ext4: use bio op helprs in ext4 crypto code
[linux-block.git] / include / linux / blkdev.h
... / ...
CommitLineData
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
28struct module;
29struct scsi_ioctl_command;
30
31struct request_queue;
32struct elevator_queue;
33struct blk_trace;
34struct request;
35struct sg_io_hdr;
36struct bsg_job;
37struct blkcg_gq;
38struct blk_flush_queue;
39struct 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
50struct request;
51typedef void (rq_end_io_fn)(struct request *, int);
52
53#define BLK_RL_SYNCFULL (1U << 0)
54#define BLK_RL_ASYNCFULL (1U << 1)
55
56struct 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 */
75enum 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 */
89struct 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#define REQ_OP_SHIFT (8 * sizeof(u64) - REQ_OP_BITS)
204#define req_op(req) ((req)->cmd_flags >> REQ_OP_SHIFT)
205
206#define req_set_op(req, op) do { \
207 WARN_ON(op >= (1 << REQ_OP_BITS)); \
208 (req)->cmd_flags &= ((1ULL << REQ_OP_SHIFT) - 1); \
209 (req)->cmd_flags |= ((u64) (op) << REQ_OP_SHIFT); \
210} while (0)
211
212#define req_set_op_attrs(req, op, flags) do { \
213 req_set_op(req, op); \
214 (req)->cmd_flags |= flags; \
215} while (0)
216
217static inline unsigned short req_get_ioprio(struct request *req)
218{
219 return req->ioprio;
220}
221
222#include <linux/elevator.h>
223
224struct blk_queue_ctx;
225
226typedef void (request_fn_proc) (struct request_queue *q);
227typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio);
228typedef int (prep_rq_fn) (struct request_queue *, struct request *);
229typedef void (unprep_rq_fn) (struct request_queue *, struct request *);
230
231struct bio_vec;
232typedef void (softirq_done_fn)(struct request *);
233typedef int (dma_drain_needed_fn)(struct request *);
234typedef int (lld_busy_fn) (struct request_queue *q);
235typedef int (bsg_job_fn) (struct bsg_job *);
236
237enum blk_eh_timer_return {
238 BLK_EH_NOT_HANDLED,
239 BLK_EH_HANDLED,
240 BLK_EH_RESET_TIMER,
241};
242
243typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
244
245enum blk_queue_state {
246 Queue_down,
247 Queue_up,
248};
249
250struct blk_queue_tag {
251 struct request **tag_index; /* map of busy tags */
252 unsigned long *tag_map; /* bit map of free/busy tags */
253 int busy; /* current depth */
254 int max_depth; /* what we will send to device */
255 int real_max_depth; /* what the array can hold */
256 atomic_t refcnt; /* map can be shared */
257 int alloc_policy; /* tag allocation policy */
258 int next_tag; /* next tag */
259};
260#define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
261#define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
262
263#define BLK_SCSI_MAX_CMDS (256)
264#define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
265
266struct queue_limits {
267 unsigned long bounce_pfn;
268 unsigned long seg_boundary_mask;
269 unsigned long virt_boundary_mask;
270
271 unsigned int max_hw_sectors;
272 unsigned int max_dev_sectors;
273 unsigned int chunk_sectors;
274 unsigned int max_sectors;
275 unsigned int max_segment_size;
276 unsigned int physical_block_size;
277 unsigned int alignment_offset;
278 unsigned int io_min;
279 unsigned int io_opt;
280 unsigned int max_discard_sectors;
281 unsigned int max_hw_discard_sectors;
282 unsigned int max_write_same_sectors;
283 unsigned int discard_granularity;
284 unsigned int discard_alignment;
285
286 unsigned short logical_block_size;
287 unsigned short max_segments;
288 unsigned short max_integrity_segments;
289
290 unsigned char misaligned;
291 unsigned char discard_misaligned;
292 unsigned char cluster;
293 unsigned char discard_zeroes_data;
294 unsigned char raid_partial_stripes_expensive;
295};
296
297struct request_queue {
298 /*
299 * Together with queue_head for cacheline sharing
300 */
301 struct list_head queue_head;
302 struct request *last_merge;
303 struct elevator_queue *elevator;
304 int nr_rqs[2]; /* # allocated [a]sync rqs */
305 int nr_rqs_elvpriv; /* # allocated rqs w/ elvpriv */
306
307 /*
308 * If blkcg is not used, @q->root_rl serves all requests. If blkcg
309 * is used, root blkg allocates from @q->root_rl and all other
310 * blkgs from their own blkg->rl. Which one to use should be
311 * determined using bio_request_list().
312 */
313 struct request_list root_rl;
314
315 request_fn_proc *request_fn;
316 make_request_fn *make_request_fn;
317 prep_rq_fn *prep_rq_fn;
318 unprep_rq_fn *unprep_rq_fn;
319 softirq_done_fn *softirq_done_fn;
320 rq_timed_out_fn *rq_timed_out_fn;
321 dma_drain_needed_fn *dma_drain_needed;
322 lld_busy_fn *lld_busy_fn;
323
324 struct blk_mq_ops *mq_ops;
325
326 unsigned int *mq_map;
327
328 /* sw queues */
329 struct blk_mq_ctx __percpu *queue_ctx;
330 unsigned int nr_queues;
331
332 /* hw dispatch queues */
333 struct blk_mq_hw_ctx **queue_hw_ctx;
334 unsigned int nr_hw_queues;
335
336 /*
337 * Dispatch queue sorting
338 */
339 sector_t end_sector;
340 struct request *boundary_rq;
341
342 /*
343 * Delayed queue handling
344 */
345 struct delayed_work delay_work;
346
347 struct backing_dev_info backing_dev_info;
348
349 /*
350 * The queue owner gets to use this for whatever they like.
351 * ll_rw_blk doesn't touch it.
352 */
353 void *queuedata;
354
355 /*
356 * various queue flags, see QUEUE_* below
357 */
358 unsigned long queue_flags;
359
360 /*
361 * ida allocated id for this queue. Used to index queues from
362 * ioctx.
