]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - include/linux/blkdev.h
Merge branch 'overlayfs-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mszer...
[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/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 list_head timeout_list;
413
414 struct list_head icq_list;
415 #ifdef CONFIG_BLK_CGROUP
416 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
417 struct blkcg_gq *root_blkg;
418 struct list_head blkg_list;
419 #endif
420
421 struct queue_limits limits;
422
423 /*
424 * sg stuff
425 */
426 unsigned int sg_timeout;
427 unsigned int sg_reserved_size;
428 int node;
429 #ifdef CONFIG_BLK_DEV_IO_TRACE
430 struct blk_trace *blk_trace;
431 #endif
432 /*
433 * for flush operations
434 */
435 unsigned int flush_flags;
436 unsigned int flush_not_queueable:1;
437 struct blk_flush_queue *fq;
438
439 struct list_head requeue_list;
440 spinlock_t requeue_lock;
441 struct work_struct requeue_work;
442
443 struct mutex sysfs_lock;
444
445 int bypass_depth;
446 atomic_t mq_freeze_depth;
447
448 #if defined(CONFIG_BLK_DEV_BSG)
449 bsg_job_fn *bsg_job_fn;
450 int bsg_job_size;
451 struct bsg_class_device bsg_dev;
452 #endif
453
454 #ifdef CONFIG_BLK_DEV_THROTTLING
455 /* Throttle data */
456 struct throtl_data *td;
457 #endif
458 struct rcu_head rcu_head;
459 wait_queue_head_t mq_freeze_wq;
460 struct percpu_ref q_usage_counter;
461 struct list_head all_q_node;
462
463 struct blk_mq_tag_set *tag_set;
464 struct list_head tag_set_list;
465 struct bio_set *bio_split;
466
467 bool mq_sysfs_init_done;
468 };
469
470 #define QUEUE_FLAG_QUEUED 1 /* uses generic tag queueing */
471 #define QUEUE_FLAG_STOPPED 2 /* queue is stopped */
472 #define QUEUE_FLAG_SYNCFULL 3 /* read queue has been filled */
473 #define QUEUE_FLAG_ASYNCFULL 4 /* write queue has been filled */
474 #define QUEUE_FLAG_DYING 5 /* queue being torn down */
475 #define QUEUE_FLAG_BYPASS 6 /* act as dumb FIFO queue */
476 #define QUEUE_FLAG_BIDI 7 /* queue supports bidi requests */
477 #define QUEUE_FLAG_NOMERGES 8 /* disable merge attempts */
478 #define QUEUE_FLAG_SAME_COMP 9 /* complete on same CPU-group */
479 #define QUEUE_FLAG_FAIL_IO 10 /* fake timeout */
480 #define QUEUE_FLAG_STACKABLE 11 /* supports request stacking */
481 #define QUEUE_FLAG_NONROT 12 /* non-rotational device (SSD) */
482 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
483 #define QUEUE_FLAG_IO_STAT 13 /* do IO stats */
484 #define QUEUE_FLAG_DISCARD 14 /* supports DISCARD */
485 #define QUEUE_FLAG_NOXMERGES 15 /* No extended merges */
486 #define QUEUE_FLAG_ADD_RANDOM 16 /* Contributes to random pool */
487 #define QUEUE_FLAG_SECDISCARD 17 /* supports SECDISCARD */
488 #define QUEUE_FLAG_SAME_FORCE 18 /* force complete on same CPU */
489 #define QUEUE_FLAG_DEAD 19 /* queue tear-down finished */
490 #define QUEUE_FLAG_INIT_DONE 20 /* queue is initialized */
491 #define QUEUE_FLAG_NO_SG_MERGE 21 /* don't attempt to merge SG segments*/
492 #define QUEUE_FLAG_POLL 22 /* IO polling enabled if set */
493
494 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
495 (1 << QUEUE_FLAG_STACKABLE) | \
496 (1 << QUEUE_FLAG_SAME_COMP) | \
497 (1 << QUEUE_FLAG_ADD_RANDOM))
498
499 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
500 (1 << QUEUE_FLAG_STACKABLE) | \
501 (1 << QUEUE_FLAG_SAME_COMP))
502
503 static inline void queue_lockdep_assert_held(struct request_queue *q)
504 {
505 if (q->queue_lock)
506 lockdep_assert_held(q->queue_lock);
507 }
508
509 static inline void queue_flag_set_unlocked(unsigned int flag,
510 struct request_queue *q)
511 {
512 __set_bit(flag, &q->queue_flags);
513 }
514
515 static inline int queue_flag_test_and_clear(unsigned int flag,
516 struct request_queue *q)
517 {
518 queue_lockdep_assert_held(q);
519
520 if (test_bit(flag, &q->queue_flags)) {
521 __clear_bit(flag, &q->queue_flags);
522 return 1;
523 }
524
525 return 0;
526 }
527
528 static inline int queue_flag_test_and_set(unsigned int flag,
529 struct request_queue *q)
530 {
531 queue_lockdep_assert_held(q);
532
533 if (!test_bit(flag, &q->queue_flags)) {
534 __set_bit(flag, &q->queue_flags);
535 return 0;
536 }
537
538 return 1;
539 }
540
541 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
542 {
543 queue_lockdep_assert_held(q);
544 __set_bit(flag, &q->queue_flags);
545 }
546
547 static inline void queue_flag_clear_unlocked(unsigned int flag,
548 struct request_queue *q)
549 {
550 __clear_bit(flag, &q->queue_flags);
551 }
552
553 static inline int queue_in_flight(struct request_queue *q)
554 {
555 return q->in_flight[0] + q->in_flight[1];
556 }
557
558 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
559 {
560 queue_lockdep_assert_held(q);
561 __clear_bit(flag, &q->queue_flags);
562 }
563
564 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
565 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
566 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
567 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
568 #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
