]>
Commit | Line | Data |
---|---|---|
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 | #include <linux/seqlock.h> | |
31 | #include <linux/u64_stats_sync.h> | |
32 | ||
33 | struct module; | |
34 | struct scsi_ioctl_command; | |
35 | ||
36 | struct request_queue; | |
37 | struct elevator_queue; | |
38 | struct blk_trace; | |
39 | struct request; | |
40 | struct sg_io_hdr; | |
41 | struct bsg_job; | |
42 | struct blkcg_gq; | |
43 | struct blk_flush_queue; | |
44 | struct pr_ops; | |
45 | struct rq_wb; | |
46 | struct blk_queue_stats; | |
47 | struct blk_stat_callback; | |
48 | ||
49 | #define BLKDEV_MIN_RQ 4 | |
50 | #define BLKDEV_MAX_RQ 128 /* Default maximum */ | |
51 | ||
52 | /* Must be consistent with blk_mq_poll_stats_bkt() */ | |
53 | #define BLK_MQ_POLL_STATS_BKTS 16 | |
54 | ||
55 | /* | |
56 | * Maximum number of blkcg policies allowed to be registered concurrently. | |
57 | * Defined here to simplify include dependency. | |
58 | */ | |
59 | #define BLKCG_MAX_POLS 3 | |
60 | ||
61 | typedef void (rq_end_io_fn)(struct request *, blk_status_t); | |
62 | ||
63 | #define BLK_RL_SYNCFULL (1U << 0) | |
64 | #define BLK_RL_ASYNCFULL (1U << 1) | |
65 | ||
66 | struct request_list { | |
67 | struct request_queue *q; /* the queue this rl belongs to */ | |
68 | #ifdef CONFIG_BLK_CGROUP | |
69 | struct blkcg_gq *blkg; /* blkg this request pool belongs to */ | |
70 | #endif | |
71 | /* | |
72 | * count[], starved[], and wait[] are indexed by | |
73 | * BLK_RW_SYNC/BLK_RW_ASYNC | |
74 | */ | |
75 | int count[2]; | |
76 | int starved[2]; | |
77 | mempool_t *rq_pool; | |
78 | wait_queue_head_t wait[2]; | |
79 | unsigned int flags; | |
80 | }; | |
81 | ||
82 | /* | |
83 | * request flags */ | |
84 | typedef __u32 __bitwise req_flags_t; | |
85 | ||
86 | /* elevator knows about this request */ | |
87 | #define RQF_SORTED ((__force req_flags_t)(1 << 0)) | |
88 | /* drive already may have started this one */ | |
89 | #define RQF_STARTED ((__force req_flags_t)(1 << 1)) | |
90 | /* uses tagged queueing */ | |
91 | #define RQF_QUEUED ((__force req_flags_t)(1 << 2)) | |
92 | /* may not be passed by ioscheduler */ | |
93 | #define RQF_SOFTBARRIER ((__force req_flags_t)(1 << 3)) | |
94 | /* request for flush sequence */ | |
95 | #define RQF_FLUSH_SEQ ((__force req_flags_t)(1 << 4)) | |
96 | /* merge of different types, fail separately */ | |
97 | #define RQF_MIXED_MERGE ((__force req_flags_t)(1 << 5)) | |
98 | /* track inflight for MQ */ | |
99 | #define RQF_MQ_INFLIGHT ((__force req_flags_t)(1 << 6)) | |
100 | /* don't call prep for this one */ | |
101 | #define RQF_DONTPREP ((__force req_flags_t)(1 << 7)) | |
102 | /* set for "ide_preempt" requests and also for requests for which the SCSI | |
103 | "quiesce" state must be ignored. */ | |
104 | #define RQF_PREEMPT ((__force req_flags_t)(1 << 8)) | |
105 | /* contains copies of user pages */ | |
106 | #define RQF_COPY_USER ((__force req_flags_t)(1 << 9)) | |
107 | /* vaguely specified driver internal error. Ignored by the block layer */ | |
108 | #define RQF_FAILED ((__force req_flags_t)(1 << 10)) | |
109 | /* don't warn about errors */ | |
110 | #define RQF_QUIET ((__force req_flags_t)(1 << 11)) | |
111 | /* elevator private data attached */ | |
112 | #define RQF_ELVPRIV ((__force req_flags_t)(1 << 12)) | |
113 | /* account I/O stat */ | |
114 | #define RQF_IO_STAT ((__force req_flags_t)(1 << 13)) | |
115 | /* request came from our alloc pool */ | |
116 | #define RQF_ALLOCED ((__force req_flags_t)(1 << 14)) | |
117 | /* runtime pm request */ | |
118 | #define RQF_PM ((__force req_flags_t)(1 << 15)) | |
119 | /* on IO scheduler merge hash */ | |
120 | #define RQF_HASHED ((__force req_flags_t)(1 << 16)) | |
121 | /* IO stats tracking on */ | |
122 | #define RQF_STATS ((__force req_flags_t)(1 << 17)) | |
123 | /* Look at ->special_vec for the actual data payload instead of the | |
124 | bio chain. */ | |
125 | #define RQF_SPECIAL_PAYLOAD ((__force req_flags_t)(1 << 18)) | |
126 | /* The per-zone write lock is held for this request */ | |
127 | #define RQF_ZONE_WRITE_LOCKED ((__force req_flags_t)(1 << 19)) | |
128 | /* timeout is expired */ | |
129 | #define RQF_MQ_TIMEOUT_EXPIRED ((__force req_flags_t)(1 << 20)) | |
130 | /* already slept for hybrid poll */ | |
131 | #define RQF_MQ_POLL_SLEPT ((__force req_flags_t)(1 << 21)) | |
132 | ||
133 | /* flags that prevent us from merging requests: */ | |
134 | #define RQF_NOMERGE_FLAGS \ | |
135 | (RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD) | |
136 | ||
137 | /* | |
138 | * Try to put the fields that are referenced together in the same cacheline. | |
139 | * | |
140 | * If you modify this structure, make sure to update blk_rq_init() and | |
141 | * especially blk_mq_rq_ctx_init() to take care of the added fields. | |
142 | */ | |
143 | struct request { | |
144 | struct request_queue *q; | |
145 | struct blk_mq_ctx *mq_ctx; | |
146 | ||
147 | int cpu; | |
148 | unsigned int cmd_flags; /* op and common flags */ | |
149 | req_flags_t rq_flags; | |
150 | ||
151 | int internal_tag; | |
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 | struct list_head queuelist; | |
162 | ||
163 | /* | |
164 | * The hash is used inside the scheduler, and killed once the | |
165 | * request reaches the dispatch list. The ipi_list is only used | |
166 | * to queue the request for softirq completion, which is long | |
167 | * after the request has been unhashed (and even removed from | |
168 | * the dispatch list). | |
169 | */ | |
170 | union { | |
171 | struct hlist_node hash; /* merge hash */ | |
172 | struct list_head ipi_list; | |
173 | }; | |
174 | ||
175 | /* | |
176 | * The rb_node is only used inside the io scheduler, requests | |
177 | * are pruned when moved to the dispatch queue. So let the | |
178 | * completion_data share space with the rb_node. | |
179 | */ | |
180 | union { | |
181 | struct rb_node rb_node; /* sort/lookup */ | |
182 | struct bio_vec special_vec; | |
183 | void *completion_data; | |
184 | int error_count; /* for legacy drivers, don't use */ | |
185 | }; | |
186 | ||
187 | /* | |
188 | * Three pointers are available for the IO schedulers, if they need | |
189 | * more they have to dynamically allocate it. Flush requests are | |
190 | * never put on the IO scheduler. So let the flush fields share | |
191 | * space with the elevator data. | |
192 | */ | |
193 | union { | |
194 | struct { | |
195 | struct io_cq *icq; | |
196 | void *priv[2]; | |
197 | } elv; | |
198 | ||
199 | struct { | |
200 | unsigned int seq; | |
201 | struct list_head list; | |
202 | rq_end_io_fn *saved_end_io; | |
203 | } flush; | |
204 | }; | |
205 | ||
206 | struct gendisk *rq_disk; | |
207 | struct hd_struct *part; | |
208 | /* Time that I/O was submitted to the kernel. */ | |
209 | u64 start_time_ns; | |
210 | /* Time that I/O was submitted to the device. */ | |
211 | u64 io_start_time_ns; | |
212 | ||
213 | #ifdef CONFIG_BLK_WBT | |
214 | unsigned short wbt_flags; | |
215 | #endif | |
216 | #ifdef CONFIG_BLK_DEV_THROTTLING_LOW | |
217 | unsigned short throtl_size; | |
218 | #endif | |
219 | ||
220 | /* | |
221 | * Number of scatter-gather DMA addr+len pairs after | |
222 | * physical address coalescing is performed. | |
223 | */ | |
224 | unsigned short nr_phys_segments; | |
225 | ||
226 | #if defined(CONFIG_BLK_DEV_INTEGRITY) | |
227 | unsigned short nr_integrity_segments; | |
228 | #endif | |
229 | ||
230 | unsigned short write_hint; | |
231 | unsigned short ioprio; | |
232 | ||
233 | unsigned int timeout; | |
234 | ||
235 | void *special; /* opaque pointer available for LLD use */ | |
236 | ||
237 | unsigned int extra_len; /* length of alignment and padding */ | |
238 | ||
239 | /* | |
240 | * On blk-mq, the lower bits of ->gstate (generation number and | |
241 | * state) carry the MQ_RQ_* state value and the upper bits the | |
242 | * generation number which is monotonically incremented and used to | |
243 | * distinguish the reuse instances. | |
244 | * | |
245 | * ->gstate_seq allows updates to ->gstate and other fields | |
246 | * (currently ->deadline) during request start to be read | |
247 | * atomically from the timeout path, so that it can operate on a | |
248 | * coherent set of information. | |
249 | */ | |
250 | seqcount_t gstate_seq; | |
251 | u64 gstate; | |
252 | ||
253 | /* | |
254 | * ->aborted_gstate is used by the timeout to claim a specific | |
255 | * recycle instance of this request. See blk_mq_timeout_work(). | |
256 | */ | |
257 | struct u64_stats_sync aborted_gstate_sync; | |
258 | u64 aborted_gstate; | |
259 | ||
260 | /* access through blk_rq_set_deadline, blk_rq_deadline */ | |
261 | unsigned long __deadline; | |
262 | ||
263 | struct list_head timeout_list; | |
264 | ||
265 | union { | |
266 | struct __call_single_data csd; | |
267 | u64 fifo_time; | |
268 | }; | |
269 | ||
270 | /* | |
271 | * completion callback. | |
272 | */ | |
273 | rq_end_io_fn *end_io; | |
274 | void *end_io_data; | |
275 | ||
276 | /* for bidi */ | |
277 | struct request *next_rq; | |
278 | ||
279 | #ifdef CONFIG_BLK_CGROUP | |
280 | struct request_list *rl; /* rl this rq is alloced from */ | |
281 | #endif | |
282 | }; | |
283 | ||
284 | static inline bool blk_op_is_scsi(unsigned int op) | |
285 | { | |
286 | return op == REQ_OP_SCSI_IN || op == REQ_OP_SCSI_OUT; | |
287 | } | |
288 | ||
289 | static inline bool blk_op_is_private(unsigned int op) | |
290 | { | |
291 | return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT; | |
292 | } | |
293 | ||
294 | static inline bool blk_rq_is_scsi(struct request *rq) | |
295 | { | |
296 | return blk_op_is_scsi(req_op(rq)); | |
297 | } | |
298 | ||
299 | static inline bool blk_rq_is_private(struct request *rq) | |
300 | { | |
301 | return blk_op_is_private(req_op(rq)); | |
302 | } | |
303 | ||
304 | static inline bool blk_rq_is_passthrough(struct request *rq) | |