363 */
364 int id;
365
366 /*
367 * queue needs bounce pages for pages above this limit
368 */
369 gfp_t bounce_gfp;
370
371 /*
372 * protects queue structures from reentrancy. ->__queue_lock should
373 * _never_ be used directly, it is queue private. always use
374 * ->queue_lock.
375 */
376 spinlock_t __queue_lock;
377 spinlock_t *queue_lock;
378
379 /*
380 * queue kobject
381 */
382 struct kobject kobj;
383
384 /*
385 * mq queue kobject
386 */
387 struct kobject mq_kobj;
388
389#ifdef CONFIG_BLK_DEV_INTEGRITY
390 struct blk_integrity integrity;
391#endif /* CONFIG_BLK_DEV_INTEGRITY */
392
393#ifdef CONFIG_PM
394 struct device *dev;
395 int rpm_status;
396 unsigned int nr_pending;
397#endif
398
399 /*
400 * queue settings
401 */
402 unsigned long nr_requests; /* Max # of requests */
403 unsigned int nr_congestion_on;
404 unsigned int nr_congestion_off;
405 unsigned int nr_batching;
406
407 unsigned int dma_drain_size;
408 void *dma_drain_buffer;
409 unsigned int dma_pad_mask;
410 unsigned int dma_alignment;
411
412 struct blk_queue_tag *queue_tags;
413 struct list_head tag_busy_list;
414
415 unsigned int nr_sorted;
416 unsigned int in_flight[2];
417 /*
418 * Number of active block driver functions for which blk_drain_queue()
419 * must wait. Must be incremented around functions that unlock the
420 * queue_lock internally, e.g. scsi_request_fn().
421 */
422 unsigned int request_fn_active;
423
424 unsigned int rq_timeout;
425 struct timer_list timeout;
426 struct work_struct timeout_work;
427 struct list_head timeout_list;
428
429 struct list_head icq_list;
430#ifdef CONFIG_BLK_CGROUP
431 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
432 struct blkcg_gq *root_blkg;
433 struct list_head blkg_list;
434#endif
435
436 struct queue_limits limits;
437
438 /*
439 * sg stuff
440 */
441 unsigned int sg_timeout;
442 unsigned int sg_reserved_size;
443 int node;
444#ifdef CONFIG_BLK_DEV_IO_TRACE
445 struct blk_trace *blk_trace;
446#endif
447 /*
448 * for flush operations
449 */
450 struct blk_flush_queue *fq;
451
452 struct list_head requeue_list;
453 spinlock_t requeue_lock;
454 struct work_struct requeue_work;
455
456 struct mutex sysfs_lock;
457
458 int bypass_depth;
459 atomic_t mq_freeze_depth;
460
461#if defined(CONFIG_BLK_DEV_BSG)
462 bsg_job_fn *bsg_job_fn;
463 int bsg_job_size;
464 struct bsg_class_device bsg_dev;
465#endif
466
467#ifdef CONFIG_BLK_DEV_THROTTLING
468 /* Throttle data */
469 struct throtl_data *td;
470#endif
471 struct rcu_head rcu_head;
472 wait_queue_head_t mq_freeze_wq;
473 struct percpu_ref q_usage_counter;
474 struct list_head all_q_node;
475
476 struct blk_mq_tag_set *tag_set;
477 struct list_head tag_set_list;
478 struct bio_set *bio_split;
479
480 bool mq_sysfs_init_done;
481};
482
483#define QUEUE_FLAG_QUEUED 1 /* uses generic tag queueing */
484#define QUEUE_FLAG_STOPPED 2 /* queue is stopped */
485#define QUEUE_FLAG_SYNCFULL 3 /* read queue has been filled */
486#define QUEUE_FLAG_ASYNCFULL 4 /* write queue has been filled */
487#define QUEUE_FLAG_DYING 5 /* queue being torn down */
488#define QUEUE_FLAG_BYPASS 6 /* act as dumb FIFO queue */
489#define QUEUE_FLAG_BIDI 7 /* queue supports bidi requests */
490#define QUEUE_FLAG_NOMERGES 8 /* disable merge attempts */
491#define QUEUE_FLAG_SAME_COMP 9 /* complete on same CPU-group */
492#define QUEUE_FLAG_FAIL_IO 10 /* fake timeout */
493#define QUEUE_FLAG_STACKABLE 11 /* supports request stacking */
494#define QUEUE_FLAG_NONROT 12 /* non-rotational device (SSD) */
495#define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
496#define QUEUE_FLAG_IO_STAT 13 /* do IO stats */
497#define QUEUE_FLAG_DISCARD 14 /* supports DISCARD */
498#define QUEUE_FLAG_NOXMERGES 15 /* No extended merges */
499#define QUEUE_FLAG_ADD_RANDOM 16 /* Contributes to random pool */
500#define QUEUE_FLAG_SECDISCARD 17 /* supports SECDISCARD */
501#define QUEUE_FLAG_SAME_FORCE 18 /* force complete on same CPU */
502#define QUEUE_FLAG_DEAD 19 /* queue tear-down finished */
503#define QUEUE_FLAG_INIT_DONE 20 /* queue is initialized */
504#define QUEUE_FLAG_NO_SG_MERGE 21 /* don't attempt to merge SG segments*/
505#define QUEUE_FLAG_POLL 22 /* IO polling enabled if set */
506#define QUEUE_FLAG_WC 23 /* Write back caching */
507#define QUEUE_FLAG_FUA 24 /* device supports FUA writes */
508#define QUEUE_FLAG_FLUSH_NQ 25 /* flush not queueuable */
509
510#define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
511 (1 << QUEUE_FLAG_STACKABLE) | \
512 (1 << QUEUE_FLAG_SAME_COMP) | \
513 (1 << QUEUE_FLAG_ADD_RANDOM))
514
515#define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
516 (1 << QUEUE_FLAG_STACKABLE) | \
517 (1 << QUEUE_FLAG_SAME_COMP) | \
518 (1 << QUEUE_FLAG_POLL))
519
520static inline void queue_lockdep_assert_held(struct request_queue *q)
521{
522 if (q->queue_lock)
523 lockdep_assert_held(q->queue_lock);
524}
525
526static inline void queue_flag_set_unlocked(unsigned int flag,
527 struct request_queue *q)
528{
529 __set_bit(flag, &q->queue_flags);
530}
531
532static inline int queue_flag_test_and_clear(unsigned int flag,
533 struct request_queue *q)
534{
535 queue_lockdep_assert_held(q);
536
537 if (test_bit(flag, &q->queue_flags)) {
538 __clear_bit(flag, &q->queue_flags);
539 return 1;
540 }
541
542 return 0;
543}
544
545static inline int queue_flag_test_and_set(unsigned int flag,
546 struct request_queue *q)
547{
548 queue_lockdep_assert_held(q);
549
550 if (!test_bit(flag, &q->queue_flags)) {
551 __set_bit(flag, &q->queue_flags);
552 return 0;
553 }
554
555 return 1;
556}
557
558static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
559{
560 queue_lockdep_assert_held(q);
561 __set_bit(flag, &q->queue_flags);
