]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blob - include/linux/blkdev.h
Merge tag 'perf-core-for-mingo-20160323' of git://git.kernel.org/pub/scm/linux/kernel...
[mirror_ubuntu-zesty-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/pfn.h>
19 #include <linux/bio.h>
20 #include <linux/stringify.h>
21 #include <linux/gfp.h>
22 #include <linux/bsg.h>
23 #include <linux/smp.h>
24 #include <linux/rcupdate.h>
25 #include <linux/percpu-refcount.h>
26 #include <linux/scatterlist.h>
27
28 struct module;
29 struct scsi_ioctl_command;
30
31 struct request_queue;
32 struct elevator_queue;
33 struct blk_trace;
34 struct request;
35 struct sg_io_hdr;
36 struct bsg_job;
37 struct blkcg_gq;
38 struct blk_flush_queue;
39 struct pr_ops;
40
41 #define BLKDEV_MIN_RQ 4
42 #define BLKDEV_MAX_RQ 128 /* Default maximum */
43
44 /*
45 * Maximum number of blkcg policies allowed to be registered concurrently.
46 * Defined here to simplify include dependency.
47 */
48 #define BLKCG_MAX_POLS 2
49
50 struct request;
51 typedef void (rq_end_io_fn)(struct request *, int);
52
53 #define BLK_RL_SYNCFULL (1U << 0)
54 #define BLK_RL_ASYNCFULL (1U << 1)
55
56 struct request_list {
57 struct request_queue *q; /* the queue this rl belongs to */
58 #ifdef CONFIG_BLK_CGROUP
59 struct blkcg_gq *blkg; /* blkg this request pool belongs to */
60 #endif
61 /*
62 * count[], starved[], and wait[] are indexed by
63 * BLK_RW_SYNC/BLK_RW_ASYNC
64 */
65 int count[2];
66 int starved[2];
67 mempool_t *rq_pool;
68 wait_queue_head_t wait[2];
69 unsigned int flags;
70 };
71
72 /*
73 * request command types
74 */
75 enum rq_cmd_type_bits {
76 REQ_TYPE_FS = 1, /* fs request */
77 REQ_TYPE_BLOCK_PC, /* scsi command */
78 REQ_TYPE_DRV_PRIV, /* driver defined types from here */
79 };
80
81 #define BLK_MAX_CDB 16
82
83 /*
84 * Try to put the fields that are referenced together in the same cacheline.
85 *
86 * If you modify this structure, make sure to update blk_rq_init() and
87 * especially blk_mq_rq_ctx_init() to take care of the added fields.
88 */
89 struct request {
90 struct list_head queuelist;
91 union {
92 struct call_single_data csd;
93 unsigned long fifo_time;
94 };
95
96 struct request_queue *q;
97 struct blk_mq_ctx *mq_ctx;
98
99 u64 cmd_flags;
100 unsigned cmd_type;
101 unsigned long atomic_flags;
102
103 int cpu;
104
105 /* the following two fields are internal, NEVER access directly */
106 unsigned int __data_len; /* total data len */
107 sector_t __sector; /* sector cursor */
108
109 struct bio *bio;
110 struct bio *biotail;
111
112 /*
113 * The hash is used inside the scheduler, and killed once the
114 * request reaches the dispatch list. The ipi_list is only used
115 * to queue the request for softirq completion, which is long
116 * after the request has been unhashed (and even removed from
117 * the dispatch list).
118 */
119 union {
120 struct hlist_node hash; /* merge hash */
121 struct list_head ipi_list;
122 };
123
124 /*
125 * The rb_node is only used inside the io scheduler, requests
126 * are pruned when moved to the dispatch queue. So let the
127 * completion_data share space with the rb_node.
128 */
129 union {
130 struct rb_node rb_node; /* sort/lookup */
131 void *completion_data;
132 };
133
134 /*
135 * Three pointers are available for the IO schedulers, if they need
136 * more they have to dynamically allocate it. Flush requests are
137 * never put on the IO scheduler. So let the flush fields share
138 * space with the elevator data.
139 */
140 union {
141 struct {
142 struct io_cq *icq;
143 void *priv[2];
144 } elv;
145
146 struct {
147 unsigned int seq;
148 struct list_head list;
149 rq_end_io_fn *saved_end_io;
150 } flush;
151 };
152
153 struct gendisk *rq_disk;
154 struct hd_struct *part;
155 unsigned long start_time;
156 #ifdef CONFIG_BLK_CGROUP
157 struct request_list *rl; /* rl this rq is alloced from */
158 unsigned long long start_time_ns;
159 unsigned long long io_start_time_ns; /* when passed to hardware */
160 #endif
161 /* Number of scatter-gather DMA addr+len pairs after
162 * physical address coalescing is performed.
163 */
164 unsigned short nr_phys_segments;
165 #if defined(CONFIG_BLK_DEV_INTEGRITY)
166 unsigned short nr_integrity_segments;
167 #endif
168
169 unsigned short ioprio;
170
171 void *special; /* opaque pointer available for LLD use */
172
173 int tag;
174 int errors;
175
176 /*
177 * when request is used as a packet command carrier
178 */
179 unsigned char __cmd[BLK_MAX_CDB];
180 unsigned char *cmd;
181 unsigned short cmd_len;
182
183 unsigned int extra_len; /* length of alignment and padding */
184 unsigned int sense_len;
185 unsigned int resid_len; /* residual count */
186 void *sense;
187
188 unsigned long deadline;
189 struct list_head timeout_list;
190 unsigned int timeout;
191 int retries;
192
193 /*
194 * completion callback.