305 | { | |
306 | return blk_rq_is_scsi(rq) || blk_rq_is_private(rq); | |
307 | } | |
308 | ||
309 | static inline bool bio_is_passthrough(struct bio *bio) | |
310 | { | |
311 | unsigned op = bio_op(bio); | |
312 | ||
313 | return blk_op_is_scsi(op) || blk_op_is_private(op); | |
314 | } | |
315 | ||
316 | static inline unsigned short req_get_ioprio(struct request *req) | |
317 | { | |
318 | return req->ioprio; | |
319 | } | |
320 | ||
321 | #include <linux/elevator.h> | |
322 | ||
323 | struct blk_queue_ctx; | |
324 | ||
325 | typedef void (request_fn_proc) (struct request_queue *q); | |
326 | typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio); | |
327 | typedef bool (poll_q_fn) (struct request_queue *q, blk_qc_t); | |
328 | typedef int (prep_rq_fn) (struct request_queue *, struct request *); | |
329 | typedef void (unprep_rq_fn) (struct request_queue *, struct request *); | |
330 | ||
331 | struct bio_vec; | |
332 | typedef void (softirq_done_fn)(struct request *); | |
333 | typedef int (dma_drain_needed_fn)(struct request *); | |
334 | typedef int (lld_busy_fn) (struct request_queue *q); | |
335 | typedef int (bsg_job_fn) (struct bsg_job *); | |
336 | typedef int (init_rq_fn)(struct request_queue *, struct request *, gfp_t); | |
337 | typedef void (exit_rq_fn)(struct request_queue *, struct request *); | |
338 | ||
339 | enum blk_eh_timer_return { | |
340 | BLK_EH_NOT_HANDLED, | |
341 | BLK_EH_HANDLED, | |
342 | BLK_EH_RESET_TIMER, | |
343 | }; | |
344 | ||
345 | typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *); | |
346 | ||
347 | enum blk_queue_state { | |
348 | Queue_down, | |
349 | Queue_up, | |
350 | }; | |
351 | ||
352 | struct blk_queue_tag { | |
353 | struct request **tag_index; /* map of busy tags */ | |
354 | unsigned long *tag_map; /* bit map of free/busy tags */ | |
355 | int max_depth; /* what we will send to device */ | |
356 | int real_max_depth; /* what the array can hold */ | |
357 | atomic_t refcnt; /* map can be shared */ | |
358 | int alloc_policy; /* tag allocation policy */ | |
359 | int next_tag; /* next tag */ | |
360 | }; | |
361 | #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */ | |
362 | #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */ | |
363 | ||
364 | #define BLK_SCSI_MAX_CMDS (256) | |
365 | #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8)) | |
366 | ||
367 | /* | |
368 | * Zoned block device models (zoned limit). | |
369 | */ | |
370 | enum blk_zoned_model { | |
371 | BLK_ZONED_NONE, /* Regular block device */ | |
372 | BLK_ZONED_HA, /* Host-aware zoned block device */ | |
373 | BLK_ZONED_HM, /* Host-managed zoned block device */ | |
374 | }; | |
375 | ||
376 | struct queue_limits { | |
377 | unsigned long bounce_pfn; | |
378 | unsigned long seg_boundary_mask; | |
379 | unsigned long virt_boundary_mask; | |
380 | ||
381 | unsigned int max_hw_sectors; | |
382 | unsigned int max_dev_sectors; | |
383 | unsigned int chunk_sectors; | |
384 | unsigned int max_sectors; | |
385 | unsigned int max_segment_size; | |
386 | unsigned int physical_block_size; | |
387 | unsigned int alignment_offset; | |
388 | unsigned int io_min; | |
389 | unsigned int io_opt; | |
390 | unsigned int max_discard_sectors; | |
391 | unsigned int max_hw_discard_sectors; | |
392 | unsigned int max_write_same_sectors; | |
393 | unsigned int max_write_zeroes_sectors; | |
394 | unsigned int discard_granularity; | |
395 | unsigned int discard_alignment; | |
396 | ||
397 | unsigned short logical_block_size; | |
398 | unsigned short max_segments; | |
399 | unsigned short max_integrity_segments; | |
400 | unsigned short max_discard_segments; | |
401 | ||
402 | unsigned char misaligned; | |
403 | unsigned char discard_misaligned; | |
404 | unsigned char cluster; | |
405 | unsigned char raid_partial_stripes_expensive; | |
406 | enum blk_zoned_model zoned; | |
407 | }; | |
408 | ||
409 | #ifdef CONFIG_BLK_DEV_ZONED | |
410 | ||
411 | struct blk_zone_report_hdr { | |
412 | unsigned int nr_zones; | |
413 | u8 padding[60]; | |
414 | }; | |
415 | ||
416 | extern int blkdev_report_zones(struct block_device *bdev, | |
417 | sector_t sector, struct blk_zone *zones, | |
418 | unsigned int *nr_zones, gfp_t gfp_mask); | |
419 | extern int blkdev_reset_zones(struct block_device *bdev, sector_t sectors, | |
420 | sector_t nr_sectors, gfp_t gfp_mask); | |
421 | ||
422 | extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode, | |
423 | unsigned int cmd, unsigned long arg); | |
424 | extern int blkdev_reset_zones_ioctl(struct block_device *bdev, fmode_t mode, | |
425 | unsigned int cmd, unsigned long arg); | |
426 | ||
427 | #else /* CONFIG_BLK_DEV_ZONED */ | |
428 | ||
429 | static inline int blkdev_report_zones_ioctl(struct block_device *bdev, | |
430 | fmode_t mode, unsigned int cmd, | |
431 | unsigned long arg) | |
432 | { | |
433 | return -ENOTTY; | |
434 | } | |
435 | ||
436 | static inline int blkdev_reset_zones_ioctl(struct block_device *bdev, | |
437 | fmode_t mode, unsigned int cmd, | |
438 | unsigned long arg) | |
439 | { | |
440 | return -ENOTTY; | |
441 | } | |
442 | ||
443 | #endif /* CONFIG_BLK_DEV_ZONED */ | |
444 | ||
445 | struct request_queue { | |
446 | /* | |
447 | * Together with queue_head for cacheline sharing | |
448 | */ | |
449 | struct list_head queue_head; | |
450 | struct request *last_merge; | |
451 | struct elevator_queue *elevator; | |
452 | int nr_rqs[2]; /* # allocated [a]sync rqs */ | |
453 | int nr_rqs_elvpriv; /* # allocated rqs w/ elvpriv */ | |
454 | ||
455 | atomic_t shared_hctx_restart; | |
456 | ||
457 | struct blk_queue_stats *stats; | |
458 | struct rq_wb *rq_wb; | |
459 | ||
460 | /* | |
461 | * If blkcg is not used, @q->root_rl serves all requests. If blkcg | |
462 | * is used, root blkg allocates from @q->root_rl and all other | |
463 | * blkgs from their own blkg->rl. Which one to use should be | |
464 | * determined using bio_request_list(). | |
465 | */ | |
466 | struct request_list root_rl; | |
467 | ||
468 | request_fn_proc *request_fn; | |
469 | make_request_fn *make_request_fn; | |
470 | poll_q_fn *poll_fn; | |
471 | prep_rq_fn *prep_rq_fn; | |
472 | unprep_rq_fn *unprep_rq_fn; | |
473 | softirq_done_fn *softirq_done_fn; | |
474 | rq_timed_out_fn *rq_timed_out_fn; | |
475 | dma_drain_needed_fn *dma_drain_needed; | |
476 | lld_busy_fn *lld_busy_fn; | |
477 | /* Called just after a request is allocated */ | |
478 | init_rq_fn *init_rq_fn; | |
479 | /* Called just before a request is freed */ | |
480 | exit_rq_fn *exit_rq_fn; | |
481 | /* Called from inside blk_get_request() */ | |
482 | void (*initialize_rq_fn)(struct request *rq); | |
483 | ||
484 | const struct blk_mq_ops *mq_ops; | |
485 | ||
486 | unsigned int *mq_map; | |
487 | ||
488 | /* sw queues */ | |
489 | struct blk_mq_ctx __percpu *queue_ctx; | |
490 | unsigned int nr_queues; | |
491 | ||
492 | unsigned int queue_depth; | |
493 | ||
494 | /* hw dispatch queues */ | |
495 | struct blk_mq_hw_ctx **queue_hw_ctx; | |
496 | unsigned int nr_hw_queues; | |
497 | ||
498 | /* | |
499 | * Dispatch queue sorting | |
500 | */ | |
501 | sector_t end_sector; | |
502 | struct request *boundary_rq; | |
503 | ||
504 | /* | |
505 | * Delayed queue handling | |
506 | */ | |
507 | struct delayed_work delay_work; | |
508 | ||
509 | struct backing_dev_info *backing_dev_info; | |
510 | ||
511 | /* | |
512 | * The queue owner gets to use this for whatever they like. | |
513 | * ll_rw_blk doesn't touch it. | |
514 | */ | |
515 | void *queuedata; | |
516 | ||
517 | /* | |
518 | * various queue flags, see QUEUE_* below | |
519 | */ | |
520 | unsigned long queue_flags; | |
521 | ||
522 | /* | |
523 | * ida allocated id for this queue. Used to index queues from | |
524 | * ioctx. | |
525 | */ | |
526 | int id; | |
527 | ||
528 | /* | |
529 | * queue needs bounce pages for pages above this limit | |
530 | */ | |
531 | gfp_t bounce_gfp; | |
532 | ||
533 | /* | |
534 | * protects queue structures from reentrancy. ->__queue_lock should | |
535 | * _never_ be used directly, it is queue private. always use | |
536 | * ->queue_lock. | |
537 | */ | |
538 | spinlock_t __queue_lock; | |
539 | spinlock_t *queue_lock; | |
540 | ||
541 | /* | |
542 | * queue kobject | |
543 | */ | |
544 | struct kobject kobj; | |
545 | ||
546 | /* | |
547 | * mq queue kobject | |
548 | */ | |
549 | struct kobject mq_kobj; | |
550 | ||
551 | #ifdef CONFIG_BLK_DEV_INTEGRITY | |
552 | struct blk_integrity integrity; | |
553 | #endif /* CONFIG_BLK_DEV_INTEGRITY */ | |
554 | ||
555 | #ifdef CONFIG_PM | |
556 | struct device *dev; | |
557 | int rpm_status; | |
558 | unsigned int nr_pending; | |
559 | #endif | |
560 | ||
561 | /* | |
562 | * queue settings | |
563 | */ | |
564 | unsigned long nr_requests; /* Max # of requests */ | |
565 | unsigned int nr_congestion_on; | |
566 | unsigned int nr_congestion_off; | |
567 | unsigned int nr_batching; | |
568 | ||
569 | unsigned int dma_drain_size; | |
570 | void *dma_drain_buffer; | |
571 | unsigned int dma_pad_mask; | |
572 | unsigned int dma_alignment; | |
573 | ||
574 | struct blk_queue_tag *queue_tags; | |
575 | struct list_head tag_busy_list; | |
576 | ||
577 | unsigned int nr_sorted; | |
578 | unsigned int in_flight[2]; | |
579 | ||
580 | /* | |
581 | * Number of active block driver functions for which blk_drain_queue() | |
582 | * must wait. Must be incremented around functions that unlock the | |
583 | * queue_lock internally, e.g. scsi_request_fn(). | |
584 | */ | |
585 | unsigned int request_fn_active; | |
586 | ||
587 | unsigned int rq_timeout; | |
588 | int poll_nsec; | |
589 | ||
590 | struct blk_stat_callback *poll_cb; | |
591 | struct blk_rq_stat poll_stat[BLK_MQ_POLL_STATS_BKTS]; | |
592 | ||
593 | struct timer_list timeout; | |
594 | struct work_struct timeout_work; | |
595 | struct list_head timeout_list; | |
596 | ||