569 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
570 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
571 #define blk_queue_noxmerges(q) \
572 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
573 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
574 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
575 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
576 #define blk_queue_stackable(q) \
577 test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
578 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
579 #define blk_queue_secdiscard(q) (blk_queue_discard(q) && \
580 test_bit(QUEUE_FLAG_SECDISCARD, &(q)->queue_flags))
581
582 #define blk_noretry_request(rq) \
583 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
584 REQ_FAILFAST_DRIVER))
585
586 #define blk_account_rq(rq) \
587 (((rq)->cmd_flags & REQ_STARTED) && \
588 ((rq)->cmd_type == REQ_TYPE_FS))
589
590 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
591 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
592 /* rq->queuelist of dequeued request must be list_empty() */
593 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
594
595 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
596
597 #define rq_data_dir(rq) ((int)((rq)->cmd_flags & 1))
598
599 /*
600 * Driver can handle struct request, if it either has an old style
601 * request_fn defined, or is blk-mq based.
602 */
603 static inline bool queue_is_rq_based(struct request_queue *q)
604 {
605 return q->request_fn || q->mq_ops;
606 }
607
608 static inline unsigned int blk_queue_cluster(struct request_queue *q)
609 {
610 return q->limits.cluster;
611 }
612
613 /*
614 * We regard a request as sync, if either a read or a sync write
615 */
616 static inline bool rw_is_sync(unsigned int rw_flags)
617 {
618 return !(rw_flags & REQ_WRITE) || (rw_flags & REQ_SYNC);
619 }
620
621 static inline bool rq_is_sync(struct request *rq)
622 {
623 return rw_is_sync(rq->cmd_flags);
624 }
625
626 static inline bool blk_rl_full(struct request_list *rl, bool sync)
627 {
628 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
629
630 return rl->flags & flag;
631 }
632
633 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
634 {
635 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
636
637 rl->flags |= flag;
638 }
639
640 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
641 {
642 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
643
644 rl->flags &= ~flag;
645 }
646
647 static inline bool rq_mergeable(struct request *rq)
648 {
649 if (rq->cmd_type != REQ_TYPE_FS)
650 return false;
651
652 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
653 return false;
654
655 return true;
656 }
657
658 static inline bool blk_check_merge_flags(unsigned int flags1,
659 unsigned int flags2)
660 {
661 if ((flags1 & REQ_DISCARD) != (flags2 & REQ_DISCARD))
662 return false;
663
664 if ((flags1 & REQ_SECURE) != (flags2 & REQ_SECURE))
665 return false;
666
667 if ((flags1 & REQ_WRITE_SAME) != (flags2 & REQ_WRITE_SAME))
668 return false;
669
670 return true;
671 }
672
673 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
674 {
675 if (bio_data(a) == bio_data(b))
676 return true;
677
678 return false;
679 }
680
681 /*
682 * q->prep_rq_fn return values
683 */
684 #define BLKPREP_OK 0 /* serve it */
685 #define BLKPREP_KILL 1 /* fatal error, kill */
686 #define BLKPREP_DEFER 2 /* leave on queue */
687
688 extern unsigned long blk_max_low_pfn, blk_max_pfn;
689
690 /*
691 * standard bounce addresses:
692 *
693 * BLK_BOUNCE_HIGH : bounce all highmem pages
694 * BLK_BOUNCE_ANY : don't bounce anything
695 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
696 */
697
698 #if BITS_PER_LONG == 32
699 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
700 #else
701 #define BLK_BOUNCE_HIGH -1ULL
702 #endif
703 #define BLK_BOUNCE_ANY (-1ULL)
704 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
705
706 /*
707 * default timeout for SG_IO if none specified
708 */
709 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
710 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
711
712 #ifdef CONFIG_BOUNCE
713 extern int init_emergency_isa_pool(void);
714 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
715 #else
716 static inline int init_emergency_isa_pool(void)
717 {
718 return 0;
719 }
720 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
721 {
722 }
723 #endif /* CONFIG_MMU */
724
725 struct rq_map_data {
726 struct page **pages;
727 int page_order;
728 int nr_entries;
729 unsigned long offset;
730 int null_mapped;
731 int from_user;
732 };
733
734 struct req_iterator {
735 struct bvec_iter iter;
736 struct bio *bio;
737 };
738
739 /* This should not be used directly - use rq_for_each_segment */
740 #define for_each_bio(_bio) \
741 for (; _bio; _bio = _bio->bi_next)
742 #define __rq_for_each_bio(_bio, rq) \
743 if ((rq->bio)) \
744 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
745
746 #define rq_for_each_segment(bvl, _rq, _iter) \
747 __rq_for_each_bio(_iter.bio, _rq) \
748 bio_for_each_segment(bvl, _iter.bio, _iter.iter)
749
750 #define rq_iter_last(bvec, _iter) \
