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,
74 * For request of type REQ_TYPE_LINUX_BLOCK, rq->cmd[0] is the opcode being
75 * sent down (similar to how REQ_TYPE_BLOCK_PC means that ->cmd[] holds a
78 * 0x00 -> 0x3f are driver private, to be used for whatever purpose they need,
79 * typically to differentiate REQ_TYPE_SPECIAL requests.
83 REQ_LB_OP_EJECT = 0x40, /* eject request */
84 REQ_LB_OP_FLUSH = 0x41, /* flush request */
88 * request type modified bits. first four bits match BIO_RW* bits, important
91 __REQ_RW, /* not set, read. set, write */
92 __REQ_FAILFAST_DEV, /* no driver retries of device errors */
93 __REQ_FAILFAST_TRANSPORT, /* no driver retries of transport errors */
94 __REQ_FAILFAST_DRIVER, /* no driver retries of driver errors */
95 /* above flags must match BIO_RW_* */
96 __REQ_DISCARD, /* request to discard sectors */
97 __REQ_SORTED, /* elevator knows about this request */
98 __REQ_SOFTBARRIER, /* may not be passed by ioscheduler */
99 __REQ_HARDBARRIER, /* may not be passed by drive either */
100 __REQ_FUA, /* forced unit access */
101 __REQ_NOMERGE, /* don't touch this for merging */
102 __REQ_STARTED, /* drive already may have started this one */
103 __REQ_DONTPREP, /* don't call prep for this one */
104 __REQ_QUEUED, /* uses queueing */
105 __REQ_ELVPRIV, /* elevator private data attached */
106 __REQ_FAILED, /* set if the request failed */
107 __REQ_QUIET, /* don't worry about errors */
108 __REQ_PREEMPT, /* set for "ide_preempt" requests */
109 __REQ_ORDERED_COLOR, /* is before or after barrier */
110 __REQ_RW_SYNC, /* request is sync (sync write or read) */
111 __REQ_ALLOCED, /* request came from our alloc pool */
112 __REQ_RW_META, /* metadata io request */
113 __REQ_COPY_USER, /* contains copies of user pages */
114 __REQ_INTEGRITY, /* integrity metadata has been remapped */
115 __REQ_NOIDLE, /* Don't anticipate more IO after this one */
116 __REQ_IO_STAT, /* account I/O stat */
117 __REQ_MIXED_MERGE, /* merge of different types, fail separately */
118 __REQ_NR_BITS, /* stops here */
121 #define REQ_RW (1 << __REQ_RW)
122 #define REQ_FAILFAST_DEV (1 << __REQ_FAILFAST_DEV)
123 #define REQ_FAILFAST_TRANSPORT (1 << __REQ_FAILFAST_TRANSPORT)
124 #define REQ_FAILFAST_DRIVER (1 << __REQ_FAILFAST_DRIVER)
125 #define REQ_DISCARD (1 << __REQ_DISCARD)
126 #define REQ_SORTED (1 << __REQ_SORTED)
127 #define REQ_SOFTBARRIER (1 << __REQ_SOFTBARRIER)
128 #define REQ_HARDBARRIER (1 << __REQ_HARDBARRIER)
129 #define REQ_FUA (1 << __REQ_FUA)
130 #define REQ_NOMERGE (1 << __REQ_NOMERGE)
131 #define REQ_STARTED (1 << __REQ_STARTED)
132 #define REQ_DONTPREP (1 << __REQ_DONTPREP)
133 #define REQ_QUEUED (1 << __REQ_QUEUED)
134 #define REQ_ELVPRIV (1 << __REQ_ELVPRIV)
135 #define REQ_FAILED (1 << __REQ_FAILED)
136 #define REQ_QUIET (1 << __REQ_QUIET)
137 #define REQ_PREEMPT (1 << __REQ_PREEMPT)
138 #define REQ_ORDERED_COLOR (1 << __REQ_ORDERED_COLOR)
139 #define REQ_RW_SYNC (1 << __REQ_RW_SYNC)
140 #define REQ_ALLOCED (1 << __REQ_ALLOCED)
141 #define REQ_RW_META (1 << __REQ_RW_META)
142 #define REQ_COPY_USER (1 << __REQ_COPY_USER)
143 #define REQ_INTEGRITY (1 << __REQ_INTEGRITY)
144 #define REQ_NOIDLE (1 << __REQ_NOIDLE)
145 #define REQ_IO_STAT (1 << __REQ_IO_STAT)
146 #define REQ_MIXED_MERGE (1 << __REQ_MIXED_MERGE)
148 #define REQ_FAILFAST_MASK (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT | \
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;
162 struct request_queue *q;
164 unsigned int cmd_flags;
165 enum rq_cmd_type_bits cmd_type;
166 unsigned long atomic_flags;
170 /* the following two fields are internal, NEVER access directly */
171 unsigned int __data_len; /* total data len */
172 sector_t __sector; /* sector cursor */
177 struct hlist_node hash; /* merge hash */
179 * The rb_node is only used inside the io scheduler, requests
180 * are pruned when moved to the dispatch queue. So let the
181 * completion_data share space with the rb_node.
