1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_BLKDEV_H
3 #define _LINUX_BLKDEV_H
4
5 #include <linux/sched.h>
6 #include <linux/sched/clock.h>
7
8 #ifdef CONFIG_BLOCK
9
10 #include <linux/major.h>
11 #include <linux/genhd.h>
12 #include <linux/list.h>
13 #include <linux/llist.h>
14 #include <linux/timer.h>
15 #include <linux/workqueue.h>
16 #include <linux/pagemap.h>
17 #include <linux/backing-dev-defs.h>
18 #include <linux/wait.h>
19 #include <linux/mempool.h>
20 #include <linux/pfn.h>
21 #include <linux/bio.h>
22 #include <linux/stringify.h>
23 #include <linux/gfp.h>
24 #include <linux/bsg.h>
25 #include <linux/smp.h>
26 #include <linux/rcupdate.h>
27 #include <linux/percpu-refcount.h>
28 #include <linux/scatterlist.h>
29 #include <linux/blkzoned.h>
30
31 struct module;
32 struct scsi_ioctl_command;
33
34 struct request_queue;
35 struct elevator_queue;
36 struct blk_trace;
37 struct request;
38 struct sg_io_hdr;
39 struct bsg_job;
40 struct blkcg_gq;
41 struct blk_flush_queue;
42 struct pr_ops;
43 struct rq_wb;
44 struct blk_queue_stats;
45 struct blk_stat_callback;
46
47 #define BLKDEV_MIN_RQ 4
48 #define BLKDEV_MAX_RQ 128 /* Default maximum */
49
50 /* Must be consisitent with blk_mq_poll_stats_bkt() */
51 #define BLK_MQ_POLL_STATS_BKTS 16
52
53 /*
54 * Maximum number of blkcg policies allowed to be registered concurrently.
55 * Defined here to simplify include dependency.
56 */
57 #define BLKCG_MAX_POLS 3
58
59 typedef void (rq_end_io_fn)(struct request *, blk_status_t);
60
61 #define BLK_RL_SYNCFULL (1U << 0)
62 #define BLK_RL_ASYNCFULL (1U << 1)
63
64 struct request_list {
65 struct request_queue *q; /* the queue this rl belongs to */
66 #ifdef CONFIG_BLK_CGROUP
67 struct blkcg_gq *blkg; /* blkg this request pool belongs to */
68 #endif
69 /*
70 * count[], starved[], and wait[] are indexed by
71 * BLK_RW_SYNC/BLK_RW_ASYNC
72 */
73 int count[2];
74 int starved[2];
75 mempool_t *rq_pool;
76 wait_queue_head_t wait[2];
77 unsigned int flags;
78 };
79
80 /*
81 * request flags */
82 typedef __u32 __bitwise req_flags_t;
83
84 /* elevator knows about this request */
85 #define RQF_SORTED ((__force req_flags_t)(1 << 0))
86 /* drive already may have started this one */
87 #define RQF_STARTED ((__force req_flags_t)(1 << 1))
88 /* uses tagged queueing */
89 #define RQF_QUEUED ((__force req_flags_t)(1 << 2))
90 /* may not be passed by ioscheduler */
91 #define RQF_SOFTBARRIER ((__force req_flags_t)(1 << 3))
92 /* request for flush sequence */
93 #define RQF_FLUSH_SEQ ((__force req_flags_t)(1 << 4))
94 /* merge of different types, fail separately */
95 #define RQF_MIXED_MERGE ((__force req_flags_t)(1 << 5))
96 /* track inflight for MQ */
97 #define RQF_MQ_INFLIGHT ((__force req_flags_t)(1 << 6))
98 /* don't call prep for this one */
99 #define RQF_DONTPREP ((__force req_flags_t)(1 << 7))
100 /* set for "ide_preempt" requests and also for requests for which the SCSI
101 "quiesce" state must be ignored. */
102 #define RQF_PREEMPT ((__force req_flags_t)(1 << 8))
103 /* contains copies of user pages */
104 #define RQF_COPY_USER ((__force req_flags_t)(1 << 9))
105 /* vaguely specified driver internal error. Ignored by the block layer */
106 #define RQF_FAILED ((__force req_flags_t)(1 << 10))
107 /* don't warn about errors */
108 #define RQF_QUIET ((__force req_flags_t)(1 << 11))
109 /* elevator private data attached */
110 #define RQF_ELVPRIV ((__force req_flags_t)(1 << 12))
111 /* account I/O stat */
112 #define RQF_IO_STAT ((__force req_flags_t)(1 << 13))
113 /* request came from our alloc pool */
114 #define RQF_ALLOCED ((__force req_flags_t)(1 << 14))
115 /* runtime pm request */
116 #define RQF_PM ((__force req_flags_t)(1 << 15))
117 /* on IO scheduler merge hash */
118 #define RQF_HASHED ((__force req_flags_t)(1 << 16))
119 /* IO stats tracking on */
120 #define RQF_STATS ((__force req_flags_t)(1 << 17))
121 /* Look at ->special_vec for the actual data payload instead of the
122 bio chain. */
123 #define RQF_SPECIAL_PAYLOAD ((__force req_flags_t)(1 << 18))
124
125 /* flags that prevent us from merging requests: */
126 #define RQF_NOMERGE_FLAGS \
127 (RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD)
128
129 /*
130 * Try to put the fields that are referenced together in the same cacheline.
131 *
132 * If you modify this structure, make sure to update blk_rq_init() and
133 * especially blk_mq_rq_ctx_init() to take care of the added fields.
134 */
135 struct request {
136 struct list_head queuelist;
137 union {
138 struct __call_single_data csd;
139 u64 fifo_time;
140 };
141
142 struct request_queue *q;
143 struct blk_mq_ctx *mq_ctx;
144
145 int cpu;
146 unsigned int cmd_flags; /* op and common flags */
147 req_flags_t rq_flags;
148
149 int internal_tag;
150
151 unsigned long atomic_flags;
152
153 /* the following two fields are internal, NEVER access directly */
154 unsigned int __data_len; /* total data len */
155 int tag;
156 sector_t __sector; /* sector cursor */
157
158 struct bio *bio;
159 struct bio *biotail;
160
161 /*
162 * The hash is used inside the scheduler, and killed once the
163 * request reaches the dispatch list. The ipi_list is only used
164 * to queue the request for softirq completion, which is long
165 * after the request has been unhashed (and even removed from
166 * the dispatch list).
167 */
168 union {
169 struct hlist_node hash; /* merge hash */
170 struct list_head ipi_list;
171 };
172
173 /*
174 * The rb_node is only used inside the io scheduler, requests
175 * are pruned when moved to the dispatch queue. So let the
176 * completion_data share space with the rb_node.
177 */
178 union {
179 struct rb_node rb_node; /* sort/lookup */
180 struct bio_vec special_vec;
181 void *completion_data;
182 int error_count; /* for legacy drivers, don't use */
183 };
184
185 /*
186 * Three pointers are available for the IO schedulers, if they need
187 * more they have to dynamically allocate it. Flush requests are
188 * never put on the IO scheduler. So let the flush fields share
189 * space with the elevator data.
190 */
191 union {
192 struct {
193 struct io_cq *icq;
194 void *priv[2];
195 } elv;
196
197 struct {
198 unsigned int seq;
199 struct list_head list;
200 rq_end_io_fn *saved_end_io;
201 } flush;
202 };
203
204 struct gendisk *rq_disk;
205 struct hd_struct *part;
206 unsigned long start_time;
207 struct blk_issue_stat issue_stat;
208 #ifdef CONFIG_BLK_CGROUP
209 struct request_list *rl; /* rl this rq is alloced from */
210 unsigned long long start_time_ns;
211 unsigned long long io_start_time_ns; /* when passed to hardware */
212 #endif
213 /* Number of scatter-gather DMA addr+len pairs after
214 * physical address coalescing is performed.
215 */
216 unsigned short nr_phys_segments;
217 #if defined(CONFIG_BLK_DEV_INTEGRITY)
218 unsigned short nr_integrity_segments;
219 #endif
220
221 unsigned short ioprio;
222
223 unsigned int timeout;
224
225 void *special; /* opaque pointer available for LLD use */
226
227 unsigned int extra_len; /* length of alignment and padding */
228
229 unsigned short write_hint;
230
231 unsigned long deadline;
232 struct list_head timeout_list;
233
234 /*
235 * completion callback.