751 (_iter.bio->bi_next == NULL && \
752 bio_iter_last(bvec, _iter.iter))
753
754 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
755 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
756 #endif
757 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
758 extern void rq_flush_dcache_pages(struct request *rq);
759 #else
760 static inline void rq_flush_dcache_pages(struct request *rq)
761 {
762 }
763 #endif
764
765 extern int blk_register_queue(struct gendisk *disk);
766 extern void blk_unregister_queue(struct gendisk *disk);
767 extern blk_qc_t generic_make_request(struct bio *bio);
768 extern void blk_rq_init(struct request_queue *q, struct request *rq);
769 extern void blk_put_request(struct request *);
770 extern void __blk_put_request(struct request_queue *, struct request *);
771 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
772 extern struct request *blk_make_request(struct request_queue *, struct bio *,
773 gfp_t);
774 extern void blk_rq_set_block_pc(struct request *);
775 extern void blk_requeue_request(struct request_queue *, struct request *);
776 extern void blk_add_request_payload(struct request *rq, struct page *page,
777 unsigned int len);
778 extern int blk_lld_busy(struct request_queue *q);
779 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
780 struct bio_set *bs, gfp_t gfp_mask,
781 int (*bio_ctr)(struct bio *, struct bio *, void *),
782 void *data);
783 extern void blk_rq_unprep_clone(struct request *rq);
784 extern int blk_insert_cloned_request(struct request_queue *q,
785 struct request *rq);
786 extern void blk_delay_queue(struct request_queue *, unsigned long);
787 extern void blk_queue_split(struct request_queue *, struct bio **,
788 struct bio_set *);
789 extern void blk_recount_segments(struct request_queue *, struct bio *);
790 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
791 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
792 unsigned int, void __user *);
793 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
794 unsigned int, void __user *);
795 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
796 struct scsi_ioctl_command __user *);
797
798 extern int blk_queue_enter(struct request_queue *q, bool nowait);
799 extern void blk_queue_exit(struct request_queue *q);
800 extern void blk_start_queue(struct request_queue *q);
801 extern void blk_start_queue_async(struct request_queue *q);
802 extern void blk_stop_queue(struct request_queue *q);
803 extern void blk_sync_queue(struct request_queue *q);
804 extern void __blk_stop_queue(struct request_queue *q);
805 extern void __blk_run_queue(struct request_queue *q);
806 extern void __blk_run_queue_uncond(struct request_queue *q);
807 extern void blk_run_queue(struct request_queue *);
808 extern void blk_run_queue_async(struct request_queue *q);
809 extern int blk_rq_map_user(struct request_queue *, struct request *,
810 struct rq_map_data *, void __user *, unsigned long,
811 gfp_t);
812 extern int blk_rq_unmap_user(struct bio *);
813 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
814 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
815 struct rq_map_data *, const struct iov_iter *,
816 gfp_t);
817 extern int blk_execute_rq(struct request_queue *, struct gendisk *,
818 struct request *, int);
819 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
820 struct request *, int, rq_end_io_fn *);
821
822 bool blk_poll(struct request_queue *q, blk_qc_t cookie);
823
824 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
825 {
826 return bdev->bd_disk->queue; /* this is never NULL */
827 }
828
829 /*
830 * blk_rq_pos() : the current sector
831 * blk_rq_bytes() : bytes left in the entire request
832 * blk_rq_cur_bytes() : bytes left in the current segment
833 * blk_rq_err_bytes() : bytes left till the next error boundary
834 * blk_rq_sectors() : sectors left in the entire request
835 * blk_rq_cur_sectors() : sectors left in the current segment
836 */
837 static inline sector_t blk_rq_pos(const struct request *rq)
838 {
839 return rq->__sector;
840 }
841
842 static inline unsigned int blk_rq_bytes(const struct request *rq)
843 {
844 return rq->__data_len;
845 }
846
847 static inline int blk_rq_cur_bytes(const struct request *rq)
848 {
849 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
850 }
851
852 extern unsigned int blk_rq_err_bytes(const struct request *rq);
853
854 static inline unsigned int blk_rq_sectors(const struct request *rq)
855 {
856 return blk_rq_bytes(rq) >> 9;
857 }
858
859 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
860 {
861 return blk_rq_cur_bytes(rq) >> 9;
862 }
863
864 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
865 unsigned int cmd_flags)
866 {
867 if (unlikely(cmd_flags & REQ_DISCARD))
868 return min(q->limits.max_discard_sectors, UINT_MAX >> 9);
869
870 if (unlikely(cmd_flags & REQ_WRITE_SAME))
871 return q->limits.max_write_same_sectors;
872
873 return q->limits.max_sectors;
874 }
875
876 /*
877 * Return maximum size of a request at given offset. Only valid for
878 * file system requests.