195 */
196 rq_end_io_fn *end_io;
197 void *end_io_data;
198
199 /* for bidi */
200 struct request *next_rq;
201 };
202
203 static inline unsigned short req_get_ioprio(struct request *req)
204 {
205 return req->ioprio;
206 }
207
208 #include <linux/elevator.h>
209
210 struct blk_queue_ctx;
211
212 typedef void (request_fn_proc) (struct request_queue *q);
213 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio);
214 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
215 typedef void (unprep_rq_fn) (struct request_queue *, struct request *);
216
217 struct bio_vec;
218 typedef void (softirq_done_fn)(struct request *);
219 typedef int (dma_drain_needed_fn)(struct request *);
220 typedef int (lld_busy_fn) (struct request_queue *q);
221 typedef int (bsg_job_fn) (struct bsg_job *);
222
223 enum blk_eh_timer_return {
224 BLK_EH_NOT_HANDLED,
225 BLK_EH_HANDLED,
226 BLK_EH_RESET_TIMER,
227 };
228
229 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
230
231 enum blk_queue_state {
232 Queue_down,
233 Queue_up,
234 };
235
236 struct blk_queue_tag {
237 struct request **tag_index; /* map of busy tags */
238 unsigned long *tag_map; /* bit map of free/busy tags */
239 int busy; /* current depth */
240 int max_depth; /* what we will send to device */
241 int real_max_depth; /* what the array can hold */
242 atomic_t refcnt; /* map can be shared */
243 int alloc_policy; /* tag allocation policy */
244 int next_tag; /* next tag */
245 };
246 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
247 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
248
249 #define BLK_SCSI_MAX_CMDS (256)
250 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
251
252 struct queue_limits {
253 unsigned long bounce_pfn;
254 unsigned long seg_boundary_mask;
255 unsigned long virt_boundary_mask;
256
257 unsigned int max_hw_sectors;
258 unsigned int max_dev_sectors;
259 unsigned int chunk_sectors;
260 unsigned int max_sectors;
261 unsigned int max_segment_size;
262 unsigned int physical_block_size;
263 unsigned int alignment_offset;
264 unsigned int io_min;
265 unsigned int io_opt;
266 unsigned int max_discard_sectors;
267 unsigned int max_hw_discard_sectors;
268 unsigned int max_write_same_sectors;
269 unsigned int discard_granularity;
270 unsigned int discard_alignment;
271
272 unsigned short logical_block_size;
273 unsigned short max_segments;
274 unsigned short max_integrity_segments;
275
276 unsigned char misaligned;
277 unsigned char discard_misaligned;
278 unsigned char cluster;
279 unsigned char discard_zeroes_data;
280 unsigned char raid_partial_stripes_expensive;
281 };
282
283 struct request_queue {
284 /*
285 * Together with queue_head for cacheline sharing
286 */
287 struct list_head queue_head;
288 struct request *last_merge;
289 struct elevator_queue *elevator;
290 int nr_rqs[2]; /* # allocated [a]sync rqs */
291 int nr_rqs_elvpriv; /* # allocated rqs w/ elvpriv */
292
293 /*
294 * If blkcg is not used, @q->root_rl serves all requests. If blkcg
295 * is used, root blkg allocates from @q->root_rl and all other
296 * blkgs from their own blkg->rl. Which one to use should be
297 * determined using bio_request_list().
298 */
299 struct request_list root_rl;
300
301 request_fn_proc *request_fn;
302 make_request_fn *make_request_fn;
303 prep_rq_fn *prep_rq_fn;
304 unprep_rq_fn *unprep_rq_fn;
305 softirq_done_fn *softirq_done_fn;
306 rq_timed_out_fn *rq_timed_out_fn;
307 dma_drain_needed_fn *dma_drain_needed;
308 lld_busy_fn *lld_busy_fn;
309
310 struct blk_mq_ops *mq_ops;
311
312 unsigned int *mq_map;
313
314 /* sw queues */
315 struct blk_mq_ctx __percpu *queue_ctx;
316 unsigned int nr_queues;
317
318 /* hw dispatch queues */
319 struct blk_mq_hw_ctx **queue_hw_ctx;
320 unsigned int nr_hw_queues;
321
322 /*
323 * Dispatch queue sorting
324 */
325 sector_t end_sector;
326 struct request *boundary_rq;
327
328 /*
329 * Delayed queue handling
330 */
331 struct delayed_work delay_work;
332
333 struct backing_dev_info backing_dev_info;
334
335 /*
336 * The queue owner gets to use this for whatever they like.
337 * ll_rw_blk doesn't touch it.
338 */
339 void *queuedata;
340
341 /*
342 * various queue flags, see QUEUE_* below
343 */
344 unsigned long queue_flags;
345
346 /*
347 * ida allocated id for this queue. Used to index queues from
348 * ioctx.
349 */
350 int id;
351
352 /*
353 * queue needs bounce pages for pages above this limit
354 */
355 gfp_t bounce_gfp;
356
357 /*
358 * protects queue structures from reentrancy. ->__queue_lock should
359 * _never_ be used directly, it is queue private. always use
360 * ->queue_lock.
361 */
362 spinlock_t __queue_lock;
363 spinlock_t *queue_lock;
364
365 /*
366 * queue kobject
367 */
368 struct kobject kobj;
369
370 /*
371 * mq queue kobject
372 */
373 struct kobject mq_kobj;
374
375 #ifdef CONFIG_BLK_DEV_INTEGRITY
376 struct blk_integrity integrity;
377 #endif /* CONFIG_BLK_DEV_INTEGRITY */
378
379 #ifdef CONFIG_PM
380 struct device *dev;
381 int rpm_status;
382 unsigned int nr_pending;
383 #endif
384
385 /*
386 * queue settings
387 */
388 unsigned long nr_requests; /* Max # of requests */
389 unsigned int nr_congestion_on;
390 unsigned int nr_congestion_off;
391 unsigned int nr_batching;
392
393 unsigned int dma_drain_size;
394 void *dma_drain_buffer;
395 unsigned int dma_pad_mask;
396 unsigned int dma_alignment;
397
398 struct blk_queue_tag *queue_tags;
399 struct list_head tag_busy_list;
400
401 unsigned int nr_sorted;
402 unsigned int in_flight[2];
403 /*
404 * Number of active block driver functions for which blk_drain_queue()
405 * must wait. Must be incremented around functions that unlock the
406 * queue_lock internally, e.g. scsi_request_fn().