562}
563
564static inline void queue_flag_clear_unlocked(unsigned int flag,
565 struct request_queue *q)
566{
567 __clear_bit(flag, &q->queue_flags);
568}
569
570static inline int queue_in_flight(struct request_queue *q)
571{
572 return q->in_flight[0] + q->in_flight[1];
573}
574
575static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
576{
577 queue_lockdep_assert_held(q);
578 __clear_bit(flag, &q->queue_flags);
579}
580
581#define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
582#define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
583#define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
584#define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
585#define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
586#define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
587#define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
588#define blk_queue_noxmerges(q) \
589 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
590#define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
591#define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
592#define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
593#define blk_queue_stackable(q) \
594 test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
595#define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
596#define blk_queue_secdiscard(q) (blk_queue_discard(q) && \
597 test_bit(QUEUE_FLAG_SECDISCARD, &(q)->queue_flags))
598
599#define blk_noretry_request(rq) \
600 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
601 REQ_FAILFAST_DRIVER))
602
603#define blk_account_rq(rq) \
604 (((rq)->cmd_flags & REQ_STARTED) && \
605 ((rq)->cmd_type == REQ_TYPE_FS))
606
607#define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
608#define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
609/* rq->queuelist of dequeued request must be list_empty() */
610#define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
611
612#define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
613
614#define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ)
615
616/*
617 * Driver can handle struct request, if it either has an old style
618 * request_fn defined, or is blk-mq based.
619 */
620static inline bool queue_is_rq_based(struct request_queue *q)
621{
622 return q->request_fn || q->mq_ops;
623}
624
625static inline unsigned int blk_queue_cluster(struct request_queue *q)
626{
627 return q->limits.cluster;
628}
629
630/*
631 * We regard a request as sync, if either a read or a sync write
632 */
633static inline bool rw_is_sync(int op, unsigned int rw_flags)
634{
635 return op == REQ_OP_READ || (rw_flags & REQ_SYNC);
636}
637
638static inline bool rq_is_sync(struct request *rq)
639{
640 return rw_is_sync(req_op(rq), rq->cmd_flags);
641}
642
643static inline bool blk_rl_full(struct request_list *rl, bool sync)
644{
645 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
646
647 return rl->flags & flag;
648}
649
650static inline void blk_set_rl_full(struct request_list *rl, bool sync)
651{
652 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
653
654 rl->flags |= flag;
655}
656
657static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
658{
659 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
660
661 rl->flags &= ~flag;
662}
663
664static inline bool rq_mergeable(struct request *rq)
665{
666 if (rq->cmd_type != REQ_TYPE_FS)
667 return false;
668
669 if (req_op(rq) == REQ_OP_FLUSH)
670 return false;
671
672 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
673 return false;
674
675 return true;
676}
677
678static inline bool blk_check_merge_flags(unsigned int flags1, unsigned int op1,
679 unsigned int flags2, unsigned int op2)
680{
681 if ((op1 == REQ_OP_DISCARD) != (op2 == REQ_OP_DISCARD))
682 return false;
683
684 if ((flags1 & REQ_SECURE) != (flags2 & REQ_SECURE))
685 return false;
686
687 if ((op1 == REQ_OP_WRITE_SAME) != (op2 == REQ_OP_WRITE_SAME))
688 return false;
689
690 return true;
691}
692
693static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
694{
695 if (bio_data(a) == bio_data(b))
696 return true;
697
698 return false;
699}
700
701/*
702 * q->prep_rq_fn return values
703 */
704enum {
705 BLKPREP_OK, /* serve it */
706 BLKPREP_KILL, /* fatal error, kill, return -EIO */
707 BLKPREP_DEFER, /* leave on queue */
708 BLKPREP_INVALID, /* invalid command, kill, return -EREMOTEIO */
709};
710
711extern unsigned long blk_max_low_pfn, blk_max_pfn;
712
713/*
714 * standard bounce addresses:
715 *
716 * BLK_BOUNCE_HIGH : bounce all highmem pages
717 * BLK_BOUNCE_ANY : don't bounce anything
718 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
719 */
720
721#if BITS_PER_LONG == 32
722#define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
723#else
724#define BLK_BOUNCE_HIGH -1ULL
725#endif
726#define BLK_BOUNCE_ANY (-1ULL)
727#define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
728
729/*
730 * default timeout for SG_IO if none specified
731 */
732#define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
733#define BLK_MIN_SG_TIMEOUT (7 * HZ)
734
735#ifdef CONFIG_BOUNCE
736extern int init_emergency_isa_pool(void);
737extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
738#else
739static inline int init_emergency_isa_pool(void)
740{
741 return 0;
742}
743static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
744{
745}
746#endif /* CONFIG_MMU */
747
748struct rq_map_data {
749 struct page **pages;
750 int page_order;
751 int nr_entries;
752 unsigned long offset;
753 int null_mapped;
754 int from_user;
755};
756
757struct req_iterator {
758 struct bvec_iter iter;
759 struct bio *bio;
760};
761
762/* This should not be used directly - use rq_for_each_segment */