184 struct rb_node rb_node; /* sort/lookup */
185 void *completion_data;
189 * two pointers are available for the IO schedulers, if they need
190 * more they have to dynamically allocate it.
192 void *elevator_private;
193 void *elevator_private2;
195 struct gendisk *rq_disk;
196 unsigned long start_time;
198 /* Number of scatter-gather DMA addr+len pairs after
199 * physical address coalescing is performed.
201 unsigned short nr_phys_segments;
203 unsigned short ioprio;
207 void *special; /* opaque pointer available for LLD use */
208 char *buffer; /* kaddr of the current segment if available */
214 * when request is used as a packet command carrier
216 unsigned char __cmd[BLK_MAX_CDB];
218 unsigned short cmd_len;
220 unsigned int extra_len; /* length of alignment and padding */
221 unsigned int sense_len;
222 unsigned int resid_len; /* residual count */
225 unsigned long deadline;
226 struct list_head timeout_list;
227 unsigned int timeout;
231 * completion callback.
233 rq_end_io_fn *end_io;
237 struct request *next_rq;
240 static inline unsigned short req_get_ioprio(struct request *req)
246 * State information carried for REQ_TYPE_PM_SUSPEND and REQ_TYPE_PM_RESUME
247 * requests. Some step values could eventually be made generic.
249 struct request_pm_state
251 /* PM state machine step value, currently driver specific */
253 /* requested PM state value (S1, S2, S3, S4, ...) */
255 void* data; /* for driver use */
258 #include <linux/elevator.h>
260 typedef void (request_fn_proc) (struct request_queue *q);
261 typedef int (make_request_fn) (struct request_queue *q, struct bio *bio);
262 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
263 typedef void (unplug_fn) (struct request_queue *);
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 queue_limits {
305 unsigned long bounce_pfn;
306 unsigned long seg_boundary_mask;
308 unsigned int max_hw_sectors;
309 unsigned int max_sectors;
310 unsigned int max_segment_size;
311 unsigned int physical_block_size;
312 unsigned int alignment_offset;
315 unsigned int max_discard_sectors;
316 unsigned int discard_granularity;
317 unsigned int discard_alignment;
319 unsigned short logical_block_size;
320 unsigned short max_segments;
322 unsigned char misaligned;
323 unsigned char discard_misaligned;
324 unsigned char no_cluster;
325 signed char discard_zeroes_data;
331 * Together with queue_head for cacheline sharing
333 struct list_head queue_head;
334 struct request *last_merge;
335 struct elevator_queue *elevator;
338 * the queue request freelist, one for reads and one for writes
340 struct request_list rq;
342 request_fn_proc *request_fn;
343 make_request_fn *make_request_fn;
344 prep_rq_fn *prep_rq_fn;
345 unplug_fn *unplug_fn;
346 merge_bvec_fn *merge_bvec_fn;
347 prepare_flush_fn *prepare_flush_fn;
348 softirq_done_fn *softirq_done_fn;
349 rq_timed_out_fn *rq_timed_out_fn;
350 dma_drain_needed_fn *dma_drain_needed;
351 lld_busy_fn *lld_busy_fn;
354 * Dispatch queue sorting
357 struct request *boundary_rq;
360 * Auto-unplugging state
362 struct timer_list unplug_timer;
363 int unplug_thresh; /* After this many requests */
364 unsigned long unplug_delay; /* After this many jiffies */
365 struct work_struct unplug_work;
367 struct backing_dev_info backing_dev_info;
370 * The queue owner gets to use this for whatever they like.
371 * ll_rw_blk doesn't touch it.