407 */
408 unsigned int request_fn_active;
409
410 unsigned int rq_timeout;
411 struct timer_list timeout;
412 struct work_struct timeout_work;
413 struct list_head timeout_list;
414
415 struct list_head icq_list;
416 #ifdef CONFIG_BLK_CGROUP
417 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
418 struct blkcg_gq *root_blkg;
419 struct list_head blkg_list;
420 #endif
421
422 struct queue_limits limits;
423
424 /*
425 * sg stuff
426 */
427 unsigned int sg_timeout;
428 unsigned int sg_reserved_size;
429 int node;
430 #ifdef CONFIG_BLK_DEV_IO_TRACE
431 struct blk_trace *blk_trace;
432 #endif
433 /*
434 * for flush operations
435 */
436 unsigned int flush_flags;
437 unsigned int flush_not_queueable:1;
438 struct blk_flush_queue *fq;
439
440 struct list_head requeue_list;
441 spinlock_t requeue_lock;
442 struct work_struct requeue_work;
443
444 struct mutex sysfs_lock;
445
446 int bypass_depth;
447 atomic_t mq_freeze_depth;
448
449 #if defined(CONFIG_BLK_DEV_BSG)
450 bsg_job_fn *bsg_job_fn;
451 int bsg_job_size;
452 struct bsg_class_device bsg_dev;
453 #endif
454
455 #ifdef CONFIG_BLK_DEV_THROTTLING
456 /* Throttle data */
457 struct throtl_data *td;
458 #endif
459 struct rcu_head rcu_head;
460 wait_queue_head_t mq_freeze_wq;
461 struct percpu_ref q_usage_counter;
462 struct list_head all_q_node;
463
464 struct blk_mq_tag_set *tag_set;
465 struct list_head tag_set_list;
466 struct bio_set *bio_split;
467
468 bool mq_sysfs_init_done;
469 };
470
471 #define QUEUE_FLAG_QUEUED 1 /* uses generic tag queueing */
472 #define QUEUE_FLAG_STOPPED 2 /* queue is stopped */
473 #define QUEUE_FLAG_SYNCFULL 3 /* read queue has been filled */
474 #define QUEUE_FLAG_ASYNCFULL 4 /* write queue has been filled */
475 #define QUEUE_FLAG_DYING 5 /* queue being torn down */
476 #define QUEUE_FLAG_BYPASS 6 /* act as dumb FIFO queue */
477 #define QUEUE_FLAG_BIDI 7 /* queue supports bidi requests */
478 #define QUEUE_FLAG_NOMERGES 8 /* disable merge attempts */
479 #define QUEUE_FLAG_SAME_COMP 9 /* complete on same CPU-group */
480 #define QUEUE_FLAG_FAIL_IO 10 /* fake timeout */
481 #define QUEUE_FLAG_STACKABLE 11 /* supports request stacking */
482 #define QUEUE_FLAG_NONROT 12 /* non-rotational device (SSD) */
483 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
484 #define QUEUE_FLAG_IO_STAT 13 /* do IO stats */
485 #define QUEUE_FLAG_DISCARD 14 /* supports DISCARD */
486 #define QUEUE_FLAG_NOXMERGES 15 /* No extended merges */
487 #define QUEUE_FLAG_ADD_RANDOM 16 /* Contributes to random pool */
488 #define QUEUE_FLAG_SECDISCARD 17 /* supports SECDISCARD */
489 #define QUEUE_FLAG_SAME_FORCE 18 /* force complete on same CPU */
490 #define QUEUE_FLAG_DEAD 19 /* queue tear-down finished */
491 #define QUEUE_FLAG_INIT_DONE 20 /* queue is initialized */
492 #define QUEUE_FLAG_NO_SG_MERGE 21 /* don't attempt to merge SG segments*/
493 #define QUEUE_FLAG_POLL 22 /* IO polling enabled if set */
494
495 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
496 (1 << QUEUE_FLAG_STACKABLE) | \
497 (1 << QUEUE_FLAG_SAME_COMP) | \
498 (1 << QUEUE_FLAG_ADD_RANDOM))
499
500 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
501 (1 << QUEUE_FLAG_STACKABLE) | \
502 (1 << QUEUE_FLAG_SAME_COMP))
503
504 static inline void queue_lockdep_assert_held(struct request_queue *q)
505 {
506 if (q->queue_lock)
507 lockdep_assert_held(q->queue_lock);
508 }
509
510 static inline void queue_flag_set_unlocked(unsigned int flag,
511 struct request_queue *q)
512 {
513 __set_bit(flag, &q->queue_flags);
514 }
515
516 static inline int queue_flag_test_and_clear(unsigned int flag,
517 struct request_queue *q)
518 {
519 queue_lockdep_assert_held(q);
520
521 if (test_bit(flag, &q->queue_flags)) {
522 __clear_bit(flag, &q->queue_flags);
523 return 1;
524 }
525
526 return 0;
527 }
528
529 static inline int queue_flag_test_and_set(unsigned int flag,
530 struct request_queue *q)
531 {
532 queue_lockdep_assert_held(q);
533
534 if (!test_bit(flag, &q->queue_flags)) {
535 __set_bit(flag, &q->queue_flags);
536 return 0;
537 }
538
539 return 1;
540 }
541
542 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
543 {
544 queue_lockdep_assert_held(q);
545 __set_bit(flag, &q->queue_flags);
546 }
547
548 static inline void queue_flag_clear_unlocked(unsigned int flag,
549 struct request_queue *q)
550 {
551 __clear_bit(flag, &q->queue_flags);
552 }
553
554 static inline int queue_in_flight(struct request_queue *q)
555 {
556 return q->in_flight[0] + q->in_flight[1];
557 }
558
559 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
560 {
561 queue_lockdep_assert_held(q);
562 __clear_bit(flag, &q->queue_flags);
563 }
564
565 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
566 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
567 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
568 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
569 #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
570 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
571 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
572 #define blk_queue_noxmerges(q) \
573 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
574 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
575 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
576 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
577 #define blk_queue_stackable(q) \
578 test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
579 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
580 #define blk_queue_secdiscard(q) (blk_queue_discard(q) && \
581 test_bit(QUEUE_FLAG_SECDISCARD, &(q)->queue_flags))
582
583 #define blk_noretry_request(rq) \
584 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
585 REQ_FAILFAST_DRIVER))
586
587 #define blk_account_rq(rq) \
588 (((rq)->cmd_flags & REQ_STARTED) && \
589 ((rq)->cmd_type == REQ_TYPE_FS))
590
591 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
592 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
593 /* rq->queuelist of dequeued request must be list_empty() */
594 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
595
596 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
597
598 #define rq_data_dir(rq) ((int)((rq)->cmd_flags & 1))
599
600 /*
601 * Driver can handle struct request, if it either has an old style
602 * request_fn defined, or is blk-mq based.