763#define for_each_bio(_bio) \
764 for (; _bio; _bio = _bio->bi_next)
765#define __rq_for_each_bio(_bio, rq) \
766 if ((rq->bio)) \
767 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
768
769#define rq_for_each_segment(bvl, _rq, _iter) \
770 __rq_for_each_bio(_iter.bio, _rq) \
771 bio_for_each_segment(bvl, _iter.bio, _iter.iter)
772
773#define rq_iter_last(bvec, _iter) \
774 (_iter.bio->bi_next == NULL && \
775 bio_iter_last(bvec, _iter.iter))
776
777#ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
778# error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
779#endif
780#if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
781extern void rq_flush_dcache_pages(struct request *rq);
782#else
783static inline void rq_flush_dcache_pages(struct request *rq)
784{
785}
786#endif
787
788#ifdef CONFIG_PRINTK
789#define vfs_msg(sb, level, fmt, ...) \
790 __vfs_msg(sb, level, fmt, ##__VA_ARGS__)
791#else
792#define vfs_msg(sb, level, fmt, ...) \
793do { \
794 no_printk(fmt, ##__VA_ARGS__); \
795 __vfs_msg(sb, "", " "); \
796} while (0)
797#endif
798
799extern int blk_register_queue(struct gendisk *disk);
800extern void blk_unregister_queue(struct gendisk *disk);
801extern blk_qc_t generic_make_request(struct bio *bio);
802extern void blk_rq_init(struct request_queue *q, struct request *rq);
803extern void blk_put_request(struct request *);
804extern void __blk_put_request(struct request_queue *, struct request *);
805extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
806extern struct request *blk_make_request(struct request_queue *, struct bio *,
807 gfp_t);
808extern void blk_rq_set_block_pc(struct request *);
809extern void blk_requeue_request(struct request_queue *, struct request *);
810extern void blk_add_request_payload(struct request *rq, struct page *page,
811 int offset, unsigned int len);
812extern int blk_lld_busy(struct request_queue *q);
813extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
814 struct bio_set *bs, gfp_t gfp_mask,
815 int (*bio_ctr)(struct bio *, struct bio *, void *),
816 void *data);
817extern void blk_rq_unprep_clone(struct request *rq);
818extern int blk_insert_cloned_request(struct request_queue *q,
819 struct request *rq);
820extern void blk_delay_queue(struct request_queue *, unsigned long);
821extern void blk_queue_split(struct request_queue *, struct bio **,
822 struct bio_set *);
823extern void blk_recount_segments(struct request_queue *, struct bio *);
824extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
825extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
826 unsigned int, void __user *);
827extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
828 unsigned int, void __user *);
829extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
830 struct scsi_ioctl_command __user *);
831
832extern int blk_queue_enter(struct request_queue *q, bool nowait);
833extern void blk_queue_exit(struct request_queue *q);
834extern void blk_start_queue(struct request_queue *q);
835extern void blk_start_queue_async(struct request_queue *q);
836extern void blk_stop_queue(struct request_queue *q);
837extern void blk_sync_queue(struct request_queue *q);
838extern void __blk_stop_queue(struct request_queue *q);
839extern void __blk_run_queue(struct request_queue *q);
840extern void __blk_run_queue_uncond(struct request_queue *q);
841extern void blk_run_queue(struct request_queue *);
842extern void blk_run_queue_async(struct request_queue *q);
843extern int blk_rq_map_user(struct request_queue *, struct request *,
844 struct rq_map_data *, void __user *, unsigned long,
845 gfp_t);
846extern int blk_rq_unmap_user(struct bio *);
847extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
848extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
849 struct rq_map_data *, const struct iov_iter *,
850 gfp_t);
851extern int blk_execute_rq(struct request_queue *, struct gendisk *,
852 struct request *, int);
853extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
854 struct request *, int, rq_end_io_fn *);
855
856bool blk_poll(struct request_queue *q, blk_qc_t cookie);
857
858static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
859{
860 return bdev->bd_disk->queue; /* this is never NULL */
861}
862
863/*
864 * blk_rq_pos() : the current sector
865 * blk_rq_bytes() : bytes left in the entire request
866 * blk_rq_cur_bytes() : bytes left in the current segment
867 * blk_rq_err_bytes() : bytes left till the next error boundary
868 * blk_rq_sectors() : sectors left in the entire request
869 * blk_rq_cur_sectors() : sectors left in the current segment
870 */
871static inline sector_t blk_rq_pos(const struct request *rq)
872{
873 return rq->__sector;
874}
875
876static inline unsigned int blk_rq_bytes(const struct request *rq)
877{
878 return rq->__data_len;
879}
880
881static inline int blk_rq_cur_bytes(const struct request *rq)
882{
883 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
884}
885
886extern unsigned int blk_rq_err_bytes(const struct request *rq);
887
888static inline unsigned int blk_rq_sectors(const struct request *rq)
889{
890 return blk_rq_bytes(rq) >> 9;
891}
892
893static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
894{
895 return blk_rq_cur_bytes(rq) >> 9;
896}
897
898static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
899 int op)
900{
901 if (unlikely(op == REQ_OP_DISCARD))
902 return min(q->limits.max_discard_sectors, UINT_MAX >> 9);
903
904 if (unlikely(op == REQ_OP_WRITE_SAME))
905 return q->limits.max_write_same_sectors;
906
907 return q->limits.max_sectors;
908}
909
910/*
911 * Return maximum size of a request at given offset. Only valid for
912 * file system requests.