376 * queue needs bounce pages for pages above this limit
381 * various queue flags, see QUEUE_* below
383 unsigned long queue_flags;
386 * protects queue structures from reentrancy. ->__queue_lock should
387 * _never_ be used directly, it is queue private. always use
390 spinlock_t __queue_lock;
391 spinlock_t *queue_lock;
401 unsigned long nr_requests; /* Max # of requests */
402 unsigned int nr_congestion_on;
403 unsigned int nr_congestion_off;
404 unsigned int nr_batching;
406 void *dma_drain_buffer;
407 unsigned int dma_drain_size;
408 unsigned int dma_pad_mask;
409 unsigned int dma_alignment;
411 struct blk_queue_tag *queue_tags;
412 struct list_head tag_busy_list;
414 unsigned int nr_sorted;
415 unsigned int in_flight[2];
417 unsigned int rq_timeout;
418 struct timer_list timeout;
419 struct list_head timeout_list;
421 struct queue_limits limits;
426 unsigned int sg_timeout;
427 unsigned int sg_reserved_size;
429 #ifdef CONFIG_BLK_DEV_IO_TRACE
430 struct blk_trace *blk_trace;
433 * reserved for flush operations
435 unsigned int ordered, next_ordered, ordseq;
436 int orderr, ordcolor;
437 struct request pre_flush_rq, bar_rq, post_flush_rq;
438 struct request *orig_bar_rq;
440 struct mutex sysfs_lock;
442 #if defined(CONFIG_BLK_DEV_BSG)
443 struct bsg_class_device bsg_dev;
447 #define QUEUE_FLAG_CLUSTER 0 /* cluster several segments into 1 */
448 #define QUEUE_FLAG_QUEUED 1 /* uses generic tag queueing */
449 #define QUEUE_FLAG_STOPPED 2 /* queue is stopped */
450 #define QUEUE_FLAG_SYNCFULL 3 /* read queue has been filled */
451 #define QUEUE_FLAG_ASYNCFULL 4 /* write queue has been filled */
452 #define QUEUE_FLAG_DEAD 5 /* queue being torn down */
453 #define QUEUE_FLAG_REENTER 6 /* Re-entrancy avoidance */
454 #define QUEUE_FLAG_PLUGGED 7 /* queue is plugged */
455 #define QUEUE_FLAG_ELVSWITCH 8 /* don't use elevator, just do FIFO */
456 #define QUEUE_FLAG_BIDI 9 /* queue supports bidi requests */
457 #define QUEUE_FLAG_NOMERGES 10 /* disable merge attempts */
458 #define QUEUE_FLAG_SAME_COMP 11 /* force complete on same CPU */
459 #define QUEUE_FLAG_FAIL_IO 12 /* fake timeout */
460 #define QUEUE_FLAG_STACKABLE 13 /* supports request stacking */
461 #define QUEUE_FLAG_NONROT 14 /* non-rotational device (SSD) */
462 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
463 #define QUEUE_FLAG_IO_STAT 15 /* do IO stats */
464 #define QUEUE_FLAG_DISCARD 16 /* supports DISCARD */
465 #define QUEUE_FLAG_NOXMERGES 17 /* No extended merges */
467 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
468 (1 << QUEUE_FLAG_CLUSTER) | \
469 (1 << QUEUE_FLAG_STACKABLE) | \
470 (1 << QUEUE_FLAG_SAME_COMP))
472 static inline int queue_is_locked(struct request_queue *q)
475 spinlock_t *lock = q->queue_lock;
476 return lock && spin_is_locked(lock);
482 static inline void queue_flag_set_unlocked(unsigned int flag,
483 struct request_queue *q)
485 __set_bit(flag, &q->queue_flags);
488 static inline int queue_flag_test_and_clear(unsigned int flag,
489 struct request_queue *q)
491 WARN_ON_ONCE(!queue_is_locked(q));
493 if (test_bit(flag, &q->queue_flags)) {
494 __clear_bit(flag, &q->queue_flags);
501 static inline int queue_flag_test_and_set(unsigned int flag,
502 struct request_queue *q)
504 WARN_ON_ONCE(!queue_is_locked(q));
506 if (!test_bit(flag, &q->queue_flags)) {
507 __set_bit(flag, &q->queue_flags);
514 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
516 WARN_ON_ONCE(!queue_is_locked(q));
517 __set_bit(flag, &q->queue_flags);
520 static inline void queue_flag_clear_unlocked(unsigned int flag,
521 struct request_queue *q)
523 __clear_bit(flag, &q->queue_flags);
526 static inline int queue_in_flight(struct request_queue *q)
528 return q->in_flight[0] + q->in_flight[1];
531 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
533 WARN_ON_ONCE(!queue_is_locked(q));
534 __clear_bit(flag, &q->queue_flags);
539 * Hardbarrier is supported with one of the following methods.
