projects/linux·blob v5.12-rc2 · refs

2034 lines · 57.0 KiB

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#include <linux/major.h>
8#include <linux/genhd.h>
9#include <linux/list.h>
10#include <linux/llist.h>
11#include <linux/minmax.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#include <linux/blkzoned.h>
28#include <linux/pm.h>
29
30struct module;
31struct scsi_ioctl_command;
32
33struct request_queue;
34struct elevator_queue;
35struct blk_trace;
36struct request;
37struct sg_io_hdr;
38struct bsg_job;
39struct blkcg_gq;
40struct blk_flush_queue;
41struct pr_ops;
42struct rq_qos;
43struct blk_queue_stats;
44struct blk_stat_callback;
45struct blk_keyslot_manager;
46
47#define BLKDEV_MIN_RQ 4
48#define BLKDEV_MAX_RQ 128 /* Default maximum */
49
50/* Must be consistent with blk_mq_poll_stats_bkt() */
51#define BLK_MQ_POLL_STATS_BKTS 16
52
53/* Doing classic polling */
54#define BLK_MQ_POLL_CLASSIC -1
55
56/*
57 * Maximum number of blkcg policies allowed to be registered concurrently.
58 * Defined here to simplify include dependency.
59 */
60#define BLKCG_MAX_POLS 5
61
62typedef void (rq_end_io_fn)(struct request *, blk_status_t);
63
64/*
65 * request flags */
66typedef __u32 __bitwise req_flags_t;
67
68/* drive already may have started this one */
69#define RQF_STARTED ((__force req_flags_t)(1 << 1))
70/* may not be passed by ioscheduler */
71#define RQF_SOFTBARRIER ((__force req_flags_t)(1 << 3))
72/* request for flush sequence */
73#define RQF_FLUSH_SEQ ((__force req_flags_t)(1 << 4))
74/* merge of different types, fail separately */
75#define RQF_MIXED_MERGE ((__force req_flags_t)(1 << 5))
76/* track inflight for MQ */
77#define RQF_MQ_INFLIGHT ((__force req_flags_t)(1 << 6))
78/* don't call prep for this one */
79#define RQF_DONTPREP ((__force req_flags_t)(1 << 7))
80/* vaguely specified driver internal error. Ignored by the block layer */
81#define RQF_FAILED ((__force req_flags_t)(1 << 10))
82/* don't warn about errors */
83#define RQF_QUIET ((__force req_flags_t)(1 << 11))
84/* elevator private data attached */
85#define RQF_ELVPRIV ((__force req_flags_t)(1 << 12))
86/* account into disk and partition IO statistics */
87#define RQF_IO_STAT ((__force req_flags_t)(1 << 13))
88/* request came from our alloc pool */
89#define RQF_ALLOCED ((__force req_flags_t)(1 << 14))
90/* runtime pm request */
91#define RQF_PM ((__force req_flags_t)(1 << 15))
92/* on IO scheduler merge hash */
93#define RQF_HASHED ((__force req_flags_t)(1 << 16))
94/* track IO completion time */
95#define RQF_STATS ((__force req_flags_t)(1 << 17))
96/* Look at ->special_vec for the actual data payload instead of the
97 bio chain. */
98#define RQF_SPECIAL_PAYLOAD ((__force req_flags_t)(1 << 18))
99/* The per-zone write lock is held for this request */
100#define RQF_ZONE_WRITE_LOCKED ((__force req_flags_t)(1 << 19))
101/* already slept for hybrid poll */
102#define RQF_MQ_POLL_SLEPT ((__force req_flags_t)(1 << 20))
103/* ->timeout has been called, don't expire again */
104#define RQF_TIMED_OUT ((__force req_flags_t)(1 << 21))
105
106/* flags that prevent us from merging requests: */
107#define RQF_NOMERGE_FLAGS \
108 (RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD)
109
110/*
111 * Request state for blk-mq.
112 */
113enum mq_rq_state {
114 MQ_RQ_IDLE = 0,
115 MQ_RQ_IN_FLIGHT = 1,
116 MQ_RQ_COMPLETE = 2,
117};
118
119/*
120 * Try to put the fields that are referenced together in the same cacheline.
121 *
122 * If you modify this structure, make sure to update blk_rq_init() and
123 * especially blk_mq_rq_ctx_init() to take care of the added fields.
124 */
125struct request {
126 struct request_queue *q;
127 struct blk_mq_ctx *mq_ctx;
128 struct blk_mq_hw_ctx *mq_hctx;
129
130 unsigned int cmd_flags; /* op and common flags */
131 req_flags_t rq_flags;
132
133 int tag;
134 int internal_tag;
135
136 /* the following two fields are internal, NEVER access directly */
137 unsigned int __data_len; /* total data len */
138 sector_t __sector; /* sector cursor */
139
140 struct bio *bio;
141 struct bio *biotail;
142
143 struct list_head queuelist;
144
145 /*
146 * The hash is used inside the scheduler, and killed once the
147 * request reaches the dispatch list. The ipi_list is only used
148 * to queue the request for softirq completion, which is long
149 * after the request has been unhashed (and even removed from
150 * the dispatch list).
151 */
152 union {
153 struct hlist_node hash; /* merge hash */
154 struct llist_node ipi_list;
155 };
156
157 /*
158 * The rb_node is only used inside the io scheduler, requests
159 * are pruned when moved to the dispatch queue. So let the
160 * completion_data share space with the rb_node.
161 */
162 union {
163 struct rb_node rb_node; /* sort/lookup */
164 struct bio_vec special_vec;
165 void *completion_data;
166 int error_count; /* for legacy drivers, don't use */
167 };
168
169 /*
170 * Three pointers are available for the IO schedulers, if they need
171 * more they have to dynamically allocate it. Flush requests are
172 * never put on the IO scheduler. So let the flush fields share
173 * space with the elevator data.
174 */
175 union {
176 struct {
177 struct io_cq *icq;
178 void *priv[2];
179 } elv;
180
181 struct {
182 unsigned int seq;
183 struct list_head list;
184 rq_end_io_fn *saved_end_io;
185 } flush;
186 };
187
188 struct gendisk *rq_disk;
189 struct block_device *part;
190#ifdef CONFIG_BLK_RQ_ALLOC_TIME
191 /* Time that the first bio started allocating this request. */
192 u64 alloc_time_ns;
193#endif
194 /* Time that this request was allocated for this IO. */
195 u64 start_time_ns;
196 /* Time that I/O was submitted to the device. */
197 u64 io_start_time_ns;
198
199#ifdef CONFIG_BLK_WBT
200 unsigned short wbt_flags;
201#endif
202 /*
203 * rq sectors used for blk stats. It has the same value
204 * with blk_rq_sectors(rq), except that it never be zeroed
205 * by completion.
206 */
207 unsigned short stats_sectors;
208
209 /*
210 * Number of scatter-gather DMA addr+len pairs after
211 * physical address coalescing is performed.
212 */
213 unsigned short nr_phys_segments;
214
215#if defined(CONFIG_BLK_DEV_INTEGRITY)
216 unsigned short nr_integrity_segments;
217#endif
218
219#ifdef CONFIG_BLK_INLINE_ENCRYPTION
220 struct bio_crypt_ctx *crypt_ctx;
221 struct blk_ksm_keyslot *crypt_keyslot;
222#endif
223
224 unsigned short write_hint;
225 unsigned short ioprio;
226
227 enum mq_rq_state state;
228 refcount_t ref;
229
230 unsigned int timeout;
231 unsigned long deadline;
232
233 union {
234 struct __call_single_data csd;
235 u64 fifo_time;
236 };
237
238 /*
239 * completion callback.