236 */
237 rq_end_io_fn *end_io;
238 void *end_io_data;
239
240 /* for bidi */
241 struct request *next_rq;
242 };
243
blk_op_is_scsi(unsigned int op)244 static inline bool blk_op_is_scsi(unsigned int op)
245 {
246 return op == REQ_OP_SCSI_IN || op == REQ_OP_SCSI_OUT;
247 }
248
blk_op_is_private(unsigned int op)249 static inline bool blk_op_is_private(unsigned int op)
250 {
251 return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT;
252 }
253
blk_rq_is_scsi(struct request * rq)254 static inline bool blk_rq_is_scsi(struct request *rq)
255 {
256 return blk_op_is_scsi(req_op(rq));
257 }
258
blk_rq_is_private(struct request * rq)259 static inline bool blk_rq_is_private(struct request *rq)
260 {
261 return blk_op_is_private(req_op(rq));
262 }
263
blk_rq_is_passthrough(struct request * rq)264 static inline bool blk_rq_is_passthrough(struct request *rq)
265 {
266 return blk_rq_is_scsi(rq) || blk_rq_is_private(rq);
267 }
268
bio_is_passthrough(struct bio * bio)269 static inline bool bio_is_passthrough(struct bio *bio)
270 {
271 unsigned op = bio_op(bio);
272
273 return blk_op_is_scsi(op) || blk_op_is_private(op);
274 }
275
req_get_ioprio(struct request * req)276 static inline unsigned short req_get_ioprio(struct request *req)
277 {
278 return req->ioprio;
279 }
280
281 #include <linux/elevator.h>
282
283 struct blk_queue_ctx;
284
285 typedef void (request_fn_proc) (struct request_queue *q);
286 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio);
287 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
288 typedef void (unprep_rq_fn) (struct request_queue *, struct request *);
289
290 struct bio_vec;
291 typedef void (softirq_done_fn)(struct request *);
292 typedef int (dma_drain_needed_fn)(struct request *);
293 typedef int (lld_busy_fn) (struct request_queue *q);
294 typedef int (bsg_job_fn) (struct bsg_job *);
295 typedef int (init_rq_fn)(struct request_queue *, struct request *, gfp_t);
296 typedef void (exit_rq_fn)(struct request_queue *, struct request *);
297
298 enum blk_eh_timer_return {
299 BLK_EH_NOT_HANDLED,
300 BLK_EH_HANDLED,
301 BLK_EH_RESET_TIMER,
302 };
303
304 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
305
306 enum blk_queue_state {
307 Queue_down,
308 Queue_up,
309 };
310
311 struct blk_queue_tag {
312 struct request **tag_index; /* map of busy tags */
313 unsigned long *tag_map; /* bit map of free/busy tags */
314 int max_depth; /* what we will send to device */
315 int real_max_depth; /* what the array can hold */
316 atomic_t refcnt; /* map can be shared */
317 int alloc_policy; /* tag allocation policy */
318 int next_tag; /* next tag */
319 };
320 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
321 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
322
323 #define BLK_SCSI_MAX_CMDS (256)
324 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
325
326 /*
327 * Zoned block device models (zoned limit).
328 */
329 enum blk_zoned_model {
330 BLK_ZONED_NONE, /* Regular block device */
331 BLK_ZONED_HA, /* Host-aware zoned block device */
332 BLK_ZONED_HM, /* Host-managed zoned block device */
333 };
334
335 struct queue_limits {
336 unsigned long bounce_pfn;
337 unsigned long seg_boundary_mask;
338 unsigned long virt_boundary_mask;
339
340 unsigned int max_hw_sectors;
341 unsigned int max_dev_sectors;
342 unsigned int chunk_sectors;
343 unsigned int max_sectors;
344 unsigned int max_segment_size;
345 unsigned int physical_block_size;
346 unsigned int logical_block_size;
347 unsigned int alignment_offset;
348 unsigned int io_min;
349 unsigned int io_opt;
350 unsigned int max_discard_sectors;
351 unsigned int max_hw_discard_sectors;
352 unsigned int max_write_same_sectors;
353 unsigned int max_write_zeroes_sectors;
354 unsigned int discard_granularity;
355 unsigned int discard_alignment;
356
357 unsigned short max_segments;
358 unsigned short max_integrity_segments;
359 unsigned short max_discard_segments;
360
361 unsigned char misaligned;
362 unsigned char discard_misaligned;
363 unsigned char cluster;
364 unsigned char raid_partial_stripes_expensive;
365 enum blk_zoned_model zoned;
366 };
367
368 #ifdef CONFIG_BLK_DEV_ZONED
369
370 struct blk_zone_report_hdr {
371 unsigned int nr_zones;
372 u8 padding[60];
373 };
374
375 extern int blkdev_report_zones(struct block_device *bdev,
376 sector_t sector, struct blk_zone *zones,
377 unsigned int *nr_zones, gfp_t gfp_mask);
378 extern int blkdev_reset_zones(struct block_device *bdev, sector_t sectors,
379 sector_t nr_sectors, gfp_t gfp_mask);
380
381 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
382 unsigned int cmd, unsigned long arg);
383 extern int blkdev_reset_zones_ioctl(struct block_device *bdev, fmode_t mode,
384 unsigned int cmd, unsigned long arg);
385
386 #else /* CONFIG_BLK_DEV_ZONED */
387
blkdev_report_zones_ioctl(struct block_device * bdev,fmode_t mode,unsigned int cmd,unsigned long arg)388 static inline int blkdev_report_zones_ioctl(struct block_device *bdev,
389 fmode_t mode, unsigned int cmd,
390 unsigned long arg)
391 {
392 return -ENOTTY;
393 }
394
blkdev_reset_zones_ioctl(struct block_device * bdev,fmode_t mode,unsigned int cmd,unsigned long arg)395 static inline int blkdev_reset_zones_ioctl(struct block_device *bdev,
396 fmode_t mode, unsigned int cmd,
397 unsigned long arg)
398 {
399 return -ENOTTY;
400 }
401
402 #endif /* CONFIG_BLK_DEV_ZONED */
403
404 struct request_queue {
405 /*
406 * Together with queue_head for cacheline sharing
407 */
408 struct list_head queue_head;
409 struct request *last_merge;
410 struct elevator_queue *elevator;
411 int nr_rqs[2]; /* # allocated [a]sync rqs */
412 int nr_rqs_elvpriv; /* # allocated rqs w/ elvpriv */
413
414 atomic_t shared_hctx_restart;
415
416 struct blk_queue_stats *stats;
417 struct rq_wb *rq_wb;
418
419 /*
420 * If blkcg is not used, @q->root_rl serves all requests. If blkcg
421 * is used, root blkg allocates from @q->root_rl and all other
422 * blkgs from their own blkg->rl. Which one to use should be
423 * determined using bio_request_list().
424 */
425 struct request_list root_rl;
426
427 request_fn_proc *request_fn;
428 make_request_fn *make_request_fn;
429 prep_rq_fn *prep_rq_fn;
430 unprep_rq_fn *unprep_rq_fn;
431 softirq_done_fn *softirq_done_fn;
432 rq_timed_out_fn *rq_timed_out_fn;
433 dma_drain_needed_fn *dma_drain_needed;
434 lld_busy_fn *lld_busy_fn;
435 /* Called just after a request is allocated */
436 init_rq_fn *init_rq_fn;
437 /* Called just before a request is freed */
438 exit_rq_fn *exit_rq_fn;
439 /* Called from inside blk_get_request() */
440 void (*initialize_rq_fn)(struct request *rq);
441
442 const struct blk_mq_ops *mq_ops;
443
444 unsigned int *mq_map;
445
446 /* sw queues */
447 struct blk_mq_ctx __percpu *queue_ctx;
448 unsigned int nr_queues;
449
450 unsigned int queue_depth;
451
452 /* hw dispatch queues */
453 struct blk_mq_hw_ctx **queue_hw_ctx;
454 unsigned int nr_hw_queues;
455
456 /*
457 * Dispatch queue sorting
458 */
459 sector_t end_sector;
460 struct request *boundary_rq;
461
462 /*
463 * Delayed queue handling
464 */
465 struct delayed_work delay_work;
466
467 struct backing_dev_info *backing_dev_info;
468
469 /*
470 * The queue owner gets to use this for whatever they like.
471 * ll_rw_blk doesn't touch it.
472 */
473 void *queuedata;
474
475 /*
476 * various queue flags, see QUEUE_* below
477 */
478 unsigned long queue_flags;
479
480 /*
481 * ida allocated id for this queue. Used to index queues from
482 * ioctx.
483 */
484 int id;
485
486 /*
487 * queue needs bounce pages for pages above this limit
488 */
489 gfp_t bounce_gfp;
490
491 /*
492 * protects queue structures from reentrancy. ->__queue_lock should
493 * _never_ be used directly, it is queue private. always use
494 * ->queue_lock.