541 * NONE : hardbarrier unsupported
542 * DRAIN : ordering by draining is enough
543 * DRAIN_FLUSH : ordering by draining w/ pre and post flushes
544 * DRAIN_FUA : ordering by draining w/ pre flush and FUA write
545 * TAG : ordering by tag is enough
546 * TAG_FLUSH : ordering by tag w/ pre and post flushes
547 * TAG_FUA : ordering by tag w/ pre flush and FUA write
549 QUEUE_ORDERED_BY_DRAIN = 0x01,
550 QUEUE_ORDERED_BY_TAG = 0x02,
551 QUEUE_ORDERED_DO_PREFLUSH = 0x10,
552 QUEUE_ORDERED_DO_BAR = 0x20,
553 QUEUE_ORDERED_DO_POSTFLUSH = 0x40,
554 QUEUE_ORDERED_DO_FUA = 0x80,
556 QUEUE_ORDERED_NONE = 0x00,
558 QUEUE_ORDERED_DRAIN = QUEUE_ORDERED_BY_DRAIN |
559 QUEUE_ORDERED_DO_BAR,
560 QUEUE_ORDERED_DRAIN_FLUSH = QUEUE_ORDERED_DRAIN |
561 QUEUE_ORDERED_DO_PREFLUSH |
562 QUEUE_ORDERED_DO_POSTFLUSH,
563 QUEUE_ORDERED_DRAIN_FUA = QUEUE_ORDERED_DRAIN |
564 QUEUE_ORDERED_DO_PREFLUSH |
565 QUEUE_ORDERED_DO_FUA,
567 QUEUE_ORDERED_TAG = QUEUE_ORDERED_BY_TAG |
568 QUEUE_ORDERED_DO_BAR,
569 QUEUE_ORDERED_TAG_FLUSH = QUEUE_ORDERED_TAG |
570 QUEUE_ORDERED_DO_PREFLUSH |
571 QUEUE_ORDERED_DO_POSTFLUSH,
572 QUEUE_ORDERED_TAG_FUA = QUEUE_ORDERED_TAG |
573 QUEUE_ORDERED_DO_PREFLUSH |
574 QUEUE_ORDERED_DO_FUA,
577 * Ordered operation sequence
579 QUEUE_ORDSEQ_STARTED = 0x01, /* flushing in progress */
580 QUEUE_ORDSEQ_DRAIN = 0x02, /* waiting for the queue to be drained */
581 QUEUE_ORDSEQ_PREFLUSH = 0x04, /* pre-flushing in progress */
582 QUEUE_ORDSEQ_BAR = 0x08, /* original barrier req in progress */
583 QUEUE_ORDSEQ_POSTFLUSH = 0x10, /* post-flushing in progress */
584 QUEUE_ORDSEQ_DONE = 0x20,
587 #define blk_queue_plugged(q) test_bit(QUEUE_FLAG_PLUGGED, &(q)->queue_flags)
588 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
589 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
590 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
591 #define blk_queue_noxmerges(q) \
592 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
593 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
594 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
595 #define blk_queue_flushing(q) ((q)->ordseq)
596 #define blk_queue_stackable(q) \
597 test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
598 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
600 #define blk_fs_request(rq) ((rq)->cmd_type == REQ_TYPE_FS)
601 #define blk_pc_request(rq) ((rq)->cmd_type == REQ_TYPE_BLOCK_PC)
602 #define blk_special_request(rq) ((rq)->cmd_type == REQ_TYPE_SPECIAL)
603 #define blk_sense_request(rq) ((rq)->cmd_type == REQ_TYPE_SENSE)
605 #define blk_failfast_dev(rq) ((rq)->cmd_flags & REQ_FAILFAST_DEV)
606 #define blk_failfast_transport(rq) ((rq)->cmd_flags & REQ_FAILFAST_TRANSPORT)
607 #define blk_failfast_driver(rq) ((rq)->cmd_flags & REQ_FAILFAST_DRIVER)
608 #define blk_noretry_request(rq) (blk_failfast_dev(rq) || \
609 blk_failfast_transport(rq) || \
610 blk_failfast_driver(rq))
611 #define blk_rq_started(rq) ((rq)->cmd_flags & REQ_STARTED)
612 #define blk_rq_io_stat(rq) ((rq)->cmd_flags & REQ_IO_STAT)
613 #define blk_rq_quiet(rq) ((rq)->cmd_flags & REQ_QUIET)
615 #define blk_account_rq(rq) (blk_rq_started(rq) && (blk_fs_request(rq) || blk_discard_rq(rq)))
617 #define blk_pm_suspend_request(rq) ((rq)->cmd_type == REQ_TYPE_PM_SUSPEND)
618 #define blk_pm_resume_request(rq) ((rq)->cmd_type == REQ_TYPE_PM_RESUME)
619 #define blk_pm_request(rq) \
620 (blk_pm_suspend_request(rq) || blk_pm_resume_request(rq))
622 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
623 #define blk_sorted_rq(rq) ((rq)->cmd_flags & REQ_SORTED)
624 #define blk_barrier_rq(rq) ((rq)->cmd_flags & REQ_HARDBARRIER)
625 #define blk_fua_rq(rq) ((rq)->cmd_flags & REQ_FUA)
626 #define blk_discard_rq(rq) ((rq)->cmd_flags & REQ_DISCARD)
627 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
628 /* rq->queuelist of dequeued request must be list_empty() */
629 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
631 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
633 #define rq_data_dir(rq) ((rq)->cmd_flags & 1)
636 * We regard a request as sync, if either a read or a sync write
638 static inline bool rw_is_sync(unsigned int rw_flags)
640 return !(rw_flags & REQ_RW) || (rw_flags & REQ_RW_SYNC);
643 static inline bool rq_is_sync(struct request *rq)
645 return rw_is_sync(rq->cmd_flags);
648 #define rq_is_meta(rq) ((rq)->cmd_flags & REQ_RW_META)
649 #define rq_noidle(rq) ((rq)->cmd_flags & REQ_NOIDLE)
651 static inline int blk_queue_full(struct request_queue *q, int sync)
654 return test_bit(QUEUE_FLAG_SYNCFULL, &q->queue_flags);
655 return test_bit(QUEUE_FLAG_ASYNCFULL, &q->queue_flags);
658 static inline void blk_set_queue_full(struct request_queue *q, int sync)
661 queue_flag_set(QUEUE_FLAG_SYNCFULL, q);
663 queue_flag_set(QUEUE_FLAG_ASYNCFULL, q);
666 static inline void blk_clear_queue_full(struct request_queue *q, int sync)
669 queue_flag_clear(QUEUE_FLAG_SYNCFULL, q);
671 queue_flag_clear(QUEUE_FLAG_ASYNCFULL, q);
676 * mergeable request must not have _NOMERGE or _BARRIER bit set, nor may
677 * it already be started by driver.
