block: use normal I/O path for discard requests
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / include / linux / blkdev.h
blobf62d45e87618d1a4c389b1e3229fd90a42202bb4
1 #ifndef _LINUX_BLKDEV_H
2 #define _LINUX_BLKDEV_H
4 #ifdef CONFIG_BLOCK
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;
27 struct request_queue;
28 struct elevator_queue;
29 struct request_pm_state;
30 struct blk_trace;
31 struct request;
32 struct sg_io_hdr;
34 #define BLKDEV_MIN_RQ 4
35 #define BLKDEV_MAX_RQ 128 /* Default maximum */
37 struct request;
38 typedef void (rq_end_io_fn)(struct request *, int);
40 struct request_list {
42 * count[], starved[], and wait[] are indexed by
43 * BLK_RW_SYNC/BLK_RW_ASYNC
45 int count[2];
46 int starved[2];
47 int elvpriv;
48 mempool_t *rq_pool;
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,
70 REQ_TYPE_ATA_PC,
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
76 * SCSI cdb.
78 * 0x00 -> 0x3f are driver private, to be used for whatever purpose they need,
79 * typically to differentiate REQ_TYPE_SPECIAL requests.
82 enum {
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
90 enum rq_flag_bits {
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 | \
149 REQ_FAILFAST_DRIVER)
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()
156 * as well!
158 struct request {
159 struct list_head queuelist;
160 struct call_single_data csd;
161 int cpu;
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 */
173 struct bio *bio;
174 struct bio *biotail;
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.
182 union {
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 */
207 int tag;
208 int errors;
210 int ref_count;
213 * when request is used as a packet command carrier
215 unsigned short cmd_len;
216 unsigned char __cmd[BLK_MAX_CDB];
217 unsigned char *cmd;
219 unsigned int extra_len; /* length of alignment and padding */
220 unsigned int sense_len;
221 unsigned int resid_len; /* residual count */
222 void *sense;
224 unsigned long deadline;
225 struct list_head timeout_list;
226 unsigned int timeout;
227 int retries;
230 * completion callback.
232 rq_end_io_fn *end_io;
233 void *end_io_data;
235 /* for bidi */
236 struct request *next_rq;
239 static inline unsigned short req_get_ioprio(struct request *req)
241 return req->ioprio;
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 */
251 int pm_step;
252 /* requested PM state value (S1, S2, S3, S4, ...) */
253 u32 pm_state;
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 *);
264 struct bio_vec;
265 struct bvec_merge_data {
266 struct block_device *bi_bdev;
267 sector_t bi_sector;
268 unsigned bi_size;
269 unsigned long bi_rw;
271 typedef int (merge_bvec_fn) (struct request_queue *, struct bvec_merge_data *,
272 struct bio_vec *);
273 typedef void (prepare_flush_fn) (struct request_queue *, struct request *);
274 typedef void (softirq_done_fn)(struct request *);
275 typedef int (dma_drain_needed_fn)(struct request *);
276 typedef int (lld_busy_fn) (struct request_queue *q);
278 enum blk_eh_timer_return {
279 BLK_EH_NOT_HANDLED,
280 BLK_EH_HANDLED,
281 BLK_EH_RESET_TIMER,
284 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
286 enum blk_queue_state {
287 Queue_down,
288 Queue_up,
291 struct blk_queue_tag {
292 struct request **tag_index; /* map of busy tags */
293 unsigned long *tag_map; /* bit map of free/busy tags */
294 int busy; /* current depth */
295 int max_depth; /* what we will send to device */
296 int real_max_depth; /* what the array can hold */
297 atomic_t refcnt; /* map can be shared */
300 #define BLK_SCSI_MAX_CMDS (256)
301 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