913 */
914static inline unsigned int blk_max_size_offset(struct request_queue *q,
915 sector_t offset)
916{
917 if (!q->limits.chunk_sectors)
918 return q->limits.max_sectors;
919
920 return q->limits.chunk_sectors -
921 (offset & (q->limits.chunk_sectors - 1));
922}
923
924static inline unsigned int blk_rq_get_max_sectors(struct request *rq)
925{
926 struct request_queue *q = rq->q;
927
928 if (unlikely(rq->cmd_type != REQ_TYPE_FS))
929 return q->limits.max_hw_sectors;
930
931 if (!q->limits.chunk_sectors || (req_op(rq) == REQ_OP_DISCARD))
932 return blk_queue_get_max_sectors(q, req_op(rq));
933
934 return min(blk_max_size_offset(q, blk_rq_pos(rq)),
935 blk_queue_get_max_sectors(q, req_op(rq)));
936}
937
938static inline unsigned int blk_rq_count_bios(struct request *rq)
939{
940 unsigned int nr_bios = 0;
941 struct bio *bio;
942
943 __rq_for_each_bio(bio, rq)
944 nr_bios++;
945
946 return nr_bios;
947}
948
949/*
950 * Request issue related functions.
951 */
952extern struct request *blk_peek_request(struct request_queue *q);
953extern void blk_start_request(struct request *rq);
954extern struct request *blk_fetch_request(struct request_queue *q);
955
956/*
957 * Request completion related functions.
958 *
959 * blk_update_request() completes given number of bytes and updates
960 * the request without completing it.
961 *
962 * blk_end_request() and friends. __blk_end_request() must be called
963 * with the request queue spinlock acquired.
964 *
965 * Several drivers define their own end_request and call
966 * blk_end_request() for parts of the original function.
967 * This prevents code duplication in drivers.
968 */
969extern bool blk_update_request(struct request *rq, int error,
970 unsigned int nr_bytes);
971extern void blk_finish_request(struct request *rq, int error);
972extern bool blk_end_request(struct request *rq, int error,
973 unsigned int nr_bytes);
974extern void blk_end_request_all(struct request *rq, int error);
975extern bool blk_end_request_cur(struct request *rq, int error);
976extern bool blk_end_request_err(struct request *rq, int error);
977extern bool __blk_end_request(struct request *rq, int error,
978 unsigned int nr_bytes);
979extern void __blk_end_request_all(struct request *rq, int error);
980extern bool __blk_end_request_cur(struct request *rq, int error);
981extern bool __blk_end_request_err(struct request *rq, int error);
982
983extern void blk_complete_request(struct request *);
984extern void __blk_complete_request(struct request *);
985extern void blk_abort_request(struct request *);
986extern void blk_unprep_request(struct request *);
987
988/*
989 * Access functions for manipulating queue properties
990 */
991extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
992 spinlock_t *lock, int node_id);
993extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
994extern struct request_queue *blk_init_allocated_queue(struct request_queue *,
995 request_fn_proc *, spinlock_t *);
996extern void blk_cleanup_queue(struct request_queue *);
997extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
998extern void blk_queue_bounce_limit(struct request_queue *, u64);
999extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1000extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1001extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1002extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1003extern void blk_queue_max_discard_sectors(struct request_queue *q,
1004 unsigned int max_discard_sectors);
1005extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1006 unsigned int max_write_same_sectors);
1007extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
1008extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1009extern void blk_queue_alignment_offset(struct request_queue *q,
1010 unsigned int alignment);
1011extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1012extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1013extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1014extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1015extern void blk_set_default_limits(struct queue_limits *lim);
1016extern void blk_set_stacking_limits(struct queue_limits *lim);
1017extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1018 sector_t offset);
1019extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
1020 sector_t offset);
1021extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1022 sector_t offset);
1023extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
1024extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
1025extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1026extern int blk_queue_dma_drain(struct request_queue *q,
1027 dma_drain_needed_fn *dma_drain_needed,
1028 void *buf, unsigned int size);
1029extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
1030extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1031extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1032extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
1033extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
1034extern void blk_queue_dma_alignment(struct request_queue *, int);
1035extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1036extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1037extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1038extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1039extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1040extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1041extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
1042
1043extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1044extern void blk_dump_rq_flags(struct request *, char *);
1045extern long nr_blockdev_pages(void);
1046
1047bool __must_check blk_get_queue(struct request_queue *);
1048struct request_queue *blk_alloc_queue(gfp_t);
1049struct request_queue *blk_alloc_queue_node(gfp_t, int);
1050extern void blk_put_queue(struct request_queue *);
1051extern void blk_set_queue_dying(struct request_queue *);
1052
1053/*
1054 * block layer runtime pm functions
1055 */
1056#ifdef CONFIG_PM
1057extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1058extern int blk_pre_runtime_suspend(struct request_queue *q);
1059extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1060extern void blk_pre_runtime_resume(struct request_queue *q);
1061extern void blk_post_runtime_resume(struct request_queue *q, int err);
1062extern void blk_set_runtime_active(struct request_queue *q);
1063#else
1064static inline void blk_pm_runtime_init(struct request_queue *q,
1065 struct device *dev) {}
1066static inline int blk_pre_runtime_suspend(struct request_queue *q)
1067{
1068 return -ENOSYS;
1069}
1070static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1071static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1072static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1073extern inline void blk_set_runtime_active(struct request_queue *q) {}
1074#endif
1075
1076/*
1077 * blk_plug permits building a queue of related requests by holding the I/O
1078 * fragments for a short period. This allows merging of sequential requests
1079 * into single larger request. As the requests are moved from a per-task list to
1080 * the device's request_queue in a batch, this results in improved scalability
1081 * as the lock contention for request_queue lock is reduced.
1082 *
1083 * It is ok not to disable preemption when adding the request to the plug list
1084 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1085 * the plug list when the task sleeps by itself. For details, please see
1086 * schedule() where blk_schedule_flush_plug() is called.
