]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - include/linux/blkdev.h
block: add block polling support
[mirror_ubuntu-artful-kernel.git] / include / linux / blkdev.h
1 #ifndef _LINUX_BLKDEV_H
2 #define _LINUX_BLKDEV_H
3
4 #include <linux/sched.h>
5
6 #ifdef CONFIG_BLOCK
7
8 #include <linux/major.h>
9 #include <linux/genhd.h>
10 #include <linux/list.h>
11 #include <linux/llist.h>
12 #include <linux/timer.h>
13 #include <linux/workqueue.h>
14 #include <linux/pagemap.h>
15 #include <linux/backing-dev-defs.h>
16 #include <linux/wait.h>
17 #include <linux/mempool.h>
18 #include <linux/bio.h>
19 #include <linux/stringify.h>
20 #include <linux/gfp.h>
21 #include <linux/bsg.h>
22 #include <linux/smp.h>
23 #include <linux/rcupdate.h>
24 #include <linux/percpu-refcount.h>
25 #include <linux/scatterlist.h>
26
27 struct module;
28 struct scsi_ioctl_command;
29
30 struct request_queue;
31 struct elevator_queue;
32 struct blk_trace;
33 struct request;
34 struct sg_io_hdr;
35 struct bsg_job;
36 struct blkcg_gq;
37 struct blk_flush_queue;
38 struct pr_ops;
39
40 #define BLKDEV_MIN_RQ 4
41 #define BLKDEV_MAX_RQ 128 /* Default maximum */
42
43 /*
44 * Maximum number of blkcg policies allowed to be registered concurrently.
45 * Defined here to simplify include dependency.
46 */
47 #define BLKCG_MAX_POLS 2
48
49 struct request;
50 typedef void (rq_end_io_fn)(struct request *, int);
51
52 #define BLK_RL_SYNCFULL (1U << 0)
53 #define BLK_RL_ASYNCFULL (1U << 1)
54
55 struct request_list {
56 struct request_queue *q; /* the queue this rl belongs to */
57 #ifdef CONFIG_BLK_CGROUP
58 struct blkcg_gq *blkg; /* blkg this request pool belongs to */
59 #endif
60 /*
61 * count[], starved[], and wait[] are indexed by
62 * BLK_RW_SYNC/BLK_RW_ASYNC
63 */
64 int count[2];
65 int starved[2];
66 mempool_t *rq_pool;
67 wait_queue_head_t wait[2];
68 unsigned int flags;
69 };
70
71 /*
72 * request command types
73 */
74 enum rq_cmd_type_bits {
75 REQ_TYPE_FS = 1, /* fs request */
76 REQ_TYPE_BLOCK_PC, /* scsi command */
77 REQ_TYPE_DRV_PRIV, /* driver defined types from here */
78 };
79
80 #define BLK_MAX_CDB 16
81
82 /*
83 * Try to put the fields that are referenced together in the same cacheline.
84 *
85 * If you modify this structure, make sure to update blk_rq_init() and
86 * especially blk_mq_rq_ctx_init() to take care of the added fields.
87 */
88 struct request {
89 struct list_head queuelist;
90 union {
91 struct call_single_data csd;
92 unsigned long fifo_time;
93 };
94
95 struct request_queue *q;
96 struct blk_mq_ctx *mq_ctx;
97
98 u64 cmd_flags;
99 unsigned cmd_type;
100 unsigned long atomic_flags;
101
102 int cpu;
103
104 /* the following two fields are internal, NEVER access directly */
105 unsigned int __data_len; /* total data len */
106 sector_t __sector; /* sector cursor */
107
108 struct bio *bio;
109 struct bio *biotail;
110
111 /*
112 * The hash is used inside the scheduler, and killed once the
113 * request reaches the dispatch list. The ipi_list is only used
114 * to queue the request for softirq completion, which is long
115 * after the request has been unhashed (and even removed from
116 * the dispatch list).
117 */
118 union {
119 struct hlist_node hash; /* merge hash */
120 struct list_head ipi_list;
121 };
122
123 /*
124 * The rb_node is only used inside the io scheduler, requests
125 * are pruned when moved to the dispatch queue. So let the
126 * completion_data share space with the rb_node.
127 */
128 union {
129 struct rb_node rb_node; /* sort/lookup */
130 void *completion_data;
131 };
132
133 /*
134 * Three pointers are available for the IO schedulers, if they need
135 * more they have to dynamically allocate it. Flush requests are
136 * never put on the IO scheduler. So let the flush fields share
137 * space with the elevator data.
138 */
139 union {
140 struct {
141 struct io_cq *icq;
142 void *priv[2];
143 } elv;
144
145 struct {
146 unsigned int seq;
147 struct list_head list;
148 rq_end_io_fn *saved_end_io;
149 } flush;
150 };
151
152 struct gendisk *rq_disk;
153 struct hd_struct *part;
154 unsigned long start_time;
155 #ifdef CONFIG_BLK_CGROUP
156 struct request_list *rl; /* rl this rq is alloced from */
157 unsigned long long start_time_ns;
158 unsigned long long io_start_time_ns; /* when passed to hardware */
159 #endif
160 /* Number of scatter-gather DMA addr+len pairs after
161 * physical address coalescing is performed.
162 */
163 unsigned short nr_phys_segments;
164 #if defined(CONFIG_BLK_DEV_INTEGRITY)
165 unsigned short nr_integrity_segments;
166 #endif
167
168 unsigned short ioprio;
169
170 void *special; /* opaque pointer available for LLD use */
171
172 int tag;
173 int errors;
174
175 /*
176 * when request is used as a packet command carrier
177 */
178 unsigned char __cmd[BLK_MAX_CDB];
179 unsigned char *cmd;
180 unsigned short cmd_len;
181
182 unsigned int extra_len; /* length of alignment and padding */
183 unsigned int sense_len;
184 unsigned int resid_len; /* residual count */
185 void *sense;
186
187 unsigned long deadline;
188 struct list_head timeout_list;
189 unsigned int timeout;
190 int retries;
191
192 /*
193 * completion callback.