303 struct queue_limits {
304 unsigned long bounce_pfn;
305 unsigned long seg_boundary_mask;
307 unsigned int max_hw_sectors;
308 unsigned int max_sectors;
309 unsigned int max_segment_size;
310 unsigned int physical_block_size;
311 unsigned int alignment_offset;
312 unsigned int io_min;
313 unsigned int io_opt;
315 unsigned short logical_block_size;
316 unsigned short max_hw_segments;
317 unsigned short max_phys_segments;
319 unsigned char misaligned;
320 unsigned char no_cluster;
323 struct request_queue
326 * Together with queue_head for cacheline sharing
328 struct list_head queue_head;
329 struct request *last_merge;
330 struct elevator_queue *elevator;
333 * the queue request freelist, one for reads and one for writes
335 struct request_list rq;
337 request_fn_proc *request_fn;
338 make_request_fn *make_request_fn;
339 prep_rq_fn *prep_rq_fn;
340 unplug_fn *unplug_fn;
341 merge_bvec_fn *merge_bvec_fn;
342 prepare_flush_fn *prepare_flush_fn;
343 softirq_done_fn *softirq_done_fn;
344 rq_timed_out_fn *rq_timed_out_fn;
345 dma_drain_needed_fn *dma_drain_needed;
346 lld_busy_fn *lld_busy_fn;
349 * Dispatch queue sorting
351 sector_t end_sector;
352 struct request *boundary_rq;
355 * Auto-unplugging state
357 struct timer_list unplug_timer;
358 int unplug_thresh; /* After this many requests */
359 unsigned long unplug_delay; /* After this many jiffies */
360 struct work_struct unplug_work;
362 struct backing_dev_info backing_dev_info;
365 * The queue owner gets to use this for whatever they like.
366 * ll_rw_blk doesn't touch it.
368 void *queuedata;
371 * queue needs bounce pages for pages above this limit
373 gfp_t bounce_gfp;
376 * various queue flags, see QUEUE_* below
378 unsigned long queue_flags;
381 * protects queue structures from reentrancy. ->__queue_lock should
382 * _never_ be used directly, it is queue private. always use
383 * ->queue_lock.
385 spinlock_t __queue_lock;
386 spinlock_t *queue_lock;
389 * queue kobject
391 struct kobject kobj;
394 * queue settings
396 unsigned long nr_requests; /* Max # of requests */
397 unsigned int nr_congestion_on;
398 unsigned int nr_congestion_off;
399 unsigned int nr_batching;
401 void *dma_drain_buffer;
402 unsigned int dma_drain_size;
403 unsigned int dma_pad_mask;
404 unsigned int dma_alignment;
406 struct blk_queue_tag *queue_tags;
407 struct list_head tag_busy_list;
409 unsigned int nr_sorted;
410 unsigned int in_flight[2];
412 unsigned int rq_timeout;
413 struct timer_list timeout;
414 struct list_head timeout_list;
416 struct queue_limits limits;
419 * sg stuff
421 unsigned int sg_timeout;
422 unsigned int sg_reserved_size;
423 int node;
424 #ifdef CONFIG_BLK_DEV_IO_TRACE
425 struct blk_trace *blk_trace;
426 #endif
428 * reserved for flush operations
430 unsigned int ordered, next_ordered, ordseq;
431 int orderr, ordcolor;
432 struct request pre_flush_rq, bar_rq, post_flush_rq;
433 struct request *orig_bar_rq;
435 struct mutex sysfs_lock;
437 #if defined(CONFIG_BLK_DEV_BSG)
438 struct bsg_class_device bsg_dev;
439 #endif
442 #define QUEUE_FLAG_CLUSTER 0 /* cluster several segments into 1 */
443 #define QUEUE_FLAG_QUEUED 1 /* uses generic tag queueing */
444 #define QUEUE_FLAG_STOPPED 2 /* queue is stopped */
445 #define QUEUE_FLAG_SYNCFULL 3 /* read queue has been filled */
446 #define QUEUE_FLAG_ASYNCFULL 4 /* write queue has been filled */
447 #define QUEUE_FLAG_DEAD 5 /* queue being torn down */
448 #define QUEUE_FLAG_REENTER 6 /* Re-entrancy avoidance */
449 #define QUEUE_FLAG_PLUGGED 7 /* queue is plugged */
450 #define QUEUE_FLAG_ELVSWITCH 8 /* don't use elevator, just do FIFO */