1087 */
1088struct blk_plug {
1089 struct list_head list; /* requests */
1090 struct list_head mq_list; /* blk-mq requests */
1091 struct list_head cb_list; /* md requires an unplug callback */
1092};
1093#define BLK_MAX_REQUEST_COUNT 16
1094
1095struct blk_plug_cb;
1096typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1097struct blk_plug_cb {
1098 struct list_head list;
1099 blk_plug_cb_fn callback;
1100 void *data;
1101};
1102extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1103 void *data, int size);
1104extern void blk_start_plug(struct blk_plug *);
1105extern void blk_finish_plug(struct blk_plug *);
1106extern void blk_flush_plug_list(struct blk_plug *, bool);
1107
1108static inline void blk_flush_plug(struct task_struct *tsk)
1109{
1110 struct blk_plug *plug = tsk->plug;
1111
1112 if (plug)
1113 blk_flush_plug_list(plug, false);
1114}
1115
1116static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1117{
1118 struct blk_plug *plug = tsk->plug;
1119
1120 if (plug)
1121 blk_flush_plug_list(plug, true);
1122}
1123
1124static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1125{
1126 struct blk_plug *plug = tsk->plug;
1127
1128 return plug &&
1129 (!list_empty(&plug->list) ||
1130 !list_empty(&plug->mq_list) ||
1131 !list_empty(&plug->cb_list));
1132}
1133
1134/*
1135 * tag stuff
1136 */
1137extern int blk_queue_start_tag(struct request_queue *, struct request *);
1138extern struct request *blk_queue_find_tag(struct request_queue *, int);
1139extern void blk_queue_end_tag(struct request_queue *, struct request *);
1140extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1141extern void blk_queue_free_tags(struct request_queue *);
1142extern int blk_queue_resize_tags(struct request_queue *, int);
1143extern void blk_queue_invalidate_tags(struct request_queue *);
1144extern struct blk_queue_tag *blk_init_tags(int, int);
1145extern void blk_free_tags(struct blk_queue_tag *);
1146
1147static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1148 int tag)
1149{
1150 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1151 return NULL;
1152 return bqt->tag_index[tag];
1153}
1154
1155#define BLKDEV_DISCARD_SECURE 0x01 /* secure discard */
1156
1157extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1158extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1159 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1160extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1161 sector_t nr_sects, gfp_t gfp_mask, int op_flags,
1162 struct bio **biop);
1163extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1164 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1165extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1166 sector_t nr_sects, gfp_t gfp_mask, bool discard);
1167static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1168 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1169{
1170 return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9),
1171 nr_blocks << (sb->s_blocksize_bits - 9),
1172 gfp_mask, flags);
1173}
1174static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1175 sector_t nr_blocks, gfp_t gfp_mask)
1176{
1177 return blkdev_issue_zeroout(sb->s_bdev,
1178 block << (sb->s_blocksize_bits - 9),
1179 nr_blocks << (sb->s_blocksize_bits - 9),
1180 gfp_mask, true);
1181}
1182
1183extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm);
1184
1185enum blk_default_limits {
1186 BLK_MAX_SEGMENTS = 128,
1187 BLK_SAFE_MAX_SECTORS = 255,
1188 BLK_DEF_MAX_SECTORS = 2560,
1189 BLK_MAX_SEGMENT_SIZE = 65536,
1190 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1191};
1192
1193#define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1194
1195static inline unsigned long queue_bounce_pfn(struct request_queue *q)
1196{
1197 return q->limits.bounce_pfn;
1198}
1199
1200static inline unsigned long queue_segment_boundary(struct request_queue *q)
1201{
1202 return q->limits.seg_boundary_mask;
1203}
1204
1205static inline unsigned long queue_virt_boundary(struct request_queue *q)
1206{
1207 return q->limits.virt_boundary_mask;
1208}
1209
1210static inline unsigned int queue_max_sectors(struct request_queue *q)
1211{
1212 return q->limits.max_sectors;
1213}
1214
1215static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1216{
1217 return q->limits.max_hw_sectors;
1218}
1219
1220static inline unsigned short queue_max_segments(struct request_queue *q)
1221{
1222 return q->limits.max_segments;
1223}
1224
1225static inline unsigned int queue_max_segment_size(struct request_queue *q)
1226{
1227 return q->limits.max_segment_size;
1228}
1229
1230static inline unsigned short queue_logical_block_size(struct request_queue *q)
1231{
1232 int retval = 512;
1233
1234 if (q && q->limits.logical_block_size)
1235 retval = q->limits.logical_block_size;
1236
1237 return retval;
1238}
1239
1240static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1241{
1242 return queue_logical_block_size(bdev_get_queue(bdev));
1243}
1244
1245static inline unsigned int queue_physical_block_size(struct request_queue *q)
1246{
1247 return q->limits.physical_block_size;
1248}
1249
1250static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1251{
1252 return queue_physical_block_size(bdev_get_queue(bdev));
1253}
1254
1255static inline unsigned int queue_io_min(struct request_queue *q)
1256{
1257 return q->limits.io_min;
1258}
1259
1260static inline int bdev_io_min(struct block_device *bdev)
1261{
1262 return queue_io_min(bdev_get_queue(bdev));
1263}
1264
1265static inline unsigned int queue_io_opt(struct request_queue *q)