194 */
195 rq_end_io_fn *end_io;
196 void *end_io_data;
197
198 /* for bidi */
199 struct request *next_rq;
200 };
201
202 static inline unsigned short req_get_ioprio(struct request *req)
203 {
204 return req->ioprio;
205 }
206
207 #include <linux/elevator.h>
208
209 struct blk_queue_ctx;
210
211 typedef void (request_fn_proc) (struct request_queue *q);
212 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio);
213 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
214 typedef void (unprep_rq_fn) (struct request_queue *, struct request *);
215
216 struct bio_vec;
217 typedef void (softirq_done_fn)(struct request *);
218 typedef int (dma_drain_needed_fn)(struct request *);
219 typedef int (lld_busy_fn) (struct request_queue *q);
220 typedef int (bsg_job_fn) (struct bsg_job *);
221
222 enum blk_eh_timer_return {
223 BLK_EH_NOT_HANDLED,
224 BLK_EH_HANDLED,
225 BLK_EH_RESET_TIMER,
226 };
227
228 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
229
230 enum blk_queue_state {
231 Queue_down,
232 Queue_up,
233 };
234
235 struct blk_queue_tag {
236 struct request **tag_index; /* map of busy tags */
237 unsigned long *tag_map; /* bit map of free/busy tags */
238 int busy; /* current depth */
239 int max_depth; /* what we will send to device */
240 int real_max_depth; /* what the array can hold */
241 atomic_t refcnt; /* map can be shared */
242 int alloc_policy; /* tag allocation policy */
243 int next_tag; /* next tag */
244 };
245 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
246 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
247
248 #define BLK_SCSI_MAX_CMDS (256)
249 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
250
251 struct queue_limits {
252 unsigned long bounce_pfn;
253 unsigned long seg_boundary_mask;
254 unsigned long virt_boundary_mask;
255
256 unsigned int max_hw_sectors;
257 unsigned int chunk_sectors;
258 unsigned int max_sectors;
259 unsigned int max_segment_size;
260 unsigned int physical_block_size;
261 unsigned int alignment_offset;
262 unsigned int io_min;
263 unsigned int io_opt;
264 unsigned int max_discard_sectors;
265 unsigned int max_hw_discard_sectors;
266 unsigned int max_write_same_sectors;
267 unsigned int discard_granularity;
268 unsigned int discard_alignment;
269
270 unsigned short logical_block_size;
271 unsigned short max_segments;
272 unsigned short max_integrity_segments;
273
274 unsigned char misaligned;
275 unsigned char discard_misaligned;
276 unsigned char cluster;
277 unsigned char discard_zeroes_data;
278 unsigned char raid_partial_stripes_expensive;
279 };
280
281 struct request_queue {
282 /*
283 * Together with queue_head for cacheline sharing
284 */
285 struct list_head queue_head;
286 struct request *last_merge;
287 struct elevator_queue *elevator;
288 int nr_rqs[2]; /* # allocated [a]sync rqs */
289 int nr_rqs_elvpriv; /* # allocated rqs w/ elvpriv */
290
291 /*
292 * If blkcg is not used, @q->root_rl serves all requests. If blkcg
293 * is used, root blkg allocates from @q->root_rl and all other
294 * blkgs from their own blkg->rl. Which one to use should be
295 * determined using bio_request_list().
296 */
297 struct request_list root_rl;
298
299 request_fn_proc *request_fn;
300 make_request_fn *make_request_fn;
301 prep_rq_fn *prep_rq_fn;
302 unprep_rq_fn *unprep_rq_fn;
303 softirq_done_fn *softirq_done_fn;
304 rq_timed_out_fn *rq_timed_out_fn;
305 dma_drain_needed_fn *dma_drain_needed;
306 lld_busy_fn *lld_busy_fn;
307
308 struct blk_mq_ops *mq_ops;
309
310 unsigned int *mq_map;
311
312 /* sw queues */
313 struct blk_mq_ctx __percpu *queue_ctx;
314 unsigned int nr_queues;
315
316 /* hw dispatch queues */
317 struct blk_mq_hw_ctx **queue_hw_ctx;
318 unsigned int nr_hw_queues;
319
320 /*
321 * Dispatch queue sorting
322 */
323 sector_t end_sector;
324 struct request *boundary_rq;
325
326 /*
327 * Delayed queue handling
328 */
329 struct delayed_work delay_work;
330
331 struct backing_dev_info backing_dev_info;
332
333 /*
334 * The queue owner gets to use this for whatever they like.
335 * ll_rw_blk doesn't touch it.
336 */
337 void *queuedata;
338
339 /*
340 * various queue flags, see QUEUE_* below
341 */
342 unsigned long queue_flags;
343
344 /*
345 * ida allocated id for this queue. Used to index queues from
346 * ioctx.
347 */
348 int id;
349
350 /*
351 * queue needs bounce pages for pages above this limit
352 */
353 gfp_t bounce_gfp;
354
355 /*
356 * protects queue structures from reentrancy. ->__queue_lock should
357 * _never_ be used directly, it is queue private. always use
358 * ->queue_lock.
359 */
360 spinlock_t __queue_lock;
361 spinlock_t *queue_lock;
362
363 /*
364 * queue kobject
365 */
366 struct kobject kobj;
367
368 /*
369 * mq queue kobject
370 */
371 struct kobject mq_kobj;
372
373 #ifdef CONFIG_BLK_DEV_INTEGRITY
374 struct blk_integrity integrity;
375 #endif /* CONFIG_BLK_DEV_INTEGRITY */
376
377 #ifdef CONFIG_PM
378 struct device *dev;
379 int rpm_status;
380 unsigned int nr_pending;
381 #endif
382
383 /*
384 * queue settings
385 */
386 unsigned long nr_requests; /* Max # of requests */
387 unsigned int nr_congestion_on;
388 unsigned int nr_congestion_off;
389 unsigned int nr_batching;
390
391 unsigned int dma_drain_size;
392 void *dma_drain_buffer;
393 unsigned int dma_pad_mask;
394 unsigned int dma_alignment;
395
396 struct blk_queue_tag *queue_tags;
397 struct list_head tag_busy_list;
398
399 unsigned int nr_sorted;
400 unsigned int in_flight[2];
401 /*
402 * Number of active block driver functions for which blk_drain_queue()
403 * must wait. Must be incremented around functions that unlock the
404 * queue_lock internally, e.g. scsi_request_fn().