240 */
241 rq_end_io_fn *end_io;
242 void *end_io_data;
243};
244
245static inline bool blk_op_is_scsi(unsigned int op)
246{
247 return op == REQ_OP_SCSI_IN || op == REQ_OP_SCSI_OUT;
248}
249
250static inline bool blk_op_is_private(unsigned int op)
251{
252 return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT;
253}
254
255static inline bool blk_rq_is_scsi(struct request *rq)
256{
257 return blk_op_is_scsi(req_op(rq));
258}
259
260static inline bool blk_rq_is_private(struct request *rq)
261{
262 return blk_op_is_private(req_op(rq));
263}
264
265static inline bool blk_rq_is_passthrough(struct request *rq)
266{
267 return blk_rq_is_scsi(rq) || blk_rq_is_private(rq);
268}
269
270static inline bool bio_is_passthrough(struct bio *bio)
271{
272 unsigned op = bio_op(bio);
273
274 return blk_op_is_scsi(op) || blk_op_is_private(op);
275}
276
277static inline unsigned short req_get_ioprio(struct request *req)
278{
279 return req->ioprio;
280}
281
282#include <linux/elevator.h>
283
284struct blk_queue_ctx;
285
286struct bio_vec;
287
288enum blk_eh_timer_return {
289 BLK_EH_DONE, /* drivers has completed the command */
290 BLK_EH_RESET_TIMER, /* reset timer and try again */
291};
292
293enum blk_queue_state {
294 Queue_down,
295 Queue_up,
296};
297
298#define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
299#define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
300
301#define BLK_SCSI_MAX_CMDS (256)
302#define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
303
304/*
305 * Zoned block device models (zoned limit).
306 *
307 * Note: This needs to be ordered from the least to the most severe
308 * restrictions for the inheritance in blk_stack_limits() to work.
309 */
310enum blk_zoned_model {
311 BLK_ZONED_NONE = 0, /* Regular block device */
312 BLK_ZONED_HA, /* Host-aware zoned block device */
313 BLK_ZONED_HM, /* Host-managed zoned block device */
314};
315
316struct queue_limits {
317 unsigned long bounce_pfn;
318 unsigned long seg_boundary_mask;
319 unsigned long virt_boundary_mask;
320
321 unsigned int max_hw_sectors;
322 unsigned int max_dev_sectors;
323 unsigned int chunk_sectors;
324 unsigned int max_sectors;
325 unsigned int max_segment_size;
326 unsigned int physical_block_size;
327 unsigned int logical_block_size;
328 unsigned int alignment_offset;
329 unsigned int io_min;
330 unsigned int io_opt;
331 unsigned int max_discard_sectors;
332 unsigned int max_hw_discard_sectors;
333 unsigned int max_write_same_sectors;
334 unsigned int max_write_zeroes_sectors;
335 unsigned int max_zone_append_sectors;
336 unsigned int discard_granularity;
337 unsigned int discard_alignment;
338 unsigned int zone_write_granularity;
339
340 unsigned short max_segments;
341 unsigned short max_integrity_segments;
342 unsigned short max_discard_segments;
343
344 unsigned char misaligned;
345 unsigned char discard_misaligned;
346 unsigned char raid_partial_stripes_expensive;
347 enum blk_zoned_model zoned;
348};
349
350typedef int (*report_zones_cb)(struct blk_zone *zone, unsigned int idx,
351 void *data);
352
353void blk_queue_set_zoned(struct gendisk *disk, enum blk_zoned_model model);
354
355#ifdef CONFIG_BLK_DEV_ZONED
356
357#define BLK_ALL_ZONES ((unsigned int)-1)
358int blkdev_report_zones(struct block_device *bdev, sector_t sector,
359 unsigned int nr_zones, report_zones_cb cb, void *data);
360unsigned int blkdev_nr_zones(struct gendisk *disk);
361extern int blkdev_zone_mgmt(struct block_device *bdev, enum req_opf op,
362 sector_t sectors, sector_t nr_sectors,
363 gfp_t gfp_mask);
364int blk_revalidate_disk_zones(struct gendisk *disk,
365 void (*update_driver_data)(struct gendisk *disk));
366
367extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
368 unsigned int cmd, unsigned long arg);
369extern int blkdev_zone_mgmt_ioctl(struct block_device *bdev, fmode_t mode,
370 unsigned int cmd, unsigned long arg);
371
372#else /* CONFIG_BLK_DEV_ZONED */
373
374static inline unsigned int blkdev_nr_zones(struct gendisk *disk)
375{
376 return 0;
377}
378
379static inline int blkdev_report_zones_ioctl(struct block_device *bdev,
380 fmode_t mode, unsigned int cmd,
381 unsigned long arg)
382{
383 return -ENOTTY;
384}
385
386static inline int blkdev_zone_mgmt_ioctl(struct block_device *bdev,
387 fmode_t mode, unsigned int cmd,
388 unsigned long arg)
389{
390 return -ENOTTY;
391}
392
393#endif /* CONFIG_BLK_DEV_ZONED */
394
395struct request_queue {
396 struct request *last_merge;
397 struct elevator_queue *elevator;
398
399 struct percpu_ref q_usage_counter;
400
401 struct blk_queue_stats *stats;
402 struct rq_qos *rq_qos;
403
404 const struct blk_mq_ops *mq_ops;
405
406 /* sw queues */
407 struct blk_mq_ctx __percpu *queue_ctx;
408
409 unsigned int queue_depth;
410
411 /* hw dispatch queues */
412 struct blk_mq_hw_ctx **queue_hw_ctx;
413 unsigned int nr_hw_queues;
414
415 struct backing_dev_info *backing_dev_info;
416
417 /*
418 * The queue owner gets to use this for whatever they like.
419 * ll_rw_blk doesn't touch it.
420 */
421 void *queuedata;
422
423 /*
424 * various queue flags, see QUEUE_* below
425 */
426 unsigned long queue_flags;
427 /*
428 * Number of contexts that have called blk_set_pm_only(). If this
429 * counter is above zero then only RQF_PM requests are processed.
430 */
431 atomic_t pm_only;
432
433 /*
434 * ida allocated id for this queue. Used to index queues from
435 * ioctx.
436 */
437 int id;
438
439 /*
440 * queue needs bounce pages for pages above this limit
441 */
442 gfp_t bounce_gfp;
443
444 spinlock_t queue_lock;
445
446 /*
447 * queue kobject
448 */
449 struct kobject kobj;
450
451 /*
452 * mq queue kobject
453 */
454 struct kobject *mq_kobj;
455
456#ifdef CONFIG_BLK_DEV_INTEGRITY
457 struct blk_integrity integrity;
458#endif /* CONFIG_BLK_DEV_INTEGRITY */
459
460#ifdef CONFIG_PM
461 struct device *dev;
462 enum rpm_status rpm_status;
463#endif
464
465 /*
466 * queue settings
467 */
468 unsigned long nr_requests; /* Max # of requests */
469
470 unsigned int dma_pad_mask;
471 unsigned int dma_alignment;
472
473#ifdef CONFIG_BLK_INLINE_ENCRYPTION
474 /* Inline crypto capabilities */
475 struct blk_keyslot_manager *ksm;
476#endif
477
478 unsigned int rq_timeout;
479 int poll_nsec;
480
481 struct blk_stat_callback *poll_cb;
482 struct blk_rq_stat poll_stat[BLK_MQ_POLL_STATS_BKTS];
483
484 struct timer_list timeout;
485 struct work_struct timeout_work;
486
487 atomic_t nr_active_requests_shared_sbitmap;
488
489 struct list_head icq_list;
490#ifdef CONFIG_BLK_CGROUP
491 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
492 struct blkcg_gq *root_blkg;
493 struct list_head blkg_list;
494#endif
495
496 struct queue_limits limits;
497
498 unsigned int required_elevator_features;
499
500#ifdef CONFIG_BLK_DEV_ZONED
501 /*
502 * Zoned block device information for request dispatch control.
503 * nr_zones is the total number of zones of the device. This is always
504 * 0 for regular block devices. conv_zones_bitmap is a bitmap of nr_zones
505 * bits which indicates if a zone is conventional (bit set) or
506 * sequential (bit clear). seq_zones_wlock is a bitmap of nr_zones
507 * bits which indicates if a zone is write locked, that is, if a write
508 * request targeting the zone was dispatched. All three fields are
509 * initialized by the low level device driver (e.g. scsi/sd.c).
510 * Stacking drivers (device mappers) may or may not initialize
511 * these fields.
512 *
513 * Reads of this information must be protected with blk_queue_enter() /
514 * blk_queue_exit(). Modifying this information is only allowed while
515 * no requests are being processed. See also blk_mq_freeze_queue() and
516 * blk_mq_unfreeze_queue().