495 */
496 spinlock_t __queue_lock;
497 spinlock_t *queue_lock;
498
499 /*
500 * queue kobject
501 */
502 struct kobject kobj;
503
504 /*
505 * mq queue kobject
506 */
507 struct kobject mq_kobj;
508
509 #ifdef CONFIG_BLK_DEV_INTEGRITY
510 struct blk_integrity integrity;
511 #endif /* CONFIG_BLK_DEV_INTEGRITY */
512
513 #ifdef CONFIG_PM
514 struct device *dev;
515 int rpm_status;
516 unsigned int nr_pending;
517 #endif
518
519 /*
520 * queue settings
521 */
522 unsigned long nr_requests; /* Max # of requests */
523 unsigned int nr_congestion_on;
524 unsigned int nr_congestion_off;
525 unsigned int nr_batching;
526
527 unsigned int dma_drain_size;
528 void *dma_drain_buffer;
529 unsigned int dma_pad_mask;
530 unsigned int dma_alignment;
531
532 struct blk_queue_tag *queue_tags;
533 struct list_head tag_busy_list;
534
535 unsigned int nr_sorted;
536 unsigned int in_flight[2];
537
538 /*
539 * Number of active block driver functions for which blk_drain_queue()
540 * must wait. Must be incremented around functions that unlock the
541 * queue_lock internally, e.g. scsi_request_fn().
542 */
543 unsigned int request_fn_active;
544
545 unsigned int rq_timeout;
546 int poll_nsec;
547
548 struct blk_stat_callback *poll_cb;
549 struct blk_rq_stat poll_stat[BLK_MQ_POLL_STATS_BKTS];
550
551 struct timer_list timeout;
552 struct work_struct timeout_work;
553 struct list_head timeout_list;
554
555 struct list_head icq_list;
556 #ifdef CONFIG_BLK_CGROUP
557 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
558 struct blkcg_gq *root_blkg;
559 struct list_head blkg_list;
560 #endif
561
562 struct queue_limits limits;
563
564 /*
565 * sg stuff
566 */
567 unsigned int sg_timeout;
568 unsigned int sg_reserved_size;
569 int node;
570 #ifdef CONFIG_BLK_DEV_IO_TRACE
571 struct blk_trace *blk_trace;
572 struct mutex blk_trace_mutex;
573 #endif
574 /*
575 * for flush operations
576 */
577 struct blk_flush_queue *fq;
578
579 struct list_head requeue_list;
580 spinlock_t requeue_lock;
581 struct delayed_work requeue_work;
582
583 struct mutex sysfs_lock;
584
585 int bypass_depth;
586 atomic_t mq_freeze_depth;
587
588 #if defined(CONFIG_BLK_DEV_BSG)
589 bsg_job_fn *bsg_job_fn;
590 struct bsg_class_device bsg_dev;
591 #endif
592
593 #ifdef CONFIG_BLK_DEV_THROTTLING
594 /* Throttle data */
595 struct throtl_data *td;
596 #endif
597 struct rcu_head rcu_head;
598 wait_queue_head_t mq_freeze_wq;
599 struct percpu_ref q_usage_counter;
600 struct list_head all_q_node;
601
602 struct blk_mq_tag_set *tag_set;
603 struct list_head tag_set_list;
604 struct bio_set *bio_split;
605
606 #ifdef CONFIG_BLK_DEBUG_FS
607 struct dentry *debugfs_dir;
608 struct dentry *sched_debugfs_dir;
609 #endif
610
611 bool mq_sysfs_init_done;
612
613 size_t cmd_size;
614 void *rq_alloc_data;
615
616 struct work_struct release_work;
617
618 #define BLK_MAX_WRITE_HINTS 5
619 u64 write_hints[BLK_MAX_WRITE_HINTS];
620 };
621
622 #define QUEUE_FLAG_QUEUED 0 /* uses generic tag queueing */
623 #define QUEUE_FLAG_STOPPED 1 /* queue is stopped */
624 #define QUEUE_FLAG_DYING 2 /* queue being torn down */
625 #define QUEUE_FLAG_BYPASS 3 /* act as dumb FIFO queue */
626 #define QUEUE_FLAG_BIDI 4 /* queue supports bidi requests */
627 #define QUEUE_FLAG_NOMERGES 5 /* disable merge attempts */
628 #define QUEUE_FLAG_SAME_COMP 6 /* complete on same CPU-group */
629 #define QUEUE_FLAG_FAIL_IO 7 /* fake timeout */
630 #define QUEUE_FLAG_STACKABLE 8 /* supports request stacking */
631 #define QUEUE_FLAG_NONROT 9 /* non-rotational device (SSD) */
632 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
633 #define QUEUE_FLAG_IO_STAT 10 /* do IO stats */
634 #define QUEUE_FLAG_DISCARD 11 /* supports DISCARD */
635 #define QUEUE_FLAG_NOXMERGES 12 /* No extended merges */
636 #define QUEUE_FLAG_ADD_RANDOM 13 /* Contributes to random pool */
637 #define QUEUE_FLAG_SECERASE 14 /* supports secure erase */
638 #define QUEUE_FLAG_SAME_FORCE 15 /* force complete on same CPU */
639 #define QUEUE_FLAG_DEAD 16 /* queue tear-down finished */
640 #define QUEUE_FLAG_INIT_DONE 17 /* queue is initialized */
641 #define QUEUE_FLAG_NO_SG_MERGE 18 /* don't attempt to merge SG segments*/
642 #define QUEUE_FLAG_POLL 19 /* IO polling enabled if set */
643 #define QUEUE_FLAG_WC 20 /* Write back caching */
644 #define QUEUE_FLAG_FUA 21 /* device supports FUA writes */
645 #define QUEUE_FLAG_FLUSH_NQ 22 /* flush not queueuable */
646 #define QUEUE_FLAG_DAX 23 /* device supports DAX */
647 #define QUEUE_FLAG_STATS 24 /* track rq completion times */
648 #define QUEUE_FLAG_POLL_STATS 25 /* collecting stats for hybrid polling */
649 #define QUEUE_FLAG_REGISTERED 26 /* queue has been registered to a disk */
650 #define QUEUE_FLAG_SCSI_PASSTHROUGH 27 /* queue supports SCSI commands */
651 #define QUEUE_FLAG_QUIESCED 28 /* queue has been quiesced */
652
653 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
654 (1 << QUEUE_FLAG_STACKABLE) | \
655 (1 << QUEUE_FLAG_SAME_COMP) | \
656 (1 << QUEUE_FLAG_ADD_RANDOM))
657
658 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
659 (1 << QUEUE_FLAG_STACKABLE) | \
660 (1 << QUEUE_FLAG_SAME_COMP) | \
661 (1 << QUEUE_FLAG_POLL))
662
663 /*
664 * @q->queue_lock is set while a queue is being initialized. Since we know
665 * that no other threads access the queue object before @q->queue_lock has
666 * been set, it is safe to manipulate queue flags without holding the
667 * queue_lock if @q->queue_lock == NULL. See also blk_alloc_queue_node() and
668 * blk_init_allocated_queue().
669 */
queue_lockdep_assert_held(struct request_queue * q)670 static inline void queue_lockdep_assert_held(struct request_queue *q)
671 {
672 if (q->queue_lock)
673 lockdep_assert_held(q->queue_lock);
674 }
675
queue_flag_set_unlocked(unsigned int flag,struct request_queue * q)676 static inline void queue_flag_set_unlocked(unsigned int flag,
677 struct request_queue *q)
678 {
679 __set_bit(flag, &q->queue_flags);
680 }
681
queue_flag_test_and_clear(unsigned int flag,struct request_queue * q)682 static inline int queue_flag_test_and_clear(unsigned int flag,
683 struct request_queue *q)
684 {
685 queue_lockdep_assert_held(q);
686
687 if (test_bit(flag, &q->queue_flags)) {
688 __clear_bit(flag, &q->queue_flags);
689 return 1;
690 }
691
692 return 0;
693 }
694
queue_flag_test_and_set(unsigned int flag,struct request_queue * q)695 static inline int queue_flag_test_and_set(unsigned int flag,
696 struct request_queue *q)
697 {
698 queue_lockdep_assert_held(q);
699
700 if (!test_bit(flag, &q->queue_flags)) {
701 __set_bit(flag, &q->queue_flags);
702 return 0;
703 }
704
705 return 1;
706 }
707
queue_flag_set(unsigned int flag,struct request_queue * q)708 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
709 {
710 queue_lockdep_assert_held(q);
711 __set_bit(flag, &q->queue_flags);
712 }
713
queue_flag_clear_unlocked(unsigned int flag,struct request_queue * q)714 static inline void queue_flag_clear_unlocked(unsigned int flag,
715 struct request_queue *q)
716 {
717 __clear_bit(flag, &q->queue_flags);
718 }
719
queue_in_flight(struct request_queue * q)720 static inline int queue_in_flight(struct request_queue *q)
721 {
722 return q->in_flight[0] + q->in_flight[1];
723 }
724
queue_flag_clear(unsigned int flag,struct request_queue * q)725 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
726 {
727 queue_lockdep_assert_held(q);
728 __clear_bit(flag, &q->queue_flags);
729 }
730
731 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
732 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
733 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
734 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
735 #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
736 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
737 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
738 #define blk_queue_noxmerges(q) \
739 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
740 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
741 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
742 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
743 #define blk_queue_stackable(q) \
744 test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
745 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
746 #define blk_queue_secure_erase(q) \
747 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
748 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
749 #define blk_queue_scsi_passthrough(q) \
750 test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags)
751
752 #define blk_noretry_request(rq) \
753 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
754 REQ_FAILFAST_DRIVER))
755 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
756
blk_account_rq(struct request * rq)757 static inline bool blk_account_rq(struct request *rq)
758 {
759 return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq);
760 }
761
762 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
763 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
764 /* rq->queuelist of dequeued request must be list_empty() */
765 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
766
767 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
768
769 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ)
770
771 /*
772 * Driver can handle struct request, if it either has an old style
773 * request_fn defined, or is blk-mq based.