405 */
406 unsigned int request_fn_active;
407
408 unsigned int rq_timeout;
409 struct timer_list timeout;
410 struct list_head timeout_list;
411
412 struct list_head icq_list;
413 #ifdef CONFIG_BLK_CGROUP
414 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
415 struct blkcg_gq *root_blkg;
416 struct list_head blkg_list;
417 #endif
418
419 struct queue_limits limits;
420
421 /*
422 * sg stuff
423 */
424 unsigned int sg_timeout;
425 unsigned int sg_reserved_size;
426 int node;
427 #ifdef CONFIG_BLK_DEV_IO_TRACE
428 struct blk_trace *blk_trace;
429 #endif
430 /*
431 * for flush operations
432 */
433 unsigned int flush_flags;
434 unsigned int flush_not_queueable:1;
435 struct blk_flush_queue *fq;
436
437 struct list_head requeue_list;
438 spinlock_t requeue_lock;
439 struct work_struct requeue_work;
440
441 struct mutex sysfs_lock;
442
443 int bypass_depth;
444 atomic_t mq_freeze_depth;
445
446 #if defined(CONFIG_BLK_DEV_BSG)
447 bsg_job_fn *bsg_job_fn;
448 int bsg_job_size;
449 struct bsg_class_device bsg_dev;
450 #endif
451
452 #ifdef CONFIG_BLK_DEV_THROTTLING
453 /* Throttle data */
454 struct throtl_data *td;
455 #endif
456 struct rcu_head rcu_head;
457 wait_queue_head_t mq_freeze_wq;
458 struct percpu_ref q_usage_counter;
459 struct list_head all_q_node;
460
461 struct blk_mq_tag_set *tag_set;
462 struct list_head tag_set_list;
463 struct bio_set *bio_split;
464
465 bool mq_sysfs_init_done;
466 };
467
468 #define QUEUE_FLAG_QUEUED 1 /* uses generic tag queueing */
469 #define QUEUE_FLAG_STOPPED 2 /* queue is stopped */
470 #define QUEUE_FLAG_SYNCFULL 3 /* read queue has been filled */
471 #define QUEUE_FLAG_ASYNCFULL 4 /* write queue has been filled */
472 #define QUEUE_FLAG_DYING 5 /* queue being torn down */
473 #define QUEUE_FLAG_BYPASS 6 /* act as dumb FIFO queue */
474 #define QUEUE_FLAG_BIDI 7 /* queue supports bidi requests */
475 #define QUEUE_FLAG_NOMERGES 8 /* disable merge attempts */
476 #define QUEUE_FLAG_SAME_COMP 9 /* complete on same CPU-group */
477 #define QUEUE_FLAG_FAIL_IO 10 /* fake timeout */
478 #define QUEUE_FLAG_STACKABLE 11 /* supports request stacking */
479 #define QUEUE_FLAG_NONROT 12 /* non-rotational device (SSD) */
480 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
481 #define QUEUE_FLAG_IO_STAT 13 /* do IO stats */
482 #define QUEUE_FLAG_DISCARD 14 /* supports DISCARD */
483 #define QUEUE_FLAG_NOXMERGES 15 /* No extended merges */
484 #define QUEUE_FLAG_ADD_RANDOM 16 /* Contributes to random pool */
485 #define QUEUE_FLAG_SECDISCARD 17 /* supports SECDISCARD */
486 #define QUEUE_FLAG_SAME_FORCE 18 /* force complete on same CPU */
487 #define QUEUE_FLAG_DEAD 19 /* queue tear-down finished */
488 #define QUEUE_FLAG_INIT_DONE 20 /* queue is initialized */
489 #define QUEUE_FLAG_NO_SG_MERGE 21 /* don't attempt to merge SG segments*/
490 #define QUEUE_FLAG_POLL 22 /* IO polling enabled if set */
491
492 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
493 (1 << QUEUE_FLAG_STACKABLE) | \
494 (1 << QUEUE_FLAG_SAME_COMP) | \
495 (1 << QUEUE_FLAG_ADD_RANDOM))
496
497 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
498 (1 << QUEUE_FLAG_STACKABLE) | \
499 (1 << QUEUE_FLAG_SAME_COMP))
500
501 static inline void queue_lockdep_assert_held(struct request_queue *q)
502 {
503 if (q->queue_lock)
504 lockdep_assert_held(q->queue_lock);
505 }
506
507 static inline void queue_flag_set_unlocked(unsigned int flag,
508 struct request_queue *q)
509 {
510 __set_bit(flag, &q->queue_flags);
511 }
512
513 static inline int queue_flag_test_and_clear(unsigned int flag,
514 struct request_queue *q)
515 {
516 queue_lockdep_assert_held(q);
517
518 if (test_bit(flag, &q->queue_flags)) {
519 __clear_bit(flag, &q->queue_flags);
520 return 1;
521 }
522
523 return 0;
524 }
525
526 static inline int queue_flag_test_and_set(unsigned int flag,
527 struct request_queue *q)
528 {
529 queue_lockdep_assert_held(q);
530
531 if (!test_bit(flag, &q->queue_flags)) {
532 __set_bit(flag, &q->queue_flags);
533 return 0;
534 }
535
536 return 1;
537 }
538
539 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
540 {
541 queue_lockdep_assert_held(q);
542 __set_bit(flag, &q->queue_flags);
543 }
544
545 static inline void queue_flag_clear_unlocked(unsigned int flag,
546 struct request_queue *q)
547 {
548 __clear_bit(flag, &q->queue_flags);
549 }
550
551 static inline int queue_in_flight(struct request_queue *q)
552 {
553 return q->in_flight[0] + q->in_flight[1];
554 }
555
556 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
557 {
558 queue_lockdep_assert_held(q);
559 __clear_bit(flag, &q->queue_flags);
560 }
561
562 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
563 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
564 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
565 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
566 #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
567 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
568 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
569 #define blk_queue_noxmerges(q) \
570 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
571 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
572 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
573 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
574 #define blk_queue_stackable(q) \
575 test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
576 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
577 #define blk_queue_secdiscard(q) (blk_queue_discard(q) && \
578 test_bit(QUEUE_FLAG_SECDISCARD, &(q)->queue_flags))
579
580 #define blk_noretry_request(rq) \
581 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
582 REQ_FAILFAST_DRIVER))
583
584 #define blk_account_rq(rq) \
585 (((rq)->cmd_flags & REQ_STARTED) && \
586 ((rq)->cmd_type == REQ_TYPE_FS))
587
588 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
589 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
590 /* rq->queuelist of dequeued request must be list_empty() */
591 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
592
593 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
594
595 #define rq_data_dir(rq) ((int)((rq)->cmd_flags & 1))
596
597 /*
598 * Driver can handle struct request, if it either has an old style
599 * request_fn defined, or is blk-mq based.
