Commit | Line | Data |
---|---|---|
b2441318 | 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
320ae51f JA |
2 | #ifndef BLK_MQ_H |
3 | #define BLK_MQ_H | |
4 | ||
5 | #include <linux/blkdev.h> | |
88459642 | 6 | #include <linux/sbitmap.h> |
6a83e74d | 7 | #include <linux/srcu.h> |
fb01a293 | 8 | #include <linux/lockdep.h> |
24b83deb | 9 | #include <linux/scatterlist.h> |
320ae51f JA |
10 | |
11 | struct blk_mq_tags; | |
f70ced09 | 12 | struct blk_flush_queue; |
320ae51f | 13 | |
24b83deb CH |
14 | #define BLKDEV_MIN_RQ 4 |
15 | #define BLKDEV_MAX_RQ 128 /* Default maximum */ | |
16 | ||
17 | typedef void (rq_end_io_fn)(struct request *, blk_status_t); | |
18 | ||
19 | /* | |
20 | * request flags */ | |
21 | typedef __u32 __bitwise req_flags_t; | |
22 | ||
23 | /* drive already may have started this one */ | |
24 | #define RQF_STARTED ((__force req_flags_t)(1 << 1)) | |
25 | /* may not be passed by ioscheduler */ | |
26 | #define RQF_SOFTBARRIER ((__force req_flags_t)(1 << 3)) | |
27 | /* request for flush sequence */ | |
28 | #define RQF_FLUSH_SEQ ((__force req_flags_t)(1 << 4)) | |
29 | /* merge of different types, fail separately */ | |
30 | #define RQF_MIXED_MERGE ((__force req_flags_t)(1 << 5)) | |
31 | /* track inflight for MQ */ | |
32 | #define RQF_MQ_INFLIGHT ((__force req_flags_t)(1 << 6)) | |
33 | /* don't call prep for this one */ | |
34 | #define RQF_DONTPREP ((__force req_flags_t)(1 << 7)) | |
35 | /* vaguely specified driver internal error. Ignored by the block layer */ | |
36 | #define RQF_FAILED ((__force req_flags_t)(1 << 10)) | |
37 | /* don't warn about errors */ | |
38 | #define RQF_QUIET ((__force req_flags_t)(1 << 11)) | |
39 | /* elevator private data attached */ | |
40 | #define RQF_ELVPRIV ((__force req_flags_t)(1 << 12)) | |
41 | /* account into disk and partition IO statistics */ | |
42 | #define RQF_IO_STAT ((__force req_flags_t)(1 << 13)) | |
43 | /* runtime pm request */ | |
44 | #define RQF_PM ((__force req_flags_t)(1 << 15)) | |
45 | /* on IO scheduler merge hash */ | |
46 | #define RQF_HASHED ((__force req_flags_t)(1 << 16)) | |
47 | /* track IO completion time */ | |
48 | #define RQF_STATS ((__force req_flags_t)(1 << 17)) | |
49 | /* Look at ->special_vec for the actual data payload instead of the | |
50 | bio chain. */ | |
51 | #define RQF_SPECIAL_PAYLOAD ((__force req_flags_t)(1 << 18)) | |
52 | /* The per-zone write lock is held for this request */ | |
53 | #define RQF_ZONE_WRITE_LOCKED ((__force req_flags_t)(1 << 19)) | |
54 | /* already slept for hybrid poll */ | |
55 | #define RQF_MQ_POLL_SLEPT ((__force req_flags_t)(1 << 20)) | |
56 | /* ->timeout has been called, don't expire again */ | |
57 | #define RQF_TIMED_OUT ((__force req_flags_t)(1 << 21)) | |
58 | ||
59 | /* flags that prevent us from merging requests: */ | |
60 | #define RQF_NOMERGE_FLAGS \ | |
61 | (RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD) | |
62 | ||
63 | enum mq_rq_state { | |
64 | MQ_RQ_IDLE = 0, | |
65 | MQ_RQ_IN_FLIGHT = 1, | |
66 | MQ_RQ_COMPLETE = 2, | |
67 | }; | |
68 | ||
69 | /* | |
70 | * Try to put the fields that are referenced together in the same cacheline. | |
71 | * | |
72 | * If you modify this structure, make sure to update blk_rq_init() and | |
73 | * especially blk_mq_rq_ctx_init() to take care of the added fields. | |
74 | */ | |
75 | struct request { | |
76 | struct request_queue *q; | |
77 | struct blk_mq_ctx *mq_ctx; | |
78 | struct blk_mq_hw_ctx *mq_hctx; | |
79 | ||
80 | unsigned int cmd_flags; /* op and common flags */ | |