774 */
queue_is_rq_based(struct request_queue * q)775 static inline bool queue_is_rq_based(struct request_queue *q)
776 {
777 return q->request_fn || q->mq_ops;
778 }
779
blk_queue_cluster(struct request_queue * q)780 static inline unsigned int blk_queue_cluster(struct request_queue *q)
781 {
782 return q->limits.cluster;
783 }
784
785 static inline enum blk_zoned_model
blk_queue_zoned_model(struct request_queue * q)786 blk_queue_zoned_model(struct request_queue *q)
787 {
788 return q->limits.zoned;
789 }
790
blk_queue_is_zoned(struct request_queue * q)791 static inline bool blk_queue_is_zoned(struct request_queue *q)
792 {
793 switch (blk_queue_zoned_model(q)) {
794 case BLK_ZONED_HA:
795 case BLK_ZONED_HM:
796 return true;
797 default:
798 return false;
799 }
800 }
801
blk_queue_zone_sectors(struct request_queue * q)802 static inline unsigned int blk_queue_zone_sectors(struct request_queue *q)
803 {
804 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
805 }
806
rq_is_sync(struct request * rq)807 static inline bool rq_is_sync(struct request *rq)
808 {
809 return op_is_sync(rq->cmd_flags);
810 }
811
blk_rl_full(struct request_list * rl,bool sync)812 static inline bool blk_rl_full(struct request_list *rl, bool sync)
813 {
814 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
815
816 return rl->flags & flag;
817 }
818
blk_set_rl_full(struct request_list * rl,bool sync)819 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
820 {
821 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
822
823 rl->flags |= flag;
824 }
825
blk_clear_rl_full(struct request_list * rl,bool sync)826 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
827 {
828 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
829
830 rl->flags &= ~flag;
831 }
832
rq_mergeable(struct request * rq)833 static inline bool rq_mergeable(struct request *rq)
834 {
835 if (blk_rq_is_passthrough(rq))
836 return false;
837
838 if (req_op(rq) == REQ_OP_FLUSH)
839 return false;
840
841 if (req_op(rq) == REQ_OP_WRITE_ZEROES)
842 return false;
843
844 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
845 return false;
846 if (rq->rq_flags & RQF_NOMERGE_FLAGS)
847 return false;
848
849 return true;
850 }
851
blk_write_same_mergeable(struct bio * a,struct bio * b)852 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
853 {
854 if (bio_page(a) == bio_page(b) &&
855 bio_offset(a) == bio_offset(b))
856 return true;
857
858 return false;
859 }
860
blk_queue_depth(struct request_queue * q)861 static inline unsigned int blk_queue_depth(struct request_queue *q)
862 {
863 if (q->queue_depth)
864 return q->queue_depth;
865
866 return q->nr_requests;
867 }
868
869 /*
870 * q->prep_rq_fn return values
871 */
872 enum {
873 BLKPREP_OK, /* serve it */
874 BLKPREP_KILL, /* fatal error, kill, return -EIO */
875 BLKPREP_DEFER, /* leave on queue */
876 BLKPREP_INVALID, /* invalid command, kill, return -EREMOTEIO */
877 };
878
879 extern unsigned long blk_max_low_pfn, blk_max_pfn;
880
881 /*
882 * standard bounce addresses:
883 *
884 * BLK_BOUNCE_HIGH : bounce all highmem pages
885 * BLK_BOUNCE_ANY : don't bounce anything
886 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
887 */
888
889 #if BITS_PER_LONG == 32
890 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
891 #else
892 #define BLK_BOUNCE_HIGH -1ULL
893 #endif
894 #define BLK_BOUNCE_ANY (-1ULL)
895 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
896
897 /*
898 * default timeout for SG_IO if none specified
899 */
900 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
901 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
902
903 struct rq_map_data {
904 struct page **pages;
905 int page_order;
906 int nr_entries;
907 unsigned long offset;
908 int null_mapped;
909 int from_user;
910 };
911
912 struct req_iterator {
913 struct bvec_iter iter;
914 struct bio *bio;
915 };
916
917 /* This should not be used directly - use rq_for_each_segment */
918 #define for_each_bio(_bio) \
919 for (; _bio; _bio = _bio->bi_next)
920 #define __rq_for_each_bio(_bio, rq) \
921 if ((rq->bio)) \
922 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
923
924 #define rq_for_each_segment(bvl, _rq, _iter) \
925 __rq_for_each_bio(_iter.bio, _rq) \
926 bio_for_each_segment(bvl, _iter.bio, _iter.iter)
927
928 #define rq_iter_last(bvec, _iter) \
929 (_iter.bio->bi_next == NULL && \
930 bio_iter_last(bvec, _iter.iter))
931
932 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
933 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
934 #endif
935 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
936 extern void rq_flush_dcache_pages(struct request *rq);
937 #else
rq_flush_dcache_pages(struct request * rq)938 static inline void rq_flush_dcache_pages(struct request *rq)
939 {
940 }
941 #endif
942
943 #ifdef CONFIG_PRINTK
944 #define vfs_msg(sb, level, fmt, ...) \
945 __vfs_msg(sb, level, fmt, ##__VA_ARGS__)
946 #else
947 #define vfs_msg(sb, level, fmt, ...) \
948 do { \
949 no_printk(fmt, ##__VA_ARGS__); \
950 __vfs_msg(sb, "", " "); \
951 } while (0)
952 #endif
953
954 extern int blk_register_queue(struct gendisk *disk);
955 extern void blk_unregister_queue(struct gendisk *disk);
956 extern blk_qc_t generic_make_request(struct bio *bio);
957 extern void blk_rq_init(struct request_queue *q, struct request *rq);
958 extern void blk_init_request_from_bio(struct request *req, struct bio *bio);
959 extern void blk_put_request(struct request *);
960 extern void __blk_put_request(struct request_queue *, struct request *);
961 extern struct request *blk_get_request(struct request_queue *, unsigned int op,
962 gfp_t gfp_mask);
963 extern void blk_requeue_request(struct request_queue *, struct request *);
964 extern int blk_lld_busy(struct request_queue *q);
965 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
966 struct bio_set *bs, gfp_t gfp_mask,
967 int (*bio_ctr)(struct bio *, struct bio *, void *),
968 void *data);
969 extern void blk_rq_unprep_clone(struct request *rq);
970 extern blk_status_t blk_insert_cloned_request(struct request_queue *q,
971 struct request *rq);
972 extern int blk_rq_append_bio(struct request *rq, struct bio **bio);
973 extern void blk_delay_queue(struct request_queue *, unsigned long);
974 extern void blk_queue_split(struct request_queue *, struct bio **);
975 extern void blk_recount_segments(struct request_queue *, struct bio *);
976 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
977 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
978 unsigned int, void __user *);
979 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
980 unsigned int, void __user *);
981 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
982 struct scsi_ioctl_command __user *);
983
984 extern int blk_queue_enter(struct request_queue *q, bool nowait);
985 extern void blk_queue_exit(struct request_queue *q);
986 extern void blk_start_queue(struct request_queue *q);
987 extern void blk_start_queue_async(struct request_queue *q);
988 extern void blk_stop_queue(struct request_queue *q);
989 extern void blk_sync_queue(struct request_queue *q);
990 extern void __blk_stop_queue(struct request_queue *q);
991 extern void __blk_run_queue(struct request_queue *q);
992 extern void __blk_run_queue_uncond(struct request_queue *q);
993 extern void blk_run_queue(struct request_queue *);
994 extern void blk_run_queue_async(struct request_queue *q);
995 extern int blk_rq_map_user(struct request_queue *, struct request *,
996 struct rq_map_data *, void __user *, unsigned long,
997 gfp_t);
998 extern int blk_rq_unmap_user(struct bio *);
999 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
1000 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
1001 struct rq_map_data *, const struct iov_iter *,
1002 gfp_t);
1003 extern void blk_execute_rq(struct request_queue *, struct gendisk *,
1004 struct request *, int);
1005 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
1006 struct request *, int, rq_end_io_fn *);
1007
1008 int blk_status_to_errno(blk_status_t status);
1009 blk_status_t errno_to_blk_status(int errno);
1010
1011 bool blk_mq_poll(struct request_queue *q, blk_qc_t cookie);
1012
bdev_get_queue(struct block_device * bdev)1013 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
1014 {
1015 return bdev->bd_disk->queue; /* this is never NULL */
1016 }
1017
1018 /*
1019 * blk_rq_pos() : the current sector
1020 * blk_rq_bytes() : bytes left in the entire request
1021 * blk_rq_cur_bytes() : bytes left in the current segment
1022 * blk_rq_err_bytes() : bytes left till the next error boundary
1023 * blk_rq_sectors() : sectors left in the entire request
1024 * blk_rq_cur_sectors() : sectors left in the current segment
1025 */
blk_rq_pos(const struct request * rq)1026 static inline sector_t blk_rq_pos(const struct request *rq)
1027 {
1028 return rq->__sector;
1029 }
1030
blk_rq_bytes(const struct request * rq)1031 static inline unsigned int blk_rq_bytes(const struct request *rq)
1032 {
1033 return rq->__data_len;
1034 }
1035
blk_rq_cur_bytes(const struct request * rq)1036 static inline int blk_rq_cur_bytes(const struct request *rq)
1037 {
1038 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
1039 }
1040
1041 extern unsigned int blk_rq_err_bytes(const struct request *rq);
1042
blk_rq_sectors(const struct request * rq)1043 static inline unsigned int blk_rq_sectors(const struct request *rq)
1044 {
1045 return blk_rq_bytes(rq) >> 9;
1046 }
1047
blk_rq_cur_sectors(const struct request * rq)1048 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
1049 {
1050 return blk_rq_cur_bytes(rq) >> 9;
1051 }
1052
1053 /*
1054 * Some commands like WRITE SAME have a payload or data transfer size which
1055 * is different from the size of the request. Any driver that supports such
1056 * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to
1057 * calculate the data transfer size.