600 */
601 static inline bool queue_is_rq_based(struct request_queue *q)
602 {
603 return q->request_fn || q->mq_ops;
604 }
605
606 static inline unsigned int blk_queue_cluster(struct request_queue *q)
607 {
608 return q->limits.cluster;
609 }
610
611 /*
612 * We regard a request as sync, if either a read or a sync write
613 */
614 static inline bool rw_is_sync(unsigned int rw_flags)
615 {
616 return !(rw_flags & REQ_WRITE) || (rw_flags & REQ_SYNC);
617 }
618
619 static inline bool rq_is_sync(struct request *rq)
620 {
621 return rw_is_sync(rq->cmd_flags);
622 }
623
624 static inline bool blk_rl_full(struct request_list *rl, bool sync)
625 {
626 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
627
628 return rl->flags & flag;
629 }
630
631 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
632 {
633 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
634
635 rl->flags |= flag;
636 }
637
638 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
639 {
640 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
641
642 rl->flags &= ~flag;
643 }
644
645 static inline bool rq_mergeable(struct request *rq)
646 {
647 if (rq->cmd_type != REQ_TYPE_FS)
648 return false;
649
650 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
651 return false;
652
653 return true;
654 }
655
656 static inline bool blk_check_merge_flags(unsigned int flags1,
657 unsigned int flags2)
658 {
659 if ((flags1 & REQ_DISCARD) != (flags2 & REQ_DISCARD))
660 return false;
661
662 if ((flags1 & REQ_SECURE) != (flags2 & REQ_SECURE))
663 return false;
664
665 if ((flags1 & REQ_WRITE_SAME) != (flags2 & REQ_WRITE_SAME))
666 return false;
667
668 return true;
669 }
670
671 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
672 {
673 if (bio_data(a) == bio_data(b))
674 return true;
675
676 return false;
677 }
678
679 /*
680 * q->prep_rq_fn return values
681 */
682 #define BLKPREP_OK 0 /* serve it */
683 #define BLKPREP_KILL 1 /* fatal error, kill */
684 #define BLKPREP_DEFER 2 /* leave on queue */
685
686 extern unsigned long blk_max_low_pfn, blk_max_pfn;
687
688 /*
689 * standard bounce addresses:
690 *
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
694 */
695
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 (DMA_BIT_MASK(24))
703
704 /*
705 * default timeout for SG_IO if none specified
706 */
707 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
708 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
709
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)
715 {
716 return 0;
717 }
718 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
719 {
720 }
721 #endif /* CONFIG_MMU */
722
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;
730 };
731
732 struct req_iterator {
733 struct bvec_iter iter;
734 struct bio *bio;
735 };
736
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)
743
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.iter)
747
748 #define rq_iter_last(bvec, _iter) \
749 (_iter.bio->bi_next == NULL && \
750 bio_iter_last(bvec, _iter.iter))
751
752 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
753 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
754 #endif
755 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
756 extern void rq_flush_dcache_pages(struct request *rq);
757 #else
758 static inline void rq_flush_dcache_pages(struct request *rq)
759 {
760 }
761 #endif
762
763 extern int blk_register_queue(struct gendisk *disk);
764 extern void blk_unregister_queue(struct gendisk *disk);
765 extern blk_qc_t generic_make_request(struct bio *bio);
766 extern void blk_rq_init(struct request_queue *q, struct request *rq);
767 extern void blk_put_request(struct request *);
768 extern void __blk_put_request(struct request_queue *, struct request *);
769 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
770 extern struct request *blk_make_request(struct request_queue *, struct bio *,
771 gfp_t);
772 extern void blk_rq_set_block_pc(struct request *);
773 extern void blk_requeue_request(struct request_queue *, struct request *);
774 extern void blk_add_request_payload(struct request *rq, struct page *page,
775 unsigned int len);
776 extern int blk_rq_check_limits(struct request_queue *q, struct request *rq);
777 extern int blk_lld_busy(struct request_queue *q);
778 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
779 struct bio_set *bs, gfp_t gfp_mask,
780 int (*bio_ctr)(struct bio *, struct bio *, void *),
781 void *data);
782 extern void blk_rq_unprep_clone(struct request *rq);
783 extern int blk_insert_cloned_request(struct request_queue *q,
784 struct request *rq);
785 extern void blk_delay_queue(struct request_queue *, unsigned long);
786 extern void blk_queue_split(struct request_queue *, struct bio **,
787 struct bio_set *);
788 extern void blk_recount_segments(struct request_queue *, struct bio *);
789 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
790 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
791 unsigned int, void __user *);
792 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
793 unsigned int, void __user *);
794 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
795 struct scsi_ioctl_command __user *);
796
797 extern void blk_start_queue(struct request_queue *q);
798 extern void blk_stop_queue(struct request_queue *q);
799 extern void blk_sync_queue(struct request_queue *q);
800 extern void __blk_stop_queue(struct request_queue *q);
801 extern void __blk_run_queue(struct request_queue *q);
802 extern void __blk_run_queue_uncond(struct request_queue *q);
803 extern void blk_run_queue(struct request_queue *);
804 extern void blk_run_queue_async(struct request_queue *q);
805 extern int blk_rq_map_user(struct request_queue *, struct request *,
806 struct rq_map_data *, void __user *, unsigned long,
807 gfp_t);
808 extern int blk_rq_unmap_user(struct bio *);
809 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
810 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
811 struct rq_map_data *, const struct iov_iter *,
812 gfp_t);
813 extern int blk_execute_rq(struct request_queue *, struct gendisk *,
814 struct request *, int);
815 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
816 struct request *, int, rq_end_io_fn *);
817
818 bool blk_poll(struct request_queue *q, blk_qc_t cookie);
819
820 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
821 {
822 return bdev->bd_disk->queue; /* this is never NULL */
823 }
824
825 /*
826 * blk_rq_pos() : the current sector
827 * blk_rq_bytes() : bytes left in the entire request
828 * blk_rq_cur_bytes() : bytes left in the current segment
829 * blk_rq_err_bytes() : bytes left till the next error boundary
830 * blk_rq_sectors() : sectors left in the entire request
831 * blk_rq_cur_sectors() : sectors left in the current segment
832 */
833 static inline sector_t blk_rq_pos(const struct request *rq)
834 {
835 return rq->__sector;
836 }
837
838 static inline unsigned int blk_rq_bytes(const struct request *rq)
839 {
840 return rq->__data_len;
841 }
842
843 static inline int blk_rq_cur_bytes(const struct request *rq)
844 {
845 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
846 }
847
848 extern unsigned int blk_rq_err_bytes(const struct request *rq);
849
850 static inline unsigned int blk_rq_sectors(const struct request *rq)
851 {
852 return blk_rq_bytes(rq) >> 9;
853 }
854
855 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
856 {
857 return blk_rq_cur_bytes(rq) >> 9;
858 }
859
860 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
861 unsigned int cmd_flags)
862 {
863 if (unlikely(cmd_flags & REQ_DISCARD))
864 return min(q->limits.max_discard_sectors, UINT_MAX >> 9);
865
866 if (unlikely(cmd_flags & REQ_WRITE_SAME))
867 return q->limits.max_write_same_sectors;
868
869 return q->limits.max_sectors;
870 }
871
872 /*
873 * Return maximum size of a request at given offset. Only valid for
874 * file system requests.