1266{
1267 return q->limits.io_opt;
1268}
1269
1270static inline int bdev_io_opt(struct block_device *bdev)
1271{
1272 return queue_io_opt(bdev_get_queue(bdev));
1273}
1274
1275static inline int queue_alignment_offset(struct request_queue *q)
1276{
1277 if (q->limits.misaligned)
1278 return -1;
1279
1280 return q->limits.alignment_offset;
1281}
1282
1283static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1284{
1285 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1286 unsigned int alignment = sector_div(sector, granularity >> 9) << 9;
1287
1288 return (granularity + lim->alignment_offset - alignment) % granularity;
1289}
1290
1291static inline int bdev_alignment_offset(struct block_device *bdev)
1292{
1293 struct request_queue *q = bdev_get_queue(bdev);
1294
1295 if (q->limits.misaligned)
1296 return -1;
1297
1298 if (bdev != bdev->bd_contains)
1299 return bdev->bd_part->alignment_offset;
1300
1301 return q->limits.alignment_offset;
1302}
1303
1304static inline int queue_discard_alignment(struct request_queue *q)
1305{
1306 if (q->limits.discard_misaligned)
1307 return -1;
1308
1309 return q->limits.discard_alignment;
1310}
1311
1312static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1313{
1314 unsigned int alignment, granularity, offset;
1315
1316 if (!lim->max_discard_sectors)
1317 return 0;
1318
1319 /* Why are these in bytes, not sectors? */
1320 alignment = lim->discard_alignment >> 9;
1321 granularity = lim->discard_granularity >> 9;
1322 if (!granularity)
1323 return 0;
1324
1325 /* Offset of the partition start in 'granularity' sectors */
1326 offset = sector_div(sector, granularity);
1327
1328 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1329 offset = (granularity + alignment - offset) % granularity;
1330
1331 /* Turn it back into bytes, gaah */
1332 return offset << 9;
1333}
1334
1335static inline int bdev_discard_alignment(struct block_device *bdev)
1336{
1337 struct request_queue *q = bdev_get_queue(bdev);
1338
1339 if (bdev != bdev->bd_contains)
1340 return bdev->bd_part->discard_alignment;
1341
1342 return q->limits.discard_alignment;
1343}
1344
1345static inline unsigned int queue_discard_zeroes_data(struct request_queue *q)
1346{
1347 if (q->limits.max_discard_sectors && q->limits.discard_zeroes_data == 1)
1348 return 1;
1349
1350 return 0;
1351}
1352
1353static inline unsigned int bdev_discard_zeroes_data(struct block_device *bdev)
1354{
1355 return queue_discard_zeroes_data(bdev_get_queue(bdev));
1356}
1357
1358static inline unsigned int bdev_write_same(struct block_device *bdev)
1359{
1360 struct request_queue *q = bdev_get_queue(bdev);
1361
1362 if (q)
1363 return q->limits.max_write_same_sectors;
1364
1365 return 0;
1366}
1367
1368static inline int queue_dma_alignment(struct request_queue *q)
1369{
1370 return q ? q->dma_alignment : 511;
1371}
1372
1373static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1374 unsigned int len)
1375{
1376 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1377 return !(addr & alignment) && !(len & alignment);
1378}
1379
1380/* assumes size > 256 */
1381static inline unsigned int blksize_bits(unsigned int size)
1382{
1383 unsigned int bits = 8;
1384 do {
1385 bits++;
1386 size >>= 1;
1387 } while (size > 256);
1388 return bits;
1389}
1390
1391static inline unsigned int block_size(struct block_device *bdev)
1392{
1393 return bdev->bd_block_size;
1394}
1395
1396static inline bool queue_flush_queueable(struct request_queue *q)
1397{
1398 return !test_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags);
1399}
1400
1401typedef struct {struct page *v;} Sector;
1402
1403unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1404
1405static inline void put_dev_sector(Sector p)
1406{
1407 put_page(p.v);
1408}
1409
1410static inline bool __bvec_gap_to_prev(struct request_queue *q,
1411 struct bio_vec *bprv, unsigned int offset)
1412{
1413 return offset ||
1414 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1415}
1416
1417/*
1418 * Check if adding a bio_vec after bprv with offset would create a gap in
1419 * the SG list. Most drivers don't care about this, but some do.
1420 */
1421static inline bool bvec_gap_to_prev(struct request_queue *q,
1422 struct bio_vec *bprv, unsigned int offset)
1423{
1424 if (!queue_virt_boundary(q))
1425 return false;
1426 return __bvec_gap_to_prev(q, bprv, offset);
1427}
1428
1429static inline bool bio_will_gap(struct request_queue *q, struct bio *prev,
1430 struct bio *next)
1431{
1432 if (bio_has_data(prev) && queue_virt_boundary(q)) {
1433 struct bio_vec pb, nb;
1434
1435 bio_get_last_bvec(prev, &pb);
1436 bio_get_first_bvec(next, &nb);
1437
1438 return __bvec_gap_to_prev(q, &pb, nb.bv_offset);
1439 }
1440
1441 return false;
1442}
1443
1444static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
1445{
1446 return bio_will_gap(req->q, req->biotail, bio);
1447}
1448
1449static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
1450{
1451 return bio_will_gap(req->q, bio, req->bio);
1452}
1453
1454struct work_struct;
1455int kblockd_schedule_work(struct work_struct *work);
1456int kblockd_schedule_delayed_work(struct delayed_work *dwork, unsigned long delay);
1457int kblockd_schedule_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1458
1459#ifdef CONFIG_BLK_CGROUP
1460/*
1461 * This should not be using sched_clock(). A real patch is in progress
1462 * to fix this up, until that is in place we need to disable preemption
1463 * around sched_clock() in this function and set_io_start_time_ns().