451 #define QUEUE_FLAG_BIDI 9 /* queue supports bidi requests */
452 #define QUEUE_FLAG_NOMERGES 10 /* disable merge attempts */
453 #define QUEUE_FLAG_SAME_COMP 11 /* force complete on same CPU */
454 #define QUEUE_FLAG_FAIL_IO 12 /* fake timeout */
455 #define QUEUE_FLAG_STACKABLE 13 /* supports request stacking */
456 #define QUEUE_FLAG_NONROT 14 /* non-rotational device (SSD) */
457 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
458 #define QUEUE_FLAG_IO_STAT 15 /* do IO stats */
459 #define QUEUE_FLAG_CQ 16 /* hardware does queuing */
460 #define QUEUE_FLAG_DISCARD 17 /* supports DISCARD */
462 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
463 (1 << QUEUE_FLAG_CLUSTER) | \
464 (1 << QUEUE_FLAG_STACKABLE) | \
465 (1 << QUEUE_FLAG_SAME_COMP))
467 static inline int queue_is_locked(struct request_queue *q)
469 #ifdef CONFIG_SMP
470 spinlock_t *lock = q->queue_lock;
471 return lock && spin_is_locked(lock);
472 #else
473 return 1;
474 #endif
477 static inline void queue_flag_set_unlocked(unsigned int flag,
478 struct request_queue *q)
480 __set_bit(flag, &q->queue_flags);
483 static inline int queue_flag_test_and_clear(unsigned int flag,
484 struct request_queue *q)
486 WARN_ON_ONCE(!queue_is_locked(q));
488 if (test_bit(flag, &q->queue_flags)) {
489 __clear_bit(flag, &q->queue_flags);
490 return 1;
493 return 0;
496 static inline int queue_flag_test_and_set(unsigned int flag,
497 struct request_queue *q)
499 WARN_ON_ONCE(!queue_is_locked(q));
501 if (!test_bit(flag, &q->queue_flags)) {
502 __set_bit(flag, &q->queue_flags);
503 return 0;
506 return 1;
509 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
511 WARN_ON_ONCE(!queue_is_locked(q));
512 __set_bit(flag, &q->queue_flags);
515 static inline void queue_flag_clear_unlocked(unsigned int flag,
516 struct request_queue *q)
518 __clear_bit(flag, &q->queue_flags);
521 static inline int queue_in_flight(struct request_queue *q)
523 return q->in_flight[0] + q->in_flight[1];
526 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
528 WARN_ON_ONCE(!queue_is_locked(q));
529 __clear_bit(flag, &q->queue_flags);
532 enum {
534 * Hardbarrier is supported with one of the following methods.
536 * NONE : hardbarrier unsupported
537 * DRAIN : ordering by draining is enough
538 * DRAIN_FLUSH : ordering by draining w/ pre and post flushes
539 * DRAIN_FUA : ordering by draining w/ pre flush and FUA write
540 * TAG : ordering by tag is enough
541 * TAG_FLUSH : ordering by tag w/ pre and post flushes
542 * TAG_FUA : ordering by tag w/ pre flush and FUA write
544 QUEUE_ORDERED_BY_DRAIN = 0x01,
545 QUEUE_ORDERED_BY_TAG = 0x02,
546 QUEUE_ORDERED_DO_PREFLUSH = 0x10,
547 QUEUE_ORDERED_DO_BAR = 0x20,
548 QUEUE_ORDERED_DO_POSTFLUSH = 0x40,
549 QUEUE_ORDERED_DO_FUA = 0x80,
551 QUEUE_ORDERED_NONE = 0x00,
553 QUEUE_ORDERED_DRAIN = QUEUE_ORDERED_BY_DRAIN |
554 QUEUE_ORDERED_DO_BAR,
555 QUEUE_ORDERED_DRAIN_FLUSH = QUEUE_ORDERED_DRAIN |
556 QUEUE_ORDERED_DO_PREFLUSH |
557 QUEUE_ORDERED_DO_POSTFLUSH,
558 QUEUE_ORDERED_DRAIN_FUA = QUEUE_ORDERED_DRAIN |
559 QUEUE_ORDERED_DO_PREFLUSH |
560 QUEUE_ORDERED_DO_FUA,
562 QUEUE_ORDERED_TAG = QUEUE_ORDERED_BY_TAG |
563 QUEUE_ORDERED_DO_BAR,
564 QUEUE_ORDERED_TAG_FLUSH = QUEUE_ORDERED_TAG |
565 QUEUE_ORDERED_DO_PREFLUSH |
566 QUEUE_ORDERED_DO_POSTFLUSH,
567 QUEUE_ORDERED_TAG_FUA = QUEUE_ORDERED_TAG |
568 QUEUE_ORDERED_DO_PREFLUSH |
569 QUEUE_ORDERED_DO_FUA,
572 * Ordered operation sequence
574 QUEUE_ORDSEQ_STARTED = 0x01, /* flushing in progress */
575 QUEUE_ORDSEQ_DRAIN = 0x02, /* waiting for the queue to be drained */
576 QUEUE_ORDSEQ_PREFLUSH = 0x04, /* pre-flushing in progress */