679 #define RQ_NOMERGE_FLAGS \
680 (REQ_NOMERGE | REQ_STARTED | REQ_HARDBARRIER | REQ_SOFTBARRIER)
681 #define rq_mergeable(rq) \
682 (!((rq)->cmd_flags & RQ_NOMERGE_FLAGS) && \
683 (blk_discard_rq(rq) || blk_fs_request((rq))))
686 * q->prep_rq_fn return values
688 #define BLKPREP_OK 0 /* serve it */
689 #define BLKPREP_KILL 1 /* fatal error, kill */
690 #define BLKPREP_DEFER 2 /* leave on queue */
692 extern unsigned long blk_max_low_pfn, blk_max_pfn;
695 * standard bounce addresses:
697 * BLK_BOUNCE_HIGH : bounce all highmem pages
698 * BLK_BOUNCE_ANY : don't bounce anything
699 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
702 #if BITS_PER_LONG == 32
703 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
705 #define BLK_BOUNCE_HIGH -1ULL
707 #define BLK_BOUNCE_ANY (-1ULL)
708 #define BLK_BOUNCE_ISA (ISA_DMA_THRESHOLD)
711 * default timeout for SG_IO if none specified
713 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
714 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
717 extern int init_emergency_isa_pool(void);
718 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
720 static inline int init_emergency_isa_pool(void)
724 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
727 #endif /* CONFIG_MMU */
733 unsigned long offset;
738 struct req_iterator {
743 /* This should not be used directly - use rq_for_each_segment */
744 #define for_each_bio(_bio) \
745 for (; _bio; _bio = _bio->bi_next)
746 #define __rq_for_each_bio(_bio, rq) \
748 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
750 #define rq_for_each_segment(bvl, _rq, _iter) \
751 __rq_for_each_bio(_iter.bio, _rq) \
752 bio_for_each_segment(bvl, _iter.bio, _iter.i)
754 #define rq_iter_last(rq, _iter) \
755 (_iter.bio->bi_next == NULL && _iter.i == _iter.bio->bi_vcnt-1)
757 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
758 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
760 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
761 extern void rq_flush_dcache_pages(struct request *rq);
763 static inline void rq_flush_dcache_pages(struct request *rq)
768 extern int blk_register_queue(struct gendisk *disk);
769 extern void blk_unregister_queue(struct gendisk *disk);
770 extern void register_disk(struct gendisk *dev);
771 extern void generic_make_request(struct bio *bio);
772 extern void blk_rq_init(struct request_queue *q, struct request *rq);
773 extern void blk_put_request(struct request *);
774 extern void __blk_put_request(struct request_queue *, struct request *);
775 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
776 extern struct request *blk_make_request(struct request_queue *, struct bio *,
778 extern void blk_insert_request(struct request_queue *, struct request *, int, void *);
779 extern void blk_requeue_request(struct request_queue *, struct request *);
780 extern int blk_rq_check_limits(struct request_queue *q, struct request *rq);
781 extern int blk_lld_busy(struct request_queue *q);
782 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
783 struct bio_set *bs, gfp_t gfp_mask,
784 int (*bio_ctr)(struct bio *, struct bio *, void *),
786 extern void blk_rq_unprep_clone(struct request *rq);
787 extern int blk_insert_cloned_request(struct request_queue *q,
789 extern void blk_plug_device(struct request_queue *);
790 extern void blk_plug_device_unlocked(struct request_queue *);
791 extern int blk_remove_plug(struct request_queue *);
792 extern void blk_recount_segments(struct request_queue *, struct bio *);
793 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
794 unsigned int, void __user *);
795 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
796 struct scsi_ioctl_command __user *);
799 * A queue has just exitted congestion. Note this in the global counter of
800 * congested queues, and wake up anyone who was waiting for requests to be
803 static inline void blk_clear_queue_congested(struct request_queue *q, int sync)
805 clear_bdi_congested(&q->backing_dev_info, sync);
809 * A queue has just entered congestion. Flag that in the queue's VM-visible
810 * state flags and increment the global gounter of congested queues.