597 | struct list_head icq_list; | |
598 | #ifdef CONFIG_BLK_CGROUP | |
599 | DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS); | |
600 | struct blkcg_gq *root_blkg; | |
601 | struct list_head blkg_list; | |
602 | #endif | |
603 | ||
604 | struct queue_limits limits; | |
605 | ||
606 | /* | |
607 | * Zoned block device information for request dispatch control. | |
608 | * nr_zones is the total number of zones of the device. This is always | |
609 | * 0 for regular block devices. seq_zones_bitmap is a bitmap of nr_zones | |
610 | * bits which indicates if a zone is conventional (bit clear) or | |
611 | * sequential (bit set). seq_zones_wlock is a bitmap of nr_zones | |
612 | * bits which indicates if a zone is write locked, that is, if a write | |
613 | * request targeting the zone was dispatched. All three fields are | |
614 | * initialized by the low level device driver (e.g. scsi/sd.c). | |
615 | * Stacking drivers (device mappers) may or may not initialize | |
616 | * these fields. | |
617 | * | |
618 | * Reads of this information must be protected with blk_queue_enter() / | |
619 | * blk_queue_exit(). Modifying this information is only allowed while | |
620 | * no requests are being processed. See also blk_mq_freeze_queue() and | |
621 | * blk_mq_unfreeze_queue(). | |
622 | */ | |
623 | unsigned int nr_zones; | |
624 | unsigned long *seq_zones_bitmap; | |
625 | unsigned long *seq_zones_wlock; | |
626 | ||
627 | /* | |
628 | * sg stuff | |
629 | */ | |
630 | unsigned int sg_timeout; | |
631 | unsigned int sg_reserved_size; | |
632 | int node; | |
633 | #ifdef CONFIG_BLK_DEV_IO_TRACE | |
634 | struct blk_trace *blk_trace; | |
635 | struct mutex blk_trace_mutex; | |
636 | #endif | |
637 | /* | |
638 | * for flush operations | |
639 | */ | |
640 | struct blk_flush_queue *fq; | |
641 | ||
642 | struct list_head requeue_list; | |
643 | spinlock_t requeue_lock; | |
644 | struct delayed_work requeue_work; | |
645 | ||
646 | struct mutex sysfs_lock; | |
647 | ||
648 | int bypass_depth; | |
649 | atomic_t mq_freeze_depth; | |
650 | ||
651 | #if defined(CONFIG_BLK_DEV_BSG) | |
652 | bsg_job_fn *bsg_job_fn; | |
653 | struct bsg_class_device bsg_dev; | |
654 | #endif | |
655 | ||
656 | #ifdef CONFIG_BLK_DEV_THROTTLING | |
657 | /* Throttle data */ | |
658 | struct throtl_data *td; | |
659 | #endif | |
660 | struct rcu_head rcu_head; | |
661 | wait_queue_head_t mq_freeze_wq; | |
662 | struct percpu_ref q_usage_counter; | |
663 | struct list_head all_q_node; | |
664 | ||
665 | struct blk_mq_tag_set *tag_set; | |
666 | struct list_head tag_set_list; | |
667 | struct bio_set *bio_split; | |
668 | ||
669 | #ifdef CONFIG_BLK_DEBUG_FS | |
670 | struct dentry *debugfs_dir; | |
671 | struct dentry *sched_debugfs_dir; | |
672 | #endif | |
673 | ||
674 | bool mq_sysfs_init_done; | |
675 | ||
676 | size_t cmd_size; | |
677 | void *rq_alloc_data; | |
678 | ||
679 | struct work_struct release_work; | |
680 | ||
681 | #define BLK_MAX_WRITE_HINTS 5 | |
682 | u64 write_hints[BLK_MAX_WRITE_HINTS]; | |
683 | }; | |
684 | ||
685 | #define QUEUE_FLAG_QUEUED 0 /* uses generic tag queueing */ | |
686 | #define QUEUE_FLAG_STOPPED 1 /* queue is stopped */ | |
687 | #define QUEUE_FLAG_DYING 2 /* queue being torn down */ | |
688 | #define QUEUE_FLAG_BYPASS 3 /* act as dumb FIFO queue */ | |
689 | #define QUEUE_FLAG_BIDI 4 /* queue supports bidi requests */ | |
690 | #define QUEUE_FLAG_NOMERGES 5 /* disable merge attempts */ | |
691 | #define QUEUE_FLAG_SAME_COMP 6 /* complete on same CPU-group */ | |
692 | #define QUEUE_FLAG_FAIL_IO 7 /* fake timeout */ | |
693 | #define QUEUE_FLAG_NONROT 9 /* non-rotational device (SSD) */ | |
694 | #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */ | |
695 | #define QUEUE_FLAG_IO_STAT 10 /* do IO stats */ | |
696 | #define QUEUE_FLAG_DISCARD 11 /* supports DISCARD */ | |
697 | #define QUEUE_FLAG_NOXMERGES 12 /* No extended merges */ | |
698 | #define QUEUE_FLAG_ADD_RANDOM 13 /* Contributes to random pool */ | |
699 | #define QUEUE_FLAG_SECERASE 14 /* supports secure erase */ | |
700 | #define QUEUE_FLAG_SAME_FORCE 15 /* force complete on same CPU */ | |
701 | #define QUEUE_FLAG_DEAD 16 /* queue tear-down finished */ | |
702 | #define QUEUE_FLAG_INIT_DONE 17 /* queue is initialized */ | |
703 | #define QUEUE_FLAG_NO_SG_MERGE 18 /* don't attempt to merge SG segments*/ | |
704 | #define QUEUE_FLAG_POLL 19 /* IO polling enabled if set */ | |
705 | #define QUEUE_FLAG_WC 20 /* Write back caching */ | |
706 | #define QUEUE_FLAG_FUA 21 /* device supports FUA writes */ | |
707 | #define QUEUE_FLAG_FLUSH_NQ 22 /* flush not queueuable */ | |
708 | #define QUEUE_FLAG_DAX 23 /* device supports DAX */ | |
709 | #define QUEUE_FLAG_STATS 24 /* track rq completion times */ | |
710 | #define QUEUE_FLAG_POLL_STATS 25 /* collecting stats for hybrid polling */ | |
711 | #define QUEUE_FLAG_REGISTERED 26 /* queue has been registered to a disk */ | |
712 | #define QUEUE_FLAG_SCSI_PASSTHROUGH 27 /* queue supports SCSI commands */ | |
713 | #define QUEUE_FLAG_QUIESCED 28 /* queue has been quiesced */ | |
714 | #define QUEUE_FLAG_PREEMPT_ONLY 29 /* only process REQ_PREEMPT requests */ | |
715 | ||
716 | #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \ | |
717 | (1 << QUEUE_FLAG_SAME_COMP) | \ | |
718 | (1 << QUEUE_FLAG_ADD_RANDOM)) | |
719 | ||
720 | #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \ | |
721 | (1 << QUEUE_FLAG_SAME_COMP) | \ | |
722 | (1 << QUEUE_FLAG_POLL)) | |
723 | ||
724 | void blk_queue_flag_set(unsigned int flag, struct request_queue *q); | |
725 | void blk_queue_flag_clear(unsigned int flag, struct request_queue *q); | |
726 | bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q); | |
727 | bool blk_queue_flag_test_and_clear(unsigned int flag, struct request_queue *q); | |
728 | ||
729 | #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags) | |
730 | #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags) | |
731 | #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags) | |
732 | #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags) | |
733 | #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags) | |
734 | #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags) | |
735 | #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags) | |
736 | #define blk_queue_noxmerges(q) \ | |
737 | test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags) | |
738 | #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags) | |
739 | #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags) | |
740 | #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags) | |
741 | #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags) | |
742 | #define blk_queue_secure_erase(q) \ | |
743 | (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags)) | |
744 | #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags) | |
745 | #define blk_queue_scsi_passthrough(q) \ | |
746 | test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags) | |
747 | ||
748 | #define blk_noretry_request(rq) \ | |
749 | ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \ | |
750 | REQ_FAILFAST_DRIVER)) | |
751 | #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags) | |
752 | #define blk_queue_preempt_only(q) \ | |
753 | test_bit(QUEUE_FLAG_PREEMPT_ONLY, &(q)->queue_flags) | |
754 | #define blk_queue_fua(q) test_bit(QUEUE_FLAG_FUA, &(q)->queue_flags) | |
755 | ||
756 | extern int blk_set_preempt_only(struct request_queue *q); | |
757 | extern void blk_clear_preempt_only(struct request_queue *q); | |
758 | ||
759 | static inline int queue_in_flight(struct request_queue *q) | |
760 | { | |
761 | return q->in_flight[0] + q->in_flight[1]; | |
762 | } | |
763 | ||
764 | static inline bool blk_account_rq(struct request *rq) | |
765 | { | |
766 | return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq); | |
767 | } | |
768 | ||
769 | #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1) | |
770 | #define blk_bidi_rq(rq) ((rq)->next_rq != NULL) | |
771 | /* rq->queuelist of dequeued request must be list_empty() */ | |
772 | #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist)) | |
773 | ||
774 | #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist) | |
775 | ||
776 | #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ) | |
777 | ||
778 | /* | |
779 | * Driver can handle struct request, if it either has an old style | |
780 | * request_fn defined, or is blk-mq based. | |
781 | */ | |
782 | static inline bool queue_is_rq_based(struct request_queue *q) | |
783 | { | |
784 | return q->request_fn || q->mq_ops; | |
785 | } | |
786 | ||
787 | static inline unsigned int blk_queue_cluster(struct request_queue *q) | |
788 | { | |
789 | return q->limits.cluster; | |
790 | } | |
791 | ||
792 | static inline enum blk_zoned_model | |
793 | blk_queue_zoned_model(struct request_queue *q) | |
794 | { | |
795 | return q->limits.zoned; | |
796 | } | |
797 | ||
798 | static inline bool blk_queue_is_zoned(struct request_queue *q) | |
799 | { | |
800 | switch (blk_queue_zoned_model(q)) { | |
801 | case BLK_ZONED_HA: | |
802 | case BLK_ZONED_HM: | |
803 | return true; | |
804 | default: | |
805 | return false; | |
806 | } | |
807 | } | |
808 | ||
809 | static inline unsigned int blk_queue_zone_sectors(struct request_queue *q) | |
810 | { | |
811 | return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0; | |
812 | } | |
813 | ||
814 | static inline unsigned int blk_queue_nr_zones(struct request_queue *q) | |
815 | { | |
816 | return q->nr_zones; | |
817 | } | |
818 | ||
819 | static inline unsigned int blk_queue_zone_no(struct request_queue *q, | |
820 | sector_t sector) | |
821 | { | |
822 | if (!blk_queue_is_zoned(q)) | |
823 | return 0; | |
824 | return sector >> ilog2(q->limits.chunk_sectors); | |
825 | } | |
826 | ||
827 | static inline bool blk_queue_zone_is_seq(struct request_queue *q, | |