1058 */
blk_rq_payload_bytes(struct request * rq)1059 static inline unsigned int blk_rq_payload_bytes(struct request *rq)
1060 {
1061 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1062 return rq->special_vec.bv_len;
1063 return blk_rq_bytes(rq);
1064 }
1065
blk_queue_get_max_sectors(struct request_queue * q,int op)1066 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
1067 int op)
1068 {
1069 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
1070 return min(q->limits.max_discard_sectors, UINT_MAX >> 9);
1071
1072 if (unlikely(op == REQ_OP_WRITE_SAME))
1073 return q->limits.max_write_same_sectors;
1074
1075 if (unlikely(op == REQ_OP_WRITE_ZEROES))
1076 return q->limits.max_write_zeroes_sectors;
1077
1078 return q->limits.max_sectors;
1079 }
1080
1081 /*
1082 * Return maximum size of a request at given offset. Only valid for
1083 * file system requests.
1084 */
blk_max_size_offset(struct request_queue * q,sector_t offset)1085 static inline unsigned int blk_max_size_offset(struct request_queue *q,
1086 sector_t offset)
1087 {
1088 if (!q->limits.chunk_sectors)
1089 return q->limits.max_sectors;
1090
1091 return min(q->limits.max_sectors, (unsigned int)(q->limits.chunk_sectors -
1092 (offset & (q->limits.chunk_sectors - 1))));
1093 }
1094
blk_rq_get_max_sectors(struct request * rq,sector_t offset)1095 static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1096 sector_t offset)
1097 {
1098 struct request_queue *q = rq->q;
1099
1100 if (blk_rq_is_passthrough(rq))
1101 return q->limits.max_hw_sectors;
1102
1103 if (!q->limits.chunk_sectors ||
1104 req_op(rq) == REQ_OP_DISCARD ||
1105 req_op(rq) == REQ_OP_SECURE_ERASE)
1106 return blk_queue_get_max_sectors(q, req_op(rq));
1107
1108 return min(blk_max_size_offset(q, offset),
1109 blk_queue_get_max_sectors(q, req_op(rq)));
1110 }
1111
blk_rq_count_bios(struct request * rq)1112 static inline unsigned int blk_rq_count_bios(struct request *rq)
1113 {
1114 unsigned int nr_bios = 0;
1115 struct bio *bio;
1116
1117 __rq_for_each_bio(bio, rq)
1118 nr_bios++;
1119
1120 return nr_bios;
1121 }
1122
1123 /*
1124 * Request issue related functions.
1125 */
1126 extern struct request *blk_peek_request(struct request_queue *q);
1127 extern void blk_start_request(struct request *rq);
1128 extern struct request *blk_fetch_request(struct request_queue *q);
1129
1130 /*
1131 * Request completion related functions.
1132 *
1133 * blk_update_request() completes given number of bytes and updates
1134 * the request without completing it.
1135 *
1136 * blk_end_request() and friends. __blk_end_request() must be called
1137 * with the request queue spinlock acquired.
1138 *
1139 * Several drivers define their own end_request and call
1140 * blk_end_request() for parts of the original function.
1141 * This prevents code duplication in drivers.
1142 */
1143 extern bool blk_update_request(struct request *rq, blk_status_t error,
1144 unsigned int nr_bytes);
1145 extern void blk_finish_request(struct request *rq, blk_status_t error);
1146 extern bool blk_end_request(struct request *rq, blk_status_t error,
1147 unsigned int nr_bytes);
1148 extern void blk_end_request_all(struct request *rq, blk_status_t error);
1149 extern bool __blk_end_request(struct request *rq, blk_status_t error,
1150 unsigned int nr_bytes);
1151 extern void __blk_end_request_all(struct request *rq, blk_status_t error);
1152 extern bool __blk_end_request_cur(struct request *rq, blk_status_t error);
1153
1154 extern void blk_complete_request(struct request *);
1155 extern void __blk_complete_request(struct request *);
1156 extern void blk_abort_request(struct request *);
1157 extern void blk_unprep_request(struct request *);
1158
1159 /*
1160 * Access functions for manipulating queue properties
1161 */
1162 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
1163 spinlock_t *lock, int node_id);
1164 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
1165 extern int blk_init_allocated_queue(struct request_queue *);
1166 extern void blk_cleanup_queue(struct request_queue *);
1167 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
1168 extern void blk_queue_bounce_limit(struct request_queue *, u64);
1169 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1170 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1171 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1172 extern void blk_queue_max_discard_segments(struct request_queue *,
1173 unsigned short);
1174 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1175 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1176 unsigned int max_discard_sectors);
1177 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1178 unsigned int max_write_same_sectors);
1179 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1180 unsigned int max_write_same_sectors);
1181 extern void blk_queue_logical_block_size(struct request_queue *, unsigned int);
1182 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1183 extern void blk_queue_alignment_offset(struct request_queue *q,
1184 unsigned int alignment);
1185 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1186 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1187 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1188 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1189 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1190 extern void blk_set_default_limits(struct queue_limits *lim);
1191 extern void blk_set_stacking_limits(struct queue_limits *lim);
1192 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1193 sector_t offset);
1194 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
1195 sector_t offset);
1196 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1197 sector_t offset);
1198 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
1199 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
1200 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1201 extern int blk_queue_dma_drain(struct request_queue *q,
1202 dma_drain_needed_fn *dma_drain_needed,
1203 void *buf, unsigned int size);
1204 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
1205 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1206 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1207 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
1208 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
1209 extern void blk_queue_dma_alignment(struct request_queue *, int);
1210 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1211 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1212 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1213 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1214 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1215 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1216
1217 /*
1218 * Number of physical segments as sent to the device.
1219 *
1220 * Normally this is the number of discontiguous data segments sent by the
1221 * submitter. But for data-less command like discard we might have no
1222 * actual data segments submitted, but the driver might have to add it's
1223 * own special payload. In that case we still return 1 here so that this
1224 * special payload will be mapped.
1225 */
blk_rq_nr_phys_segments(struct request * rq)1226 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1227 {
1228 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1229 return 1;
1230 return rq->nr_phys_segments;
1231 }
1232
1233 /*
1234 * Number of discard segments (or ranges) the driver needs to fill in.
1235 * Each discard bio merged into a request is counted as one segment.