577 QUEUE_ORDSEQ_BAR = 0x08, /* original barrier req in progress */
578 QUEUE_ORDSEQ_POSTFLUSH = 0x10, /* post-flushing in progress */
579 QUEUE_ORDSEQ_DONE = 0x20,
582 #define blk_queue_plugged(q) test_bit(QUEUE_FLAG_PLUGGED, &(q)->queue_flags)
583 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
584 #define blk_queue_queuing(q) test_bit(QUEUE_FLAG_CQ, &(q)->queue_flags)
585 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
586 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
587 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
588 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
589 #define blk_queue_flushing(q) ((q)->ordseq)
590 #define blk_queue_stackable(q) \
591 test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
592 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
594 #define blk_fs_request(rq) ((rq)->cmd_type == REQ_TYPE_FS)
595 #define blk_pc_request(rq) ((rq)->cmd_type == REQ_TYPE_BLOCK_PC)
596 #define blk_special_request(rq) ((rq)->cmd_type == REQ_TYPE_SPECIAL)
597 #define blk_sense_request(rq) ((rq)->cmd_type == REQ_TYPE_SENSE)
599 #define blk_failfast_dev(rq) ((rq)->cmd_flags & REQ_FAILFAST_DEV)
600 #define blk_failfast_transport(rq) ((rq)->cmd_flags & REQ_FAILFAST_TRANSPORT)
601 #define blk_failfast_driver(rq) ((rq)->cmd_flags & REQ_FAILFAST_DRIVER)
602 #define blk_noretry_request(rq) (blk_failfast_dev(rq) || \
603 blk_failfast_transport(rq) || \
604 blk_failfast_driver(rq))
605 #define blk_rq_started(rq) ((rq)->cmd_flags & REQ_STARTED)
606 #define blk_rq_io_stat(rq) ((rq)->cmd_flags & REQ_IO_STAT)
607 #define blk_rq_quiet(rq) ((rq)->cmd_flags & REQ_QUIET)
609 #define blk_account_rq(rq) (blk_rq_started(rq) && (blk_fs_request(rq) || blk_discard_rq(rq)))
611 #define blk_pm_suspend_request(rq) ((rq)->cmd_type == REQ_TYPE_PM_SUSPEND)
612 #define blk_pm_resume_request(rq) ((rq)->cmd_type == REQ_TYPE_PM_RESUME)
613 #define blk_pm_request(rq) \
614 (blk_pm_suspend_request(rq) || blk_pm_resume_request(rq))
616 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
617 #define blk_sorted_rq(rq) ((rq)->cmd_flags & REQ_SORTED)
618 #define blk_barrier_rq(rq) ((rq)->cmd_flags & REQ_HARDBARRIER)
619 #define blk_fua_rq(rq) ((rq)->cmd_flags & REQ_FUA)
620 #define blk_discard_rq(rq) ((rq)->cmd_flags & REQ_DISCARD)
621 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
622 /* rq->queuelist of dequeued request must be list_empty() */
623 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
625 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
627 #define rq_data_dir(rq) ((rq)->cmd_flags & 1)
630 * We regard a request as sync, if either a read or a sync write
632 static inline bool rw_is_sync(unsigned int rw_flags)
634 return !(rw_flags & REQ_RW) || (rw_flags & REQ_RW_SYNC);
637 static inline bool rq_is_sync(struct request *rq)
639 return rw_is_sync(rq->cmd_flags);
642 #define rq_is_meta(rq) ((rq)->cmd_flags & REQ_RW_META)
643 #define rq_noidle(rq) ((rq)->cmd_flags & REQ_NOIDLE)
645 static inline int blk_queue_full(struct request_queue *q, int sync)
647 if (sync)
648 return test_bit(QUEUE_FLAG_SYNCFULL, &q->queue_flags);
649 return test_bit(QUEUE_FLAG_ASYNCFULL, &q->queue_flags);
652 static inline void blk_set_queue_full(struct request_queue *q, int sync)
654 if (sync)
655 queue_flag_set(QUEUE_FLAG_SYNCFULL, q);
656 else
657 queue_flag_set(QUEUE_FLAG_ASYNCFULL, q);
660 static inline void blk_clear_queue_full(struct request_queue *q, int sync)
662 if (sync)
663 queue_flag_clear(QUEUE_FLAG_SYNCFULL, q);
664 else
665 queue_flag_clear(QUEUE_FLAG_ASYNCFULL, q);
670 * mergeable request must not have _NOMERGE or _BARRIER bit set, nor may
671 * it already be started by driver.