828 | sector_t sector) | |
829 | { | |
830 | if (!blk_queue_is_zoned(q) || !q->seq_zones_bitmap) | |
831 | return false; | |
832 | return test_bit(blk_queue_zone_no(q, sector), q->seq_zones_bitmap); | |
833 | } | |
834 | ||
835 | static inline bool rq_is_sync(struct request *rq) | |
836 | { | |
837 | return op_is_sync(rq->cmd_flags); | |
838 | } | |
839 | ||
840 | static inline bool blk_rl_full(struct request_list *rl, bool sync) | |
841 | { | |
842 | unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL; | |
843 | ||
844 | return rl->flags & flag; | |
845 | } | |
846 | ||
847 | static inline void blk_set_rl_full(struct request_list *rl, bool sync) | |
848 | { | |
849 | unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL; | |
850 | ||
851 | rl->flags |= flag; | |
852 | } | |
853 | ||
854 | static inline void blk_clear_rl_full(struct request_list *rl, bool sync) | |
855 | { | |
856 | unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL; | |
857 | ||
858 | rl->flags &= ~flag; | |
859 | } | |
860 | ||
861 | static inline bool rq_mergeable(struct request *rq) | |
862 | { | |
863 | if (blk_rq_is_passthrough(rq)) | |
864 | return false; | |
865 | ||
866 | if (req_op(rq) == REQ_OP_FLUSH) | |
867 | return false; | |
868 | ||
869 | if (req_op(rq) == REQ_OP_WRITE_ZEROES) | |
870 | return false; | |
871 | ||
872 | if (rq->cmd_flags & REQ_NOMERGE_FLAGS) | |
873 | return false; | |
874 | if (rq->rq_flags & RQF_NOMERGE_FLAGS) | |
875 | return false; | |
876 | ||
877 | return true; | |
878 | } | |
879 | ||
880 | static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b) | |
881 | { | |
882 | if (bio_page(a) == bio_page(b) && | |
883 | bio_offset(a) == bio_offset(b)) | |
884 | return true; | |
885 | ||
886 | return false; | |
887 | } | |
888 | ||
889 | static inline unsigned int blk_queue_depth(struct request_queue *q) | |
890 | { | |
891 | if (q->queue_depth) | |
892 | return q->queue_depth; | |
893 | ||
894 | return q->nr_requests; | |
895 | } | |
896 | ||
897 | /* | |
898 | * q->prep_rq_fn return values | |
899 | */ | |
900 | enum { | |
901 | BLKPREP_OK, /* serve it */ | |
902 | BLKPREP_KILL, /* fatal error, kill, return -EIO */ | |
903 | BLKPREP_DEFER, /* leave on queue */ | |
904 | BLKPREP_INVALID, /* invalid command, kill, return -EREMOTEIO */ | |
905 | }; | |
906 | ||
907 | extern unsigned long blk_max_low_pfn, blk_max_pfn; | |
908 | ||
909 | /* | |
910 | * standard bounce addresses: | |
911 | * | |
912 | * BLK_BOUNCE_HIGH : bounce all highmem pages | |
913 | * BLK_BOUNCE_ANY : don't bounce anything | |
914 | * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary | |
915 | */ | |
916 | ||
917 | #if BITS_PER_LONG == 32 | |
918 | #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT) | |
919 | #else | |
920 | #define BLK_BOUNCE_HIGH -1ULL | |
921 | #endif | |
922 | #define BLK_BOUNCE_ANY (-1ULL) | |
923 | #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24)) | |
924 | ||
925 | /* | |
926 | * default timeout for SG_IO if none specified | |
927 | */ | |
928 | #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ) | |
929 | #define BLK_MIN_SG_TIMEOUT (7 * HZ) | |
930 | ||
931 | struct rq_map_data { | |
932 | struct page **pages; | |
933 | int page_order; | |
934 | int nr_entries; | |
935 | unsigned long offset; | |
936 | int null_mapped; | |
937 | int from_user; | |
938 | }; | |
939 | ||
940 | struct req_iterator { | |
941 | struct bvec_iter iter; | |
942 | struct bio *bio; | |
943 | }; | |
944 | ||
945 | /* This should not be used directly - use rq_for_each_segment */ | |
946 | #define for_each_bio(_bio) \ | |
947 | for (; _bio; _bio = _bio->bi_next) | |
948 | #define __rq_for_each_bio(_bio, rq) \ | |
949 | if ((rq->bio)) \ | |
950 | for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next) | |
951 | ||
952 | #define rq_for_each_segment(bvl, _rq, _iter) \ | |
953 | __rq_for_each_bio(_iter.bio, _rq) \ | |
954 | bio_for_each_segment(bvl, _iter.bio, _iter.iter) | |
955 | ||
956 | #define rq_iter_last(bvec, _iter) \ | |
957 | (_iter.bio->bi_next == NULL && \ | |
958 | bio_iter_last(bvec, _iter.iter)) | |
959 | ||
960 | #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE | |
961 | # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform" | |
962 | #endif | |
963 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE | |
964 | extern void rq_flush_dcache_pages(struct request *rq); | |
965 | #else | |
966 | static inline void rq_flush_dcache_pages(struct request *rq) | |
967 | { | |
968 | } | |
969 | #endif | |
970 | ||
971 | extern int blk_register_queue(struct gendisk *disk); | |
972 | extern void blk_unregister_queue(struct gendisk *disk); | |
973 | extern blk_qc_t generic_make_request(struct bio *bio); | |
974 | extern blk_qc_t direct_make_request(struct bio *bio); | |
975 | extern void blk_rq_init(struct request_queue *q, struct request *rq); | |
976 | extern void blk_init_request_from_bio(struct request *req, struct bio *bio); | |
977 | extern void blk_put_request(struct request *); | |
978 | extern void __blk_put_request(struct request_queue *, struct request *); | |
979 | extern struct request *blk_get_request_flags(struct request_queue *, | |
980 | unsigned int op, | |
981 | blk_mq_req_flags_t flags); | |
982 | extern struct request *blk_get_request(struct request_queue *, unsigned int op, | |
983 | gfp_t gfp_mask); | |
984 | extern void blk_requeue_request(struct request_queue *, struct request *); | |
985 | extern int blk_lld_busy(struct request_queue *q); | |
986 | extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src, | |
987 | struct bio_set *bs, gfp_t gfp_mask, | |
988 | int (*bio_ctr)(struct bio *, struct bio *, void *), | |
989 | void *data); | |
990 | extern void blk_rq_unprep_clone(struct request *rq); | |
991 | extern blk_status_t blk_insert_cloned_request(struct request_queue *q, | |
992 | struct request *rq); | |
993 | extern int blk_rq_append_bio(struct request *rq, struct bio **bio); | |
994 | extern void blk_delay_queue(struct request_queue *, unsigned long); | |
995 | extern void blk_queue_split(struct request_queue *, struct bio **); | |
996 | extern void blk_recount_segments(struct request_queue *, struct bio *); | |
997 | extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int); | |
998 | extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t, | |
999 | unsigned int, void __user *); | |
1000 | extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t, | |
1001 | unsigned int, void __user *); | |
1002 | extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t, | |
1003 | struct scsi_ioctl_command __user *); | |
1004 | ||
1005 | extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags); | |
1006 | extern void blk_queue_exit(struct request_queue *q); | |
1007 | extern void blk_start_queue(struct request_queue *q); | |
1008 | extern void blk_start_queue_async(struct request_queue *q); | |
1009 | extern void blk_stop_queue(struct request_queue *q); | |
1010 | extern void blk_sync_queue(struct request_queue *q); | |
1011 | extern void __blk_stop_queue(struct request_queue *q); | |
1012 | extern void __blk_run_queue(struct request_queue *q); | |
1013 | extern void __blk_run_queue_uncond(struct request_queue *q); | |
1014 | extern void blk_run_queue(struct request_queue *); | |
1015 | extern void blk_run_queue_async(struct request_queue *q); | |
1016 | extern int blk_rq_map_user(struct request_queue *, struct request *, | |
1017 | struct rq_map_data *, void __user *, unsigned long, | |
1018 | gfp_t); | |
1019 | extern int blk_rq_unmap_user(struct bio *); | |
1020 | extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t); | |
1021 | extern int blk_rq_map_user_iov(struct request_queue *, struct request *, | |
1022 | struct rq_map_data *, const struct iov_iter *, | |
1023 | gfp_t); | |
1024 | extern void blk_execute_rq(struct request_queue *, struct gendisk *, | |
1025 | struct request *, int); | |
1026 | extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *, | |
1027 | struct request *, int, rq_end_io_fn *); | |
1028 | ||
1029 | int blk_status_to_errno(blk_status_t status); | |
1030 | blk_status_t errno_to_blk_status(int errno); | |
1031 | ||
1032 | bool blk_poll(struct request_queue *q, blk_qc_t cookie); | |
1033 | ||
1034 | static inline struct request_queue *bdev_get_queue(struct block_device *bdev) | |
1035 | { | |
1036 | return bdev->bd_disk->queue; /* this is never NULL */ | |
1037 | } | |
1038 | ||
1039 | /* | |
1040 | * The basic unit of block I/O is a sector. It is used in a number of contexts | |
1041 | * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9 | |
1042 | * bytes. Variables of type sector_t represent an offset or size that is a | |
1043 | * multiple of 512 bytes. Hence these two constants. | |
1044 | */ | |
1045 | #ifndef SECTOR_SHIFT | |
1046 | #define SECTOR_SHIFT 9 | |
1047 | #endif | |
1048 | #ifndef SECTOR_SIZE | |
1049 | #define SECTOR_SIZE (1 << SECTOR_SHIFT) | |
1050 | #endif | |
1051 | ||
1052 | /* | |
1053 | * blk_rq_pos() : the current sector | |
1054 | * blk_rq_bytes() : bytes left in the entire request | |
1055 | * blk_rq_cur_bytes() : bytes left in the current segment | |
1056 | * blk_rq_err_bytes() : bytes left till the next error boundary | |
1057 | * blk_rq_sectors() : sectors left in the entire request | |
1058 | * blk_rq_cur_sectors() : sectors left in the current segment | |
1059 | */ | |
1060 | static inline sector_t blk_rq_pos(const struct request *rq) | |
1061 | { | |
1062 | return rq->__sector; | |
1063 | } | |
1064 | ||
1065 | static inline unsigned int blk_rq_bytes(const struct request *rq) | |
1066 | { | |
1067 | return rq->__data_len; | |
1068 | } | |
1069 | ||
1070 | static inline int blk_rq_cur_bytes(const struct request *rq) | |
1071 | { | |
1072 | return rq->bio ? bio_cur_bytes(rq->bio) : 0; | |
1073 | } | |
1074 | ||
1075 | extern unsigned int blk_rq_err_bytes(const struct request *rq); | |