875 */
876 static inline unsigned int blk_max_size_offset(struct request_queue *q,
877 sector_t offset)
878 {
879 if (!q->limits.chunk_sectors)
880 return q->limits.max_sectors;
881
882 return q->limits.chunk_sectors -
883 (offset & (q->limits.chunk_sectors - 1));
884 }
885
886 static inline unsigned int blk_rq_get_max_sectors(struct request *rq)
887 {
888 struct request_queue *q = rq->q;
889
890 if (unlikely(rq->cmd_type == REQ_TYPE_BLOCK_PC))
891 return q->limits.max_hw_sectors;
892
893 if (!q->limits.chunk_sectors || (rq->cmd_flags & REQ_DISCARD))
894 return blk_queue_get_max_sectors(q, rq->cmd_flags);
895
896 return min(blk_max_size_offset(q, blk_rq_pos(rq)),
897 blk_queue_get_max_sectors(q, rq->cmd_flags));
898 }
899
900 static inline unsigned int blk_rq_count_bios(struct request *rq)
901 {
902 unsigned int nr_bios = 0;
903 struct bio *bio;
904
905 __rq_for_each_bio(bio, rq)
906 nr_bios++;
907
908 return nr_bios;
909 }
910
911 /*
912 * Request issue related functions.
913 */
914 extern struct request *blk_peek_request(struct request_queue *q);
915 extern void blk_start_request(struct request *rq);
916 extern struct request *blk_fetch_request(struct request_queue *q);
917
918 /*
919 * Request completion related functions.
920 *
921 * blk_update_request() completes given number of bytes and updates
922 * the request without completing it.
923 *
924 * blk_end_request() and friends. __blk_end_request() must be called
925 * with the request queue spinlock acquired.
926 *
927 * Several drivers define their own end_request and call
928 * blk_end_request() for parts of the original function.
929 * This prevents code duplication in drivers.
930 */
931 extern bool blk_update_request(struct request *rq, int error,
932 unsigned int nr_bytes);
933 extern void blk_finish_request(struct request *rq, int error);
934 extern bool blk_end_request(struct request *rq, int error,
935 unsigned int nr_bytes);
936 extern void blk_end_request_all(struct request *rq, int error);
937 extern bool blk_end_request_cur(struct request *rq, int error);
938 extern bool blk_end_request_err(struct request *rq, int error);
939 extern bool __blk_end_request(struct request *rq, int error,
940 unsigned int nr_bytes);
941 extern void __blk_end_request_all(struct request *rq, int error);
942 extern bool __blk_end_request_cur(struct request *rq, int error);
943 extern bool __blk_end_request_err(struct request *rq, int error);
944
945 extern void blk_complete_request(struct request *);
946 extern void __blk_complete_request(struct request *);
947 extern void blk_abort_request(struct request *);
948 extern void blk_unprep_request(struct request *);
949
950 /*
951 * Access functions for manipulating queue properties
952 */
953 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
954 spinlock_t *lock, int node_id);
955 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
956 extern struct request_queue *blk_init_allocated_queue(struct request_queue *,
957 request_fn_proc *, spinlock_t *);
958 extern void blk_cleanup_queue(struct request_queue *);
959 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
960 extern void blk_queue_bounce_limit(struct request_queue *, u64);
961 extern void blk_limits_max_hw_sectors(struct queue_limits *, unsigned int);
962 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
963 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
964 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
965 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
966 extern void blk_queue_max_discard_sectors(struct request_queue *q,
967 unsigned int max_discard_sectors);
968 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
969 unsigned int max_write_same_sectors);
970 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
971 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
972 extern void blk_queue_alignment_offset(struct request_queue *q,
973 unsigned int alignment);
974 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
975 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
976 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
977 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
978 extern void blk_set_default_limits(struct queue_limits *lim);
979 extern void blk_set_stacking_limits(struct queue_limits *lim);
980 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
981 sector_t offset);
982 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
983 sector_t offset);
984 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
985 sector_t offset);
986 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
987 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
988 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
989 extern int blk_queue_dma_drain(struct request_queue *q,
990 dma_drain_needed_fn *dma_drain_needed,
991 void *buf, unsigned int size);
992 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
993 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
994 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
995 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
996 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
997 extern void blk_queue_dma_alignment(struct request_queue *, int);
998 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
999 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1000 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1001 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1002 extern void blk_queue_flush(struct request_queue *q, unsigned int flush);
1003 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1004 extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
1005
1006 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1007 extern void blk_dump_rq_flags(struct request *, char *);
1008 extern long nr_blockdev_pages(void);
1009
1010 bool __must_check blk_get_queue(struct request_queue *);
1011 struct request_queue *blk_alloc_queue(gfp_t);
1012 struct request_queue *blk_alloc_queue_node(gfp_t, int);
1013 extern void blk_put_queue(struct request_queue *);
1014 extern void blk_set_queue_dying(struct request_queue *);
1015
1016 /*
1017 * block layer runtime pm functions
1018 */
1019 #ifdef CONFIG_PM
1020 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1021 extern int blk_pre_runtime_suspend(struct request_queue *q);
1022 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1023 extern void blk_pre_runtime_resume(struct request_queue *q);
1024 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1025 #else
1026 static inline void blk_pm_runtime_init(struct request_queue *q,
1027 struct device *dev) {}
1028 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1029 {
1030 return -ENOSYS;
1031 }
1032 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1033 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1034 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1035 #endif
1036
1037 /*
1038 * blk_plug permits building a queue of related requests by holding the I/O
1039 * fragments for a short period. This allows merging of sequential requests
1040 * into single larger request. As the requests are moved from a per-task list to
1041 * the device's request_queue in a batch, this results in improved scalability
1042 * as the lock contention for request_queue lock is reduced.