517 */
518 unsigned int nr_zones;
519 unsigned long *conv_zones_bitmap;
520 unsigned long *seq_zones_wlock;
521 unsigned int max_open_zones;
522 unsigned int max_active_zones;
523#endif /* CONFIG_BLK_DEV_ZONED */
524
525 /*
526 * sg stuff
527 */
528 unsigned int sg_timeout;
529 unsigned int sg_reserved_size;
530 int node;
531 struct mutex debugfs_mutex;
532#ifdef CONFIG_BLK_DEV_IO_TRACE
533 struct blk_trace __rcu *blk_trace;
534#endif
535 /*
536 * for flush operations
537 */
538 struct blk_flush_queue *fq;
539
540 struct list_head requeue_list;
541 spinlock_t requeue_lock;
542 struct delayed_work requeue_work;
543
544 struct mutex sysfs_lock;
545 struct mutex sysfs_dir_lock;
546
547 /*
548 * for reusing dead hctx instance in case of updating
549 * nr_hw_queues
550 */
551 struct list_head unused_hctx_list;
552 spinlock_t unused_hctx_lock;
553
554 int mq_freeze_depth;
555
556#if defined(CONFIG_BLK_DEV_BSG)
557 struct bsg_class_device bsg_dev;
558#endif
559
560#ifdef CONFIG_BLK_DEV_THROTTLING
561 /* Throttle data */
562 struct throtl_data *td;
563#endif
564 struct rcu_head rcu_head;
565 wait_queue_head_t mq_freeze_wq;
566 /*
567 * Protect concurrent access to q_usage_counter by
568 * percpu_ref_kill() and percpu_ref_reinit().
569 */
570 struct mutex mq_freeze_lock;
571
572 struct blk_mq_tag_set *tag_set;
573 struct list_head tag_set_list;
574 struct bio_set bio_split;
575
576 struct dentry *debugfs_dir;
577
578#ifdef CONFIG_BLK_DEBUG_FS
579 struct dentry *sched_debugfs_dir;
580 struct dentry *rqos_debugfs_dir;
581#endif
582
583 bool mq_sysfs_init_done;
584
585 size_t cmd_size;
586
587#define BLK_MAX_WRITE_HINTS 5
588 u64 write_hints[BLK_MAX_WRITE_HINTS];
589};
590
591/* Keep blk_queue_flag_name[] in sync with the definitions below */
592#define QUEUE_FLAG_STOPPED 0 /* queue is stopped */
593#define QUEUE_FLAG_DYING 1 /* queue being torn down */
594#define QUEUE_FLAG_NOMERGES 3 /* disable merge attempts */
595#define QUEUE_FLAG_SAME_COMP 4 /* complete on same CPU-group */
596#define QUEUE_FLAG_FAIL_IO 5 /* fake timeout */
597#define QUEUE_FLAG_NONROT 6 /* non-rotational device (SSD) */
598#define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
599#define QUEUE_FLAG_IO_STAT 7 /* do disk/partitions IO accounting */
600#define QUEUE_FLAG_DISCARD 8 /* supports DISCARD */
601#define QUEUE_FLAG_NOXMERGES 9 /* No extended merges */
602#define QUEUE_FLAG_ADD_RANDOM 10 /* Contributes to random pool */
603#define QUEUE_FLAG_SECERASE 11 /* supports secure erase */
604#define QUEUE_FLAG_SAME_FORCE 12 /* force complete on same CPU */
605#define QUEUE_FLAG_DEAD 13 /* queue tear-down finished */
606#define QUEUE_FLAG_INIT_DONE 14 /* queue is initialized */
607#define QUEUE_FLAG_STABLE_WRITES 15 /* don't modify blks until WB is done */
608#define QUEUE_FLAG_POLL 16 /* IO polling enabled if set */
609#define QUEUE_FLAG_WC 17 /* Write back caching */
610#define QUEUE_FLAG_FUA 18 /* device supports FUA writes */
611#define QUEUE_FLAG_DAX 19 /* device supports DAX */
612#define QUEUE_FLAG_STATS 20 /* track IO start and completion times */
613#define QUEUE_FLAG_POLL_STATS 21 /* collecting stats for hybrid polling */
614#define QUEUE_FLAG_REGISTERED 22 /* queue has been registered to a disk */
615#define QUEUE_FLAG_SCSI_PASSTHROUGH 23 /* queue supports SCSI commands */
616#define QUEUE_FLAG_QUIESCED 24 /* queue has been quiesced */
617#define QUEUE_FLAG_PCI_P2PDMA 25 /* device supports PCI p2p requests */
618#define QUEUE_FLAG_ZONE_RESETALL 26 /* supports Zone Reset All */
619#define QUEUE_FLAG_RQ_ALLOC_TIME 27 /* record rq->alloc_time_ns */
620#define QUEUE_FLAG_HCTX_ACTIVE 28 /* at least one blk-mq hctx is active */
621#define QUEUE_FLAG_NOWAIT 29 /* device supports NOWAIT */
622
623#define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
624 (1 << QUEUE_FLAG_SAME_COMP) | \
625 (1 << QUEUE_FLAG_NOWAIT))
626
627void blk_queue_flag_set(unsigned int flag, struct request_queue *q);
628void blk_queue_flag_clear(unsigned int flag, struct request_queue *q);
629bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q);
630
631#define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
632#define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
633#define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
634#define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
635#define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
636#define blk_queue_noxmerges(q) \
637 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
638#define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
639#define blk_queue_stable_writes(q) \
640 test_bit(QUEUE_FLAG_STABLE_WRITES, &(q)->queue_flags)
641#define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
642#define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
643#define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
644#define blk_queue_zone_resetall(q) \
645 test_bit(QUEUE_FLAG_ZONE_RESETALL, &(q)->queue_flags)
646#define blk_queue_secure_erase(q) \
647 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
648#define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
649#define blk_queue_scsi_passthrough(q) \
650 test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags)
651#define blk_queue_pci_p2pdma(q) \
652 test_bit(QUEUE_FLAG_PCI_P2PDMA, &(q)->queue_flags)
653#ifdef CONFIG_BLK_RQ_ALLOC_TIME
654#define blk_queue_rq_alloc_time(q) \
655 test_bit(QUEUE_FLAG_RQ_ALLOC_TIME, &(q)->queue_flags)
656#else
657#define blk_queue_rq_alloc_time(q) false
658#endif
659
660#define blk_noretry_request(rq) \
661 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
662 REQ_FAILFAST_DRIVER))
663#define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
664#define blk_queue_pm_only(q) atomic_read(&(q)->pm_only)
665#define blk_queue_fua(q) test_bit(QUEUE_FLAG_FUA, &(q)->queue_flags)
666#define blk_queue_registered(q) test_bit(QUEUE_FLAG_REGISTERED, &(q)->queue_flags)
667#define blk_queue_nowait(q) test_bit(QUEUE_FLAG_NOWAIT, &(q)->queue_flags)
668
669extern void blk_set_pm_only(struct request_queue *q);
670extern void blk_clear_pm_only(struct request_queue *q);
671
672static inline bool blk_account_rq(struct request *rq)
673{
674 return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq);
675}
676
677#define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
678
679#define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ)
680
681#define rq_dma_dir(rq) \
682 (op_is_write(req_op(rq)) ? DMA_TO_DEVICE : DMA_FROM_DEVICE)
683
684#define dma_map_bvec(dev, bv, dir, attrs) \
685 dma_map_page_attrs(dev, (bv)->bv_page, (bv)->bv_offset, (bv)->bv_len, \
686 (dir), (attrs))
687
688static inline bool queue_is_mq(struct request_queue *q)
689{
690 return q->mq_ops;
691}
692
693#ifdef CONFIG_PM
694static inline enum rpm_status queue_rpm_status(struct request_queue *q)
695{
696 return q->rpm_status;
697}
698#else
699static inline enum rpm_status queue_rpm_status(struct request_queue *q)
700{
701 return RPM_ACTIVE;
702}
703#endif
704
705static inline enum blk_zoned_model
706blk_queue_zoned_model(struct request_queue *q)
707{
708 if (IS_ENABLED(CONFIG_BLK_DEV_ZONED))
709 return q->limits.zoned;
710 return BLK_ZONED_NONE;
711}
712
713static inline bool blk_queue_is_zoned(struct request_queue *q)
714{
715 switch (blk_queue_zoned_model(q)) {
716 case BLK_ZONED_HA:
717 case BLK_ZONED_HM:
718 return true;
719 default:
720 return false;
721 }
722}
723
724static inline sector_t blk_queue_zone_sectors(struct request_queue *q)
725{
726 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
727}
728
