
1235 extern int __blkdev_driver_ioctl(struct block_device
*, fmode_t
, unsigned int,
1237 #else /* CONFIG_BLOCK */
1239 * stubs for when the block layer is configured out
1241 #define buffer_heads_over_limit 0
1243 static inline long nr_blockdev_pages(void)
1248 #endif /* CONFIG_BLOCK */