879 */
880 static inline unsigned int blk_max_size_offset(struct request_queue *q,
881 sector_t offset)
882 {
883 if (!q->limits.chunk_sectors)
884 return q->limits.max_sectors;
885
886 return q->limits.chunk_sectors -
887 (offset & (q->limits.chunk_sectors - 1));
888 }
889
890 static inline unsigned int blk_rq_get_max_sectors(struct request *rq)
891 {
892 struct request_queue *q = rq->q;
893
894 if (unlikely(rq->cmd_type == REQ_TYPE_BLOCK_PC))
895 return q->limits.max_hw_sectors;
896
897 if (!q->limits.chunk_sectors || (rq->cmd_flags & REQ_DISCARD))
898 return blk_queue_get_max_sectors(q, rq->cmd_flags);
899
900 return min(blk_max_size_offset(q, blk_rq_pos(rq)),
901 blk_queue_get_max_sectors(q, rq->cmd_flags));
902 }
903
904 static inline unsigned int blk_rq_count_bios(struct request *rq)
905 {
906 unsigned int nr_bios = 0;
907 struct bio *bio;
908
909 __rq_for_each_bio(bio, rq)
910 nr_bios++;
911
912 return nr_bios;
913 }
914
915 /*
916 * Request issue related functions.
917 */
918 extern struct request *blk_peek_request(struct request_queue *q);
919 extern void blk_start_request(struct request *rq);
920 extern struct request *blk_fetch_request(struct request_queue *q);
921
922 /*
923 * Request completion related functions.
924 *
925 * blk_update_request() completes given number of bytes and updates
926 * the request without completing it.
927 *
928 * blk_end_request() and friends. __blk_end_request() must be called
929 * with the request queue spinlock acquired.
930 *
931 * Several drivers define their own end_request and call
932 * blk_end_request() for parts of the original function.
933 * This prevents code duplication in drivers.
934 */
935 extern bool blk_update_request(struct request *rq, int error,
936 unsigned int nr_bytes);
937 extern void blk_finish_request(struct request *rq, int error);
938 extern bool blk_end_request(struct request *rq, int error,
939 unsigned int nr_bytes);
940 extern void blk_end_request_all(struct request *rq, int error);
941 extern bool blk_end_request_cur(struct request *rq, int error);
942 extern bool blk_end_request_err(struct request *rq, int error);
943 extern bool __blk_end_request(struct request *rq, int error,
944 unsigned int nr_bytes);
945 extern void __blk_end_request_all(struct request *rq, int error);
946 extern bool __blk_end_request_cur(struct request *rq, int error);
947 extern bool __blk_end_request_err(struct request *rq, int error);
948
949 extern void blk_complete_request(struct request *);
950 extern void __blk_complete_request(struct request *);
951 extern void blk_abort_request(struct request *);
952 extern void blk_unprep_request(struct request *);
953
954 /*
955 * Access functions for manipulating queue properties
956 */
957 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
958 spinlock_t *lock, int node_id);
959 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
960 extern struct request_queue *blk_init_allocated_queue(struct request_queue *,
961 request_fn_proc *, spinlock_t *);
962 extern void blk_cleanup_queue(struct request_queue *);
963 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
964 extern void blk_queue_bounce_limit(struct request_queue *, u64);
965 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
966 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
967 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
968 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
969 extern void blk_queue_max_discard_sectors(struct request_queue *q,
970 unsigned int max_discard_sectors);
971 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
972 unsigned int max_write_same_sectors);
973 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
974 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
975 extern void blk_queue_alignment_offset(struct request_queue *q,
976 unsigned int alignment);
977 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
978 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
979 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
980 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
981 extern void blk_set_default_limits(struct queue_limits *lim);
982 extern void blk_set_stacking_limits(struct queue_limits *lim);
983 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
984 sector_t offset);
985 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
986 sector_t offset);
987 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
988 sector_t offset);
989 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
990 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
991 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
992 extern int blk_queue_dma_drain(struct request_queue *q,
993 dma_drain_needed_fn *dma_drain_needed,
994 void *buf, unsigned int size);
995 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
996 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
997 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
998 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
999 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
1000 extern void blk_queue_dma_alignment(struct request_queue *, int);
1001 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1002 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1003 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1004 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1005 extern void blk_queue_flush(struct request_queue *q, unsigned int flush);
1006 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1007 extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
1008
1009 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1010 extern void blk_dump_rq_flags(struct request *, char *);
1011 extern long nr_blockdev_pages(void);
1012
1013 bool __must_check blk_get_queue(struct request_queue *);
1014 struct request_queue *blk_alloc_queue(gfp_t);
1015 struct request_queue *blk_alloc_queue_node(gfp_t, int);
1016 extern void blk_put_queue(struct request_queue *);
1017 extern void blk_set_queue_dying(struct request_queue *);
1018
1019 /*
1020 * block layer runtime pm functions
1021 */
1022 #ifdef CONFIG_PM
1023 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1024 extern int blk_pre_runtime_suspend(struct request_queue *q);
1025 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1026 extern void blk_pre_runtime_resume(struct request_queue *q);
1027 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1028 #else
1029 static inline void blk_pm_runtime_init(struct request_queue *q,
1030 struct device *dev) {}
1031 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1032 {
1033 return -ENOSYS;
1034 }
1035 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1036 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1037 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1038 #endif
1039
1040 /*
1041 * blk_plug permits building a queue of related requests by holding the I/O
1042 * fragments for a short period. This allows merging of sequential requests
1043 * into single larger request. As the requests are moved from a per-task list to
1044 * the device's request_queue in a batch, this results in improved scalability
1045 * as the lock contention for request_queue lock is reduced.