673 #define RQ_NOMERGE_FLAGS \
674 (REQ_NOMERGE | REQ_STARTED | REQ_HARDBARRIER | REQ_SOFTBARRIER)
675 #define rq_mergeable(rq) \
676 (!((rq)->cmd_flags & RQ_NOMERGE_FLAGS) && \
677 (blk_discard_rq(rq) || blk_fs_request((rq))))
680 * q->prep_rq_fn return values
682 #define BLKPREP_OK 0 /* serve it */
683 #define BLKPREP_KILL 1 /* fatal error, kill */
684 #define BLKPREP_DEFER 2 /* leave on queue */
686 extern unsigned long blk_max_low_pfn, blk_max_pfn;
689 * standard bounce addresses:
691 * BLK_BOUNCE_HIGH : bounce all highmem pages
692 * BLK_BOUNCE_ANY : don't bounce anything
693 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
696 #if BITS_PER_LONG == 32
697 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
698 #else
699 #define BLK_BOUNCE_HIGH -1ULL
700 #endif
701 #define BLK_BOUNCE_ANY (-1ULL)
702 #define BLK_BOUNCE_ISA (ISA_DMA_THRESHOLD)
705 * default timeout for SG_IO if none specified
707 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
708 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
710 #ifdef CONFIG_BOUNCE
711 extern int init_emergency_isa_pool(void);
712 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
713 #else
714 static inline int init_emergency_isa_pool(void)
716 return 0;
718 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
721 #endif /* CONFIG_MMU */
723 struct rq_map_data {
724 struct page **pages;
725 int page_order;
726 int nr_entries;
727 unsigned long offset;
728 int null_mapped;
729 int from_user;
732 struct req_iterator {
733 int i;
734 struct bio *bio;
737 /* This should not be used directly - use rq_for_each_segment */
738 #define for_each_bio(_bio) \
739 for (; _bio; _bio = _bio->bi_next)
740 #define __rq_for_each_bio(_bio, rq) \
741 if ((rq->bio)) \
742 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
744 #define rq_for_each_segment(bvl, _rq, _iter) \
745 __rq_for_each_bio(_iter.bio, _rq) \
746 bio_for_each_segment(bvl, _iter.bio, _iter.i)
748 #define rq_iter_last(rq, _iter) \
749 (_iter.bio->bi_next == NULL && _iter.i == _iter.bio->bi_vcnt-1)
751 extern int blk_register_queue(struct gendisk *disk);
752 extern void blk_unregister_queue(struct gendisk *disk);
753 extern void register_disk(struct gendisk *dev);
754 extern void generic_make_request(struct bio *bio);
755 extern void blk_rq_init(struct request_queue *q, struct request *rq);
756 extern void blk_put_request(struct request *);
757 extern void __blk_put_request(struct request_queue *, struct request *);
758 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
759 extern struct request *blk_make_request(struct request_queue *, struct bio *,
760 gfp_t);
761 extern void blk_insert_request(struct request_queue *, struct request *, int, void *);
762 extern void blk_requeue_request(struct request_queue *, struct request *);
763 extern int blk_rq_check_limits(struct request_queue *q, struct request *rq);
764 extern int blk_lld_busy(struct request_queue *q);
765 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
766 struct bio_set *bs, gfp_t gfp_mask,
767 int (*bio_ctr)(struct bio *, struct bio *, void *),
768 void *data);
769 extern void blk_rq_unprep_clone(struct request *rq);
770 extern int blk_insert_cloned_request(struct request_queue *q,
771 struct request *rq);
772 extern void blk_plug_device(struct request_queue *);
773 extern void blk_plug_device_unlocked(struct request_queue *);
774 extern int blk_remove_plug(struct request_queue *);
775 extern void blk_recount_segments(struct request_queue *, struct bio *);
776 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
777 unsigned int, void __user *);
778 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
779 struct scsi_ioctl_command __user *);
782 * A queue has just exitted congestion. Note this in the global counter of
783 * congested queues, and wake up anyone who was waiting for requests to be
784 * put back.