1043 *
1044 * It is ok not to disable preemption when adding the request to the plug list
1045 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1046 * the plug list when the task sleeps by itself. For details, please see
1047 * schedule() where blk_schedule_flush_plug() is called.
1048 */
1049 struct blk_plug {
1050 struct list_head list; /* requests */
1051 struct list_head mq_list; /* blk-mq requests */
1052 struct list_head cb_list; /* md requires an unplug callback */
1053 };
1054 #define BLK_MAX_REQUEST_COUNT 16
1055
1056 struct blk_plug_cb;
1057 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1058 struct blk_plug_cb {
1059 struct list_head list;
1060 blk_plug_cb_fn callback;
1061 void *data;
1062 };
1063 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1064 void *data, int size);
1065 extern void blk_start_plug(struct blk_plug *);
1066 extern void blk_finish_plug(struct blk_plug *);
1067 extern void blk_flush_plug_list(struct blk_plug *, bool);
1068
1069 static inline void blk_flush_plug(struct task_struct *tsk)
1070 {
1071 struct blk_plug *plug = tsk->plug;
1072
1073 if (plug)
1074 blk_flush_plug_list(plug, false);
1075 }
1076
1077 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1078 {
1079 struct blk_plug *plug = tsk->plug;
1080
1081 if (plug)
1082 blk_flush_plug_list(plug, true);
1083 }
1084
1085 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1086 {
1087 struct blk_plug *plug = tsk->plug;
1088
1089 return plug &&
1090 (!list_empty(&plug->list) ||
1091 !list_empty(&plug->mq_list) ||
1092 !list_empty(&plug->cb_list));
1093 }
1094
1095 /*
1096 * tag stuff
1097 */
1098 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1099 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1100 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1101 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1102 extern void blk_queue_free_tags(struct request_queue *);
1103 extern int blk_queue_resize_tags(struct request_queue *, int);
1104 extern void blk_queue_invalidate_tags(struct request_queue *);
1105 extern struct blk_queue_tag *blk_init_tags(int, int);
1106 extern void blk_free_tags(struct blk_queue_tag *);
1107
1108 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1109 int tag)
1110 {
1111 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1112 return NULL;
1113 return bqt->tag_index[tag];
1114 }
1115
1116 #define BLKDEV_DISCARD_SECURE 0x01 /* secure discard */
1117
1118 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1119 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1120 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1121 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1122 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1123 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1124 sector_t nr_sects, gfp_t gfp_mask, bool discard);
1125 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1126 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1127 {
1128 return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9),
1129 nr_blocks << (sb->s_blocksize_bits - 9),
1130 gfp_mask, flags);
1131 }
1132 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1133 sector_t nr_blocks, gfp_t gfp_mask)
1134 {
1135 return blkdev_issue_zeroout(sb->s_bdev,
1136 block << (sb->s_blocksize_bits - 9),
1137 nr_blocks << (sb->s_blocksize_bits - 9),
1138 gfp_mask, true);
1139 }
1140
1141 extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm);
1142
1143 enum blk_default_limits {
1144 BLK_MAX_SEGMENTS = 128,
1145 BLK_SAFE_MAX_SECTORS = 255,
1146 BLK_DEF_MAX_SECTORS = 2560,
1147 BLK_MAX_SEGMENT_SIZE = 65536,
1148 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1149 };
1150
1151 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1152
1153 static inline unsigned long queue_bounce_pfn(struct request_queue *q)
1154 {
1155 return q->limits.bounce_pfn;
1156 }
1157
1158 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1159 {
1160 return q->limits.seg_boundary_mask;
1161 }
1162
1163 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1164 {
1165 return q->limits.virt_boundary_mask;
1166 }
1167
1168 static inline unsigned int queue_max_sectors(struct request_queue *q)
1169 {
1170 return q->limits.max_sectors;
1171 }
1172
1173 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1174 {
1175 return q->limits.max_hw_sectors;
1176 }
1177
1178 static inline unsigned short queue_max_segments(struct request_queue *q)
1179 {
1180 return q->limits.max_segments;
1181 }
1182
1183 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1184 {
1185 return q->limits.max_segment_size;
1186 }
1187
1188 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1189 {
1190 int retval = 512;
1191
1192 if (q && q->limits.logical_block_size)
1193 retval = q->limits.logical_block_size;
1194
1195 return retval;
1196 }
1197
1198 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1199 {
1200 return queue_logical_block_size(bdev_get_queue(bdev));
1201 }
1202
1203 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1204 {
1205 return q->limits.physical_block_size;
1206 }
1207
1208 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1209 {
1210 return queue_physical_block_size(bdev_get_queue(bdev));
1211 }
1212
1213 static inline unsigned int queue_io_min(struct request_queue *q)
1214 {
1215 return q->limits.io_min;
1216 }
1217
1218 static inline int bdev_io_min(struct block_device *bdev)