1046 *
1047 * It is ok not to disable preemption when adding the request to the plug list
1048 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1049 * the plug list when the task sleeps by itself. For details, please see
1050 * schedule() where blk_schedule_flush_plug() is called.
1051 */
1052 struct blk_plug {
1053 struct list_head list; /* requests */
1054 struct list_head mq_list; /* blk-mq requests */
1055 struct list_head cb_list; /* md requires an unplug callback */
1056 };
1057 #define BLK_MAX_REQUEST_COUNT 16
1058
1059 struct blk_plug_cb;
1060 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1061 struct blk_plug_cb {
1062 struct list_head list;
1063 blk_plug_cb_fn callback;
1064 void *data;
1065 };
1066 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1067 void *data, int size);
1068 extern void blk_start_plug(struct blk_plug *);
1069 extern void blk_finish_plug(struct blk_plug *);
1070 extern void blk_flush_plug_list(struct blk_plug *, bool);
1071
1072 static inline void blk_flush_plug(struct task_struct *tsk)
1073 {
1074 struct blk_plug *plug = tsk->plug;
1075
1076 if (plug)
1077 blk_flush_plug_list(plug, false);
1078 }
1079
1080 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1081 {
1082 struct blk_plug *plug = tsk->plug;
1083
1084 if (plug)
1085 blk_flush_plug_list(plug, true);
1086 }
1087
1088 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1089 {
1090 struct blk_plug *plug = tsk->plug;
1091
1092 return plug &&
1093 (!list_empty(&plug->list) ||
1094 !list_empty(&plug->mq_list) ||
1095 !list_empty(&plug->cb_list));
1096 }
1097
1098 /*
1099 * tag stuff
1100 */
1101 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1102 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1103 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1104 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1105 extern void blk_queue_free_tags(struct request_queue *);
1106 extern int blk_queue_resize_tags(struct request_queue *, int);
1107 extern void blk_queue_invalidate_tags(struct request_queue *);
1108 extern struct blk_queue_tag *blk_init_tags(int, int);
1109 extern void blk_free_tags(struct blk_queue_tag *);
1110
1111 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1112 int tag)
1113 {
1114 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1115 return NULL;
1116 return bqt->tag_index[tag];
1117 }
1118
1119 #define BLKDEV_DISCARD_SECURE 0x01 /* secure discard */
1120
1121 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1122 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1123 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1124 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1125 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1126 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1127 sector_t nr_sects, gfp_t gfp_mask, bool discard);
1128 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1129 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1130 {
1131 return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9),
1132 nr_blocks << (sb->s_blocksize_bits - 9),
1133 gfp_mask, flags);
1134 }
1135 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1136 sector_t nr_blocks, gfp_t gfp_mask)
1137 {
1138 return blkdev_issue_zeroout(sb->s_bdev,
1139 block << (sb->s_blocksize_bits - 9),
1140 nr_blocks << (sb->s_blocksize_bits - 9),
1141 gfp_mask, true);
1142 }
1143
1144 extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm);
1145
1146 enum blk_default_limits {
1147 BLK_MAX_SEGMENTS = 128,
1148 BLK_SAFE_MAX_SECTORS = 255,
1149 BLK_DEF_MAX_SECTORS = 2560,
1150 BLK_MAX_SEGMENT_SIZE = 65536,
1151 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1152 };
1153
1154 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1155
1156 static inline unsigned long queue_bounce_pfn(struct request_queue *q)
1157 {
1158 return q->limits.bounce_pfn;
1159 }
1160
1161 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1162 {
1163 return q->limits.seg_boundary_mask;
1164 }
1165
1166 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1167 {
1168 return q->limits.virt_boundary_mask;
1169 }
1170
1171 static inline unsigned int queue_max_sectors(struct request_queue *q)
1172 {
1173 return q->limits.max_sectors;
1174 }
1175
1176 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1177 {
1178 return q->limits.max_hw_sectors;
1179 }
1180
1181 static inline unsigned short queue_max_segments(struct request_queue *q)
1182 {
1183 return q->limits.max_segments;
1184 }
1185
1186 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1187 {
1188 return q->limits.max_segment_size;
1189 }
1190
1191 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1192 {
1193 int retval = 512;
1194
1195 if (q && q->limits.logical_block_size)