786 static inline void blk_clear_queue_congested(struct request_queue *q, int sync)
788 clear_bdi_congested(&q->backing_dev_info, sync);
792 * A queue has just entered congestion. Flag that in the queue's VM-visible
793 * state flags and increment the global gounter of congested queues.
795 static inline void blk_set_queue_congested(struct request_queue *q, int sync)
797 set_bdi_congested(&q->backing_dev_info, sync);
800 extern void blk_start_queue(struct request_queue *q);
801 extern void blk_stop_queue(struct request_queue *q);
802 extern void blk_sync_queue(struct request_queue *q);
803 extern void __blk_stop_queue(struct request_queue *q);
804 extern void __blk_run_queue(struct request_queue *);
805 extern void blk_run_queue(struct request_queue *);
806 extern int blk_rq_map_user(struct request_queue *, struct request *,
807 struct rq_map_data *, void __user *, unsigned long,
808 gfp_t);
809 extern int blk_rq_unmap_user(struct bio *);
810 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
811 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
812 struct rq_map_data *, struct sg_iovec *, int,
813 unsigned int, gfp_t);
814 extern int blk_execute_rq(struct request_queue *, struct gendisk *,
815 struct request *, int);
816 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
817 struct request *, int, rq_end_io_fn *);
818 extern void blk_unplug(struct request_queue *q);
820 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
822 return bdev->bd_disk->queue;
825 static inline void blk_run_backing_dev(struct backing_dev_info *bdi,
826 struct page *page)
828 if (bdi && bdi->unplug_io_fn)
829 bdi->unplug_io_fn(bdi, page);
832 static inline void blk_run_address_space(struct address_space *mapping)
834 if (mapping)
835 blk_run_backing_dev(mapping->backing_dev_info, NULL);
839 * blk_rq_pos() : the current sector
840 * blk_rq_bytes() : bytes left in the entire request
841 * blk_rq_cur_bytes() : bytes left in the current segment
842 * blk_rq_err_bytes() : bytes left till the next error boundary
843 * blk_rq_sectors() : sectors left in the entire request
844 * blk_rq_cur_sectors() : sectors left in the current segment
845 * blk_rq_err_sectors() : sectors left till the next error boundary
847 static inline sector_t blk_rq_pos(const struct request *rq)
849 return rq->__sector;
852 static inline unsigned int blk_rq_bytes(const struct request *rq)
854 return rq->__data_len;
857 static inline int blk_rq_cur_bytes(const struct request *rq)
859 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
862 extern unsigned int blk_rq_err_bytes(const struct request *rq);
864 static inline unsigned int blk_rq_sectors(const struct request *rq)
866 return blk_rq_bytes(rq) >> 9;
869 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
871 return blk_rq_cur_bytes(rq) >> 9;
874 static inline unsigned int blk_rq_err_sectors(const struct request *rq)
876 return blk_rq_err_bytes(rq) >> 9;
880 * Request issue related functions.
882 extern struct request *blk_peek_request(struct request_queue *q);
883 extern void blk_start_request(struct request *rq);
884 extern struct request *blk_fetch_request(struct request_queue *q);
887 * Request completion related functions.
889 * blk_update_request() completes given number of bytes and updates
890 * the request without completing it.
892 * blk_end_request() and friends. __blk_end_request() must be called
893 * with the request queue spinlock acquired.