729#ifdef CONFIG_BLK_DEV_ZONED
730static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
731{
732 return blk_queue_is_zoned(q) ? q->nr_zones : 0;
733}
734
735static inline unsigned int blk_queue_zone_no(struct request_queue *q,
736 sector_t sector)
737{
738 if (!blk_queue_is_zoned(q))
739 return 0;
740 return sector >> ilog2(q->limits.chunk_sectors);
741}
742
743static inline bool blk_queue_zone_is_seq(struct request_queue *q,
744 sector_t sector)
745{
746 if (!blk_queue_is_zoned(q))
747 return false;
748 if (!q->conv_zones_bitmap)
749 return true;
750 return !test_bit(blk_queue_zone_no(q, sector), q->conv_zones_bitmap);
751}
752
753static inline void blk_queue_max_open_zones(struct request_queue *q,
754 unsigned int max_open_zones)
755{
756 q->max_open_zones = max_open_zones;
757}
758
759static inline unsigned int queue_max_open_zones(const struct request_queue *q)
760{
761 return q->max_open_zones;
762}
763
764static inline void blk_queue_max_active_zones(struct request_queue *q,
765 unsigned int max_active_zones)
766{
767 q->max_active_zones = max_active_zones;
768}
769
770static inline unsigned int queue_max_active_zones(const struct request_queue *q)
771{
772 return q->max_active_zones;
773}
774#else /* CONFIG_BLK_DEV_ZONED */
775static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
776{
777 return 0;
778}
779static inline bool blk_queue_zone_is_seq(struct request_queue *q,
780 sector_t sector)
781{
782 return false;
783}
784static inline unsigned int blk_queue_zone_no(struct request_queue *q,
785 sector_t sector)
786{
787 return 0;
788}
789static inline unsigned int queue_max_open_zones(const struct request_queue *q)
790{
791 return 0;
792}
793static inline unsigned int queue_max_active_zones(const struct request_queue *q)
794{
795 return 0;
796}
797#endif /* CONFIG_BLK_DEV_ZONED */
798
799static inline bool rq_is_sync(struct request *rq)
800{
801 return op_is_sync(rq->cmd_flags);
802}
803
804static inline bool rq_mergeable(struct request *rq)
805{
806 if (blk_rq_is_passthrough(rq))
807 return false;
808
809 if (req_op(rq) == REQ_OP_FLUSH)
810 return false;
811
812 if (req_op(rq) == REQ_OP_WRITE_ZEROES)
813 return false;
814
815 if (req_op(rq) == REQ_OP_ZONE_APPEND)
816 return false;
817
818 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
819 return false;
820 if (rq->rq_flags & RQF_NOMERGE_FLAGS)
821 return false;
822
823 return true;
824}
825
826static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
827{
828 if (bio_page(a) == bio_page(b) &&
829 bio_offset(a) == bio_offset(b))
830 return true;
831
832 return false;
833}
834
835static inline unsigned int blk_queue_depth(struct request_queue *q)
836{
837 if (q->queue_depth)
838 return q->queue_depth;
839
840 return q->nr_requests;
841}
842
843extern unsigned long blk_max_low_pfn, blk_max_pfn;
844
845/*
846 * standard bounce addresses:
847 *
848 * BLK_BOUNCE_HIGH : bounce all highmem pages
849 * BLK_BOUNCE_ANY : don't bounce anything
850 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
851 */
852
853#if BITS_PER_LONG == 32
854#define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
855#else
856#define BLK_BOUNCE_HIGH -1ULL
857#endif
858#define BLK_BOUNCE_ANY (-1ULL)
859#define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
860
861/*
862 * default timeout for SG_IO if none specified
863 */
864#define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
865#define BLK_MIN_SG_TIMEOUT (7 * HZ)
866
867struct rq_map_data {
868 struct page **pages;
869 int page_order;
870 int nr_entries;
871 unsigned long offset;
872 int null_mapped;
873 int from_user;
874};
875
876struct req_iterator {
877 struct bvec_iter iter;
878 struct bio *bio;
879};
880
881/* This should not be used directly - use rq_for_each_segment */
882#define for_each_bio(_bio) \
883 for (; _bio; _bio = _bio->bi_next)
884#define __rq_for_each_bio(_bio, rq) \
885 if ((rq->bio)) \
886 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
887
888#define rq_for_each_segment(bvl, _rq, _iter) \
889 __rq_for_each_bio(_iter.bio, _rq) \
890 bio_for_each_segment(bvl, _iter.bio, _iter.iter)
891
892#define rq_for_each_bvec(bvl, _rq, _iter) \
893 __rq_for_each_bio(_iter.bio, _rq) \
894 bio_for_each_bvec(bvl, _iter.bio, _iter.iter)
895
896#define rq_iter_last(bvec, _iter) \
897 (_iter.bio->bi_next == NULL && \
898 bio_iter_last(bvec, _iter.iter))
899
900#ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
901# error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
902#endif
903#if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
904extern void rq_flush_dcache_pages(struct request *rq);
905#else
906static inline void rq_flush_dcache_pages(struct request *rq)
907{
908}
909#endif
910
911extern int blk_register_queue(struct gendisk *disk);
912extern void blk_unregister_queue(struct gendisk *disk);
913blk_qc_t submit_bio_noacct(struct bio *bio);
914extern void blk_rq_init(struct request_queue *q, struct request *rq);
915extern void blk_put_request(struct request *);
916extern struct request *blk_get_request(struct request_queue *, unsigned int op,
917 blk_mq_req_flags_t flags);
918extern int blk_lld_busy(struct request_queue *q);
919extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
920 struct bio_set *bs, gfp_t gfp_mask,
921 int (*bio_ctr)(struct bio *, struct bio *, void *),
922 void *data);
923extern void blk_rq_unprep_clone(struct request *rq);
924extern blk_status_t blk_insert_cloned_request(struct request_queue *q,
925 struct request *rq);
926extern int blk_rq_append_bio(struct request *rq, struct bio **bio);
927extern void blk_queue_split(struct bio **);
928extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
929extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
930 unsigned int, void __user *);
931extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
932 unsigned int, void __user *);
933extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
934 struct scsi_ioctl_command __user *);
935extern int get_sg_io_hdr(struct sg_io_hdr *hdr, const void __user *argp);
936extern int put_sg_io_hdr(const struct sg_io_hdr *hdr, void __user *argp);
937
938extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags);
939extern void blk_queue_exit(struct request_queue *q);
940extern void blk_sync_queue(struct request_queue *q);
941extern int blk_rq_map_user(struct request_queue *, struct request *,
942 struct rq_map_data *, void __user *, unsigned long,
943 gfp_t);
944extern int blk_rq_unmap_user(struct bio *);
945extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
946extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
947 struct rq_map_data *, const struct iov_iter *,
948 gfp_t);
949extern void blk_execute_rq(struct gendisk *, struct request *, int);
950extern void blk_execute_rq_nowait(struct gendisk *,
951 struct request *, int, rq_end_io_fn *);
952
953/* Helper to convert REQ_OP_XXX to its string format XXX */
954extern const char *blk_op_str(unsigned int op);
955
956int blk_status_to_errno(blk_status_t status);
957blk_status_t errno_to_blk_status(int errno);
958
959int blk_poll(struct request_queue *q, blk_qc_t cookie, bool spin);
960
961static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
962{
963 return bdev->bd_disk->queue; /* this is never NULL */
964}
965
966/*
967 * The basic unit of block I/O is a sector. It is used in a number of contexts
968 * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9
969 * bytes. Variables of type sector_t represent an offset or size that is a
970 * multiple of 512 bytes. Hence these two constants.