812 static inline void blk_set_queue_congested(struct request_queue *q, int sync)
814 set_bdi_congested(&q->backing_dev_info, sync);
817 extern void blk_start_queue(struct request_queue *q);
818 extern void blk_stop_queue(struct request_queue *q);
819 extern void blk_sync_queue(struct request_queue *q);
820 extern void __blk_stop_queue(struct request_queue *q);
821 extern void __blk_run_queue(struct request_queue *);
822 extern void blk_run_queue(struct request_queue *);
823 extern int blk_rq_map_user(struct request_queue *, struct request *,
824 struct rq_map_data *, void __user *, unsigned long,
826 extern int blk_rq_unmap_user(struct bio *);
827 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
828 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
829 struct rq_map_data *, struct sg_iovec *, int,
830 unsigned int, gfp_t);
831 extern int blk_execute_rq(struct request_queue *, struct gendisk *,
832 struct request *, int);
833 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
834 struct request *, int, rq_end_io_fn *);
835 extern void blk_unplug(struct request_queue *q);
837 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
839 return bdev->bd_disk->queue;
843 * blk_rq_pos() : the current sector
844 * blk_rq_bytes() : bytes left in the entire request
845 * blk_rq_cur_bytes() : bytes left in the current segment
846 * blk_rq_err_bytes() : bytes left till the next error boundary
847 * blk_rq_sectors() : sectors left in the entire request
848 * blk_rq_cur_sectors() : sectors left in the current segment
850 static inline sector_t blk_rq_pos(const struct request *rq)
855 static inline unsigned int blk_rq_bytes(const struct request *rq)
857 return rq->__data_len;
860 static inline int blk_rq_cur_bytes(const struct request *rq)
862 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
865 extern unsigned int blk_rq_err_bytes(const struct request *rq);
867 static inline unsigned int blk_rq_sectors(const struct request *rq)
869 return blk_rq_bytes(rq) >> 9;
872 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
874 return blk_rq_cur_bytes(rq) >> 9;
878 * Request issue related functions.
880 extern struct request *blk_peek_request(struct request_queue *q);
881 extern void blk_start_request(struct request *rq);
882 extern struct request *blk_fetch_request(struct request_queue *q);
885 * Request completion related functions.
887 * blk_update_request() completes given number of bytes and updates
888 * the request without completing it.
890 * blk_end_request() and friends. __blk_end_request() must be called
891 * with the request queue spinlock acquired.
893 * Several drivers define their own end_request and call
894 * blk_end_request() for parts of the original function.
895 * This prevents code duplication in drivers.
897 extern bool blk_update_request(struct request *rq, int error,
898 unsigned int nr_bytes);
899 extern bool blk_end_request(struct request *rq, int error,
900 unsigned int nr_bytes);
901 extern void blk_end_request_all(struct request *rq, int error);
902 extern bool blk_end_request_cur(struct request *rq, int error);
903 extern bool blk_end_request_err(struct request *rq, int error);
904 extern bool __blk_end_request(struct request *rq, int error,
905 unsigned int nr_bytes);
906 extern void __blk_end_request_all(struct request *rq, int error);
907 extern bool __blk_end_request_cur(struct request *rq, int error);
908 extern bool __blk_end_request_err(struct request *rq, int error);
910 extern void blk_complete_request(struct request *);
911 extern void __blk_complete_request(struct request *);
912 extern void blk_abort_request(struct request *);
913 extern void blk_abort_queue(struct request_queue *);
916 * Access functions for manipulating queue properties
918 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
919 spinlock_t *lock, int node_id);
920 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
921 extern void blk_cleanup_queue(struct request_queue *);
922 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
923 extern void blk_queue_bounce_limit(struct request_queue *, u64);
924 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
925 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
926 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
927 extern void blk_queue_max_discard_sectors(struct request_queue *q,
928 unsigned int max_discard_sectors);
929 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
930 extern void blk_queue_physical_block_size(struct request_queue *, unsigned short);
931 extern void blk_queue_alignment_offset(struct request_queue *q,
932 unsigned int alignment);
933 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
934 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
935 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
936 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
937 extern void blk_set_default_limits(struct queue_limits *lim);
938 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
940 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
942 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
944 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
945 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