1219 {
1220 return queue_io_min(bdev_get_queue(bdev));
1221 }
1222
1223 static inline unsigned int queue_io_opt(struct request_queue *q)
1224 {
1225 return q->limits.io_opt;
1226 }
1227
1228 static inline int bdev_io_opt(struct block_device *bdev)
1229 {
1230 return queue_io_opt(bdev_get_queue(bdev));
1231 }
1232
1233 static inline int queue_alignment_offset(struct request_queue *q)
1234 {
1235 if (q->limits.misaligned)
1236 return -1;
1237
1238 return q->limits.alignment_offset;
1239 }
1240
1241 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1242 {
1243 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1244 unsigned int alignment = sector_div(sector, granularity >> 9) << 9;
1245
1246 return (granularity + lim->alignment_offset - alignment) % granularity;
1247 }
1248
1249 static inline int bdev_alignment_offset(struct block_device *bdev)
1250 {
1251 struct request_queue *q = bdev_get_queue(bdev);
1252
1253 if (q->limits.misaligned)
1254 return -1;
1255
1256 if (bdev != bdev->bd_contains)
1257 return bdev->bd_part->alignment_offset;
1258
1259 return q->limits.alignment_offset;
1260 }
1261
1262 static inline int queue_discard_alignment(struct request_queue *q)
1263 {
1264 if (q->limits.discard_misaligned)
1265 return -1;
1266
1267 return q->limits.discard_alignment;
1268 }
1269
1270 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1271 {
1272 unsigned int alignment, granularity, offset;
1273
1274 if (!lim->max_discard_sectors)
1275 return 0;
1276
1277 /* Why are these in bytes, not sectors? */
1278 alignment = lim->discard_alignment >> 9;
1279 granularity = lim->discard_granularity >> 9;
1280 if (!granularity)
1281 return 0;
1282
1283 /* Offset of the partition start in 'granularity' sectors */
1284 offset = sector_div(sector, granularity);
1285
1286 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1287 offset = (granularity + alignment - offset) % granularity;
1288
1289 /* Turn it back into bytes, gaah */
1290 return offset << 9;
1291 }
1292
1293 static inline int bdev_discard_alignment(struct block_device *bdev)
1294 {
1295 struct request_queue *q = bdev_get_queue(bdev);
1296
1297 if (bdev != bdev->bd_contains)
1298 return bdev->bd_part->discard_alignment;
1299
1300 return q->limits.discard_alignment;
1301 }
1302
1303 static inline unsigned int queue_discard_zeroes_data(struct request_queue *q)
1304 {
1305 if (q->limits.max_discard_sectors && q->limits.discard_zeroes_data == 1)
1306 return 1;
1307
1308 return 0;
1309 }
1310
1311 static inline unsigned int bdev_discard_zeroes_data(struct block_device *bdev)
1312 {
1313 return queue_discard_zeroes_data(bdev_get_queue(bdev));
1314 }
1315
1316 static inline unsigned int bdev_write_same(struct block_device *bdev)
1317 {
1318 struct request_queue *q = bdev_get_queue(bdev);
1319
1320 if (q)
1321 return q->limits.max_write_same_sectors;
1322
1323 return 0;
1324 }
1325
1326 static inline int queue_dma_alignment(struct request_queue *q)
1327 {
1328 return q ? q->dma_alignment : 511;
1329 }
1330
1331 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1332 unsigned int len)
1333 {
1334 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1335 return !(addr & alignment) && !(len & alignment);
1336 }
1337
1338 /* assumes size > 256 */
1339 static inline unsigned int blksize_bits(unsigned int size)
1340 {
1341 unsigned int bits = 8;
1342 do {
1343 bits++;
1344 size >>= 1;
1345 } while (size > 256);
1346 return bits;
1347 }
1348
1349 static inline unsigned int block_size(struct block_device *bdev)
1350 {
1351 return bdev->bd_block_size;
1352 }
1353
1354 static inline bool queue_flush_queueable(struct request_queue *q)
1355 {
1356 return !q->flush_not_queueable;
1357 }
1358
1359 typedef struct {struct page *v;} Sector;
1360
1361 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1362
1363 static inline void put_dev_sector(Sector p)
1364 {
1365 page_cache_release(p.v);
1366 }
1367
1368 /*
1369 * Check if adding a bio_vec after bprv with offset would create a gap in
1370 * the SG list. Most drivers don't care about this, but some do.
1371 */
1372 static inline bool bvec_gap_to_prev(struct request_queue *q,
1373 struct bio_vec *bprv, unsigned int offset)
1374 {
1375 if (!queue_virt_boundary(q))
1376 return false;
1377 return offset ||
1378 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1379 }
1380
1381 static inline bool bio_will_gap(struct request_queue *q, struct bio *prev,
1382 struct bio *next)
1383 {
1384 if (!bio_has_data(prev))
1385 return false;
1386
1387 return bvec_gap_to_prev(q, &prev->bi_io_vec[prev->bi_vcnt - 1],
1388 next->bi_io_vec[0].bv_offset);
1389 }
1390
1391 static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
1392 {
1393 return bio_will_gap(req->q, req->biotail, bio);
1394 }
1395
1396 static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
1397 {
1398 return bio_will_gap(req->q, bio, req->bio);
1399 }
1400
1401 struct work_struct;
1402 int kblockd_schedule_work(struct work_struct *work);
1403 int kblockd_schedule_delayed_work(struct delayed_work *dwork, unsigned long delay);
1404 int kblockd_schedule_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1405
1406 #ifdef CONFIG_BLK_CGROUP
1407 /*
1408 * This should not be using sched_clock(). A real patch is in progress
1409 * to fix this up, until that is in place we need to disable preemption
1410 * around sched_clock() in this function and set_io_start_time_ns().