1236 */
blk_rq_nr_discard_segments(struct request * rq)1237 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq)
1238 {
1239 return max_t(unsigned short, rq->nr_phys_segments, 1);
1240 }
1241
1242 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1243 extern void blk_dump_rq_flags(struct request *, char *);
1244 extern long nr_blockdev_pages(void);
1245
1246 bool __must_check blk_get_queue(struct request_queue *);
1247 struct request_queue *blk_alloc_queue(gfp_t);
1248 struct request_queue *blk_alloc_queue_node(gfp_t, int);
1249 extern void blk_put_queue(struct request_queue *);
1250 extern void blk_set_queue_dying(struct request_queue *);
1251
1252 /*
1253 * block layer runtime pm functions
1254 */
1255 #ifdef CONFIG_PM
1256 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1257 extern int blk_pre_runtime_suspend(struct request_queue *q);
1258 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1259 extern void blk_pre_runtime_resume(struct request_queue *q);
1260 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1261 extern void blk_set_runtime_active(struct request_queue *q);
1262 #else
blk_pm_runtime_init(struct request_queue * q,struct device * dev)1263 static inline void blk_pm_runtime_init(struct request_queue *q,
1264 struct device *dev) {}
blk_pre_runtime_suspend(struct request_queue * q)1265 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1266 {
1267 return -ENOSYS;
1268 }
blk_post_runtime_suspend(struct request_queue * q,int err)1269 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
blk_pre_runtime_resume(struct request_queue * q)1270 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
blk_post_runtime_resume(struct request_queue * q,int err)1271 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
blk_set_runtime_active(struct request_queue * q)1272 static inline void blk_set_runtime_active(struct request_queue *q) {}
1273 #endif
1274
1275 /*
1276 * blk_plug permits building a queue of related requests by holding the I/O
1277 * fragments for a short period. This allows merging of sequential requests
1278 * into single larger request. As the requests are moved from a per-task list to
1279 * the device's request_queue in a batch, this results in improved scalability
1280 * as the lock contention for request_queue lock is reduced.
1281 *
1282 * It is ok not to disable preemption when adding the request to the plug list
1283 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1284 * the plug list when the task sleeps by itself. For details, please see
1285 * schedule() where blk_schedule_flush_plug() is called.
1286 */
1287 struct blk_plug {
1288 struct list_head list; /* requests */
1289 struct list_head mq_list; /* blk-mq requests */
1290 struct list_head cb_list; /* md requires an unplug callback */
1291 };
1292 #define BLK_MAX_REQUEST_COUNT 16
1293 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1294
1295 struct blk_plug_cb;
1296 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1297 struct blk_plug_cb {
1298 struct list_head list;
1299 blk_plug_cb_fn callback;
1300 void *data;
1301 };
1302 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1303 void *data, int size);
1304 extern void blk_start_plug(struct blk_plug *);
1305 extern void blk_finish_plug(struct blk_plug *);
1306 extern void blk_flush_plug_list(struct blk_plug *, bool);
1307
blk_flush_plug(struct task_struct * tsk)1308 static inline void blk_flush_plug(struct task_struct *tsk)
1309 {
1310 struct blk_plug *plug = tsk->plug;
1311
1312 if (plug)
1313 blk_flush_plug_list(plug, false);
1314 }
1315
blk_schedule_flush_plug(struct task_struct * tsk)1316 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1317 {
1318 struct blk_plug *plug = tsk->plug;
1319
1320 if (plug)
1321 blk_flush_plug_list(plug, true);
1322 }
1323
blk_needs_flush_plug(struct task_struct * tsk)1324 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1325 {
1326 struct blk_plug *plug = tsk->plug;
1327
1328 return plug &&
1329 (!list_empty(&plug->list) ||
1330 !list_empty(&plug->mq_list) ||
1331 !list_empty(&plug->cb_list));
1332 }
1333
1334 /*
1335 * tag stuff
1336 */
1337 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1338 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1339 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1340 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1341 extern void blk_queue_free_tags(struct request_queue *);
1342 extern int blk_queue_resize_tags(struct request_queue *, int);
1343 extern void blk_queue_invalidate_tags(struct request_queue *);
1344 extern struct blk_queue_tag *blk_init_tags(int, int);
1345 extern void blk_free_tags(struct blk_queue_tag *);
1346
blk_map_queue_find_tag(struct blk_queue_tag * bqt,int tag)1347 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1348 int tag)
1349 {
1350 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1351 return NULL;
1352 return bqt->tag_index[tag];
1353 }
1354
1355 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1356 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1357 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1358
1359 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */
1360
1361 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1362 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1363 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1364 sector_t nr_sects, gfp_t gfp_mask, int flags,
1365 struct bio **biop);
1366
1367 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */
1368 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */
1369
1370 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1371 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1372 unsigned flags);
1373 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1374 sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
1375
sb_issue_discard(struct super_block * sb,sector_t block,sector_t nr_blocks,gfp_t gfp_mask,unsigned long flags)1376 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1377 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1378 {
1379 return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9),
1380 nr_blocks << (sb->s_blocksize_bits - 9),
1381 gfp_mask, flags);
1382 }
sb_issue_zeroout(struct super_block * sb,sector_t block,sector_t nr_blocks,gfp_t gfp_mask)1383 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1384 sector_t nr_blocks, gfp_t gfp_mask)
1385 {
1386 return blkdev_issue_zeroout(sb->s_bdev,
1387 block << (sb->s_blocksize_bits - 9),
1388 nr_blocks << (sb->s_blocksize_bits - 9),
1389 gfp_mask, 0);
1390 }
1391
1392 extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm);
1393
1394 enum blk_default_limits {
1395 BLK_MAX_SEGMENTS = 128,
1396 BLK_SAFE_MAX_SECTORS = 255,
1397 BLK_DEF_MAX_SECTORS = 2560,
1398 BLK_MAX_SEGMENT_SIZE = 65536,
1399 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1400 };
1401
1402 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1403
queue_segment_boundary(struct request_queue * q)1404 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1405 {
1406 return q->limits.seg_boundary_mask;
1407 }
1408
queue_virt_boundary(struct request_queue * q)1409 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1410 {
1411 return q->limits.virt_boundary_mask;
1412 }
1413
queue_max_sectors(struct request_queue * q)1414 static inline unsigned int queue_max_sectors(struct request_queue *q)
1415 {
1416 return q->limits.max_sectors;
1417 }
1418
queue_max_hw_sectors(struct request_queue * q)1419 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1420 {
1421 return q->limits.max_hw_sectors;
1422 }
1423
queue_max_segments(struct request_queue * q)1424 static inline unsigned short queue_max_segments(struct request_queue *q)
1425 {
1426 return q->limits.max_segments;
1427 }
1428
queue_max_discard_segments(struct request_queue * q)1429 static inline unsigned short queue_max_discard_segments(struct request_queue *q)
1430 {
1431 return q->limits.max_discard_segments;
1432 }
1433
queue_max_segment_size(struct request_queue * q)1434 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1435 {
1436 return q->limits.max_segment_size;
1437 }
1438
queue_logical_block_size(struct request_queue * q)1439 static inline unsigned queue_logical_block_size(struct request_queue *q)
1440 {
1441 int retval = 512;
1442
1443 if (q && q->limits.logical_block_size)
1444 retval = q->limits.logical_block_size;
1445
1446 return retval;
1447 }
1448
bdev_logical_block_size(struct block_device * bdev)1449 static inline unsigned int bdev_logical_block_size(struct block_device *bdev)
1450 {
1451 return queue_logical_block_size(bdev_get_queue(bdev));
1452 }
1453
queue_physical_block_size(struct request_queue * q)1454 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1455 {
1456 return q->limits.physical_block_size;
1457 }
1458
bdev_physical_block_size(struct block_device * bdev)1459 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1460 {
1461 return queue_physical_block_size(bdev_get_queue(bdev));
1462 }
1463
queue_io_min(struct request_queue * q)1464 static inline unsigned int queue_io_min(struct request_queue *q)
1465 {
1466 return q->limits.io_min;
1467 }
1468
bdev_io_min(struct block_device * bdev)1469 static inline int bdev_io_min(struct block_device *bdev)