603 */
604 static inline bool queue_is_rq_based(struct request_queue *q)
605 {
606 return q->request_fn || q->mq_ops;
607 }
608
609 static inline unsigned int blk_queue_cluster(struct request_queue *q)
610 {
611 return q->limits.cluster;
612 }
613
614 /*
615 * We regard a request as sync, if either a read or a sync write
616 */
617 static inline bool rw_is_sync(unsigned int rw_flags)
618 {
619 return !(rw_flags & REQ_WRITE) || (rw_flags & REQ_SYNC);
620 }
621
622 static inline bool rq_is_sync(struct request *rq)
623 {
624 return rw_is_sync(rq->cmd_flags);
625 }
626
627 static inline bool blk_rl_full(struct request_list *rl, bool sync)
628 {
629 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
630
631 return rl->flags & flag;
632 }
633
634 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
635 {
636 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
637
638 rl->flags |= flag;
639 }
640
641 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
642 {
643 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
644
645 rl->flags &= ~flag;
646 }
647
648 static inline bool rq_mergeable(struct request *rq)
649 {
650 if (rq->cmd_type != REQ_TYPE_FS)
651 return false;
652
653 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
654 return false;
655
656 return true;
657 }
658
659 static inline bool blk_check_merge_flags(unsigned int flags1,
660 unsigned int flags2)
661 {
662 if ((flags1 & REQ_DISCARD) != (flags2 & REQ_DISCARD))
663 return false;
664
665 if ((flags1 & REQ_SECURE) != (flags2 & REQ_SECURE))
666 return false;
667
668 if ((flags1 & REQ_WRITE_SAME) != (flags2 & REQ_WRITE_SAME))
669 return false;
670
671 return true;
672 }
673
674 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
675 {
676 if (bio_data(a) == bio_data(b))
677 return true;
678
679 return false;
680 }
681
682 /*
683 * q->prep_rq_fn return values
684 */
685 enum {
686 BLKPREP_OK, /* serve it */
687 BLKPREP_KILL, /* fatal error, kill, return -EIO */
688 BLKPREP_DEFER, /* leave on queue */
689 BLKPREP_INVALID, /* invalid command, kill, return -EREMOTEIO */
690 };
691
692 extern unsigned long blk_max_low_pfn, blk_max_pfn;
693
694 /*
695 * standard bounce addresses:
696 *
697 * BLK_BOUNCE_HIGH : bounce all highmem pages
698 * BLK_BOUNCE_ANY : don't bounce anything
699 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
700 */
701
702 #if BITS_PER_LONG == 32
703 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
704 #else
705 #define BLK_BOUNCE_HIGH -1ULL
706 #endif
707 #define BLK_BOUNCE_ANY (-1ULL)
708 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
709
710 /*
711 * default timeout for SG_IO if none specified
712 */
713 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
714 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
715
716 #ifdef CONFIG_BOUNCE
717 extern int init_emergency_isa_pool(void);
718 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
719 #else
720 static inline int init_emergency_isa_pool(void)
721 {
722 return 0;
723 }
724 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
725 {
726 }
727 #endif /* CONFIG_MMU */
728
729 struct rq_map_data {
730 struct page **pages;
731 int page_order;
732 int nr_entries;
733 unsigned long offset;
734 int null_mapped;
735 int from_user;
736 };
737
738 struct req_iterator {
739 struct bvec_iter iter;
740 struct bio *bio;
741 };
742
743 /* This should not be used directly - use rq_for_each_segment */
744 #define for_each_bio(_bio) \
745 for (; _bio; _bio = _bio->bi_next)
746 #define __rq_for_each_bio(_bio, rq) \
747 if ((rq->bio)) \
748 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
749
750 #define rq_for_each_segment(bvl, _rq, _iter) \
751 __rq_for_each_bio(_iter.bio, _rq) \
752 bio_for_each_segment(bvl, _iter.bio, _iter.iter)
753
754 #define rq_iter_last(bvec, _iter) \
755 (_iter.bio->bi_next == NULL && \
756 bio_iter_last(bvec, _iter.iter))
757
758 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
759 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
760 #endif
761 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
762 extern void rq_flush_dcache_pages(struct request *rq);
763 #else
764 static inline void rq_flush_dcache_pages(struct request *rq)
765 {
766 }
767 #endif
768
769 extern int blk_register_queue(struct gendisk *disk);
770 extern void blk_unregister_queue(struct gendisk *disk);
771 extern blk_qc_t generic_make_request(struct bio *bio);
772 extern void blk_rq_init(struct request_queue *q, struct request *rq);
773 extern void blk_put_request(struct request *);
774 extern void __blk_put_request(struct request_queue *, struct request *);
775 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
776 extern struct request *blk_make_request(struct request_queue *, struct bio *,
777 gfp_t);
778 extern void blk_rq_set_block_pc(struct request *);
779 extern void blk_requeue_request(struct request_queue *, struct request *);
780 extern void blk_add_request_payload(struct request *rq, struct page *page,
781 unsigned int len);
782 extern int blk_lld_busy(struct request_queue *q);
783 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
784 struct bio_set *bs, gfp_t gfp_mask,
785 int (*bio_ctr)(struct bio *, struct bio *, void *),
786 void *data);
787 extern void blk_rq_unprep_clone(struct request *rq);
788 extern int blk_insert_cloned_request(struct request_queue *q,
789 struct request *rq);
790 extern void blk_delay_queue(struct request_queue *, unsigned long);
791 extern void blk_queue_split(struct request_queue *, struct bio **,
792 struct bio_set *);
793 extern void blk_recount_segments(struct request_queue *, struct bio *);
794 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
795 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
796 unsigned int, void __user *);
797 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
798 unsigned int, void __user *);
799 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
800 struct scsi_ioctl_command __user *);
801
802 extern int blk_queue_enter(struct request_queue *q, bool nowait);
803 extern void blk_queue_exit(struct request_queue *q);
804 extern void blk_start_queue(struct request_queue *q);
805 extern void blk_start_queue_async(struct request_queue *q);
806 extern void blk_stop_queue(struct request_queue *q);
807 extern void blk_sync_queue(struct request_queue *q);
808 extern void __blk_stop_queue(struct request_queue *q);
809 extern void __blk_run_queue(struct request_queue *q);
810 extern void __blk_run_queue_uncond(struct request_queue *q);
811 extern void blk_run_queue(struct request_queue *);
812 extern void blk_run_queue_async(struct request_queue *q);
813 extern int blk_rq_map_user(struct request_queue *, struct request *,
814 struct rq_map_data *, void __user *, unsigned long,
815 gfp_t);
816 extern int blk_rq_unmap_user(struct bio *);
817 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
818 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
819 struct rq_map_data *, const struct iov_iter *,
820 gfp_t);
821 extern int blk_execute_rq(struct request_queue *, struct gendisk *,
822 struct request *, int);
823 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
824 struct request *, int, rq_end_io_fn *);
825
826 bool blk_poll(struct request_queue *q, blk_qc_t cookie);
827
828 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
829 {
830 return bdev->bd_disk->queue; /* this is never NULL */
831 }
832
833 /*
834 * blk_rq_pos() : the current sector
835 * blk_rq_bytes() : bytes left in the entire request
836 * blk_rq_cur_bytes() : bytes left in the current segment
837 * blk_rq_err_bytes() : bytes left till the next error boundary
838 * blk_rq_sectors() : sectors left in the entire request
839 * blk_rq_cur_sectors() : sectors left in the current segment
840 */
841 static inline sector_t blk_rq_pos(const struct request *rq)
842 {
843 return rq->__sector;
844 }
845
846 static inline unsigned int blk_rq_bytes(const struct request *rq)
847 {
848 return rq->__data_len;
849 }
850
851 static inline int blk_rq_cur_bytes(const struct request *rq)
852 {
853 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
854 }
855
856 extern unsigned int blk_rq_err_bytes(const struct request *rq);
857
858 static inline unsigned int blk_rq_sectors(const struct request *rq)
859 {
860 return blk_rq_bytes(rq) >> 9;
861 }
862
863 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
864 {
865 return blk_rq_cur_bytes(rq) >> 9;
866 }
867
868 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
869 unsigned int cmd_flags)
870 {
871 if (unlikely(cmd_flags & REQ_DISCARD))
872 return min(q->limits.max_discard_sectors, UINT_MAX >> 9);
873
874 if (unlikely(cmd_flags & REQ_WRITE_SAME))
875 return q->limits.max_write_same_sectors;
876
877 return q->limits.max_sectors;
878 }
879
880 /*
881 * Return maximum size of a request at given offset. Only valid for
882 * file system requests.