1076 | ||
1077 | static inline unsigned int blk_rq_sectors(const struct request *rq) | |
1078 | { | |
1079 | return blk_rq_bytes(rq) >> SECTOR_SHIFT; | |
1080 | } | |
1081 | ||
1082 | static inline unsigned int blk_rq_cur_sectors(const struct request *rq) | |
1083 | { | |
1084 | return blk_rq_cur_bytes(rq) >> SECTOR_SHIFT; | |
1085 | } | |
1086 | ||
1087 | static inline unsigned int blk_rq_zone_no(struct request *rq) | |
1088 | { | |
1089 | return blk_queue_zone_no(rq->q, blk_rq_pos(rq)); | |
1090 | } | |
1091 | ||
1092 | static inline unsigned int blk_rq_zone_is_seq(struct request *rq) | |
1093 | { | |
1094 | return blk_queue_zone_is_seq(rq->q, blk_rq_pos(rq)); | |
1095 | } | |
1096 | ||
1097 | /* | |
1098 | * Some commands like WRITE SAME have a payload or data transfer size which | |
1099 | * is different from the size of the request. Any driver that supports such | |
1100 | * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to | |
1101 | * calculate the data transfer size. | |
1102 | */ | |
1103 | static inline unsigned int blk_rq_payload_bytes(struct request *rq) | |
1104 | { | |
1105 | if (rq->rq_flags & RQF_SPECIAL_PAYLOAD) | |
1106 | return rq->special_vec.bv_len; | |
1107 | return blk_rq_bytes(rq); | |
1108 | } | |
1109 | ||
1110 | static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q, | |
1111 | int op) | |
1112 | { | |
1113 | if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE)) | |
1114 | return min(q->limits.max_discard_sectors, | |
1115 | UINT_MAX >> SECTOR_SHIFT); | |
1116 | ||
1117 | if (unlikely(op == REQ_OP_WRITE_SAME)) | |
1118 | return q->limits.max_write_same_sectors; | |
1119 | ||
1120 | if (unlikely(op == REQ_OP_WRITE_ZEROES)) | |
1121 | return q->limits.max_write_zeroes_sectors; | |
1122 | ||
1123 | return q->limits.max_sectors; | |
1124 | } | |
1125 | ||
1126 | /* | |
1127 | * Return maximum size of a request at given offset. Only valid for | |
1128 | * file system requests. | |
1129 | */ | |
1130 | static inline unsigned int blk_max_size_offset(struct request_queue *q, | |
1131 | sector_t offset) | |
1132 | { | |
1133 | if (!q->limits.chunk_sectors) | |
1134 | return q->limits.max_sectors; | |
1135 | ||
1136 | return q->limits.chunk_sectors - | |
1137 | (offset & (q->limits.chunk_sectors - 1)); | |
1138 | } | |
1139 | ||
1140 | static inline unsigned int blk_rq_get_max_sectors(struct request *rq, | |
1141 | sector_t offset) | |
1142 | { | |
1143 | struct request_queue *q = rq->q; | |
1144 | ||
1145 | if (blk_rq_is_passthrough(rq)) | |
1146 | return q->limits.max_hw_sectors; | |
1147 | ||
1148 | if (!q->limits.chunk_sectors || | |
1149 | req_op(rq) == REQ_OP_DISCARD || | |
1150 | req_op(rq) == REQ_OP_SECURE_ERASE) | |
1151 | return blk_queue_get_max_sectors(q, req_op(rq)); | |
1152 | ||
1153 | return min(blk_max_size_offset(q, offset), | |
1154 | blk_queue_get_max_sectors(q, req_op(rq))); | |
1155 | } | |
1156 | ||
1157 | static inline unsigned int blk_rq_count_bios(struct request *rq) | |
1158 | { | |
1159 | unsigned int nr_bios = 0; | |
1160 | struct bio *bio; | |
1161 | ||
1162 | __rq_for_each_bio(bio, rq) | |
1163 | nr_bios++; | |
1164 | ||
1165 | return nr_bios; | |
1166 | } | |
1167 | ||
1168 | /* | |
1169 | * Request issue related functions. | |
1170 | */ | |
1171 | extern struct request *blk_peek_request(struct request_queue *q); | |
1172 | extern void blk_start_request(struct request *rq); | |
1173 | extern struct request *blk_fetch_request(struct request_queue *q); | |
1174 | ||
1175 | void blk_steal_bios(struct bio_list *list, struct request *rq); | |
1176 | ||
1177 | /* | |
1178 | * Request completion related functions. | |
1179 | * | |
1180 | * blk_update_request() completes given number of bytes and updates | |
1181 | * the request without completing it. | |
1182 | * | |
1183 | * blk_end_request() and friends. __blk_end_request() must be called | |
1184 | * with the request queue spinlock acquired. | |
1185 | * | |
1186 | * Several drivers define their own end_request and call | |
1187 | * blk_end_request() for parts of the original function. | |
1188 | * This prevents code duplication in drivers. | |
1189 | */ | |
1190 | extern bool blk_update_request(struct request *rq, blk_status_t error, | |
1191 | unsigned int nr_bytes); | |
1192 | extern void blk_finish_request(struct request *rq, blk_status_t error); | |
1193 | extern bool blk_end_request(struct request *rq, blk_status_t error, | |
1194 | unsigned int nr_bytes); | |
1195 | extern void blk_end_request_all(struct request *rq, blk_status_t error); | |
1196 | extern bool __blk_end_request(struct request *rq, blk_status_t error, | |
1197 | unsigned int nr_bytes); | |
1198 | extern void __blk_end_request_all(struct request *rq, blk_status_t error); | |
1199 | extern bool __blk_end_request_cur(struct request *rq, blk_status_t error); | |
1200 | ||
1201 | extern void blk_complete_request(struct request *); | |
1202 | extern void __blk_complete_request(struct request *); | |
1203 | extern void blk_abort_request(struct request *); | |
1204 | extern void blk_unprep_request(struct request *); | |
1205 | ||
1206 | /* | |
1207 | * Access functions for manipulating queue properties | |
1208 | */ | |
1209 | extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn, | |
1210 | spinlock_t *lock, int node_id); | |
1211 | extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *); | |
1212 | extern int blk_init_allocated_queue(struct request_queue *); | |
1213 | extern void blk_cleanup_queue(struct request_queue *); | |
1214 | extern void blk_queue_make_request(struct request_queue *, make_request_fn *); | |
1215 | extern void blk_queue_bounce_limit(struct request_queue *, u64); | |
1216 | extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int); | |
1217 | extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int); | |
1218 | extern void blk_queue_max_segments(struct request_queue *, unsigned short); | |
1219 | extern void blk_queue_max_discard_segments(struct request_queue *, | |
1220 | unsigned short); | |
1221 | extern void blk_queue_max_segment_size(struct request_queue *, unsigned int); | |
1222 | extern void blk_queue_max_discard_sectors(struct request_queue *q, | |
1223 | unsigned int max_discard_sectors); | |
1224 | extern void blk_queue_max_write_same_sectors(struct request_queue *q, | |
1225 | unsigned int max_write_same_sectors); | |
1226 | extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q, | |
1227 | unsigned int max_write_same_sectors); | |
1228 | extern void blk_queue_logical_block_size(struct request_queue *, unsigned short); | |
1229 | extern void blk_queue_physical_block_size(struct request_queue *, unsigned int); | |
1230 | extern void blk_queue_alignment_offset(struct request_queue *q, | |
1231 | unsigned int alignment); | |
1232 | extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min); | |
1233 | extern void blk_queue_io_min(struct request_queue *q, unsigned int min); | |
1234 | extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt); | |
1235 | extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt); | |
1236 | extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth); | |
1237 | extern void blk_set_default_limits(struct queue_limits *lim); | |
1238 | extern void blk_set_stacking_limits(struct queue_limits *lim); | |
1239 | extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, | |
1240 | sector_t offset); | |
1241 | extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev, | |
1242 | sector_t offset); | |
1243 | extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev, | |
1244 | sector_t offset); | |
1245 | extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b); | |
1246 | extern void blk_queue_dma_pad(struct request_queue *, unsigned int); | |
1247 | extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int); | |
1248 | extern int blk_queue_dma_drain(struct request_queue *q, | |
1249 | dma_drain_needed_fn *dma_drain_needed, | |
1250 | void *buf, unsigned int size); | |
1251 | extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn); | |
1252 | extern void blk_queue_segment_boundary(struct request_queue *, unsigned long); | |
1253 | extern void blk_queue_virt_boundary(struct request_queue *, unsigned long); | |
1254 | extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn); | |
1255 | extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn); | |
1256 | extern void blk_queue_dma_alignment(struct request_queue *, int); | |
1257 | extern void blk_queue_update_dma_alignment(struct request_queue *, int); | |
1258 | extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *); | |
1259 | extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *); | |
1260 | extern void blk_queue_rq_timeout(struct request_queue *, unsigned int); | |
1261 | extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable); | |
1262 | extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua); | |
1263 | ||
1264 | /* | |
1265 | * Number of physical segments as sent to the device. | |
1266 | * | |
1267 | * Normally this is the number of discontiguous data segments sent by the | |
1268 | * submitter. But for data-less command like discard we might have no | |
1269 | * actual data segments submitted, but the driver might have to add it's | |
1270 | * own special payload. In that case we still return 1 here so that this | |
1271 | * special payload will be mapped. | |
1272 | */ | |
1273 | static inline unsigned short blk_rq_nr_phys_segments(struct request *rq) | |
1274 | { | |
1275 | if (rq->rq_flags & RQF_SPECIAL_PAYLOAD) | |
1276 | return 1; | |
1277 | return rq->nr_phys_segments; | |
1278 | } | |
1279 | ||
1280 | /* | |
1281 | * Number of discard segments (or ranges) the driver needs to fill in. | |
1282 | * Each discard bio merged into a request is counted as one segment. | |
1283 | */ | |