971 */
972#ifndef SECTOR_SHIFT
973#define SECTOR_SHIFT 9
974#endif
975#ifndef SECTOR_SIZE
976#define SECTOR_SIZE (1 << SECTOR_SHIFT)
977#endif
978
979/*
980 * blk_rq_pos() : the current sector
981 * blk_rq_bytes() : bytes left in the entire request
982 * blk_rq_cur_bytes() : bytes left in the current segment
983 * blk_rq_err_bytes() : bytes left till the next error boundary
984 * blk_rq_sectors() : sectors left in the entire request
985 * blk_rq_cur_sectors() : sectors left in the current segment
986 * blk_rq_stats_sectors() : sectors of the entire request used for stats
987 */
988static inline sector_t blk_rq_pos(const struct request *rq)
989{
990 return rq->__sector;
991}
992
993static inline unsigned int blk_rq_bytes(const struct request *rq)
994{
995 return rq->__data_len;
996}
997
998static inline int blk_rq_cur_bytes(const struct request *rq)
999{
1000 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
1001}
1002
1003extern unsigned int blk_rq_err_bytes(const struct request *rq);
1004
1005static inline unsigned int blk_rq_sectors(const struct request *rq)
1006{
1007 return blk_rq_bytes(rq) >> SECTOR_SHIFT;
1008}
1009
1010static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
1011{
1012 return blk_rq_cur_bytes(rq) >> SECTOR_SHIFT;
1013}
1014
1015static inline unsigned int blk_rq_stats_sectors(const struct request *rq)
1016{
1017 return rq->stats_sectors;
1018}
1019
1020#ifdef CONFIG_BLK_DEV_ZONED
1021
1022/* Helper to convert BLK_ZONE_ZONE_XXX to its string format XXX */
1023const char *blk_zone_cond_str(enum blk_zone_cond zone_cond);
1024
1025static inline unsigned int blk_rq_zone_no(struct request *rq)
1026{
1027 return blk_queue_zone_no(rq->q, blk_rq_pos(rq));
1028}
1029
1030static inline unsigned int blk_rq_zone_is_seq(struct request *rq)
1031{
1032 return blk_queue_zone_is_seq(rq->q, blk_rq_pos(rq));
1033}
1034#endif /* CONFIG_BLK_DEV_ZONED */
1035
1036/*
1037 * Some commands like WRITE SAME have a payload or data transfer size which
1038 * is different from the size of the request. Any driver that supports such
1039 * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to
1040 * calculate the data transfer size.
1041 */
1042static inline unsigned int blk_rq_payload_bytes(struct request *rq)
1043{
1044 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1045 return rq->special_vec.bv_len;
1046 return blk_rq_bytes(rq);
1047}
1048
1049/*
1050 * Return the first full biovec in the request. The caller needs to check that
1051 * there are any bvecs before calling this helper.
1052 */
1053static inline struct bio_vec req_bvec(struct request *rq)
1054{
1055 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1056 return rq->special_vec;
1057 return mp_bvec_iter_bvec(rq->bio->bi_io_vec, rq->bio->bi_iter);
1058}
1059
1060static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
1061 int op)
1062{
1063 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
1064 return min(q->limits.max_discard_sectors,
1065 UINT_MAX >> SECTOR_SHIFT);
1066
1067 if (unlikely(op == REQ_OP_WRITE_SAME))
1068 return q->limits.max_write_same_sectors;
1069
1070 if (unlikely(op == REQ_OP_WRITE_ZEROES))
1071 return q->limits.max_write_zeroes_sectors;
1072
1073 return q->limits.max_sectors;
1074}
1075
1076/*
1077 * Return maximum size of a request at given offset. Only valid for
1078 * file system requests.
1079 */
1080static inline unsigned int blk_max_size_offset(struct request_queue *q,
1081 sector_t offset,
1082 unsigned int chunk_sectors)
1083{
1084 if (!chunk_sectors) {
1085 if (q->limits.chunk_sectors)
1086 chunk_sectors = q->limits.chunk_sectors;
1087 else
1088 return q->limits.max_sectors;
1089 }
1090
1091 if (likely(is_power_of_2(chunk_sectors)))
1092 chunk_sectors -= offset & (chunk_sectors - 1);
1093 else
1094 chunk_sectors -= sector_div(offset, chunk_sectors);
1095
1096 return min(q->limits.max_sectors, chunk_sectors);
1097}
1098
1099static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1100 sector_t offset)
1101{
1102 struct request_queue *q = rq->q;
1103
1104 if (blk_rq_is_passthrough(rq))
1105 return q->limits.max_hw_sectors;
1106
1107 if (!q->limits.chunk_sectors ||
1108 req_op(rq) == REQ_OP_DISCARD ||
1109 req_op(rq) == REQ_OP_SECURE_ERASE)
1110 return blk_queue_get_max_sectors(q, req_op(rq));
1111
1112 return min(blk_max_size_offset(q, offset, 0),
1113 blk_queue_get_max_sectors(q, req_op(rq)));
1114}
1115
1116static inline unsigned int blk_rq_count_bios(struct request *rq)
1117{
1118 unsigned int nr_bios = 0;
1119 struct bio *bio;
1120
1121 __rq_for_each_bio(bio, rq)
1122 nr_bios++;
1123
1124 return nr_bios;
1125}
1126
1127void blk_steal_bios(struct bio_list *list, struct request *rq);
1128
1129/*
1130 * Request completion related functions.
1131 *
1132 * blk_update_request() completes given number of bytes and updates
1133 * the request without completing it.
1134 */
1135extern bool blk_update_request(struct request *rq, blk_status_t error,
1136 unsigned int nr_bytes);
1137
1138extern void blk_abort_request(struct request *);
1139
1140/*
1141 * Access functions for manipulating queue properties
1142 */
1143extern void blk_cleanup_queue(struct request_queue *);
1144extern void blk_queue_bounce_limit(struct request_queue *, u64);
1145extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1146extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1147extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1148extern void blk_queue_max_discard_segments(struct request_queue *,
1149 unsigned short);
1150extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1151extern void blk_queue_max_discard_sectors(struct request_queue *q,
1152 unsigned int max_discard_sectors);
1153extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1154 unsigned int max_write_same_sectors);
1155extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1156 unsigned int max_write_same_sectors);
1157extern void blk_queue_logical_block_size(struct request_queue *, unsigned int);
1158extern void blk_queue_max_zone_append_sectors(struct request_queue *q,
1159 unsigned int max_zone_append_sectors);
1160extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1161void blk_queue_zone_write_granularity(struct request_queue *q,
1162 unsigned int size);
1163extern void blk_queue_alignment_offset(struct request_queue *q,
1164 unsigned int alignment);
1165void blk_queue_update_readahead(struct request_queue *q);
1166extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1167extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1168extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1169extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1170extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1171extern void blk_set_default_limits(struct queue_limits *lim);
1172extern void blk_set_stacking_limits(struct queue_limits *lim);
1173extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1174 sector_t offset);
1175extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1176 sector_t offset);
1177extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1178extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1179extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1180extern void blk_queue_dma_alignment(struct request_queue *, int);
1181extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1182extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1183extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1184extern void blk_queue_required_elevator_features(struct request_queue *q,
1185 unsigned int features);
1186extern bool blk_queue_can_use_dma_map_merging(struct request_queue *q,
1187 struct device *dev);
1188
1189/*
1190 * Number of physical segments as sent to the device.
1191 *
1192 * Normally this is the number of discontiguous data segments sent by the
1193 * submitter. But for data-less command like discard we might have no
1194 * actual data segments submitted, but the driver might have to add it's
1195 * own special payload. In that case we still return 1 here so that this
1196 * special payload will be mapped.
1197 */
1198static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1199{
1200 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1201 return 1;
1202 return rq->nr_phys_segments;
1203}
1204
1205/*
1206 * Number of discard segments (or ranges) the driver needs to fill in.
1207 * Each discard bio merged into a request is counted as one segment.