81 | req_flags_t rq_flags; | |
82 | ||
83 | int tag; | |
84 | int internal_tag; | |
85 | ||
86 | /* the following two fields are internal, NEVER access directly */ | |
87 | unsigned int __data_len; /* total data len */ | |
88 | sector_t __sector; /* sector cursor */ | |
89 | ||
90 | struct bio *bio; | |
91 | struct bio *biotail; | |
92 | ||
93 | struct list_head queuelist; | |
94 | ||
95 | /* | |
96 | * The hash is used inside the scheduler, and killed once the | |
97 | * request reaches the dispatch list. The ipi_list is only used | |
98 | * to queue the request for softirq completion, which is long | |
99 | * after the request has been unhashed (and even removed from | |
100 | * the dispatch list). | |
101 | */ | |
102 | union { | |
103 | struct hlist_node hash; /* merge hash */ | |
104 | struct llist_node ipi_list; | |
105 | }; | |
106 | ||
107 | /* | |
108 | * The rb_node is only used inside the io scheduler, requests | |
109 | * are pruned when moved to the dispatch queue. So let the | |
110 | * completion_data share space with the rb_node. | |
111 | */ | |
112 | union { | |
113 | struct rb_node rb_node; /* sort/lookup */ | |
114 | struct bio_vec special_vec; | |
115 | void *completion_data; | |
116 | int error_count; /* for legacy drivers, don't use */ | |
117 | }; | |
118 | ||
119 | /* | |
120 | * Three pointers are available for the IO schedulers, if they need | |
121 | * more they have to dynamically allocate it. Flush requests are | |
122 | * never put on the IO scheduler. So let the flush fields share | |
123 | * space with the elevator data. | |
124 | */ | |
125 | union { | |
126 | struct { | |
127 | struct io_cq *icq; | |
128 | void *priv[2]; | |
129 | } elv; | |
130 | ||
131 | struct { | |
132 | unsigned int seq; | |
133 | struct list_head list; | |
134 | rq_end_io_fn *saved_end_io; | |
135 | } flush; | |
136 | }; | |
137 | ||
138 | struct gendisk *rq_disk; | |
139 | struct block_device *part; | |
140 | #ifdef CONFIG_BLK_RQ_ALLOC_TIME | |
141 | /* Time that the first bio started allocating this request. */ | |
142 | u64 alloc_time_ns; | |
143 | #endif | |
144 | /* Time that this request was allocated for this IO. */ | |
145 | u64 start_time_ns; | |
146 | /* Time that I/O was submitted to the device. */ | |
147 | u64 io_start_time_ns; | |
148 | ||
149 | #ifdef CONFIG_BLK_WBT | |
150 | unsigned short wbt_flags; | |
151 | #endif | |
152 | /* | |
153 | * rq sectors used for blk stats. It has the same value | |
154 | * with blk_rq_sectors(rq), except that it never be zeroed | |
155 | * by completion. | |
156 | */ | |
157 | unsigned short stats_sectors; | |
158 | ||
159 | /* | |
160 | * Number of scatter-gather DMA addr+len pairs after | |
161 | * physical address coalescing is performed. | |
162 | */ | |
163 | unsigned short nr_phys_segments; | |
164 | ||
165 | #ifdef CONFIG_BLK_DEV_INTEGRITY | |
166 | unsigned short nr_integrity_segments; | |
167 | #endif | |
168 | ||
169 | #ifdef CONFIG_BLK_INLINE_ENCRYPTION | |
170 | struct bio_crypt_ctx *crypt_ctx; | |
171 | struct blk_ksm_keyslot *crypt_keyslot; | |
172 | #endif | |
173 | ||
174 | unsigned short write_hint; | |
175 | unsigned short ioprio; | |
176 | ||
177 | enum mq_rq_state state; | |
178 | refcount_t ref; | |
179 | ||
180 | unsigned int timeout; | |
181 | unsigned long deadline; | |
182 | ||
183 | union { | |
184 | struct __call_single_data csd; | |
185 | u64 fifo_time; | |
186 | }; | |
187 | ||
188 | /* | |
189 | * completion callback. | |
190 | */ | |
191 | rq_end_io_fn *end_io; | |