883 */
884 static inline unsigned int blk_max_size_offset(struct request_queue *q,
885 sector_t offset)
886 {
887 if (!q->limits.chunk_sectors)
888 return q->limits.max_sectors;
889
890 return q->limits.chunk_sectors -
891 (offset & (q->limits.chunk_sectors - 1));
892 }
893
894 static inline unsigned int blk_rq_get_max_sectors(struct request *rq)
895 {
896 struct request_queue *q = rq->q;
897
898 if (unlikely(rq->cmd_type != REQ_TYPE_FS))
899 return q->limits.max_hw_sectors;
900
901 if (!q->limits.chunk_sectors || (rq->cmd_flags & REQ_DISCARD))
902 return blk_queue_get_max_sectors(q, rq->cmd_flags);
903
904 return min(blk_max_size_offset(q, blk_rq_pos(rq)),
905 blk_queue_get_max_sectors(q, rq->cmd_flags));
906 }
907
908 static inline unsigned int blk_rq_count_bios(struct request *rq)
909 {
910 unsigned int nr_bios = 0;
911 struct bio *bio;
912
913 __rq_for_each_bio(bio, rq)
914 nr_bios++;
915
916 return nr_bios;
917 }
918
919 /*
920 * Request issue related functions.
921 */
922 extern struct request *blk_peek_request(struct request_queue *q);
923 extern void blk_start_request(struct request *rq);
924 extern struct request *blk_fetch_request(struct request_queue *q);
925
926 /*
927 * Request completion related functions.
928 *
929 * blk_update_request() completes given number of bytes and updates
930 * the request without completing it.
931 *
932 * blk_end_request() and friends. __blk_end_request() must be called
933 * with the request queue spinlock acquired.
934 *
935 * Several drivers define their own end_request and call
936 * blk_end_request() for parts of the original function.
937 * This prevents code duplication in drivers.
938 */
939 extern bool blk_update_request(struct request *rq, int error,
940 unsigned int nr_bytes);
941 extern void blk_finish_request(struct request *rq, int error);
942 extern bool blk_end_request(struct request *rq, int error,
943 unsigned int nr_bytes);
944 extern void blk_end_request_all(struct request *rq, int error);
945 extern bool blk_end_request_cur(struct request *rq, int error);
946 extern bool blk_end_request_err(struct request *rq, int error);
947 extern bool __blk_end_request(struct request *rq, int error,
948 unsigned int nr_bytes);
949 extern void __blk_end_request_all(struct request *rq, int error);
950 extern bool __blk_end_request_cur(struct request *rq, int error);
951 extern bool __blk_end_request_err(struct request *rq, int error);
952
953 extern void blk_complete_request(struct request *);
954 extern void __blk_complete_request(struct request *);
955 extern void blk_abort_request(struct request *);
956 extern void blk_unprep_request(struct request *);
957
958 /*
959 * Access functions for manipulating queue properties
960 */
961 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
962 spinlock_t *lock, int node_id);
963 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
964 extern struct request_queue *blk_init_allocated_queue(struct request_queue *,
965 request_fn_proc *, spinlock_t *);
966 extern void blk_cleanup_queue(struct request_queue *);
967 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
968 extern void blk_queue_bounce_limit(struct request_queue *, u64);
969 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
970 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
971 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
972 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
973 extern void blk_queue_max_discard_sectors(struct request_queue *q,
974 unsigned int max_discard_sectors);
975 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
976 unsigned int max_write_same_sectors);
977 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
978 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
979 extern void blk_queue_alignment_offset(struct request_queue *q,
980 unsigned int alignment);
981 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
982 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
983 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
984 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
985 extern void blk_set_default_limits(struct queue_limits *lim);
986 extern void blk_set_stacking_limits(struct queue_limits *lim);
987 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
988 sector_t offset);
989 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
990 sector_t offset);
991 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
992 sector_t offset);
993 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
994 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
995 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
996 extern int blk_queue_dma_drain(struct request_queue *q,
997 dma_drain_needed_fn *dma_drain_needed,
998 void *buf, unsigned int size);
999 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
1000 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1001 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1002 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
1003 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
1004 extern void blk_queue_dma_alignment(struct request_queue *, int);
1005 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1006 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1007 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1008 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1009 extern void blk_queue_flush(struct request_queue *q, unsigned int flush);
1010 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1011 extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
1012
1013 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1014 extern void blk_dump_rq_flags(struct request *, char *);
1015 extern long nr_blockdev_pages(void);
1016
1017 bool __must_check blk_get_queue(struct request_queue *);
1018 struct request_queue *blk_alloc_queue(gfp_t);
1019 struct request_queue *blk_alloc_queue_node(gfp_t, int);
1020 extern void blk_put_queue(struct request_queue *);
1021 extern void blk_set_queue_dying(struct request_queue *);
1022
1023 /*
1024 * block layer runtime pm functions
1025 */
1026 #ifdef CONFIG_PM
1027 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1028 extern int blk_pre_runtime_suspend(struct request_queue *q);
1029 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1030 extern void blk_pre_runtime_resume(struct request_queue *q);
1031 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1032 #else
1033 static inline void blk_pm_runtime_init(struct request_queue *q,
1034 struct device *dev) {}
1035 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1036 {
1037 return -ENOSYS;
1038 }
1039 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1040 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1041 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1042 #endif
1043
1044 /*
1045 * blk_plug permits building a queue of related requests by holding the I/O
1046 * fragments for a short period. This allows merging of sequential requests
1047 * into single larger request. As the requests are moved from a per-task list to
1048 * the device's request_queue in a batch, this results in improved scalability
1049 * as the lock contention for request_queue lock is reduced.