1208 */
1209static inline unsigned short blk_rq_nr_discard_segments(struct request *rq)
1210{
1211 return max_t(unsigned short, rq->nr_phys_segments, 1);
1212}
1213
1214int __blk_rq_map_sg(struct request_queue *q, struct request *rq,
1215 struct scatterlist *sglist, struct scatterlist **last_sg);
1216static inline int blk_rq_map_sg(struct request_queue *q, struct request *rq,
1217 struct scatterlist *sglist)
1218{
1219 struct scatterlist *last_sg = NULL;
1220
1221 return __blk_rq_map_sg(q, rq, sglist, &last_sg);
1222}
1223extern void blk_dump_rq_flags(struct request *, char *);
1224
1225bool __must_check blk_get_queue(struct request_queue *);
1226struct request_queue *blk_alloc_queue(int node_id);
1227extern void blk_put_queue(struct request_queue *);
1228extern void blk_set_queue_dying(struct request_queue *);
1229
1230#ifdef CONFIG_BLOCK
1231/*
1232 * blk_plug permits building a queue of related requests by holding the I/O
1233 * fragments for a short period. This allows merging of sequential requests
1234 * into single larger request. As the requests are moved from a per-task list to
1235 * the device's request_queue in a batch, this results in improved scalability
1236 * as the lock contention for request_queue lock is reduced.
1237 *
1238 * It is ok not to disable preemption when adding the request to the plug list
1239 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1240 * the plug list when the task sleeps by itself. For details, please see
1241 * schedule() where blk_schedule_flush_plug() is called.
1242 */
1243struct blk_plug {
1244 struct list_head mq_list; /* blk-mq requests */
1245 struct list_head cb_list; /* md requires an unplug callback */
1246 unsigned short rq_count;
1247 bool multiple_queues;
1248 bool nowait;
1249};
1250#define BLK_MAX_REQUEST_COUNT 16
1251#define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1252
1253struct blk_plug_cb;
1254typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1255struct blk_plug_cb {
1256 struct list_head list;
1257 blk_plug_cb_fn callback;
1258 void *data;
1259};
1260extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1261 void *data, int size);
1262extern void blk_start_plug(struct blk_plug *);
1263extern void blk_finish_plug(struct blk_plug *);
1264extern void blk_flush_plug_list(struct blk_plug *, bool);
1265
1266static inline void blk_flush_plug(struct task_struct *tsk)
1267{
1268 struct blk_plug *plug = tsk->plug;
1269
1270 if (plug)
1271 blk_flush_plug_list(plug, false);
1272}
1273
1274static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1275{
1276 struct blk_plug *plug = tsk->plug;
1277
1278 if (plug)
1279 blk_flush_plug_list(plug, true);
1280}
1281
1282static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1283{
1284 struct blk_plug *plug = tsk->plug;
1285
1286 return plug &&
1287 (!list_empty(&plug->mq_list) ||
1288 !list_empty(&plug->cb_list));
1289}
1290
1291int blkdev_issue_flush(struct block_device *bdev);
1292long nr_blockdev_pages(void);
1293#else /* CONFIG_BLOCK */
1294struct blk_plug {
1295};
1296
1297static inline void blk_start_plug(struct blk_plug *plug)
1298{
1299}
1300
1301static inline void blk_finish_plug(struct blk_plug *plug)
1302{
1303}
1304
1305static inline void blk_flush_plug(struct task_struct *task)
1306{
1307}
1308
1309static inline void blk_schedule_flush_plug(struct task_struct *task)
1310{
1311}
1312
1313
1314static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1315{
1316 return false;
1317}
1318
1319static inline int blkdev_issue_flush(struct block_device *bdev)
1320{
1321 return 0;
1322}
1323
1324static inline long nr_blockdev_pages(void)
1325{
1326 return 0;
1327}
1328#endif /* CONFIG_BLOCK */
1329
1330extern void blk_io_schedule(void);
1331
1332extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1333 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1334
1335#define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */
1336
1337extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1338 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1339extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1340 sector_t nr_sects, gfp_t gfp_mask, int flags,
1341 struct bio **biop);
1342
1343#define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */
1344#define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */
1345
1346extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1347 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1348 unsigned flags);
1349extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1350 sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
1351
1352static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1353 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1354{
1355 return blkdev_issue_discard(sb->s_bdev,
1356 block << (sb->s_blocksize_bits -
1357 SECTOR_SHIFT),
1358 nr_blocks << (sb->s_blocksize_bits -
1359 SECTOR_SHIFT),
1360 gfp_mask, flags);
1361}
1362static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1363 sector_t nr_blocks, gfp_t gfp_mask)
1364{
1365 return blkdev_issue_zeroout(sb->s_bdev,
1366 block << (sb->s_blocksize_bits -
1367 SECTOR_SHIFT),
1368 nr_blocks << (sb->s_blocksize_bits -
1369 SECTOR_SHIFT),
1370 gfp_mask, 0);
1371}
1372
1373extern int blk_verify_command(unsigned char *cmd, fmode_t mode);
1374
1375static inline bool bdev_is_partition(struct block_device *bdev)
1376{
1377 return bdev->bd_partno;
1378}
1379
1380enum blk_default_limits {
1381 BLK_MAX_SEGMENTS = 128,
1382 BLK_SAFE_MAX_SECTORS = 255,
1383 BLK_DEF_MAX_SECTORS = 2560,
1384 BLK_MAX_SEGMENT_SIZE = 65536,
1385 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1386};
1387
1388static inline unsigned long queue_segment_boundary(const struct request_queue *q)
1389{
1390 return q->limits.seg_boundary_mask;
1391}
1392
1393static inline unsigned long queue_virt_boundary(const struct request_queue *q)
1394{
1395 return q->limits.virt_boundary_mask;
1396}
1397
1398static inline unsigned int queue_max_sectors(const struct request_queue *q)
1399{
1400 return q->limits.max_sectors;
1401}
1402
1403static inline unsigned int queue_max_hw_sectors(const struct request_queue *q)
1404{
1405 return q->limits.max_hw_sectors;
1406}
1407
1408static inline unsigned short queue_max_segments(const struct request_queue *q)
1409{
1410 return q->limits.max_segments;
1411}
1412
1413static inline unsigned short queue_max_discard_segments(const struct request_queue *q)
1414{
1415 return q->limits.max_discard_segments;
1416}
1417
1418static inline unsigned int queue_max_segment_size(const struct request_queue *q)
1419{
1420 return q->limits.max_segment_size;
1421}
1422
1423static inline unsigned int queue_max_zone_append_sectors(const struct request_queue *q)
1424{
1425
1426 const struct queue_limits *l = &q->limits;
1427
1428 return min(l->max_zone_append_sectors, l->max_sectors);
1429}
1430
1431static inline unsigned queue_logical_block_size(const struct request_queue *q)
1432{
1433 int retval = 512;
1434
1435 if (q && q->limits.logical_block_size)
1436 retval = q->limits.logical_block_size;
1437
1438 return retval;
1439}
1440
1441static inline unsigned int bdev_logical_block_size(struct block_device *bdev)
1442{
1443 return queue_logical_block_size(bdev_get_queue(bdev));
1444}
1445
1446static inline unsigned int queue_physical_block_size(const struct request_queue *q)
1447{
1448 return q->limits.physical_block_size;
1449}
1450
1451static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1452{
1453 return queue_physical_block_size(bdev_get_queue(bdev));
1454}
1455
1456static inline unsigned int queue_io_min(const struct request_queue *q)
1457{
1458 return q->limits.io_min;
1459}
1460
1461static inline int bdev_io_min(struct block_device *bdev)
1462{
1463 return queue_io_min(bdev_get_queue(bdev));
1464}
1465
1466static inline unsigned int queue_io_opt(const struct request_queue *q)
1467{
1468 return q->limits.io_opt;
1469}
1470
1471static inline int bdev_io_opt(struct block_device *bdev)
1472{
1473 return queue_io_opt(bdev_get_queue(bdev));
1474}
1475
1476static inline unsigned int