1196 retval = q->limits.logical_block_size;
1197
1198 return retval;
1199 }
1200
1201 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1202 {
1203 return queue_logical_block_size(bdev_get_queue(bdev));
1204 }
1205
1206 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1207 {
1208 return q->limits.physical_block_size;
1209 }
1210
1211 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1212 {
1213 return queue_physical_block_size(bdev_get_queue(bdev));
1214 }
1215
1216 static inline unsigned int queue_io_min(struct request_queue *q)
1217 {
1218 return q->limits.io_min;
1219 }
1220
1221 static inline int bdev_io_min(struct block_device *bdev)
1222 {
1223 return queue_io_min(bdev_get_queue(bdev));
1224 }
1225
1226 static inline unsigned int queue_io_opt(struct request_queue *q)
1227 {
1228 return q->limits.io_opt;
1229 }
1230
1231 static inline int bdev_io_opt(struct block_device *bdev)
1232 {
1233 return queue_io_opt(bdev_get_queue(bdev));
1234 }
1235
1236 static inline int queue_alignment_offset(struct request_queue *q)
1237 {
1238 if (q->limits.misaligned)
1239 return -1;
1240
1241 return q->limits.alignment_offset;
1242 }
1243
1244 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1245 {
1246 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1247 unsigned int alignment = sector_div(sector, granularity >> 9) << 9;
1248
1249 return (granularity + lim->alignment_offset - alignment) % granularity;
1250 }
1251
1252 static inline int bdev_alignment_offset(struct block_device *bdev)
1253 {
1254 struct request_queue *q = bdev_get_queue(bdev);
1255
1256 if (q->limits.misaligned)
1257 return -1;
1258
1259 if (bdev != bdev->bd_contains)
1260 return bdev->bd_part->alignment_offset;
1261
1262 return q->limits.alignment_offset;
1263 }
1264
1265 static inline int queue_discard_alignment(struct request_queue *q)
1266 {
1267 if (q->limits.discard_misaligned)
1268 return -1;
1269
1270 return q->limits.discard_alignment;
1271 }
1272
1273 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1274 {
1275 unsigned int alignment, granularity, offset;
1276
1277 if (!lim->max_discard_sectors)
1278 return 0;
1279
1280 /* Why are these in bytes, not sectors? */
1281 alignment = lim->discard_alignment >> 9;
1282 granularity = lim->discard_granularity >> 9;
1283 if (!granularity)
1284 return 0;
1285
1286 /* Offset of the partition start in 'granularity' sectors */
1287 offset = sector_div(sector, granularity);
1288
1289 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1290 offset = (granularity + alignment - offset) % granularity;
1291
1292 /* Turn it back into bytes, gaah */
1293 return offset << 9;
1294 }
1295
1296 static inline int bdev_discard_alignment(struct block_device *bdev)
1297 {
1298 struct request_queue *q = bdev_get_queue(bdev);
1299
1300 if (bdev != bdev->bd_contains)
1301 return bdev->bd_part->discard_alignment;
1302
1303 return q->limits.discard_alignment;
1304 }
1305
1306 static inline unsigned int queue_discard_zeroes_data(struct request_queue *q)
1307 {
1308 if (q->limits.max_discard_sectors && q->limits.discard_zeroes_data == 1)
1309 return 1;
1310
1311 return 0;
1312 }
1313
1314 static inline unsigned int bdev_discard_zeroes_data(struct block_device *bdev)
1315 {
1316 return queue_discard_zeroes_data(bdev_get_queue(bdev));
1317 }
1318
1319 static inline unsigned int bdev_write_same(struct block_device *bdev)
1320 {
1321 struct request_queue *q = bdev_get_queue(bdev);
1322
1323 if (q)
1324 return q->limits.max_write_same_sectors;
1325
1326 return 0;
1327 }
1328
1329 static inline int queue_dma_alignment(struct request_queue *q)
1330 {
1331 return q ? q->dma_alignment : 511;
1332 }
1333
1334 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1335 unsigned int len)
1336 {
1337 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1338 return !(addr & alignment) && !(len & alignment);
1339 }
1340
1341 /* assumes size > 256 */
1342 static inline unsigned int blksize_bits(unsigned int size)
1343 {
1344 unsigned int bits = 8;
1345 do {
1346 bits++;
1347 size >>= 1;
1348 } while (size > 256);
1349 return bits;
1350 }
1351
1352 static inline unsigned int block_size(struct block_device *bdev)
1353 {
1354 return bdev->bd_block_size;
1355 }
1356
1357 static inline bool queue_flush_queueable(struct request_queue *q)
1358 {
1359 return !q->flush_not_queueable;
1360 }
1361
1362 typedef struct {struct page *v;} Sector;
1363
1364 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1365
1366 static inline void put_dev_sector(Sector p)
1367 {
1368 page_cache_release(p.v);
1369 }
1370
1371 /*
1372 * Check if adding a bio_vec after bprv with offset would create a gap in
1373 * the SG list. Most drivers don't care about this, but some do.