946 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
947 extern int blk_queue_dma_drain(struct request_queue *q,
948 dma_drain_needed_fn *dma_drain_needed,
949 void *buf, unsigned int size);
950 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
951 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
952 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
953 extern void blk_queue_merge_bvec(struct request_queue *, merge_bvec_fn *);
954 extern void blk_queue_dma_alignment(struct request_queue *, int);
955 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
956 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
957 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
958 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
959 extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
960 extern int blk_queue_ordered(struct request_queue *, unsigned, prepare_flush_fn *);
961 extern bool blk_do_ordered(struct request_queue *, struct request **);
962 extern unsigned blk_ordered_cur_seq(struct request_queue *);
963 extern unsigned blk_ordered_req_seq(struct request *);
964 extern bool blk_ordered_complete_seq(struct request_queue *, unsigned, int);
966 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
967 extern void blk_dump_rq_flags(struct request *, char *);
968 extern void generic_unplug_device(struct request_queue *);
969 extern long nr_blockdev_pages(void);
971 int blk_get_queue(struct request_queue *);
972 struct request_queue *blk_alloc_queue(gfp_t);
973 struct request_queue *blk_alloc_queue_node(gfp_t, int);
974 extern void blk_put_queue(struct request_queue *);
979 #define blk_rq_tagged(rq) ((rq)->cmd_flags & REQ_QUEUED)
980 extern int blk_queue_start_tag(struct request_queue *, struct request *);
981 extern struct request *blk_queue_find_tag(struct request_queue *, int);
982 extern void blk_queue_end_tag(struct request_queue *, struct request *);
983 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *);
984 extern void blk_queue_free_tags(struct request_queue *);
985 extern int blk_queue_resize_tags(struct request_queue *, int);
986 extern void blk_queue_invalidate_tags(struct request_queue *);
987 extern struct blk_queue_tag *blk_init_tags(int);
988 extern void blk_free_tags(struct blk_queue_tag *);
990 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
993 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
995 return bqt->tag_index[tag];
998 extern int blkdev_issue_flush(struct block_device *, sector_t *);
999 #define DISCARD_FL_WAIT 0x01 /* wait for completion */
1000 #define DISCARD_FL_BARRIER 0x02 /* issue DISCARD_BARRIER request */
1001 extern int blkdev_issue_discard(struct block_device *, sector_t sector,
1002 sector_t nr_sects, gfp_t, int flags);
1004 static inline int sb_issue_discard(struct super_block *sb,
1005 sector_t block, sector_t nr_blocks)
1007 block <<= (sb->s_blocksize_bits - 9);
1008 nr_blocks <<= (sb->s_blocksize_bits - 9);
1009 return blkdev_issue_discard(sb->s_bdev, block, nr_blocks, GFP_KERNEL,
1010 DISCARD_FL_BARRIER);
1013 extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm);
1015 enum blk_default_limits {
1016 BLK_MAX_SEGMENTS = 128,
1017 BLK_SAFE_MAX_SECTORS = 255,
1018 BLK_DEF_MAX_SECTORS = 1024,
1019 BLK_MAX_SEGMENT_SIZE = 65536,
1020 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1023 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1025 static inline unsigned long queue_bounce_pfn(struct request_queue *q)
1027 return q->limits.bounce_pfn;
1030 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1032 return q->limits.seg_boundary_mask;
1035 static inline unsigned int queue_max_sectors(struct request_queue *q)
1037 return q->limits.max_sectors;
1040 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1042 return q->limits.max_hw_sectors;
1045 static inline unsigned short queue_max_segments(struct request_queue *q)
1047 return q->limits.max_segments;
1050 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1052 return q->limits.max_segment_size;
1055 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1059 if (q && q->limits.logical_block_size)
1060 retval = q->limits.logical_block_size;
1065 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1067 return queue_logical_block_size(bdev_get_queue(bdev));
1070 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1072 return q->limits.physical_block_size;
1075 static inline int bdev_physical_block_size(struct block_device *bdev)
1077 return queue_physical_block_size(bdev_get_queue(bdev));
1080 static inline unsigned int queue_io_min(struct request_queue *q)
1082 return q->limits.io_min;
1085 static inline int bdev_io_min(struct block_device *bdev)
1087 return queue_io_min(bdev_get_queue(bdev));
1090 static inline unsigned int queue_io_opt(struct request_queue *q)
1092 return q->limits.io_opt;
1095 static inline int bdev_io_opt(struct block_device *bdev)
1097 return queue_io_opt(bdev_get_queue(bdev));
1100 static inline int queue_alignment_offset(struct request_queue *q)
1102 if (q->limits.misaligned)
1105 return q->limits.alignment_offset;
1108 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1110 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1111 unsigned int alignment = (sector << 9) & (granularity - 1);