1470 {
1471 return queue_io_min(bdev_get_queue(bdev));
1472 }
1473
queue_io_opt(struct request_queue * q)1474 static inline unsigned int queue_io_opt(struct request_queue *q)
1475 {
1476 return q->limits.io_opt;
1477 }
1478
bdev_io_opt(struct block_device * bdev)1479 static inline int bdev_io_opt(struct block_device *bdev)
1480 {
1481 return queue_io_opt(bdev_get_queue(bdev));
1482 }
1483
queue_alignment_offset(struct request_queue * q)1484 static inline int queue_alignment_offset(struct request_queue *q)
1485 {
1486 if (q->limits.misaligned)
1487 return -1;
1488
1489 return q->limits.alignment_offset;
1490 }
1491
queue_limit_alignment_offset(struct queue_limits * lim,sector_t sector)1492 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1493 {
1494 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1495 unsigned int alignment = sector_div(sector, granularity >> 9) << 9;
1496
1497 return (granularity + lim->alignment_offset - alignment) % granularity;
1498 }
1499
bdev_alignment_offset(struct block_device * bdev)1500 static inline int bdev_alignment_offset(struct block_device *bdev)
1501 {
1502 struct request_queue *q = bdev_get_queue(bdev);
1503
1504 if (q->limits.misaligned)
1505 return -1;
1506
1507 if (bdev != bdev->bd_contains)
1508 return bdev->bd_part->alignment_offset;
1509
1510 return q->limits.alignment_offset;
1511 }
1512
queue_discard_alignment(struct request_queue * q)1513 static inline int queue_discard_alignment(struct request_queue *q)
1514 {
1515 if (q->limits.discard_misaligned)
1516 return -1;
1517
1518 return q->limits.discard_alignment;
1519 }
1520
queue_limit_discard_alignment(struct queue_limits * lim,sector_t sector)1521 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1522 {
1523 unsigned int alignment, granularity, offset;
1524
1525 if (!lim->max_discard_sectors)
1526 return 0;
1527
1528 /* Why are these in bytes, not sectors? */
1529 alignment = lim->discard_alignment >> 9;
1530 granularity = lim->discard_granularity >> 9;
1531 if (!granularity)
1532 return 0;
1533
1534 /* Offset of the partition start in 'granularity' sectors */
1535 offset = sector_div(sector, granularity);
1536
1537 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1538 offset = (granularity + alignment - offset) % granularity;
1539
1540 /* Turn it back into bytes, gaah */
1541 return offset << 9;
1542 }
1543
bdev_discard_alignment(struct block_device * bdev)1544 static inline int bdev_discard_alignment(struct block_device *bdev)
1545 {
1546 struct request_queue *q = bdev_get_queue(bdev);
1547
1548 if (bdev != bdev->bd_contains)
1549 return bdev->bd_part->discard_alignment;
1550
1551 return q->limits.discard_alignment;
1552 }
1553
bdev_write_same(struct block_device * bdev)1554 static inline unsigned int bdev_write_same(struct block_device *bdev)
1555 {
1556 struct request_queue *q = bdev_get_queue(bdev);
1557
1558 if (q)
1559 return q->limits.max_write_same_sectors;
1560
1561 return 0;
1562 }
1563
bdev_write_zeroes_sectors(struct block_device * bdev)1564 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1565 {
1566 struct request_queue *q = bdev_get_queue(bdev);
1567
1568 if (q)
1569 return q->limits.max_write_zeroes_sectors;
1570
1571 return 0;
1572 }
1573
bdev_zoned_model(struct block_device * bdev)1574 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1575 {
1576 struct request_queue *q = bdev_get_queue(bdev);
1577
1578 if (q)
1579 return blk_queue_zoned_model(q);
1580
1581 return BLK_ZONED_NONE;
1582 }
1583
bdev_is_zoned(struct block_device * bdev)1584 static inline bool bdev_is_zoned(struct block_device *bdev)
1585 {
1586 struct request_queue *q = bdev_get_queue(bdev);
1587
1588 if (q)
1589 return blk_queue_is_zoned(q);
1590
1591 return false;
1592 }
1593
bdev_zone_sectors(struct block_device * bdev)1594 static inline unsigned int bdev_zone_sectors(struct block_device *bdev)
1595 {
1596 struct request_queue *q = bdev_get_queue(bdev);
1597
1598 if (q)
1599 return blk_queue_zone_sectors(q);
1600
1601 return 0;
1602 }
1603
queue_dma_alignment(struct request_queue * q)1604 static inline int queue_dma_alignment(struct request_queue *q)
1605 {
1606 return q ? q->dma_alignment : 511;
1607 }
1608
blk_rq_aligned(struct request_queue * q,unsigned long addr,unsigned int len)1609 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1610 unsigned int len)
1611 {
1612 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1613 return !(addr & alignment) && !(len & alignment);
1614 }
1615
1616 /* assumes size > 256 */
blksize_bits(unsigned int size)1617 static inline unsigned int blksize_bits(unsigned int size)
1618 {
1619 unsigned int bits = 8;
1620 do {
1621 bits++;
1622 size >>= 1;
1623 } while (size > 256);
1624 return bits;
1625 }
1626
block_size(struct block_device * bdev)1627 static inline unsigned int block_size(struct block_device *bdev)
1628 {
1629 return bdev->bd_block_size;
1630 }
1631
queue_flush_queueable(struct request_queue * q)1632 static inline bool queue_flush_queueable(struct request_queue *q)
1633 {
1634 return !test_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags);
1635 }
1636
1637 typedef struct {struct page *v;} Sector;
1638
1639 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1640
put_dev_sector(Sector p)1641 static inline void put_dev_sector(Sector p)
1642 {
1643 put_page(p.v);
1644 }
1645
__bvec_gap_to_prev(struct request_queue * q,struct bio_vec * bprv,unsigned int offset)1646 static inline bool __bvec_gap_to_prev(struct request_queue *q,
1647 struct bio_vec *bprv, unsigned int offset)
1648 {
1649 return offset ||
1650 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1651 }
1652
1653 /*
1654 * Check if adding a bio_vec after bprv with offset would create a gap in
1655 * the SG list. Most drivers don't care about this, but some do.
1656 */
bvec_gap_to_prev(struct request_queue * q,struct bio_vec * bprv,unsigned int offset)1657 static inline bool bvec_gap_to_prev(struct request_queue *q,
1658 struct bio_vec *bprv, unsigned int offset)
1659 {
1660 if (!queue_virt_boundary(q))
1661 return false;
1662 return __bvec_gap_to_prev(q, bprv, offset);
1663 }
1664
1665 /*
1666 * Check if the two bvecs from two bios can be merged to one segment.
1667 * If yes, no need to check gap between the two bios since the 1st bio
1668 * and the 1st bvec in the 2nd bio can be handled in one segment.
1669 */
bios_segs_mergeable(struct request_queue * q,struct bio * prev,struct bio_vec * prev_last_bv,struct bio_vec * next_first_bv)1670 static inline bool bios_segs_mergeable(struct request_queue *q,
1671 struct bio *prev, struct bio_vec *prev_last_bv,
1672 struct bio_vec *next_first_bv)
1673 {
1674 if (!BIOVEC_PHYS_MERGEABLE(prev_last_bv, next_first_bv))
1675 return false;
1676 if (!BIOVEC_SEG_BOUNDARY(q, prev_last_bv, next_first_bv))
1677 return false;
1678 if (prev->bi_seg_back_size + next_first_bv->bv_len >
1679 queue_max_segment_size(q))
1680 return false;
1681 return true;
1682 }
1683
bio_will_gap(struct request_queue * q,struct request * prev_rq,struct bio * prev,struct bio * next)1684 static inline bool bio_will_gap(struct request_queue *q,
1685 struct request *prev_rq,
1686 struct bio *prev,
1687 struct bio *next)
1688 {
1689 if (bio_has_data(prev) && queue_virt_boundary(q)) {
1690 struct bio_vec pb, nb;
1691
1692 /*
1693 * don't merge if the 1st bio starts with non-zero
1694 * offset, otherwise it is quite difficult to respect
1695 * sg gap limit. We work hard to merge a huge number of small
1696 * single bios in case of mkfs.
1697 */
1698 if (prev_rq)
1699 bio_get_first_bvec(prev_rq->bio, &pb);
1700 else
1701 bio_get_first_bvec(prev, &pb);
1702 if (pb.bv_offset)
1703 return true;
1704
1705 /*
1706 * We don't need to worry about the situation that the
1707 * merged segment ends in unaligned virt boundary:
1708 *
1709 * - if 'pb' ends aligned, the merged segment ends aligned
1710 * - if 'pb' ends unaligned, the next bio must include
1711 * one single bvec of 'nb', otherwise the 'nb' can't
1712 * merge with 'pb'
1713 */
1714 bio_get_last_bvec(prev, &pb);
1715 bio_get_first_bvec(next, &nb);
1716
1717 if (!bios_segs_mergeable(q, prev, &pb, &nb))
1718 return __bvec_gap_to_prev(q, &pb, nb.bv_offset);
1719 }
1720
1721 return false;
1722 }
1723
req_gap_back_merge(struct request * req,struct bio * bio)1724 static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
1725 {
1726 return bio_will_gap(req->q, req, req->biotail, bio);
1727 }
1728
req_gap_front_merge(struct request * req,struct bio * bio)1729 static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
1730 {
1731 return bio_will_gap(req->q, NULL, bio, req->bio);
1732 }
1733
1734 int kblockd_schedule_work(struct work_struct *work);
1735 int kblockd_schedule_work_on(int cpu, struct work_struct *work);
1736 int kblockd_schedule_delayed_work(struct delayed_work *dwork, unsigned long delay);
1737 int kblockd_schedule_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1738 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1739
1740 #ifdef CONFIG_BLK_CGROUP
1741 /*
1742 * This should not be using sched_clock(). A real patch is in progress
1743 * to fix this up, until that is in place we need to disable preemption
1744 * around sched_clock() in this function and set_io_start_time_ns().