1374 */
1375 static inline bool bvec_gap_to_prev(struct request_queue *q,
1376 struct bio_vec *bprv, unsigned int offset)
1377 {
1378 if (!queue_virt_boundary(q))
1379 return false;
1380 return offset ||
1381 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1382 }
1383
1384 static inline bool bio_will_gap(struct request_queue *q, struct bio *prev,
1385 struct bio *next)
1386 {
1387 if (!bio_has_data(prev))
1388 return false;
1389
1390 return bvec_gap_to_prev(q, &prev->bi_io_vec[prev->bi_vcnt - 1],
1391 next->bi_io_vec[0].bv_offset);
1392 }
1393
1394 static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
1395 {
1396 return bio_will_gap(req->q, req->biotail, bio);
1397 }
1398
1399 static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
1400 {
1401 return bio_will_gap(req->q, bio, req->bio);
1402 }
1403
1404 struct work_struct;
1405 int kblockd_schedule_work(struct work_struct *work);
1406 int kblockd_schedule_delayed_work(struct delayed_work *dwork, unsigned long delay);
1407 int kblockd_schedule_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1408
1409 #ifdef CONFIG_BLK_CGROUP
1410 /*
1411 * This should not be using sched_clock(). A real patch is in progress
1412 * to fix this up, until that is in place we need to disable preemption
1413 * around sched_clock() in this function and set_io_start_time_ns().
1414 */
1415 static inline void set_start_time_ns(struct request *req)
1416 {
1417 preempt_disable();
1418 req->start_time_ns = sched_clock();
1419 preempt_enable();
1420 }
1421
1422 static inline void set_io_start_time_ns(struct request *req)
1423 {
1424 preempt_disable();
1425 req->io_start_time_ns = sched_clock();
1426 preempt_enable();
1427 }
1428
1429 static inline uint64_t rq_start_time_ns(struct request *req)
1430 {
1431 return req->start_time_ns;
1432 }
1433
1434 static inline uint64_t rq_io_start_time_ns(struct request *req)
1435 {
1436 return req->io_start_time_ns;
1437 }
1438 #else
1439 static inline void set_start_time_ns(struct request *req) {}
1440 static inline void set_io_start_time_ns(struct request *req) {}
1441 static inline uint64_t rq_start_time_ns(struct request *req)
1442 {
1443 return 0;
1444 }
1445 static inline uint64_t rq_io_start_time_ns(struct request *req)
1446 {
1447 return 0;
1448 }
1449 #endif
1450
1451 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1452 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1453 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1454 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1455
1456 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1457
1458 enum blk_integrity_flags {
1459 BLK_INTEGRITY_VERIFY = 1 << 0,
1460 BLK_INTEGRITY_GENERATE = 1 << 1,
1461 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2,
1462 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3,
1463 };
1464
1465 struct blk_integrity_iter {
1466 void *prot_buf;
1467 void *data_buf;
1468 sector_t seed;
1469 unsigned int data_size;
1470 unsigned short interval;
1471 const char *disk_name;
1472 };
1473
1474 typedef int (integrity_processing_fn) (struct blk_integrity_iter *);
1475
1476 struct blk_integrity_profile {
1477 integrity_processing_fn *generate_fn;
1478 integrity_processing_fn *verify_fn;
1479 const char *name;
1480 };
1481
1482 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1483 extern void blk_integrity_unregister(struct gendisk *);
1484 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1485 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1486 struct scatterlist *);
1487 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1488 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1489 struct request *);
1490 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1491 struct bio *);
1492
1493 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1494 {
1495 struct blk_integrity *bi = &disk->queue->integrity;
1496
1497 if (!bi->profile)
1498 return NULL;
1499
1500 return bi;
1501 }
1502
1503 static inline
1504 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1505 {
1506 return blk_get_integrity(bdev->bd_disk);
1507 }
1508
1509 static inline bool blk_integrity_rq(struct request *rq)
1510 {
1511 return rq->cmd_flags & REQ_INTEGRITY;
1512 }
1513
1514 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1515 unsigned int segs)
1516 {
1517 q->limits.max_integrity_segments = segs;
1518 }
1519
1520 static inline unsigned short
1521 queue_max_integrity_segments(struct request_queue *q)
1522 {
1523 return q->limits.max_integrity_segments;
1524 }
1525
1526 static inline bool integrity_req_gap_back_merge(struct request *req,
1527 struct bio *next)
1528 {
1529 struct bio_integrity_payload *bip = bio_integrity(req->bio);
1530 struct bio_integrity_payload *bip_next = bio_integrity(next);
1531