1284 | static inline unsigned short blk_rq_nr_discard_segments(struct request *rq) | |
1285 | { | |
1286 | return max_t(unsigned short, rq->nr_phys_segments, 1); | |
1287 | } | |
1288 | ||
1289 | extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *); | |
1290 | extern void blk_dump_rq_flags(struct request *, char *); | |
1291 | extern long nr_blockdev_pages(void); | |
1292 | ||
1293 | bool __must_check blk_get_queue(struct request_queue *); | |
1294 | struct request_queue *blk_alloc_queue(gfp_t); | |
1295 | struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id, | |
1296 | spinlock_t *lock); | |
1297 | extern void blk_put_queue(struct request_queue *); | |
1298 | extern void blk_set_queue_dying(struct request_queue *); | |
1299 | ||
1300 | /* | |
1301 | * block layer runtime pm functions | |
1302 | */ | |
1303 | #ifdef CONFIG_PM | |
1304 | extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev); | |
1305 | extern int blk_pre_runtime_suspend(struct request_queue *q); | |
1306 | extern void blk_post_runtime_suspend(struct request_queue *q, int err); | |
1307 | extern void blk_pre_runtime_resume(struct request_queue *q); | |
1308 | extern void blk_post_runtime_resume(struct request_queue *q, int err); | |
1309 | extern void blk_set_runtime_active(struct request_queue *q); | |
1310 | #else | |
1311 | static inline void blk_pm_runtime_init(struct request_queue *q, | |
1312 | struct device *dev) {} | |
1313 | static inline int blk_pre_runtime_suspend(struct request_queue *q) | |
1314 | { | |
1315 | return -ENOSYS; | |
1316 | } | |
1317 | static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {} | |
1318 | static inline void blk_pre_runtime_resume(struct request_queue *q) {} | |
1319 | static inline void blk_post_runtime_resume(struct request_queue *q, int err) {} | |
1320 | static inline void blk_set_runtime_active(struct request_queue *q) {} | |
1321 | #endif | |
1322 | ||
1323 | /* | |
1324 | * blk_plug permits building a queue of related requests by holding the I/O | |
1325 | * fragments for a short period. This allows merging of sequential requests | |
1326 | * into single larger request. As the requests are moved from a per-task list to | |
1327 | * the device's request_queue in a batch, this results in improved scalability | |
1328 | * as the lock contention for request_queue lock is reduced. | |
1329 | * | |
1330 | * It is ok not to disable preemption when adding the request to the plug list | |
1331 | * or when attempting a merge, because blk_schedule_flush_list() will only flush | |
1332 | * the plug list when the task sleeps by itself. For details, please see | |
1333 | * schedule() where blk_schedule_flush_plug() is called. | |
1334 | */ | |
1335 | struct blk_plug { | |
1336 | struct list_head list; /* requests */ | |
1337 | struct list_head mq_list; /* blk-mq requests */ | |
1338 | struct list_head cb_list; /* md requires an unplug callback */ | |
1339 | }; | |
1340 | #define BLK_MAX_REQUEST_COUNT 16 | |
1341 | #define BLK_PLUG_FLUSH_SIZE (128 * 1024) | |
1342 | ||
1343 | struct blk_plug_cb; | |
1344 | typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool); | |
1345 | struct blk_plug_cb { | |
1346 | struct list_head list; | |
1347 | blk_plug_cb_fn callback; | |
1348 | void *data; | |
1349 | }; | |
1350 | extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug, | |
1351 | void *data, int size); | |
1352 | extern void blk_start_plug(struct blk_plug *); | |
1353 | extern void blk_finish_plug(struct blk_plug *); | |
1354 | extern void blk_flush_plug_list(struct blk_plug *, bool); | |
1355 | ||
1356 | static inline void blk_flush_plug(struct task_struct *tsk) | |
1357 | { | |
1358 | struct blk_plug *plug = tsk->plug; | |
1359 | ||
1360 | if (plug) | |
1361 | blk_flush_plug_list(plug, false); | |
1362 | } | |
1363 | ||
1364 | static inline void blk_schedule_flush_plug(struct task_struct *tsk) | |
1365 | { | |
1366 | struct blk_plug *plug = tsk->plug; | |
1367 | ||
1368 | if (plug) | |
1369 | blk_flush_plug_list(plug, true); | |
1370 | } | |
1371 | ||
1372 | static inline bool blk_needs_flush_plug(struct task_struct *tsk) | |
1373 | { | |
1374 | struct blk_plug *plug = tsk->plug; | |
1375 | ||
1376 | return plug && | |
1377 | (!list_empty(&plug->list) || | |
1378 | !list_empty(&plug->mq_list) || | |
1379 | !list_empty(&plug->cb_list)); | |
1380 | } | |
1381 | ||
1382 | /* | |
1383 | * tag stuff | |
1384 | */ | |
1385 | extern int blk_queue_start_tag(struct request_queue *, struct request *); | |
1386 | extern struct request *blk_queue_find_tag(struct request_queue *, int); | |
1387 | extern void blk_queue_end_tag(struct request_queue *, struct request *); | |
1388 | extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int); | |
1389 | extern void blk_queue_free_tags(struct request_queue *); | |
1390 | extern int blk_queue_resize_tags(struct request_queue *, int); | |
1391 | extern void blk_queue_invalidate_tags(struct request_queue *); | |
1392 | extern struct blk_queue_tag *blk_init_tags(int, int); | |
1393 | extern void blk_free_tags(struct blk_queue_tag *); | |
1394 | ||
1395 | static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt, | |
1396 | int tag) | |
1397 | { | |
1398 | if (unlikely(bqt == NULL || tag >= bqt->real_max_depth)) | |
1399 | return NULL; | |
1400 | return bqt->tag_index[tag]; | |
1401 | } | |
1402 | ||
1403 | extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *); | |
1404 | extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector, | |
1405 | sector_t nr_sects, gfp_t gfp_mask, struct page *page); | |
1406 | ||
1407 | #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */ | |
1408 | ||
1409 | extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector, | |
1410 | sector_t nr_sects, gfp_t gfp_mask, unsigned long flags); | |
1411 | extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector, | |
1412 | sector_t nr_sects, gfp_t gfp_mask, int flags, | |
1413 | struct bio **biop); | |
1414 | ||
1415 | #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */ | |
1416 | #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */ | |
1417 | ||
1418 | extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, | |
1419 | sector_t nr_sects, gfp_t gfp_mask, struct bio **biop, | |
1420 | unsigned flags); | |
1421 | extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, | |
1422 | sector_t nr_sects, gfp_t gfp_mask, unsigned flags); | |
1423 | ||
1424 | static inline int sb_issue_discard(struct super_block *sb, sector_t block, | |
1425 | sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags) | |
1426 | { | |
1427 | return blkdev_issue_discard(sb->s_bdev, | |
1428 | block << (sb->s_blocksize_bits - | |
1429 | SECTOR_SHIFT), | |
1430 | nr_blocks << (sb->s_blocksize_bits - | |
1431 | SECTOR_SHIFT), | |
1432 | gfp_mask, flags); | |
1433 | } | |
1434 | static inline int sb_issue_zeroout(struct super_block *sb, sector_t block, | |
1435 | sector_t nr_blocks, gfp_t gfp_mask) | |
1436 | { | |
1437 | return blkdev_issue_zeroout(sb->s_bdev, | |
1438 | block << (sb->s_blocksize_bits - | |
1439 | SECTOR_SHIFT), | |
1440 | nr_blocks << (sb->s_blocksize_bits - | |
1441 | SECTOR_SHIFT), | |
1442 | gfp_mask, 0); | |
1443 | } | |
1444 | ||
1445 | extern int blk_verify_command(unsigned char *cmd, fmode_t mode); | |
1446 | ||
1447 | enum blk_default_limits { | |
1448 | BLK_MAX_SEGMENTS = 128, | |
1449 | BLK_SAFE_MAX_SECTORS = 255, | |
1450 | BLK_DEF_MAX_SECTORS = 2560, | |
1451 | BLK_MAX_SEGMENT_SIZE = 65536, | |
1452 | BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL, | |
1453 | }; | |
1454 | ||
1455 | #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist) | |
1456 | ||
1457 | static inline unsigned long queue_segment_boundary(struct request_queue *q) | |
1458 | { | |
1459 | return q->limits.seg_boundary_mask; | |
1460 | } | |
1461 | ||
1462 | static inline unsigned long queue_virt_boundary(struct request_queue *q) | |
1463 | { | |
1464 | return q->limits.virt_boundary_mask; | |
1465 | } | |
1466 | ||
1467 | static inline unsigned int queue_max_sectors(struct request_queue *q) | |
1468 | { | |
1469 | return q->limits.max_sectors; | |
1470 | } | |
1471 | ||
1472 | static inline unsigned int queue_max_hw_sectors(struct request_queue *q) | |
1473 | { | |
1474 | return q->limits.max_hw_sectors; | |
1475 | } | |
1476 | ||
1477 | static inline unsigned short queue_max_segments(struct request_queue *q) | |
1478 | { | |
1479 | return q->limits.max_segments; | |
1480 | } | |
1481 | ||
1482 | static inline unsigned short queue_max_discard_segments(struct request_queue *q) | |
1483 | { | |
1484 | return q->limits.max_discard_segments; | |
1485 | } | |
1486 | ||
1487 | static inline unsigned int queue_max_segment_size(struct request_queue *q) | |
1488 | { | |
1489 | return q->limits.max_segment_size; | |
1490 | } | |
1491 | ||
1492 | static inline unsigned short queue_logical_block_size(struct request_queue *q) | |
1493 | { | |
1494 | int retval = 512; | |
1495 | ||
1496 | if (q && q->limits.logical_block_size) | |
1497 | retval = q->limits.logical_block_size; | |
1498 | ||
1499 | return retval; | |
1500 | } | |
1501 | ||
1502 | static inline unsigned short bdev_logical_block_size(struct block_device *bdev) | |
1503 | { | |
1504 | return queue_logical_block_size(bdev_get_queue(bdev)); | |
1505 | } | |
1506 | ||
1507 | static inline unsigned int queue_physical_block_size(struct request_queue *q) | |
1508 | { | |
1509 | return q->limits.physical_block_size; | |
1510 | } | |
1511 | ||
1512 | static inline unsigned int bdev_physical_block_size(struct block_device *bdev) | |
1513 | { | |
1514 | return queue_physical_block_size(bdev_get_queue(bdev)); | |
1515 | } | |
1516 | ||
1517 | static inline unsigned int queue_io_min(struct request_queue *q) | |
1518 | { | |
1519 | return q->limits.io_min; | |
1520 | } | |
1521 | ||
1522 | static inline int bdev_io_min(struct block_device *bdev) | |