1745 */
set_start_time_ns(struct request * req)1746 static inline void set_start_time_ns(struct request *req)
1747 {
1748 preempt_disable();
1749 req->start_time_ns = sched_clock();
1750 preempt_enable();
1751 }
1752
set_io_start_time_ns(struct request * req)1753 static inline void set_io_start_time_ns(struct request *req)
1754 {
1755 preempt_disable();
1756 req->io_start_time_ns = sched_clock();
1757 preempt_enable();
1758 }
1759
rq_start_time_ns(struct request * req)1760 static inline uint64_t rq_start_time_ns(struct request *req)
1761 {
1762 return req->start_time_ns;
1763 }
1764
rq_io_start_time_ns(struct request * req)1765 static inline uint64_t rq_io_start_time_ns(struct request *req)
1766 {
1767 return req->io_start_time_ns;
1768 }
1769 #else
set_start_time_ns(struct request * req)1770 static inline void set_start_time_ns(struct request *req) {}
set_io_start_time_ns(struct request * req)1771 static inline void set_io_start_time_ns(struct request *req) {}
rq_start_time_ns(struct request * req)1772 static inline uint64_t rq_start_time_ns(struct request *req)
1773 {
1774 return 0;
1775 }
rq_io_start_time_ns(struct request * req)1776 static inline uint64_t rq_io_start_time_ns(struct request *req)
1777 {
1778 return 0;
1779 }
1780 #endif
1781
1782 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1783 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1784 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1785 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1786
1787 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1788
1789 enum blk_integrity_flags {
1790 BLK_INTEGRITY_VERIFY = 1 << 0,
1791 BLK_INTEGRITY_GENERATE = 1 << 1,
1792 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2,
1793 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3,
1794 };
1795
1796 struct blk_integrity_iter {
1797 void *prot_buf;
1798 void *data_buf;
1799 sector_t seed;
1800 unsigned int data_size;
1801 unsigned short interval;
1802 const char *disk_name;
1803 };
1804
1805 typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *);
1806
1807 struct blk_integrity_profile {
1808 integrity_processing_fn *generate_fn;
1809 integrity_processing_fn *verify_fn;
1810 const char *name;
1811 };
1812
1813 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1814 extern void blk_integrity_unregister(struct gendisk *);
1815 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1816 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1817 struct scatterlist *);
1818 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1819 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1820 struct request *);
1821 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1822 struct bio *);
1823
blk_get_integrity(struct gendisk * disk)1824 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1825 {
1826 struct blk_integrity *bi = &disk->queue->integrity;
1827
1828 if (!bi->profile)
1829 return NULL;
1830
1831 return bi;
1832 }
1833
1834 static inline
bdev_get_integrity(struct block_device * bdev)1835 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1836 {
1837 return blk_get_integrity(bdev->bd_disk);
1838 }
1839
blk_integrity_rq(struct request * rq)1840 static inline bool blk_integrity_rq(struct request *rq)
1841 {
1842 return rq->cmd_flags & REQ_INTEGRITY;
1843 }
1844
blk_queue_max_integrity_segments(struct request_queue * q,unsigned int segs)1845 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1846 unsigned int segs)
1847 {
1848 q->limits.max_integrity_segments = segs;
1849 }
1850
1851 static inline unsigned short
queue_max_integrity_segments(struct request_queue * q)1852 queue_max_integrity_segments(struct request_queue *q)
1853 {
1854 return q->limits.max_integrity_segments;
1855 }
1856
integrity_req_gap_back_merge(struct request * req,struct bio * next)1857 static inline bool integrity_req_gap_back_merge(struct request *req,
1858 struct bio *next)
1859 {
1860 struct bio_integrity_payload *bip = bio_integrity(req->bio);
1861 struct bio_integrity_payload *bip_next = bio_integrity(next);
1862
1863 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1864 bip_next->bip_vec[0].bv_offset);
1865 }
1866
integrity_req_gap_front_merge(struct request * req,struct bio * bio)1867 static inline bool integrity_req_gap_front_merge(struct request *req,
1868 struct bio *bio)
1869 {
1870 struct bio_integrity_payload *bip = bio_integrity(bio);
1871 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
1872
1873 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1874 bip_next->bip_vec[0].bv_offset);
1875 }
1876
1877 #else /* CONFIG_BLK_DEV_INTEGRITY */
1878
1879 struct bio;
1880 struct block_device;
1881 struct gendisk;
1882 struct blk_integrity;
1883
blk_integrity_rq(struct request * rq)1884 static inline int blk_integrity_rq(struct request *rq)
1885 {
1886 return 0;
1887 }
blk_rq_count_integrity_sg(struct request_queue * q,struct bio * b)1888 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1889 struct bio *b)
1890 {
1891 return 0;
1892 }
blk_rq_map_integrity_sg(struct request_queue * q,struct bio * b,struct scatterlist * s)1893 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1894 struct bio *b,
1895 struct scatterlist *s)
1896 {
1897 return 0;
1898 }
bdev_get_integrity(struct block_device * b)1899 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1900 {
1901 return NULL;
1902 }
blk_get_integrity(struct gendisk * disk)1903 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1904 {
1905 return NULL;
1906 }
blk_integrity_compare(struct gendisk * a,struct gendisk * b)1907 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1908 {
1909 return 0;
1910 }
blk_integrity_register(struct gendisk * d,struct blk_integrity * b)1911 static inline void blk_integrity_register(struct gendisk *d,
1912 struct blk_integrity *b)
1913 {
1914 }
blk_integrity_unregister(struct gendisk * d)1915 static inline void blk_integrity_unregister(struct gendisk *d)
1916 {
1917 }
blk_queue_max_integrity_segments(struct request_queue * q,unsigned int segs)1918 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1919 unsigned int segs)
1920 {
1921 }
queue_max_integrity_segments(struct request_queue * q)1922 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1923 {
1924 return 0;
1925 }
blk_integrity_merge_rq(struct request_queue * rq,struct request * r1,struct request * r2)1926 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1927 struct request *r1,
1928 struct request *r2)
1929 {
1930 return true;
1931 }
blk_integrity_merge_bio(struct request_queue * rq,struct request * r,struct bio * b)1932 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1933 struct request *r,
1934 struct bio *b)
1935 {
1936 return true;
1937 }
1938
integrity_req_gap_back_merge(struct request * req,struct bio * next)1939 static inline bool integrity_req_gap_back_merge(struct request *req,
1940 struct bio *next)
1941 {
1942 return false;
1943 }
integrity_req_gap_front_merge(struct request * req,struct bio * bio)1944 static inline bool integrity_req_gap_front_merge(struct request *req,
1945 struct bio *bio)
1946 {
1947 return false;
1948 }
1949
1950 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1951
1952 struct block_device_operations {
1953 int (*open) (struct block_device *, fmode_t);
1954 void (*release) (struct gendisk *, fmode_t);
1955 int (*rw_page)(struct block_device *, sector_t, struct page *, bool);
1956 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1957 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1958 unsigned int (*check_events) (struct gendisk *disk,
1959 unsigned int clearing);
1960 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1961 int (*media_changed) (struct gendisk *);
1962 void (*unlock_native_capacity) (struct gendisk *);
1963 int (*revalidate_disk) (struct gendisk *);
1964 int (*getgeo)(struct block_device *, struct hd_geometry *);
1965 /* this callback is with swap_lock and sometimes page table lock held */
1966 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1967 struct module *owner;
1968 const struct pr_ops *pr_ops;
1969 };
1970
1971 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1972 unsigned long);
1973 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1974 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1975 struct writeback_control *);
1976 #else /* CONFIG_BLOCK */
1977
1978 struct block_device;
1979
1980 /*
1981 * stubs for when the block layer is configured out
1982 */
1983 #define buffer_heads_over_limit 0
1984
nr_blockdev_pages(void)1985 static inline long nr_blockdev_pages(void)
1986 {
1987 return 0;
1988 }
1989
1990 struct blk_plug {
1991 };
1992
blk_start_plug(struct blk_plug * plug)1993 static inline void blk_start_plug(struct blk_plug *plug)
1994 {
1995 }
1996
blk_finish_plug(struct blk_plug * plug)1997 static inline void blk_finish_plug(struct blk_plug *plug)
1998 {
1999 }
2000
blk_flush_plug(struct task_struct * task)2001 static inline void blk_flush_plug(struct task_struct *task)
2002 {
2003 }
2004
blk_schedule_flush_plug(struct task_struct * task)2005 static inline void blk_schedule_flush_plug(struct task_struct *task)
2006 {
2007 }
2008
2009
blk_needs_flush_plug(struct task_struct * tsk)2010 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
2011 {
2012 return false;
2013 }
2014
blkdev_issue_flush(struct block_device * bdev,gfp_t gfp_mask,sector_t * error_sector)2015 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
2016 sector_t *error_sector)
2017 {
2018 return 0;
2019 }
2020
2021 #endif /* CONFIG_BLOCK */
2022
2023 #endif
2024