1464 */
1465static inline void set_start_time_ns(struct request *req)
1466{
1467 preempt_disable();
1468 req->start_time_ns = sched_clock();
1469 preempt_enable();
1470}
1471
1472static inline void set_io_start_time_ns(struct request *req)
1473{
1474 preempt_disable();
1475 req->io_start_time_ns = sched_clock();
1476 preempt_enable();
1477}
1478
1479static inline uint64_t rq_start_time_ns(struct request *req)
1480{
1481 return req->start_time_ns;
1482}
1483
1484static inline uint64_t rq_io_start_time_ns(struct request *req)
1485{
1486 return req->io_start_time_ns;
1487}
1488#else
1489static inline void set_start_time_ns(struct request *req) {}
1490static inline void set_io_start_time_ns(struct request *req) {}
1491static inline uint64_t rq_start_time_ns(struct request *req)
1492{
1493 return 0;
1494}
1495static inline uint64_t rq_io_start_time_ns(struct request *req)
1496{
1497 return 0;
1498}
1499#endif
1500
1501#define MODULE_ALIAS_BLOCKDEV(major,minor) \
1502 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1503#define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1504 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1505
1506#if defined(CONFIG_BLK_DEV_INTEGRITY)
1507
1508enum blk_integrity_flags {
1509 BLK_INTEGRITY_VERIFY = 1 << 0,
1510 BLK_INTEGRITY_GENERATE = 1 << 1,
1511 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2,
1512 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3,
1513};
1514
1515struct blk_integrity_iter {
1516 void *prot_buf;
1517 void *data_buf;
1518 sector_t seed;
1519 unsigned int data_size;
1520 unsigned short interval;
1521 const char *disk_name;
1522};
1523
1524typedef int (integrity_processing_fn) (struct blk_integrity_iter *);
1525
1526struct blk_integrity_profile {
1527 integrity_processing_fn *generate_fn;
1528 integrity_processing_fn *verify_fn;
1529 const char *name;
1530};
1531
1532extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1533extern void blk_integrity_unregister(struct gendisk *);
1534extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1535extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1536 struct scatterlist *);
1537extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1538extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1539 struct request *);
1540extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1541 struct bio *);
1542
1543static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1544{
1545 struct blk_integrity *bi = &disk->queue->integrity;
1546
1547 if (!bi->profile)
1548 return NULL;
1549
1550 return bi;
1551}
1552
1553static inline
1554struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1555{
1556 return blk_get_integrity(bdev->bd_disk);
1557}
1558
1559static inline bool blk_integrity_rq(struct request *rq)
1560{
1561 return rq->cmd_flags & REQ_INTEGRITY;
1562}
1563
1564static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1565 unsigned int segs)
1566{
1567 q->limits.max_integrity_segments = segs;
1568}
1569
1570static inline unsigned short
1571queue_max_integrity_segments(struct request_queue *q)
1572{
1573 return q->limits.max_integrity_segments;
1574}
1575
1576static inline bool integrity_req_gap_back_merge(struct request *req,
1577 struct bio *next)
1578{
1579 struct bio_integrity_payload *bip = bio_integrity(req->bio);
1580 struct bio_integrity_payload *bip_next = bio_integrity(next);
1581
1582 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1583 bip_next->bip_vec[0].bv_offset);
1584}
1585
1586static inline bool integrity_req_gap_front_merge(struct request *req,
1587 struct bio *bio)
1588{
1589 struct bio_integrity_payload *bip = bio_integrity(bio);
1590 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
1591
1592 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1593 bip_next->bip_vec[0].bv_offset);
1594}
1595
1596#else /* CONFIG_BLK_DEV_INTEGRITY */
1597
1598struct bio;
1599struct block_device;
1600struct gendisk;
1601struct blk_integrity;
1602
1603static inline int blk_integrity_rq(struct request *rq)
1604{
1605 return 0;
1606}
1607static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1608 struct bio *b)
1609{
1610 return 0;
1611}
1612static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1613 struct bio *b,
1614 struct scatterlist *s)
1615{
1616 return 0;
1617}
1618static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1619{
1620 return NULL;
1621}
1622static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1623{
1624 return NULL;
1625}
1626static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1627{
1628 return 0;
1629}
1630static inline void blk_integrity_register(struct gendisk *d,
1631 struct blk_integrity *b)
1632{
1633}
1634static inline void blk_integrity_unregister(struct gendisk *d)
1635{
1636}
1637static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1638 unsigned int segs)
1639{
1640}
1641static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1642{
1643 return 0;
1644}
1645static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1646 struct request *r1,
1647 struct request *r2)
1648{
1649 return true;
1650}
1651static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1652 struct request *r,
1653 struct bio *b)
1654{
1655 return true;
1656}
1657
1658static inline bool integrity_req_gap_back_merge(struct request *req,
1659 struct bio *next)
1660{
1661 return false;
1662}
1663static inline bool integrity_req_gap_front_merge(struct request *req,
1664 struct bio *bio)
1665{
1666 return false;
1667}
1668
1669#endif /* CONFIG_BLK_DEV_INTEGRITY */
1670
1671/**
1672 * struct blk_dax_ctl - control and output parameters for ->direct_access
1673 * @sector: (input) offset relative to a block_device
1674 * @addr: (output) kernel virtual address for @sector populated by driver
1675 * @pfn: (output) page frame number for @addr populated by driver
1676 * @size: (input) number of bytes requested
1677 */
1678struct blk_dax_ctl {
1679 sector_t sector;
1680 void __pmem *addr;
1681 long size;
1682 pfn_t pfn;
1683};
1684
1685struct block_device_operations {
1686 int (*open) (struct block_device *, fmode_t);
1687 void (*release) (struct gendisk *, fmode_t);
1688 int (*rw_page)(struct block_device *, sector_t, struct page *, int rw);
1689 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1690 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1691 long (*direct_access)(struct block_device *, sector_t, void __pmem **,
1692 pfn_t *, long);
1693 unsigned int (*check_events) (struct gendisk *disk,
1694 unsigned int clearing);
1695 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1696 int (*media_changed) (struct gendisk *);
1697 void (*unlock_native_capacity) (struct gendisk *);
1698 int (*revalidate_disk) (struct gendisk *);
1699 int (*getgeo)(struct block_device *, struct hd_geometry *);
1700 /* this callback is with swap_lock and sometimes page table lock held */
1701 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1702 struct module *owner;
1703 const struct pr_ops *pr_ops;
1704};
1705
1706extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1707 unsigned long);
1708extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1709extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1710 struct writeback_control *);
1711extern long bdev_direct_access(struct block_device *, struct blk_dax_ctl *);
1712extern int bdev_dax_supported(struct super_block *, int);
1713extern bool bdev_dax_capable(struct block_device *);
1714#else /* CONFIG_BLOCK */
1715
1716struct block_device;
1717
1718/*
1719 * stubs for when the block layer is configured out
1720 */
1721#define buffer_heads_over_limit 0
1722
1723static inline long nr_blockdev_pages(void)
1724{
1725 return 0;
1726}
1727
1728struct blk_plug {
1729};
1730
1731static inline void blk_start_plug(struct blk_plug *plug)
1732{
1733}
1734
1735static inline void blk_finish_plug(struct blk_plug *plug)
1736{
1737}
1738
1739static inline void blk_flush_plug(struct task_struct *task)
1740{
1741}
1742
1743static inline void blk_schedule_flush_plug(struct task_struct *task)
1744{
1745}
1746
1747
1748static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1749{
1750 return false;
1751}
1752
1753static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
1754 sector_t *error_sector)
1755{
1756 return 0;
1757}
1758
1759#endif /* CONFIG_BLOCK */
1760
1761#endif