1532 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1533 bip_next->bip_vec[0].bv_offset);
1534 }
1535
1536 static inline bool integrity_req_gap_front_merge(struct request *req,
1537 struct bio *bio)
1538 {
1539 struct bio_integrity_payload *bip = bio_integrity(bio);
1540 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
1541
1542 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1543 bip_next->bip_vec[0].bv_offset);
1544 }
1545
1546 #else /* CONFIG_BLK_DEV_INTEGRITY */
1547
1548 struct bio;
1549 struct block_device;
1550 struct gendisk;
1551 struct blk_integrity;
1552
1553 static inline int blk_integrity_rq(struct request *rq)
1554 {
1555 return 0;
1556 }
1557 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1558 struct bio *b)
1559 {
1560 return 0;
1561 }
1562 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1563 struct bio *b,
1564 struct scatterlist *s)
1565 {
1566 return 0;
1567 }
1568 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1569 {
1570 return NULL;
1571 }
1572 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1573 {
1574 return NULL;
1575 }
1576 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1577 {
1578 return 0;
1579 }
1580 static inline void blk_integrity_register(struct gendisk *d,
1581 struct blk_integrity *b)
1582 {
1583 }
1584 static inline void blk_integrity_unregister(struct gendisk *d)
1585 {
1586 }
1587 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1588 unsigned int segs)
1589 {
1590 }
1591 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1592 {
1593 return 0;
1594 }
1595 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1596 struct request *r1,
1597 struct request *r2)
1598 {
1599 return true;
1600 }
1601 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1602 struct request *r,
1603 struct bio *b)
1604 {
1605 return true;
1606 }
1607
1608 static inline bool integrity_req_gap_back_merge(struct request *req,
1609 struct bio *next)
1610 {
1611 return false;
1612 }
1613 static inline bool integrity_req_gap_front_merge(struct request *req,
1614 struct bio *bio)
1615 {
1616 return false;
1617 }
1618
1619 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1620
1621 /**
1622 * struct blk_dax_ctl - control and output parameters for ->direct_access
1623 * @sector: (input) offset relative to a block_device
1624 * @addr: (output) kernel virtual address for @sector populated by driver
1625 * @pfn: (output) page frame number for @addr populated by driver
1626 * @size: (input) number of bytes requested
1627 */
1628 struct blk_dax_ctl {
1629 sector_t sector;
1630 void __pmem *addr;
1631 long size;
1632 pfn_t pfn;
1633 };
1634
1635 struct block_device_operations {
1636 int (*open) (struct block_device *, fmode_t);
1637 void (*release) (struct gendisk *, fmode_t);
1638 int (*rw_page)(struct block_device *, sector_t, struct page *, int rw);
1639 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1640 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1641 long (*direct_access)(struct block_device *, sector_t, void __pmem **,
1642 pfn_t *);
1643 unsigned int (*check_events) (struct gendisk *disk,
1644 unsigned int clearing);
1645 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1646 int (*media_changed) (struct gendisk *);
1647 void (*unlock_native_capacity) (struct gendisk *);
1648 int (*revalidate_disk) (struct gendisk *);
1649 int (*getgeo)(struct block_device *, struct hd_geometry *);
1650 /* this callback is with swap_lock and sometimes page table lock held */
1651 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1652 struct module *owner;
1653 const struct pr_ops *pr_ops;
1654 };
1655
1656 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1657 unsigned long);
1658 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1659 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1660 struct writeback_control *);
1661 extern long bdev_direct_access(struct block_device *, struct blk_dax_ctl *);
1662 #else /* CONFIG_BLOCK */
1663
1664 struct block_device;
1665
1666 /*
1667 * stubs for when the block layer is configured out
1668 */
1669 #define buffer_heads_over_limit 0
1670
1671 static inline long nr_blockdev_pages(void)
1672 {
1673 return 0;
1674 }
1675
1676 struct blk_plug {
1677 };
1678
1679 static inline void blk_start_plug(struct blk_plug *plug)
1680 {
1681 }
1682
1683 static inline void blk_finish_plug(struct blk_plug *plug)
1684 {
1685 }
1686
1687 static inline void blk_flush_plug(struct task_struct *task)
1688 {
1689 }
1690
1691 static inline void blk_schedule_flush_plug(struct task_struct *task)
1692 {
1693 }
1694
1695
1696 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1697 {
1698 return false;
1699 }
1700
1701 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
1702 sector_t *error_sector)
1703 {
1704 return 0;
1705 }
1706
1707 #endif /* CONFIG_BLOCK */
1708
1709 #endif