1050 *
1051 * It is ok not to disable preemption when adding the request to the plug list
1052 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1053 * the plug list when the task sleeps by itself. For details, please see
1054 * schedule() where blk_schedule_flush_plug() is called.
1055 */
1056 struct blk_plug {
1057 struct list_head list; /* requests */
1058 struct list_head mq_list; /* blk-mq requests */
1059 struct list_head cb_list; /* md requires an unplug callback */
1060 };
1061 #define BLK_MAX_REQUEST_COUNT 16
1062
1063 struct blk_plug_cb;
1064 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1065 struct blk_plug_cb {
1066 struct list_head list;
1067 blk_plug_cb_fn callback;
1068 void *data;
1069 };
1070 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1071 void *data, int size);
1072 extern void blk_start_plug(struct blk_plug *);
1073 extern void blk_finish_plug(struct blk_plug *);
1074 extern void blk_flush_plug_list(struct blk_plug *, bool);
1075
1076 static inline void blk_flush_plug(struct task_struct *tsk)
1077 {
1078 struct blk_plug *plug = tsk->plug;
1079
1080 if (plug)
1081 blk_flush_plug_list(plug, false);
1082 }
1083
1084 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1085 {
1086 struct blk_plug *plug = tsk->plug;
1087
1088 if (plug)
1089 blk_flush_plug_list(plug, true);
1090 }
1091
1092 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1093 {
1094 struct blk_plug *plug = tsk->plug;
1095
1096 return plug &&
1097 (!list_empty(&plug->list) ||
1098 !list_empty(&plug->mq_list) ||
1099 !list_empty(&plug->cb_list));
1100 }
1101
1102 /*
1103 * tag stuff
1104 */
1105 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1106 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1107 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1108 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1109 extern void blk_queue_free_tags(struct request_queue *);
1110 extern int blk_queue_resize_tags(struct request_queue *, int);
1111 extern void blk_queue_invalidate_tags(struct request_queue *);
1112 extern struct blk_queue_tag *blk_init_tags(int, int);
1113 extern void blk_free_tags(struct blk_queue_tag *);
1114
1115 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1116 int tag)
1117 {
1118 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1119 return NULL;
1120 return bqt->tag_index[tag];
1121 }
1122
1123 #define BLKDEV_DISCARD_SECURE 0x01 /* secure discard */
1124
1125 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1126 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1127 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1128 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1129 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1130 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1131 sector_t nr_sects, gfp_t gfp_mask, bool discard);
1132 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1133 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1134 {
1135 return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9),
1136 nr_blocks << (sb->s_blocksize_bits - 9),
1137 gfp_mask, flags);
1138 }
1139 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1140 sector_t nr_blocks, gfp_t gfp_mask)
1141 {
1142 return blkdev_issue_zeroout(sb->s_bdev,
1143 block << (sb->s_blocksize_bits - 9),
1144 nr_blocks << (sb->s_blocksize_bits - 9),
1145 gfp_mask, true);
1146 }
1147
1148 extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm);
1149
1150 enum blk_default_limits {
1151 BLK_MAX_SEGMENTS = 128,
1152 BLK_SAFE_MAX_SECTORS = 255,
1153 BLK_DEF_MAX_SECTORS = 2560,
1154 BLK_MAX_SEGMENT_SIZE = 65536,
1155 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1156 };
1157
1158 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1159
1160 static inline unsigned long queue_bounce_pfn(struct request_queue *q)
1161 {
1162 return q->limits.bounce_pfn;
1163 }
1164
1165 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1166 {
1167 return q->limits.seg_boundary_mask;
1168 }
1169
1170 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1171 {
1172 return q->limits.virt_boundary_mask;
1173 }
1174
1175 static inline unsigned int queue_max_sectors(struct request_queue *q)
1176 {
1177 return q->limits.max_sectors;
1178 }
1179
1180 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1181 {
1182 return q->limits.max_hw_sectors;
1183 }
1184
1185 static inline unsigned short queue_max_segments(struct request_queue *q)
1186 {
1187 return q->limits.max_segments;
1188 }
1189
1190 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1191 {
1192 return q->limits.max_segment_size;
1193 }
1194
1195 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1196 {
1197 int retval = 512;
1198
1199 if (q && q->limits.logical_block_size)
1200 retval = q->limits.logical_block_size;
1201
1202 return retval;
1203 }
1204
1205 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1206 {
1207 return queue_logical_block_size(bdev_get_queue(bdev));
1208 }
1209
1210 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1211 {
1212 return q->limits.physical_block_size;
1213 }
1214
1215 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1216 {
1217 return queue_physical_block_size(bdev_get_queue(bdev));
1218 }
1219
1220 static inline unsigned int queue_io_min(struct request_queue *q)
1221 {
1222 return q->limits.io_min;
1223 }
1224
1225 static inline int bdev_io_min(struct block_device *bdev)
1226 {
1227 return queue_io_min(bdev_get_queue(bdev));
1228 }
1229
1230 static inline unsigned int queue_io_opt(struct request_queue *q)
1231 {