1113 return (granularity + lim->alignment_offset - alignment)
1114 & (granularity - 1);
1117 static inline int bdev_alignment_offset(struct block_device *bdev)
1119 struct request_queue *q = bdev_get_queue(bdev);
1121 if (q->limits.misaligned)
1124 if (bdev != bdev->bd_contains)
1125 return bdev->bd_part->alignment_offset;
1127 return q->limits.alignment_offset;
1130 static inline int queue_discard_alignment(struct request_queue *q)
1132 if (q->limits.discard_misaligned)
1135 return q->limits.discard_alignment;
1138 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1140 unsigned int alignment = (sector << 9) & (lim->discard_granularity - 1);
1142 return (lim->discard_granularity + lim->discard_alignment - alignment)
1143 & (lim->discard_granularity - 1);
1146 static inline unsigned int queue_discard_zeroes_data(struct request_queue *q)
1148 if (q->limits.discard_zeroes_data == 1)
1154 static inline unsigned int bdev_discard_zeroes_data(struct block_device *bdev)
1156 return queue_discard_zeroes_data(bdev_get_queue(bdev));
1159 static inline int queue_dma_alignment(struct request_queue *q)
1161 return q ? q->dma_alignment : 511;
1164 static inline int blk_rq_aligned(struct request_queue *q, void *addr,
1167 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1168 return !((unsigned long)addr & alignment) && !(len & alignment);
1171 /* assumes size > 256 */
1172 static inline unsigned int blksize_bits(unsigned int size)
1174 unsigned int bits = 8;
1178 } while (size > 256);
1182 static inline unsigned int block_size(struct block_device *bdev)
1184 return bdev->bd_block_size;
1187 typedef struct {struct page *v;} Sector;
1189 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1191 static inline void put_dev_sector(Sector p)
1193 page_cache_release(p.v);
1197 int kblockd_schedule_work(struct request_queue *q, struct work_struct *work);
1199 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1200 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1201 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1202 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1204 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1206 #define INTEGRITY_FLAG_READ 2 /* verify data integrity on read */
1207 #define INTEGRITY_FLAG_WRITE 4 /* generate data integrity on write */
1209 struct blk_integrity_exchg {
1213 unsigned int data_size;
1214 unsigned short sector_size;
1215 const char *disk_name;
1218 typedef void (integrity_gen_fn) (struct blk_integrity_exchg *);
1219 typedef int (integrity_vrfy_fn) (struct blk_integrity_exchg *);
1220 typedef void (integrity_set_tag_fn) (void *, void *, unsigned int);
1221 typedef void (integrity_get_tag_fn) (void *, void *, unsigned int);
1223 struct blk_integrity {
1224 integrity_gen_fn *generate_fn;
1225 integrity_vrfy_fn *verify_fn;
1226 integrity_set_tag_fn *set_tag_fn;
1227 integrity_get_tag_fn *get_tag_fn;
1229 unsigned short flags;
1230 unsigned short tuple_size;
1231 unsigned short sector_size;
1232 unsigned short tag_size;
1236 struct kobject kobj;
1239 extern int blk_integrity_register(struct gendisk *, struct blk_integrity *);
1240 extern void blk_integrity_unregister(struct gendisk *);
1241 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1242 extern int blk_rq_map_integrity_sg(struct request *, struct scatterlist *);
1243 extern int blk_rq_count_integrity_sg(struct request *);
1246 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1248 return bdev->bd_disk->integrity;
1251 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1253 return disk->integrity;
1256 static inline int blk_integrity_rq(struct request *rq)
1258 if (rq->bio == NULL)
1261 return bio_integrity(rq->bio);
1264 #else /* CONFIG_BLK_DEV_INTEGRITY */
1266 #define blk_integrity_rq(rq) (0)
1267 #define blk_rq_count_integrity_sg(a) (0)
1268 #define blk_rq_map_integrity_sg(a, b) (0)
1269 #define bdev_get_integrity(a) (0)
1270 #define blk_get_integrity(a) (0)
1271 #define blk_integrity_compare(a, b) (0)
1272 #define blk_integrity_register(a, b) (0)
1273 #define blk_integrity_unregister(a) do { } while (0);
1275 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1277 struct block_device_operations {
1278 int (*open) (struct block_device *, fmode_t);
1279 int (*release) (struct gendisk *, fmode_t);
1280 int (*locked_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1281 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1282 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1283 int (*direct_access) (struct block_device *, sector_t,
1284 void **, unsigned long *);
1285 int (*media_changed) (struct gendisk *);
1286 unsigned long long (*set_capacity) (struct gendisk *,
1287 unsigned long long);
1288 int (*revalidate_disk) (struct gendisk *);
1289 int (*getgeo)(struct block_device *, struct hd_geometry *);
1290 struct module *owner;
1293 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1295 #else /* CONFIG_BLOCK */
1297 * stubs for when the block layer is configured out
1299 #define buffer_heads_over_limit 0
1301 static inline long nr_blockdev_pages(void)
1306 #endif /* CONFIG_BLOCK */