1523 | { | |
1524 | return queue_io_min(bdev_get_queue(bdev)); | |
1525 | } | |
1526 | ||
1527 | static inline unsigned int queue_io_opt(struct request_queue *q) | |
1528 | { | |
1529 | return q->limits.io_opt; | |
1530 | } | |
1531 | ||
1532 | static inline int bdev_io_opt(struct block_device *bdev) | |
1533 | { | |
1534 | return queue_io_opt(bdev_get_queue(bdev)); | |
1535 | } | |
1536 | ||
1537 | static inline int queue_alignment_offset(struct request_queue *q) | |
1538 | { | |
1539 | if (q->limits.misaligned) | |
1540 | return -1; | |
1541 | ||
1542 | return q->limits.alignment_offset; | |
1543 | } | |
1544 | ||
1545 | static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector) | |
1546 | { | |
1547 | unsigned int granularity = max(lim->physical_block_size, lim->io_min); | |
1548 | unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT) | |
1549 | << SECTOR_SHIFT; | |
1550 | ||
1551 | return (granularity + lim->alignment_offset - alignment) % granularity; | |
1552 | } | |
1553 | ||
1554 | static inline int bdev_alignment_offset(struct block_device *bdev) | |
1555 | { | |
1556 | struct request_queue *q = bdev_get_queue(bdev); | |
1557 | ||
1558 | if (q->limits.misaligned) | |
1559 | return -1; | |
1560 | ||
1561 | if (bdev != bdev->bd_contains) | |
1562 | return bdev->bd_part->alignment_offset; | |
1563 | ||
1564 | return q->limits.alignment_offset; | |
1565 | } | |
1566 | ||
1567 | static inline int queue_discard_alignment(struct request_queue *q) | |
1568 | { | |
1569 | if (q->limits.discard_misaligned) | |
1570 | return -1; | |
1571 | ||
1572 | return q->limits.discard_alignment; | |
1573 | } | |
1574 | ||
1575 | static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector) | |
1576 | { | |
1577 | unsigned int alignment, granularity, offset; | |
1578 | ||
1579 | if (!lim->max_discard_sectors) | |
1580 | return 0; | |
1581 | ||
1582 | /* Why are these in bytes, not sectors? */ | |
1583 | alignment = lim->discard_alignment >> SECTOR_SHIFT; | |
1584 | granularity = lim->discard_granularity >> SECTOR_SHIFT; | |
1585 | if (!granularity) | |
1586 | return 0; | |
1587 | ||
1588 | /* Offset of the partition start in 'granularity' sectors */ | |
1589 | offset = sector_div(sector, granularity); | |
1590 | ||
1591 | /* And why do we do this modulus *again* in blkdev_issue_discard()? */ | |
1592 | offset = (granularity + alignment - offset) % granularity; | |
1593 | ||
1594 | /* Turn it back into bytes, gaah */ | |
1595 | return offset << SECTOR_SHIFT; | |
1596 | } | |
1597 | ||
1598 | static inline int bdev_discard_alignment(struct block_device *bdev) | |
1599 | { | |
1600 | struct request_queue *q = bdev_get_queue(bdev); | |
1601 | ||
1602 | if (bdev != bdev->bd_contains) | |
1603 | return bdev->bd_part->discard_alignment; | |
1604 | ||
1605 | return q->limits.discard_alignment; | |
1606 | } | |
1607 | ||
1608 | static inline unsigned int bdev_write_same(struct block_device *bdev) | |
1609 | { | |
1610 | struct request_queue *q = bdev_get_queue(bdev); | |
1611 | ||
1612 | if (q) | |
1613 | return q->limits.max_write_same_sectors; | |
1614 | ||
1615 | return 0; | |
1616 | } | |
1617 | ||
1618 | static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev) | |
1619 | { | |
1620 | struct request_queue *q = bdev_get_queue(bdev); | |
1621 | ||
1622 | if (q) | |
1623 | return q->limits.max_write_zeroes_sectors; | |
1624 | ||
1625 | return 0; | |
1626 | } | |
1627 | ||
1628 | static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev) | |
1629 | { | |
1630 | struct request_queue *q = bdev_get_queue(bdev); | |
1631 | ||
1632 | if (q) | |
1633 | return blk_queue_zoned_model(q); | |
1634 | ||
1635 | return BLK_ZONED_NONE; | |
1636 | } | |
1637 | ||
1638 | static inline bool bdev_is_zoned(struct block_device *bdev) | |
1639 | { | |
1640 | struct request_queue *q = bdev_get_queue(bdev); | |
1641 | ||
1642 | if (q) | |
1643 | return blk_queue_is_zoned(q); | |
1644 | ||
1645 | return false; | |
1646 | } | |
1647 | ||
1648 | static inline unsigned int bdev_zone_sectors(struct block_device *bdev) | |
1649 | { | |
1650 | struct request_queue *q = bdev_get_queue(bdev); | |
1651 | ||
1652 | if (q) | |
1653 | return blk_queue_zone_sectors(q); | |
1654 | return 0; | |
1655 | } | |
1656 | ||
1657 | static inline unsigned int bdev_nr_zones(struct block_device *bdev) | |
1658 | { | |
1659 | struct request_queue *q = bdev_get_queue(bdev); | |
1660 | ||
1661 | if (q) | |
1662 | return blk_queue_nr_zones(q); | |
1663 | return 0; | |
1664 | } | |
1665 | ||
1666 | static inline int queue_dma_alignment(struct request_queue *q) | |
1667 | { | |
1668 | return q ? q->dma_alignment : 511; | |
1669 | } | |
1670 | ||
1671 | static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr, | |
1672 | unsigned int len) | |
1673 | { | |
1674 | unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask; | |
1675 | return !(addr & alignment) && !(len & alignment); | |
1676 | } | |
1677 | ||
1678 | /* assumes size > 256 */ | |
1679 | static inline unsigned int blksize_bits(unsigned int size) | |
1680 | { | |
1681 | unsigned int bits = 8; | |
1682 | do { | |
1683 | bits++; | |
1684 | size >>= 1; | |
1685 | } while (size > 256); | |
1686 | return bits; | |
1687 | } | |
1688 | ||
1689 | static inline unsigned int block_size(struct block_device *bdev) | |
1690 | { | |
1691 | return bdev->bd_block_size; | |
1692 | } | |
1693 | ||
1694 | static inline bool queue_flush_queueable(struct request_queue *q) | |
1695 | { | |
1696 | return !test_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags); | |
1697 | } | |
1698 | ||
1699 | typedef struct {struct page *v;} Sector; | |
1700 | ||
1701 | unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *); | |
1702 | ||
1703 | static inline void put_dev_sector(Sector p) | |
1704 | { | |
1705 | put_page(p.v); | |
1706 | } | |
1707 | ||
1708 | static inline bool __bvec_gap_to_prev(struct request_queue *q, | |
1709 | struct bio_vec *bprv, unsigned int offset) | |
1710 | { | |
1711 | return offset || | |
1712 | ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q)); | |
1713 | } | |
1714 | ||
1715 | /* | |
1716 | * Check if adding a bio_vec after bprv with offset would create a gap in | |
1717 | * the SG list. Most drivers don't care about this, but some do. | |
1718 | */ | |
1719 | static inline bool bvec_gap_to_prev(struct request_queue *q, | |
1720 | struct bio_vec *bprv, unsigned int offset) | |
1721 | { | |
1722 | if (!queue_virt_boundary(q)) | |
1723 | return false; | |
1724 | return __bvec_gap_to_prev(q, bprv, offset); | |
1725 | } | |
1726 | ||
1727 | /* | |
1728 | * Check if the two bvecs from two bios can be merged to one segment. | |
1729 | * If yes, no need to check gap between the two bios since the 1st bio | |
1730 | * and the 1st bvec in the 2nd bio can be handled in one segment. | |
1731 | */ | |
1732 | static inline bool bios_segs_mergeable(struct request_queue *q, | |
1733 | struct bio *prev, struct bio_vec *prev_last_bv, | |
1734 | struct bio_vec *next_first_bv) | |
1735 | { | |
1736 | if (!BIOVEC_PHYS_MERGEABLE(prev_last_bv, next_first_bv)) | |
1737 | return false; | |
1738 | if (!BIOVEC_SEG_BOUNDARY(q, prev_last_bv, next_first_bv)) | |
1739 | return false; | |
1740 | if (prev->bi_seg_back_size + next_first_bv->bv_len > | |
1741 | queue_max_segment_size(q)) | |
1742 | return false; | |
1743 | return true; | |
1744 | } | |
1745 | ||
1746 | static inline bool bio_will_gap(struct request_queue *q, | |
1747 | struct request *prev_rq, | |
1748 | struct bio *prev, | |
1749 | struct bio *next) | |
1750 | { | |
1751 | if (bio_has_data(prev) && queue_virt_boundary(q)) { | |
1752 | struct bio_vec pb, nb; | |
1753 | ||
1754 | /* | |
1755 | * don't merge if the 1st bio starts with non-zero | |
1756 | * offset, otherwise it is quite difficult to respect | |
1757 | * sg gap limit. We work hard to merge a huge number of small | |
1758 | * single bios in case of mkfs. | |
1759 | */ | |
1760 | if (prev_rq) | |
1761 | bio_get_first_bvec(prev_rq->bio, &pb); | |
1762 | else | |
1763 | bio_get_first_bvec(prev, &pb); | |
1764 | if (pb.bv_offset) | |
1765 | return true; | |
1766 | ||
1767 | /* | |
1768 | * We don't need to worry about the situation that the | |
1769 | * merged segment ends in unaligned virt boundary: | |
1770 | * | |
1771 | * - if 'pb' ends aligned, the merged segment ends aligned | |
1772 | * - if 'pb' ends unaligned, the next bio must include | |
1773 | * one single bvec of 'nb', otherwise the 'nb' can't | |
1774 | * merge with 'pb' | |
1775 | */ | |
1776 | bio_get_last_bvec(prev, &pb); | |
1777 | bio_get_first_bvec(next, &nb); | |
1778 | ||
1779 | if (!bios_segs_mergeable(q, prev, &pb, &nb)) | |
1780 | return __bvec_gap_to_prev(q, &pb, nb.bv_offset); | |
1781 | } | |
1782 | ||
1783 | return false; | |
1784 | } | |
1785 | ||
1786 | static inline bool req_gap_back_merge(struct request *req, struct bio *bio) | |
1787 | { | |
1788 | return bio_will_gap(req->q, req, req->biotail, bio); | |
1789 | } | |
1790 | ||
1791 | static inline bool req_gap_front_merge(struct request *req, struct bio *bio) | |
1792 | { | |
1793 | return bio_will_gap(req->q, NULL, bio, req->bio); | |
1794 | } | |
1795 | ||
1796 | int kblockd_schedule_work(struct work_struct *work); | |
1797 | int kblockd_schedule_work_on(int cpu, struct work_struct *work); | |
1798 | int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay); | |
1799 | ||
1800 | #define MODULE_ALIAS_BLOCKDEV(major,minor) \ | |
1801 | MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor)) | |
1802 | #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \ | |
1803 | MODULE_ALIAS("block-major-" __stringify(major) "-*") | |
1804 | ||
1805 | #if defined(CONFIG_BLK_DEV_INTEGRITY) | |
1806 | ||
1807 | enum blk_integrity_flags { | |
1808 | BLK_INTEGRITY_VERIFY = 1 << 0, | |
1809 | BLK_INTEGRITY_GENERATE = 1 << 1, | |