895 * Several drivers define their own end_request and call
896 * blk_end_request() for parts of the original function.
897 * This prevents code duplication in drivers.
899 extern bool blk_update_request(struct request *rq, int error,
900 unsigned int nr_bytes);
901 extern bool blk_end_request(struct request *rq, int error,
902 unsigned int nr_bytes);
903 extern void blk_end_request_all(struct request *rq, int error);
904 extern bool blk_end_request_cur(struct request *rq, int error);
905 extern bool blk_end_request_err(struct request *rq, int error);
906 extern bool __blk_end_request(struct request *rq, int error,
907 unsigned int nr_bytes);
908 extern void __blk_end_request_all(struct request *rq, int error);
909 extern bool __blk_end_request_cur(struct request *rq, int error);
910 extern bool __blk_end_request_err(struct request *rq, int error);
912 extern void blk_complete_request(struct request *);
913 extern void __blk_complete_request(struct request *);
914 extern void blk_abort_request(struct request *);
915 extern void blk_abort_queue(struct request_queue *);
918 * Access functions for manipulating queue properties
920 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
921 spinlock_t *lock, int node_id);
922 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
923 extern void blk_cleanup_queue(struct request_queue *);
924 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
925 extern void blk_queue_bounce_limit(struct request_queue *, u64);
926 extern void blk_queue_max_sectors(struct request_queue *, unsigned int);
927 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
928 extern void blk_queue_max_phys_segments(struct request_queue *, unsigned short);
929 extern void blk_queue_max_hw_segments(struct request_queue *, unsigned short);
930 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
931 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
932 extern void blk_queue_physical_block_size(struct request_queue *, unsigned short);
933 extern void blk_queue_alignment_offset(struct request_queue *q,
934 unsigned int alignment);
935 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
936 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
937 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
938 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
939 extern void blk_set_default_limits(struct queue_limits *lim);
940 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
941 sector_t offset);
942 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
943 sector_t offset);
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 *);
977 * tag stuff
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,
991 int tag)
993 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
994 return NULL;
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 #define MAX_PHYS_SEGMENTS 128
1016 #define MAX_HW_SEGMENTS 128
1017 #define SAFE_MAX_SECTORS 255
1018 #define BLK_DEF_MAX_SECTORS 1024
1020 #define MAX_SEGMENT_SIZE 65536
1022 #define BLK_SEG_BOUNDARY_MASK 0xFFFFFFFFUL
1024 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1026 static inline unsigned long queue_bounce_pfn(struct request_queue *q)
1028 return q->limits.bounce_pfn;
1031 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1033 return q->limits.seg_boundary_mask;
1036 static inline unsigned int queue_max_sectors(struct request_queue *q)
1038 return q->limits.max_sectors;
1041 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1043 return q->limits.max_hw_sectors;
1046 static inline unsigned short queue_max_hw_segments(struct request_queue *q)
1048 return q->limits.max_hw_segments;
1051 static inline unsigned short queue_max_phys_segments(struct request_queue *q)
1053 return q->limits.max_phys_segments;
1056 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1058 return q->limits.max_segment_size;
1061 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1063 int retval = 512;
1065 if (q && q->limits.logical_block_size)
1066 retval = q->limits.logical_block_size;
1068 return retval;
1071 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1073 return queue_logical_block_size(bdev_get_queue(bdev));
1076 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1078 return q->limits.physical_block_size;
1081 static inline unsigned int queue_io_min(struct request_queue *q)
1083 return q->limits.io_min;
1086 static inline unsigned int queue_io_opt(struct request_queue *q)
1088 return q->limits.io_opt;
1091 static inline int queue_alignment_offset(struct request_queue *q)
1093 if (q && q->limits.misaligned)
1094 return -1;