1232 return q->limits.io_opt;
1233 }
1234
1235 static inline int bdev_io_opt(struct block_device *bdev)
1236 {
1237 return queue_io_opt(bdev_get_queue(bdev));
1238 }
1239
1240 static inline int queue_alignment_offset(struct request_queue *q)
1241 {
1242 if (q->limits.misaligned)
1243 return -1;
1244
1245 return q->limits.alignment_offset;
1246 }
1247
1248 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1249 {
1250 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1251 unsigned int alignment = sector_div(sector, granularity >> 9) << 9;
1252
1253 return (granularity + lim->alignment_offset - alignment) % granularity;
1254 }
1255
1256 static inline int bdev_alignment_offset(struct block_device *bdev)
1257 {
1258 struct request_queue *q = bdev_get_queue(bdev);
1259
1260 if (q->limits.misaligned)
1261 return -1;
1262
1263 if (bdev != bdev->bd_contains)
1264 return bdev->bd_part->alignment_offset;
1265
1266 return q->limits.alignment_offset;
1267 }
1268
1269 static inline int queue_discard_alignment(struct request_queue *q)
1270 {
1271 if (q->limits.discard_misaligned)
1272 return -1;
1273
1274 return q->limits.discard_alignment;
1275 }
1276
1277 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1278 {
1279 unsigned int alignment, granularity, offset;
1280
1281 if (!lim->max_discard_sectors)
1282 return 0;
1283
1284 /* Why are these in bytes, not sectors? */
1285 alignment = lim->discard_alignment >> 9;
1286 granularity = lim->discard_granularity >> 9;
1287 if (!granularity)
1288 return 0;
1289
1290 /* Offset of the partition start in 'granularity' sectors */
1291 offset = sector_div(sector, granularity);
1292
1293 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1294 offset = (granularity + alignment - offset) % granularity;
1295
1296 /* Turn it back into bytes, gaah */
1297 return offset << 9;
1298 }
1299
1300 static inline int bdev_discard_alignment(struct block_device *bdev)
1301 {
1302 struct request_queue *q = bdev_get_queue(bdev);
1303
1304 if (bdev != bdev->bd_contains)
1305 return bdev->bd_part->discard_alignment;
1306
1307 return q->limits.discard_alignment;
1308 }
1309
1310 static inline unsigned int queue_discard_zeroes_data(struct request_queue *q)
1311 {
1312 if (q->limits.max_discard_sectors && q->limits.discard_zeroes_data == 1)
1313 return 1;
1314
1315 return 0;
1316 }
1317
1318 static inline unsigned int bdev_discard_zeroes_data(struct block_device *bdev)
1319 {
1320 return queue_discard_zeroes_data(bdev_get_queue(bdev));
1321 }
1322
1323 static inline unsigned int bdev_write_same(struct block_device *bdev)
1324 {
1325 struct request_queue *q = bdev_get_queue(bdev);
1326
1327 if (q)
1328 return q->limits.max_write_same_sectors;
1329
1330 return 0;
1331 }
1332
1333 static inline int queue_dma_alignment(struct request_queue *q)
1334 {
1335 return q ? q->dma_alignment : 511;
1336 }
1337
1338 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1339 unsigned int len)
1340 {
1341 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1342 return !(addr & alignment) && !(len & alignment);
1343 }
1344
1345 /* assumes size > 256 */
1346 static inline unsigned int blksize_bits(unsigned int size)
1347 {
1348 unsigned int bits = 8;
1349 do {
1350 bits++;
1351 size >>= 1;
1352 } while (size > 256);
1353 return bits;
1354 }
1355
1356 static inline unsigned int block_size(struct block_device *bdev)
1357 {
1358 return bdev->bd_block_size;
1359 }
1360
1361 static inline bool queue_flush_queueable(struct request_queue *q)
1362 {
1363 return !q->flush_not_queueable;
1364 }
1365
1366 typedef struct {struct page *v;} Sector;
1367
1368 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1369
1370 static inline void put_dev_sector(Sector p)
1371 {
1372 page_cache_release(p.v);
1373 }
1374
1375 static inline bool __bvec_gap_to_prev(struct request_queue *q,
1376 struct bio_vec *bprv, unsigned int offset)
1377 {
1378 return offset ||
1379 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1380 }
1381
1382 /*
1383 * Check if adding a bio_vec after bprv with offset would create a gap in
1384 * the SG list. Most drivers don't care about this, but some do.
1385 */
1386 static inline bool bvec_gap_to_prev(struct request_queue *q,
1387 struct bio_vec *bprv, unsigned int offset)
1388 {
1389 if (!queue_virt_boundary(q))
1390 return false;
1391 return __bvec_gap_to_prev(q, bprv, offset);
1392 }
1393
1394 static inline bool bio_will_gap(struct request_queue *q, struct bio *prev,
1395 struct bio *next)
1396 {
1397 if (bio_has_data(prev) && queue_virt_boundary(q)) {
1398 struct bio_vec pb, nb;
1399
1400 bio_get_last_bvec(prev, &pb);
1401 bio_get_first_bvec(next, &nb);
1402
1403 return __bvec_gap_to_prev(q, &pb, nb.bv_offset);
1404 }
1405
1406 return false;
1407 }
1408
1409 static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
1410 {
1411 return bio_will_gap(req->q, req->biotail, bio);
1412 }
1413
1414 static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
1415 {
1416 return bio_will_gap(req->q, bio, req->bio);
1417 }
1418
1419 struct work_struct;
1420 int kblockd_schedule_work(struct work_struct *work);
1421 int kblockd_schedule_delayed_work(struct delayed_work *dwork, unsigned long delay);
1422 int kblockd_schedule_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1423
1424 #ifdef CONFIG_BLK_CGROUP
1425 /*
1426 * This should not be using sched_clock(). A real patch is in progress
1427 * to fix this up, until that is in place we need to disable preemption
1428 * around sched_clock() in this function and set_io_start_time_ns().