192 | void *end_io_data; | |
193 | }; | |
194 | ||
195 | #define req_op(req) \ | |
196 | ((req)->cmd_flags & REQ_OP_MASK) | |
197 | ||
198 | static inline bool blk_rq_is_passthrough(struct request *rq) | |
199 | { | |
200 | return blk_op_is_passthrough(req_op(rq)); | |
201 | } | |
202 | ||
203 | static inline unsigned short req_get_ioprio(struct request *req) | |
204 | { | |
205 | return req->ioprio; | |
206 | } | |
207 | ||
208 | #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ) | |
209 | ||
210 | #define rq_dma_dir(rq) \ | |
211 | (op_is_write(req_op(rq)) ? DMA_TO_DEVICE : DMA_FROM_DEVICE) | |
212 | ||
213 | enum blk_eh_timer_return { | |
214 | BLK_EH_DONE, /* drivers has completed the command */ | |
215 | BLK_EH_RESET_TIMER, /* reset timer and try again */ | |
216 | }; | |
217 | ||
218 | #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */ | |
219 | #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */ | |
220 | ||
fe644072 | 221 | /** |
d386732b AA |
222 | * struct blk_mq_hw_ctx - State for a hardware queue facing the hardware |
223 | * block device | |
fe644072 | 224 | */ |
320ae51f JA |
225 | struct blk_mq_hw_ctx { |
226 | struct { | |
d386732b | 227 | /** @lock: Protects the dispatch list. */ |
320ae51f | 228 | spinlock_t lock; |
d386732b AA |
229 | /** |
230 | * @dispatch: Used for requests that are ready to be | |
231 | * dispatched to the hardware but for some reason (e.g. lack of | |
232 | * resources) could not be sent to the hardware. As soon as the | |
233 | * driver can send new requests, requests at this list will | |
234 | * be sent first for a fairer dispatch. | |
235 | */ | |
320ae51f | 236 | struct list_head dispatch; |
d386732b AA |
237 | /** |
238 | * @state: BLK_MQ_S_* flags. Defines the state of the hw | |
239 | * queue (active, scheduled to restart, stopped). | |
240 | */ | |
241 | unsigned long state; | |
320ae51f JA |
242 | } ____cacheline_aligned_in_smp; |
243 | ||
d386732b AA |
244 | /** |
245 | * @run_work: Used for scheduling a hardware queue run at a later time. | |
246 | */ | |
9f993737 | 247 | struct delayed_work run_work; |
d386732b | 248 | /** @cpumask: Map of available CPUs where this hctx can run. */ |
e4043dcf | 249 | cpumask_var_t cpumask; |
d386732b AA |
250 | /** |
251 | * @next_cpu: Used by blk_mq_hctx_next_cpu() for round-robin CPU | |
252 | * selection from @cpumask. | |
253 | */ | |
506e931f | 254 | int next_cpu; |
d386732b AA |
255 | /** |
256 | * @next_cpu_batch: Counter of how many works left in the batch before | |
257 | * changing to the next CPU. | |
258 | */ | |
506e931f | 259 | int next_cpu_batch; |
320ae51f | 260 | |
d386732b AA |
261 | /** @flags: BLK_MQ_F_* flags. Defines the behaviour of the queue. */ |
262 | unsigned long flags; | |
320ae51f | 263 | |
d386732b AA |
264 | /** |
265 | * @sched_data: Pointer owned by the IO scheduler attached to a request | |
266 | * queue. It's up to the IO scheduler how to use this pointer. | |
267 | */ | |
bd166ef1 | 268 | void *sched_data; |
d386732b AA |
269 | /** |
270 | * @queue: Pointer to the request queue that owns this hardware context. | |
271 | */ | |
320ae51f | 272 | struct request_queue *queue; |
d386732b | 273 | /** @fq: Queue of requests that need to perform a flush operation. */ |
f70ced09 | 274 | struct blk_flush_queue *fq; |
320ae51f | 275 | |
d386732b AA |
276 | /** |
277 | * @driver_data: Pointer to data owned by the block driver that created | |
278 | * this hctx | |
279 | */ | |
320ae51f JA |
280 | void *driver_data; |
281 | ||
d386732b AA |
282 | /** |