1810 | BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2, | |
1811 | BLK_INTEGRITY_IP_CHECKSUM = 1 << 3, | |
1812 | }; | |
1813 | ||
1814 | struct blk_integrity_iter { | |
1815 | void *prot_buf; | |
1816 | void *data_buf; | |
1817 | sector_t seed; | |
1818 | unsigned int data_size; | |
1819 | unsigned short interval; | |
1820 | const char *disk_name; | |
1821 | }; | |
1822 | ||
1823 | typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *); | |
1824 | ||
1825 | struct blk_integrity_profile { | |
1826 | integrity_processing_fn *generate_fn; | |
1827 | integrity_processing_fn *verify_fn; | |
1828 | const char *name; | |
1829 | }; | |
1830 | ||
1831 | extern void blk_integrity_register(struct gendisk *, struct blk_integrity *); | |
1832 | extern void blk_integrity_unregister(struct gendisk *); | |
1833 | extern int blk_integrity_compare(struct gendisk *, struct gendisk *); | |
1834 | extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *, | |
1835 | struct scatterlist *); | |
1836 | extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *); | |
1837 | extern bool blk_integrity_merge_rq(struct request_queue *, struct request *, | |
1838 | struct request *); | |
1839 | extern bool blk_integrity_merge_bio(struct request_queue *, struct request *, | |
1840 | struct bio *); | |
1841 | ||
1842 | static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk) | |
1843 | { | |
1844 | struct blk_integrity *bi = &disk->queue->integrity; | |
1845 | ||
1846 | if (!bi->profile) | |
1847 | return NULL; | |
1848 | ||
1849 | return bi; | |
1850 | } | |
1851 | ||
1852 | static inline | |
1853 | struct blk_integrity *bdev_get_integrity(struct block_device *bdev) | |
1854 | { | |
1855 | return blk_get_integrity(bdev->bd_disk); | |
1856 | } | |
1857 | ||
1858 | static inline bool blk_integrity_rq(struct request *rq) | |
1859 | { | |
1860 | return rq->cmd_flags & REQ_INTEGRITY; | |
1861 | } | |
1862 | ||
1863 | static inline void blk_queue_max_integrity_segments(struct request_queue *q, | |
1864 | unsigned int segs) | |
1865 | { | |
1866 | q->limits.max_integrity_segments = segs; | |
1867 | } | |
1868 | ||
1869 | static inline unsigned short | |
1870 | queue_max_integrity_segments(struct request_queue *q) | |
1871 | { | |
1872 | return q->limits.max_integrity_segments; | |
1873 | } | |
1874 | ||
1875 | static inline bool integrity_req_gap_back_merge(struct request *req, | |
1876 | struct bio *next) | |
1877 | { | |
1878 | struct bio_integrity_payload *bip = bio_integrity(req->bio); | |
1879 | struct bio_integrity_payload *bip_next = bio_integrity(next); | |
1880 | ||
1881 | return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1], | |
1882 | bip_next->bip_vec[0].bv_offset); | |
1883 | } | |
1884 | ||
1885 | static inline bool integrity_req_gap_front_merge(struct request *req, | |
1886 | struct bio *bio) | |
1887 | { | |
1888 | struct bio_integrity_payload *bip = bio_integrity(bio); | |
1889 | struct bio_integrity_payload *bip_next = bio_integrity(req->bio); | |
1890 | ||
1891 | return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1], | |
1892 | bip_next->bip_vec[0].bv_offset); | |
1893 | } | |
1894 | ||
1895 | #else /* CONFIG_BLK_DEV_INTEGRITY */ | |
1896 | ||
1897 | struct bio; | |
1898 | struct block_device; | |
1899 | struct gendisk; | |
1900 | struct blk_integrity; | |
1901 | ||
1902 | static inline int blk_integrity_rq(struct request *rq) | |
1903 | { | |
1904 | return 0; | |
1905 | } | |
1906 | static inline int blk_rq_count_integrity_sg(struct request_queue *q, | |
1907 | struct bio *b) | |
1908 | { | |
1909 | return 0; | |
1910 | } | |
1911 | static inline int blk_rq_map_integrity_sg(struct request_queue *q, | |
1912 | struct bio *b, | |
1913 | struct scatterlist *s) | |
1914 | { | |
1915 | return 0; | |
1916 | } | |
1917 | static inline struct blk_integrity *bdev_get_integrity(struct block_device *b) | |
1918 | { | |
1919 | return NULL; | |
1920 | } | |
1921 | static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk) | |
1922 | { | |
1923 | return NULL; | |
1924 | } | |
1925 | static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b) | |
1926 | { | |
1927 | return 0; | |
1928 | } | |
1929 | static inline void blk_integrity_register(struct gendisk *d, | |
1930 | struct blk_integrity *b) | |
1931 | { | |
1932 | } | |
1933 | static inline void blk_integrity_unregister(struct gendisk *d) | |
1934 | { | |
1935 | } | |
1936 | static inline void blk_queue_max_integrity_segments(struct request_queue *q, | |
1937 | unsigned int segs) | |
1938 | { | |
1939 | } | |
1940 | static inline unsigned short queue_max_integrity_segments(struct request_queue *q) | |
1941 | { | |
1942 | return 0; | |
1943 | } | |
1944 | static inline bool blk_integrity_merge_rq(struct request_queue *rq, | |
1945 | struct request *r1, | |
1946 | struct request *r2) | |
1947 | { | |
1948 | return true; | |
1949 | } | |
1950 | static inline bool blk_integrity_merge_bio(struct request_queue *rq, | |
1951 | struct request *r, | |
1952 | struct bio *b) | |
1953 | { | |
1954 | return true; | |
1955 | } | |
1956 | ||
1957 | static inline bool integrity_req_gap_back_merge(struct request *req, | |
1958 | struct bio *next) | |
1959 | { | |
1960 | return false; | |
1961 | } | |
1962 | static inline bool integrity_req_gap_front_merge(struct request *req, | |
1963 | struct bio *bio) | |
1964 | { | |
1965 | return false; | |
1966 | } | |
1967 | ||
1968 | #endif /* CONFIG_BLK_DEV_INTEGRITY */ | |
1969 | ||
1970 | struct block_device_operations { | |
1971 | int (*open) (struct block_device *, fmode_t); | |
1972 | void (*release) (struct gendisk *, fmode_t); | |
1973 | int (*rw_page)(struct block_device *, sector_t, struct page *, bool); | |
1974 | int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long); | |
1975 | int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long); | |
1976 | unsigned int (*check_events) (struct gendisk *disk, | |
1977 | unsigned int clearing); | |
1978 | /* ->media_changed() is DEPRECATED, use ->check_events() instead */ | |
1979 | int (*media_changed) (struct gendisk *); | |
1980 | void (*unlock_native_capacity) (struct gendisk *); | |
1981 | int (*revalidate_disk) (struct gendisk *); | |
1982 | int (*getgeo)(struct block_device *, struct hd_geometry *); | |
1983 | /* this callback is with swap_lock and sometimes page table lock held */ | |
1984 | void (*swap_slot_free_notify) (struct block_device *, unsigned long); | |
1985 | struct module *owner; | |
1986 | const struct pr_ops *pr_ops; | |
1987 | }; | |
1988 | ||
1989 | extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int, | |
1990 | unsigned long); | |
1991 | extern int bdev_read_page(struct block_device *, sector_t, struct page *); | |
1992 | extern int bdev_write_page(struct block_device *, sector_t, struct page *, | |
1993 | struct writeback_control *); | |
1994 | ||
1995 | #ifdef CONFIG_BLK_DEV_ZONED | |
1996 | bool blk_req_needs_zone_write_lock(struct request *rq); | |
1997 | void __blk_req_zone_write_lock(struct request *rq); | |
1998 | void __blk_req_zone_write_unlock(struct request *rq); | |
1999 | ||
2000 | static inline void blk_req_zone_write_lock(struct request *rq) | |
2001 | { | |
2002 | if (blk_req_needs_zone_write_lock(rq)) | |
2003 | __blk_req_zone_write_lock(rq); | |
2004 | } | |
2005 | ||
2006 | static inline void blk_req_zone_write_unlock(struct request *rq) | |
2007 | { | |
2008 | if (rq->rq_flags & RQF_ZONE_WRITE_LOCKED) | |
2009 | __blk_req_zone_write_unlock(rq); | |
2010 | } | |
2011 | ||
2012 | static inline bool blk_req_zone_is_write_locked(struct request *rq) | |
2013 | { | |
2014 | return rq->q->seq_zones_wlock && | |
2015 | test_bit(blk_rq_zone_no(rq), rq->q->seq_zones_wlock); | |
2016 | } | |
2017 | ||
2018 | static inline bool blk_req_can_dispatch_to_zone(struct request *rq) | |
2019 | { | |
2020 | if (!blk_req_needs_zone_write_lock(rq)) | |
2021 | return true; | |
2022 | return !blk_req_zone_is_write_locked(rq); | |
2023 | } | |
2024 | #else | |
2025 | static inline bool blk_req_needs_zone_write_lock(struct request *rq) | |
2026 | { | |
2027 | return false; | |
2028 | } | |
2029 | ||
2030 | static inline void blk_req_zone_write_lock(struct request *rq) | |
2031 | { | |
2032 | } | |
2033 | ||
2034 | static inline void blk_req_zone_write_unlock(struct request *rq) | |
2035 | { | |
2036 | } | |
2037 | static inline bool blk_req_zone_is_write_locked(struct request *rq) | |
2038 | { | |
2039 | return false; | |
2040 | } | |
2041 | ||
2042 | static inline bool blk_req_can_dispatch_to_zone(struct request *rq) | |
2043 | { | |
2044 | return true; | |
2045 | } | |
2046 | #endif /* CONFIG_BLK_DEV_ZONED */ | |
2047 | ||
2048 | #else /* CONFIG_BLOCK */ | |
2049 | ||
2050 | struct block_device; | |
2051 | ||
2052 | /* | |
2053 | * stubs for when the block layer is configured out | |
2054 | */ | |
2055 | #define buffer_heads_over_limit 0 | |
2056 | ||
2057 | static inline long nr_blockdev_pages(void) | |
2058 | { | |
2059 | return 0; | |
2060 | } | |
2061 | ||
2062 | struct blk_plug { | |
2063 | }; | |
2064 | ||
2065 | static inline void blk_start_plug(struct blk_plug *plug) | |
2066 | { | |
2067 | } | |
2068 | ||
2069 | static inline void blk_finish_plug(struct blk_plug *plug) | |
2070 | { | |
2071 | } | |
2072 | ||
2073 | static inline void blk_flush_plug(struct task_struct *task) | |
2074 | { | |
2075 | } | |
2076 | ||
2077 | static inline void blk_schedule_flush_plug(struct task_struct *task) | |
2078 | { | |
2079 | } | |
2080 | ||
2081 | ||
2082 | static inline bool blk_needs_flush_plug(struct task_struct *tsk) | |
2083 | { | |
2084 | return false; | |
2085 | } | |
2086 | ||
2087 | static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask, | |
2088 | sector_t *error_sector) | |
2089 | { | |
2090 | return 0; | |
2091 | } | |
2092 | ||
2093 | #endif /* CONFIG_BLOCK */ | |
2094 | ||
2095 | #endif |