1411 */
1412 static inline void set_start_time_ns(struct request *req)
1413 {
1414 preempt_disable();
1415 req->start_time_ns = sched_clock();
1416 preempt_enable();
1417 }
1418
1419 static inline void set_io_start_time_ns(struct request *req)
1420 {
1421 preempt_disable();
1422 req->io_start_time_ns = sched_clock();
1423 preempt_enable();
1424 }
1425
1426 static inline uint64_t rq_start_time_ns(struct request *req)
1427 {
1428 return req->start_time_ns;
1429 }
1430
1431 static inline uint64_t rq_io_start_time_ns(struct request *req)
1432 {
1433 return req->io_start_time_ns;
1434 }
1435 #else
1436 static inline void set_start_time_ns(struct request *req) {}
1437 static inline void set_io_start_time_ns(struct request *req) {}
1438 static inline uint64_t rq_start_time_ns(struct request *req)
1439 {
1440 return 0;
1441 }
1442 static inline uint64_t rq_io_start_time_ns(struct request *req)
1443 {
1444 return 0;
1445 }
1446 #endif
1447
1448 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1449 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1450 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1451 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1452
1453 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1454
1455 enum blk_integrity_flags {
1456 BLK_INTEGRITY_VERIFY = 1 << 0,
1457 BLK_INTEGRITY_GENERATE = 1 << 1,
1458 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2,
1459 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3,
1460 };
1461
1462 struct blk_integrity_iter {
1463 void *prot_buf;
1464 void *data_buf;
1465 sector_t seed;
1466 unsigned int data_size;
1467 unsigned short interval;
1468 const char *disk_name;
1469 };
1470
1471 typedef int (integrity_processing_fn) (struct blk_integrity_iter *);
1472
1473 struct blk_integrity_profile {
1474 integrity_processing_fn *generate_fn;
1475 integrity_processing_fn *verify_fn;
1476 const char *name;
1477 };
1478
1479 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1480 extern void blk_integrity_unregister(struct gendisk *);
1481 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1482 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1483 struct scatterlist *);
1484 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1485 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1486 struct request *);
1487 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1488 struct bio *);
1489
1490 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1491 {
1492 struct blk_integrity *bi = &disk->queue->integrity;
1493
1494 if (!bi->profile)
1495 return NULL;
1496
1497 return bi;
1498 }
1499
1500 static inline
1501 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1502 {
1503 return blk_get_integrity(bdev->bd_disk);
1504 }
1505
1506 static inline bool blk_integrity_rq(struct request *rq)
1507 {
1508 return rq->cmd_flags & REQ_INTEGRITY;
1509 }
1510
1511 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1512 unsigned int segs)
1513 {
1514 q->limits.max_integrity_segments = segs;
1515 }
1516
1517 static inline unsigned short
1518 queue_max_integrity_segments(struct request_queue *q)
1519 {
1520 return q->limits.max_integrity_segments;
1521 }
1522
1523 static inline bool integrity_req_gap_back_merge(struct request *req,
1524 struct bio *next)
1525 {
1526 struct bio_integrity_payload *bip = bio_integrity(req->bio);
1527 struct bio_integrity_payload *bip_next = bio_integrity(next);
1528
1529 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1530 bip_next->bip_vec[0].bv_offset);
1531 }
1532
1533 static inline bool integrity_req_gap_front_merge(struct request *req,
1534 struct bio *bio)
1535 {
1536 struct bio_integrity_payload *bip = bio_integrity(bio);
1537 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
1538
1539 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1540 bip_next->bip_vec[0].bv_offset);
1541 }
1542
1543 #else /* CONFIG_BLK_DEV_INTEGRITY */
1544
1545 struct bio;
1546 struct block_device;
1547 struct gendisk;
1548 struct blk_integrity;
1549
1550 static inline int blk_integrity_rq(struct request *rq)
1551 {
1552 return 0;
1553 }
1554 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1555 struct bio *b)
1556 {
1557 return 0;
1558 }
1559 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1560 struct bio *b,
1561 struct scatterlist *s)
1562 {
1563 return 0;
1564 }
1565 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1566 {
1567 return NULL;
1568 }
1569 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1570 {
1571 return NULL;
1572 }
1573 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1574 {
1575 return 0;
1576 }
1577 static inline void blk_integrity_register(struct gendisk *d,
1578 struct blk_integrity *b)
1579 {
1580 }
1581 static inline void blk_integrity_unregister(struct gendisk *d)
1582 {
1583 }
1584 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1585 unsigned int segs)
1586 {
1587 }
1588 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1589 {
1590 return 0;
1591 }
1592 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1593 struct request *r1,
1594 struct request *r2)
1595 {
1596 return true;
1597 }
1598 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1599 struct request *r,
1600 struct bio *b)
1601 {
1602 return true;
1603 }
1604
1605 static inline bool integrity_req_gap_back_merge(struct request *req,
1606 struct bio *next)
1607 {
1608 return false;
1609 }
1610 static inline bool integrity_req_gap_front_merge(struct request *req,
1611 struct bio *bio)
1612 {
1613 return false;
1614 }
1615
1616 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1617
1618 struct block_device_operations {
1619 int (*open) (struct block_device *, fmode_t);
1620 void (*release) (struct gendisk *, fmode_t);
1621 int (*rw_page)(struct block_device *, sector_t, struct page *, int rw);
1622 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1623 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1624 long (*direct_access)(struct block_device *, sector_t, void __pmem **,
1625 unsigned long *pfn);
1626 unsigned int (*check_events) (struct gendisk *disk,
1627 unsigned int clearing);
1628 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1629 int (*media_changed) (struct gendisk *);
1630 void (*unlock_native_capacity) (struct gendisk *);
1631 int (*revalidate_disk) (struct gendisk *);
1632 int (*getgeo)(struct block_device *, struct hd_geometry *);
1633 /* this callback is with swap_lock and sometimes page table lock held */
1634 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1635 struct module *owner;
1636 const struct pr_ops *pr_ops;
1637 };
1638
1639 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1640 unsigned long);
1641 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1642 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1643 struct writeback_control *);
1644 extern long bdev_direct_access(struct block_device *, sector_t,
1645 void __pmem **addr, unsigned long *pfn, long size);
1646 #else /* CONFIG_BLOCK */
1647
1648 struct block_device;
1649
1650 /*
1651 * stubs for when the block layer is configured out
1652 */
1653 #define buffer_heads_over_limit 0
1654
1655 static inline long nr_blockdev_pages(void)
1656 {
1657 return 0;
1658 }
1659
1660 struct blk_plug {
1661 };
1662
1663 static inline void blk_start_plug(struct blk_plug *plug)
1664 {
1665 }
1666
1667 static inline void blk_finish_plug(struct blk_plug *plug)
1668 {
1669 }
1670
1671 static inline void blk_flush_plug(struct task_struct *task)
1672 {
1673 }
1674
1675 static inline void blk_schedule_flush_plug(struct task_struct *task)
1676 {
1677 }
1678
1679
1680 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1681 {
1682 return false;
1683 }
1684
1685 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
1686 sector_t *error_sector)
1687 {
1688 return 0;
1689 }
1690
1691 #endif /* CONFIG_BLOCK */
1692
1693 #endif