1429 */
1430 static inline void set_start_time_ns(struct request *req)
1431 {
1432 preempt_disable();
1433 req->start_time_ns = sched_clock();
1434 preempt_enable();
1435 }
1436
1437 static inline void set_io_start_time_ns(struct request *req)
1438 {
1439 preempt_disable();
1440 req->io_start_time_ns = sched_clock();
1441 preempt_enable();
1442 }
1443
1444 static inline uint64_t rq_start_time_ns(struct request *req)
1445 {
1446 return req->start_time_ns;
1447 }
1448
1449 static inline uint64_t rq_io_start_time_ns(struct request *req)
1450 {
1451 return req->io_start_time_ns;
1452 }
1453 #else
1454 static inline void set_start_time_ns(struct request *req) {}
1455 static inline void set_io_start_time_ns(struct request *req) {}
1456 static inline uint64_t rq_start_time_ns(struct request *req)
1457 {
1458 return 0;
1459 }
1460 static inline uint64_t rq_io_start_time_ns(struct request *req)
1461 {
1462 return 0;
1463 }
1464 #endif
1465
1466 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1467 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1468 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1469 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1470
1471 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1472
1473 enum blk_integrity_flags {
1474 BLK_INTEGRITY_VERIFY = 1 << 0,
1475 BLK_INTEGRITY_GENERATE = 1 << 1,
1476 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2,
1477 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3,
1478 };
1479
1480 struct blk_integrity_iter {
1481 void *prot_buf;
1482 void *data_buf;
1483 sector_t seed;
1484 unsigned int data_size;
1485 unsigned short interval;
1486 const char *disk_name;
1487 };
1488
1489 typedef int (integrity_processing_fn) (struct blk_integrity_iter *);
1490
1491 struct blk_integrity_profile {
1492 integrity_processing_fn *generate_fn;
1493 integrity_processing_fn *verify_fn;
1494 const char *name;
1495 };
1496
1497 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1498 extern void blk_integrity_unregister(struct gendisk *);
1499 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1500 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1501 struct scatterlist *);
1502 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1503 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1504 struct request *);
1505 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1506 struct bio *);
1507
1508 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1509 {
1510 struct blk_integrity *bi = &disk->queue->integrity;
1511
1512 if (!bi->profile)
1513 return NULL;
1514
1515 return bi;
1516 }
1517
1518 static inline
1519 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1520 {
1521 return blk_get_integrity(bdev->bd_disk);
1522 }
1523
1524 static inline bool blk_integrity_rq(struct request *rq)
1525 {
1526 return rq->cmd_flags & REQ_INTEGRITY;
1527 }
1528
1529 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1530 unsigned int segs)
1531 {
1532 q->limits.max_integrity_segments = segs;
1533 }
1534
1535 static inline unsigned short
1536 queue_max_integrity_segments(struct request_queue *q)
1537 {
1538 return q->limits.max_integrity_segments;
1539 }
1540
1541 static inline bool integrity_req_gap_back_merge(struct request *req,
1542 struct bio *next)
1543 {
1544 struct bio_integrity_payload *bip = bio_integrity(req->bio);
1545 struct bio_integrity_payload *bip_next = bio_integrity(next);
1546
1547 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1548 bip_next->bip_vec[0].bv_offset);
1549 }
1550
1551 static inline bool integrity_req_gap_front_merge(struct request *req,
1552 struct bio *bio)
1553 {
1554 struct bio_integrity_payload *bip = bio_integrity(bio);
1555 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
1556
1557 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1558 bip_next->bip_vec[0].bv_offset);
1559 }
1560
1561 #else /* CONFIG_BLK_DEV_INTEGRITY */
1562
1563 struct bio;
1564 struct block_device;
1565 struct gendisk;
1566 struct blk_integrity;
1567
1568 static inline int blk_integrity_rq(struct request *rq)
1569 {
1570 return 0;
1571 }
1572 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1573 struct bio *b)
1574 {
1575 return 0;
1576 }
1577 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1578 struct bio *b,
1579 struct scatterlist *s)
1580 {
1581 return 0;
1582 }
1583 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1584 {
1585 return NULL;
1586 }
1587 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1588 {
1589 return NULL;
1590 }
1591 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1592 {
1593 return 0;
1594 }
1595 static inline void blk_integrity_register(struct gendisk *d,
1596 struct blk_integrity *b)
1597 {
1598 }
1599 static inline void blk_integrity_unregister(struct gendisk *d)
1600 {
1601 }
1602 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1603 unsigned int segs)
1604 {
1605 }
1606 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1607 {
1608 return 0;
1609 }
1610 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1611 struct request *r1,
1612 struct request *r2)
1613 {
1614 return true;
1615 }
1616 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1617 struct request *r,
1618 struct bio *b)
1619 {
1620 return true;
1621 }
1622
1623 static inline bool integrity_req_gap_back_merge(struct request *req,
1624 struct bio *next)
1625 {
1626 return false;
1627 }
1628 static inline bool integrity_req_gap_front_merge(struct request *req,
1629 struct bio *bio)
1630 {
1631 return false;
1632 }
1633
1634 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1635
1636 /**
1637 * struct blk_dax_ctl - control and output parameters for ->direct_access
1638 * @sector: (input) offset relative to a block_device
1639 * @addr: (output) kernel virtual address for @sector populated by driver
1640 * @pfn: (output) page frame number for @addr populated by driver
1641 * @size: (input) number of bytes requested
1642 */
1643 struct blk_dax_ctl {
1644 sector_t sector;
1645 void __pmem *addr;
1646 long size;
1647 pfn_t pfn;
1648 };
1649
1650 struct block_device_operations {
1651 int (*open) (struct block_device *, fmode_t);
1652 void (*release) (struct gendisk *, fmode_t);
1653 int (*rw_page)(struct block_device *, sector_t, struct page *, int rw);
1654 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1655 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1656 long (*direct_access)(struct block_device *, sector_t, void __pmem **,
1657 pfn_t *);
1658 unsigned int (*check_events) (struct gendisk *disk,
1659 unsigned int clearing);
1660 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1661 int (*media_changed) (struct gendisk *);
1662 void (*unlock_native_capacity) (struct gendisk *);
1663 int (*revalidate_disk) (struct gendisk *);
1664 int (*getgeo)(struct block_device *, struct hd_geometry *);
1665 /* this callback is with swap_lock and sometimes page table lock held */
1666 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1667 struct module *owner;
1668 const struct pr_ops *pr_ops;
1669 };
1670
1671 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1672 unsigned long);
1673 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1674 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1675 struct writeback_control *);
1676 extern long bdev_direct_access(struct block_device *, struct blk_dax_ctl *);
1677 #else /* CONFIG_BLOCK */
1678
1679 struct block_device;
1680
1681 /*
1682 * stubs for when the block layer is configured out
1683 */
1684 #define buffer_heads_over_limit 0
1685
1686 static inline long nr_blockdev_pages(void)
1687 {
1688 return 0;
1689 }
1690
1691 struct blk_plug {
1692 };
1693
1694 static inline void blk_start_plug(struct blk_plug *plug)
1695 {
1696 }
1697
1698 static inline void blk_finish_plug(struct blk_plug *plug)
1699 {
1700 }
1701
1702 static inline void blk_flush_plug(struct task_struct *task)
1703 {
1704 }
1705
1706 static inline void blk_schedule_flush_plug(struct task_struct *task)
1707 {
1708 }
1709
1710
1711 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1712 {
1713 return false;
1714 }
1715
1716 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
1717 sector_t *error_sector)
1718 {
1719 return 0;
1720 }
1721
1722 #endif /* CONFIG_BLOCK */
1723
1724 #endif