1096 if (q && q->limits.alignment_offset)
1097 return q->limits.alignment_offset;
1099 return 0;
1102 static inline int queue_sector_alignment_offset(struct request_queue *q,
1103 sector_t sector)
1105 return ((sector << 9) - q->limits.alignment_offset)
1106 & (q->limits.io_min - 1);
1109 static inline int queue_dma_alignment(struct request_queue *q)
1111 return q ? q->dma_alignment : 511;
1114 static inline int blk_rq_aligned(struct request_queue *q, void *addr,
1115 unsigned int len)
1117 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1118 return !((unsigned long)addr & alignment) && !(len & alignment);
1121 /* assumes size > 256 */
1122 static inline unsigned int blksize_bits(unsigned int size)
1124 unsigned int bits = 8;
1125 do {
1126 bits++;
1127 size >>= 1;
1128 } while (size > 256);
1129 return bits;
1132 static inline unsigned int block_size(struct block_device *bdev)
1134 return bdev->bd_block_size;
1137 typedef struct {struct page *v;} Sector;
1139 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1141 static inline void put_dev_sector(Sector p)
1143 page_cache_release(p.v);
1146 struct work_struct;
1147 int kblockd_schedule_work(struct request_queue *q, struct work_struct *work);
1149 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1150 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1151 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1152 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1154 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1156 #define INTEGRITY_FLAG_READ 2 /* verify data integrity on read */
1157 #define INTEGRITY_FLAG_WRITE 4 /* generate data integrity on write */
1159 struct blk_integrity_exchg {
1160 void *prot_buf;
1161 void *data_buf;
1162 sector_t sector;
1163 unsigned int data_size;
1164 unsigned short sector_size;
1165 const char *disk_name;
1168 typedef void (integrity_gen_fn) (struct blk_integrity_exchg *);
1169 typedef int (integrity_vrfy_fn) (struct blk_integrity_exchg *);
1170 typedef void (integrity_set_tag_fn) (void *, void *, unsigned int);
1171 typedef void (integrity_get_tag_fn) (void *, void *, unsigned int);
1173 struct blk_integrity {
1174 integrity_gen_fn *generate_fn;
1175 integrity_vrfy_fn *verify_fn;
1176 integrity_set_tag_fn *set_tag_fn;
1177 integrity_get_tag_fn *get_tag_fn;
1179 unsigned short flags;
1180 unsigned short tuple_size;
1181 unsigned short sector_size;
1182 unsigned short tag_size;
1184 const char *name;
1186 struct kobject kobj;
1189 extern int blk_integrity_register(struct gendisk *, struct blk_integrity *);
1190 extern void blk_integrity_unregister(struct gendisk *);
1191 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1192 extern int blk_rq_map_integrity_sg(struct request *, struct scatterlist *);
1193 extern int blk_rq_count_integrity_sg(struct request *);
1195 static inline
1196 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1198 return bdev->bd_disk->integrity;
1201 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1203 return disk->integrity;
1206 static inline int blk_integrity_rq(struct request *rq)
1208 if (rq->bio == NULL)
1209 return 0;
1211 return bio_integrity(rq->bio);
1214 #else /* CONFIG_BLK_DEV_INTEGRITY */
1216 #define blk_integrity_rq(rq) (0)
1217 #define blk_rq_count_integrity_sg(a) (0)
1218 #define blk_rq_map_integrity_sg(a, b) (0)
1219 #define bdev_get_integrity(a) (0)
1220 #define blk_get_integrity(a) (0)
1221 #define blk_integrity_compare(a, b) (0)
1222 #define blk_integrity_register(a, b) (0)
1223 #define blk_integrity_unregister(a) do { } while (0);
1225 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1227 struct block_device_operations {
1228 int (*open) (struct block_device *, fmode_t);
1229 int (*release) (struct gendisk *, fmode_t);
1230 int (*locked_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1231 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1232 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1233 int (*direct_access) (struct block_device *, sector_t,
1234 void **, unsigned long *);
1235 int (*media_changed) (struct gendisk *);
1236 unsigned long long (*set_capacity) (struct gendisk *,
1237 unsigned long long);
1238 int (*revalidate_disk) (struct gendisk *);
1239 int (*getgeo)(struct block_device *, struct hd_geometry *);
1240 struct module *owner;
1243 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1244 unsigned long);
1245 #else /* CONFIG_BLOCK */
1247 * stubs for when the block layer is configured out
1249 #define buffer_heads_over_limit 0
1251 static inline long nr_blockdev_pages(void)
1253 return 0;
1256 #endif /* CONFIG_BLOCK */
1258 #endif