1477queue_zone_write_granularity(const struct request_queue *q)
1478{
1479 return q->limits.zone_write_granularity;
1480}
1481
1482static inline unsigned int
1483bdev_zone_write_granularity(struct block_device *bdev)
1484{
1485 return queue_zone_write_granularity(bdev_get_queue(bdev));
1486}
1487
1488static inline int queue_alignment_offset(const struct request_queue *q)
1489{
1490 if (q->limits.misaligned)
1491 return -1;
1492
1493 return q->limits.alignment_offset;
1494}
1495
1496static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1497{
1498 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1499 unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT)
1500 << SECTOR_SHIFT;
1501
1502 return (granularity + lim->alignment_offset - alignment) % granularity;
1503}
1504
1505static inline int bdev_alignment_offset(struct block_device *bdev)
1506{
1507 struct request_queue *q = bdev_get_queue(bdev);
1508
1509 if (q->limits.misaligned)
1510 return -1;
1511 if (bdev_is_partition(bdev))
1512 return queue_limit_alignment_offset(&q->limits,
1513 bdev->bd_start_sect);
1514 return q->limits.alignment_offset;
1515}
1516
1517static inline int queue_discard_alignment(const struct request_queue *q)
1518{
1519 if (q->limits.discard_misaligned)
1520 return -1;
1521
1522 return q->limits.discard_alignment;
1523}
1524
1525static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1526{
1527 unsigned int alignment, granularity, offset;
1528
1529 if (!lim->max_discard_sectors)
1530 return 0;
1531
1532 /* Why are these in bytes, not sectors? */
1533 alignment = lim->discard_alignment >> SECTOR_SHIFT;
1534 granularity = lim->discard_granularity >> SECTOR_SHIFT;
1535 if (!granularity)
1536 return 0;
1537
1538 /* Offset of the partition start in 'granularity' sectors */
1539 offset = sector_div(sector, granularity);
1540
1541 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1542 offset = (granularity + alignment - offset) % granularity;
1543
1544 /* Turn it back into bytes, gaah */
1545 return offset << SECTOR_SHIFT;
1546}
1547
1548static inline int bdev_discard_alignment(struct block_device *bdev)
1549{
1550 struct request_queue *q = bdev_get_queue(bdev);
1551
1552 if (bdev_is_partition(bdev))
1553 return queue_limit_discard_alignment(&q->limits,
1554 bdev->bd_start_sect);
1555 return q->limits.discard_alignment;
1556}
1557
1558static inline unsigned int bdev_write_same(struct block_device *bdev)
1559{
1560 struct request_queue *q = bdev_get_queue(bdev);
1561
1562 if (q)
1563 return q->limits.max_write_same_sectors;
1564
1565 return 0;
1566}
1567
1568static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1569{
1570 struct request_queue *q = bdev_get_queue(bdev);
1571
1572 if (q)
1573 return q->limits.max_write_zeroes_sectors;
1574
1575 return 0;
1576}
1577
1578static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1579{
1580 struct request_queue *q = bdev_get_queue(bdev);
1581
1582 if (q)
1583 return blk_queue_zoned_model(q);
1584
1585 return BLK_ZONED_NONE;
1586}
1587
1588static inline bool bdev_is_zoned(struct block_device *bdev)
1589{
1590 struct request_queue *q = bdev_get_queue(bdev);
1591
1592 if (q)
1593 return blk_queue_is_zoned(q);
1594
1595 return false;
1596}
1597
1598static inline sector_t bdev_zone_sectors(struct block_device *bdev)
1599{
1600 struct request_queue *q = bdev_get_queue(bdev);
1601
1602 if (q)
1603 return blk_queue_zone_sectors(q);
1604 return 0;
1605}
1606
1607static inline unsigned int bdev_max_open_zones(struct block_device *bdev)
1608{
1609 struct request_queue *q = bdev_get_queue(bdev);
1610
1611 if (q)
1612 return queue_max_open_zones(q);
1613 return 0;
1614}
1615
1616static inline unsigned int bdev_max_active_zones(struct block_device *bdev)
1617{
1618 struct request_queue *q = bdev_get_queue(bdev);
1619
1620 if (q)
1621 return queue_max_active_zones(q);
1622 return 0;
1623}
1624
1625static inline int queue_dma_alignment(const struct request_queue *q)
1626{
1627 return q ? q->dma_alignment : 511;
1628}
1629
1630static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1631 unsigned int len)
1632{
1633 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1634 return !(addr & alignment) && !(len & alignment);
1635}
1636
1637/* assumes size > 256 */
1638static inline unsigned int blksize_bits(unsigned int size)
1639{
1640 unsigned int bits = 8;
1641 do {
1642 bits++;
1643 size >>= 1;
1644 } while (size > 256);
1645 return bits;
1646}
1647
1648static inline unsigned int block_size(struct block_device *bdev)
1649{
1650 return 1 << bdev->bd_inode->i_blkbits;
1651}
1652
1653int kblockd_schedule_work(struct work_struct *work);
1654int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1655
1656#define MODULE_ALIAS_BLOCKDEV(major,minor) \
1657 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1658#define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1659 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1660
1661#if defined(CONFIG_BLK_DEV_INTEGRITY)
1662
1663enum blk_integrity_flags {
1664 BLK_INTEGRITY_VERIFY = 1 << 0,
1665 BLK_INTEGRITY_GENERATE = 1 << 1,
1666 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2,
1667 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3,
1668};
1669
1670struct blk_integrity_iter {
1671 void *prot_buf;
1672 void *data_buf;
1673 sector_t seed;
1674 unsigned int data_size;
1675 unsigned short interval;
1676 const char *disk_name;
1677};
1678
1679typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *);
1680typedef void (integrity_prepare_fn) (struct request *);
1681typedef void (integrity_complete_fn) (struct request *, unsigned int);
1682
1683struct blk_integrity_profile {
1684 integrity_processing_fn *generate_fn;
1685 integrity_processing_fn *verify_fn;
1686 integrity_prepare_fn *prepare_fn;
1687 integrity_complete_fn *complete_fn;
1688 const char *name;
1689};
1690
1691extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1692extern void blk_integrity_unregister(struct gendisk *);
1693extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1694extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1695 struct scatterlist *);
1696extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1697
1698static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1699{
1700 struct blk_integrity *bi = &disk->queue->integrity;
1701
1702 if (!bi->profile)
1703 return NULL;
1704
1705 return bi;
1706}
1707
1708static inline
1709struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1710{
1711 return blk_get_integrity(bdev->bd_disk);
1712}
1713
1714static inline bool
1715blk_integrity_queue_supports_integrity(struct request_queue *q)
1716{
1717 return q->integrity.profile;
1718}
1719
1720static inline bool blk_integrity_rq(struct request *rq)
1721{
1722 return rq->cmd_flags & REQ_INTEGRITY;
1723}
1724
1725static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1726 unsigned int segs)
1727{
1728 q->limits.max_integrity_segments = segs;
1729}
1730
1731static inline unsigned short
1732queue_max_integrity_segments(const struct request_queue *q)
1733{
1734 return q->limits.max_integrity_segments;
1735}
1736
1737/**
1738 * bio_integrity_intervals - Return number of integrity intervals for a bio
1739 * @bi: blk_integrity profile for device
1740 * @sectors: Size of the bio in 512-byte sectors
1741 *
1742 * Description: The block layer calculates everything in 512 byte
1743 * sectors but integrity metadata is done in terms of the data integrity
1744 * interval size of the storage device. Convert the block layer sectors
1745 * to the appropriate number of integrity intervals.
1746 */
1747static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi,
1748 unsigned int sectors)
1749{
1750 return sectors >> (bi->interval_exp - 9);
1751}
1752
1753static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
1754 unsigned int sectors)
1755{
1756 return bio_integrity_intervals(bi, sectors) * bi->tuple_size;
1757}
1758
1759/*
1760 * Return the first bvec that contains integrity data. Only drivers that are
1761 * limited to a single integrity segment should use this helper.
1762 */
1763static inline struct bio_vec *rq_integrity_vec(struct request *rq)
1764{
1765 if (WARN_ON_ONCE(queue_max_integrity_segments(rq->q) > 1))
1766 return NULL;
1767 return rq->bio->bi_integrity->bip_vec;
1768}
1769
1770#else /* CONFIG_BLK_DEV_INTEGRITY */
1771
1772struct bio;
1773struct block_device;
1774struct gendisk;
1775struct blk_integrity;
1776
1777static inline int blk_integrity_rq(struct request *rq)
1778{
1779 return 0;
1780}
1781static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1782 struct bio *b)
1783{
1784 return 0;
1785}
1786static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1787 struct bio *b,
1788 struct scatterlist *s)
1789{
1790 return 0;
1791}
1792static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1793{
1794 return NULL;
1795}
1796static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1797{
1798 return NULL;
1799}
1800static inline bool
1801blk_integrity_queue_supports_integrity(struct request_queue *q)
1802{
1803 return false;
1804}
1805static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1806{
1807 return 0;
1808}
1809static inline void blk_integrity_register(struct gendisk *d,
1810 struct blk_integrity *b)
1811{
1812}
1813static inline void blk_integrity_unregister(struct gendisk *d)
1814{
1815}
1816static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1817 unsigned int segs)
1818{
1819}
1820static inline unsigned short queue_max_integrity_segments(const struct request_queue *q)
1821{
1822 return 0;
1823}
1824
1825static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi,
1826 unsigned int sectors)
1827{
1828 return 0;
1829}
1830
1831static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
1832 unsigned int sectors)
1833{
1834 return 0;
1835}
1836
1837static inline struct bio_vec *rq_integrity_vec(struct request *rq)
1838{
1839 return NULL;
1840}
1841
1842#endif /* CONFIG_BLK_DEV_INTEGRITY */
1843
1844#ifdef CONFIG_BLK_INLINE_ENCRYPTION
1845
1846bool blk_ksm_register(struct blk_keyslot_manager *ksm, struct request_queue *q);
1847
1848void blk_ksm_unregister(struct request_queue *q);
1849
1850#else /* CONFIG_BLK_INLINE_ENCRYPTION */
1851
1852static inline bool blk_ksm_register(struct blk_keyslot_manager *ksm,
1853 struct request_queue *q)
1854{
1855 return true;
1856}
1857
1858static inline void blk_ksm_unregister(struct request_queue *q) { }
1859
1860#endif /* CONFIG_BLK_INLINE_ENCRYPTION */
1861
1862
1863struct block_device_operations {
1864 blk_qc_t (*submit_bio) (struct bio *bio);
1865 int (*open) (struct block_device *, fmode_t);
1866 void (*release) (struct gendisk *, fmode_t);
1867 int (*rw_page)(struct block_device *, sector_t, struct page *, unsigned int);
1868 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1869 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1870 unsigned int (*check_events) (struct gendisk *disk,
1871 unsigned int clearing);
1872 void (*unlock_native_capacity) (struct gendisk *);
1873 int (*revalidate_disk) (struct gendisk *);
1874 int (*getgeo)(struct block_device *, struct hd_geometry *);
1875 int (*set_read_only)(struct block_device *bdev, bool ro);
1876 /* this callback is with swap_lock and sometimes page table lock held */
1877 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1878 int (*report_zones)(struct gendisk *, sector_t sector,
1879 unsigned int nr_zones, report_zones_cb cb, void *data);
1880 char *(*devnode)(struct gendisk *disk, umode_t *mode);
1881 struct module *owner;
1882 const struct pr_ops *pr_ops;
1883};
1884
1885#ifdef CONFIG_COMPAT
1886extern int blkdev_compat_ptr_ioctl(struct block_device *, fmode_t,
1887 unsigned int, unsigned long);
1888#else
1889#define blkdev_compat_ptr_ioctl NULL
1890#endif
1891
1892extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1893extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1894 struct writeback_control *);
1895
1896#ifdef CONFIG_BLK_DEV_ZONED
1897bool blk_req_needs_zone_write_lock(struct request *rq);
1898bool blk_req_zone_write_trylock(struct request *rq);
1899void __blk_req_zone_write_lock(struct request *rq);
1900void __blk_req_zone_write_unlock(struct request *rq);
1901
1902static inline void blk_req_zone_write_lock(struct request *rq)
1903{
1904 if (blk_req_needs_zone_write_lock(rq))
1905 __blk_req_zone_write_lock(rq);
1906}
1907
1908static inline void blk_req_zone_write_unlock(struct request *rq)
1909{
1910 if (rq->rq_flags & RQF_ZONE_WRITE_LOCKED)
1911 __blk_req_zone_write_unlock(rq);
1912}
1913
1914static inline bool blk_req_zone_is_write_locked(struct request *rq)
1915{
1916 return rq->q->seq_zones_wlock &&
1917 test_bit(blk_rq_zone_no(rq), rq->q->seq_zones_wlock);
1918}
1919
1920static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
1921{
1922 if (!blk_req_needs_zone_write_lock(rq))
1923 return true;
1924 return !blk_req_zone_is_write_locked(rq);
1925}
1926#else
1927static inline bool blk_req_needs_zone_write_lock(struct request *rq)
1928{
1929 return false;
1930}
1931
1932static inline void blk_req_zone_write_lock(struct request *rq)
1933{
1934}
1935
1936static inline void blk_req_zone_write_unlock(struct request *rq)
1937{
1938}
1939static inline bool blk_req_zone_is_write_locked(struct request *rq)
1940{
1941 return false;
1942}
1943
1944static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
1945{
1946 return true;
1947}
1948#endif /* CONFIG_BLK_DEV_ZONED */
1949
1950static inline void blk_wake_io_task(struct task_struct *waiter)
1951{
1952 /*
1953 * If we're polling, the task itself is doing the completions. For
1954 * that case, we don't need to signal a wakeup, it's enough to just
1955 * mark us as RUNNING.
1956 */
1957 if (waiter == current)
1958 __set_current_state(TASK_RUNNING);
1959 else
1960 wake_up_process(waiter);
1961}
1962
1963unsigned long disk_start_io_acct(struct gendisk *disk, unsigned int sectors,
1964 unsigned int op);
1965void disk_end_io_acct(struct gendisk *disk, unsigned int op,
1966 unsigned long start_time);
1967
1968unsigned long bio_start_io_acct(struct bio *bio);
1969void bio_end_io_acct_remapped(struct bio *bio, unsigned long start_time,
1970 struct block_device *orig_bdev);
1971
1972/**
1973 * bio_end_io_acct - end I/O accounting for bio based drivers
1974 * @bio: bio to end account for
1975 * @start: start time returned by bio_start_io_acct()
1976 */
1977static inline void bio_end_io_acct(struct bio *bio, unsigned long start_time)
1978{
1979 return bio_end_io_acct_remapped(bio, start_time, bio->bi_bdev);
1980}
1981
1982int bdev_read_only(struct block_device *bdev);
1983int set_blocksize(struct block_device *bdev, int size);
1984
1985const char *bdevname(struct block_device *bdev, char *buffer);
1986int lookup_bdev(const char *pathname, dev_t *dev);
1987
1988void blkdev_show(struct seq_file *seqf, off_t offset);
1989
1990#define BDEVNAME_SIZE 32 /* Largest string for a blockdev identifier */
1991#define BDEVT_SIZE 10 /* Largest string for MAJ:MIN for blkdev */
1992#ifdef CONFIG_BLOCK
1993#define BLKDEV_MAJOR_MAX 512
1994#else
1995#define BLKDEV_MAJOR_MAX 0
1996#endif
1997
1998struct block_device *blkdev_get_by_path(const char *path, fmode_t mode,
1999 void *holder);
2000struct block_device *blkdev_get_by_dev(dev_t dev, fmode_t mode, void *holder);
2001int bd_prepare_to_claim(struct block_device *bdev, void *holder);
2002void bd_abort_claiming(struct block_device *bdev, void *holder);
2003void blkdev_put(struct block_device *bdev, fmode_t mode);
2004
2005/* just for blk-cgroup, don't use elsewhere */
2006struct block_device *blkdev_get_no_open(dev_t dev);
2007void blkdev_put_no_open(struct block_device *bdev);
2008
2009struct block_device *bdev_alloc(struct gendisk *disk, u8 partno);
2010void bdev_add(struct block_device *bdev, dev_t dev);
2011struct block_device *I_BDEV(struct inode *inode);
2012struct block_device *bdgrab(struct block_device *bdev);
2013void bdput(struct block_device *);
2014int truncate_bdev_range(struct block_device *bdev, fmode_t mode, loff_t lstart,
2015 loff_t lend);
2016
2017#ifdef CONFIG_BLOCK
2018void invalidate_bdev(struct block_device *bdev);
2019int sync_blockdev(struct block_device *bdev);
2020#else
2021static inline void invalidate_bdev(struct block_device *bdev)
2022{
2023}
2024static inline int sync_blockdev(struct block_device *bdev)
2025{
2026 return 0;
2027}
2028#endif
2029int fsync_bdev(struct block_device *bdev);
2030
2031int freeze_bdev(struct block_device *bdev);
2032int thaw_bdev(struct block_device *bdev);
2033
2034#endif /* _LINUX_BLKDEV_H */