283 | * @ctx_map: Bitmap for each software queue. If bit is on, there is a | |
284 | * pending request in that software queue. | |
285 | */ | |
88459642 | 286 | struct sbitmap ctx_map; |
1429d7c9 | 287 | |
d386732b AA |
288 | /** |
289 | * @dispatch_from: Software queue to be used when no scheduler was | |
290 | * selected. | |
291 | */ | |
b347689f | 292 | struct blk_mq_ctx *dispatch_from; |
d386732b AA |
293 | /** |
294 | * @dispatch_busy: Number used by blk_mq_update_dispatch_busy() to | |
295 | * decide if the hw_queue is busy using Exponential Weighted Moving | |
296 | * Average algorithm. | |
297 | */ | |
6e768717 | 298 | unsigned int dispatch_busy; |
b347689f | 299 | |
d386732b | 300 | /** @type: HCTX_TYPE_* flags. Type of hardware queue. */ |
f31967f0 | 301 | unsigned short type; |
d386732b | 302 | /** @nr_ctx: Number of software queues. */ |
f31967f0 | 303 | unsigned short nr_ctx; |
d386732b | 304 | /** @ctxs: Array of software queues. */ |
6e768717 | 305 | struct blk_mq_ctx **ctxs; |
4bb659b1 | 306 | |
d386732b | 307 | /** @dispatch_wait_lock: Lock for dispatch_wait queue. */ |
5815839b | 308 | spinlock_t dispatch_wait_lock; |
d386732b AA |
309 | /** |
310 | * @dispatch_wait: Waitqueue to put requests when there is no tag | |
311 | * available at the moment, to wait for another try in the future. | |
312 | */ | |
eb619fdb | 313 | wait_queue_entry_t dispatch_wait; |
d386732b AA |
314 | |
315 | /** | |
316 | * @wait_index: Index of next available dispatch_wait queue to insert | |
317 | * requests. | |
318 | */ | |
8537b120 | 319 | atomic_t wait_index; |
320ae51f | 320 | |
d386732b AA |
321 | /** |
322 | * @tags: Tags owned by the block driver. A tag at this set is only | |
323 | * assigned when a request is dispatched from a hardware queue. | |
324 | */ | |
320ae51f | 325 | struct blk_mq_tags *tags; |
d386732b AA |
326 | /** |
327 | * @sched_tags: Tags owned by I/O scheduler. If there is an I/O | |
328 | * scheduler associated with a request queue, a tag is assigned when | |
329 | * that request is allocated. Else, this member is not used. | |
330 | */ | |
bd166ef1 | 331 | struct blk_mq_tags *sched_tags; |
320ae51f | 332 | |
d386732b | 333 | /** @queued: Number of queued requests. */ |
320ae51f | 334 | unsigned long queued; |
d386732b | 335 | /** @run: Number of dispatched requests. */ |
320ae51f | 336 | unsigned long run; |
8d354f13 | 337 | #define BLK_MQ_MAX_DISPATCH_ORDER 7 |
d386732b | 338 | /** @dispatched: Number of dispatch requests by queue. */ |
320ae51f JA |
339 | unsigned long dispatched[BLK_MQ_MAX_DISPATCH_ORDER]; |
340 | ||
d386732b | 341 | /** @numa_node: NUMA node the storage adapter has been connected to. */ |
320ae51f | 342 | unsigned int numa_node; |
d386732b | 343 | /** @queue_num: Index of this hardware queue. */ |
17ded320 | 344 | unsigned int queue_num; |
320ae51f | 345 | |
d386732b AA |
346 | /** |
347 | * @nr_active: Number of active requests. Only used when a tag set is | |
348 | * shared across request queues. | |
349 | */ | |
0d2602ca JA |
350 | atomic_t nr_active; |
351 | ||
bf0beec0 ML |
352 | /** @cpuhp_online: List to store request if CPU is going to die */ |
353 | struct hlist_node cpuhp_online; | |
d386732b | 354 | /** @cpuhp_dead: List to store request if some CPU die. */ |
9467f859 | 355 | struct hlist_node cpuhp_dead; |
d386732b | 356 | /** @kobj: Kernel object for sysfs. */ |
320ae51f | 357 | struct kobject kobj; |
05229bee | 358 | |
d386732b | 359 | /** @poll_considered: Count times blk_poll() was called. */ |
6e219353 | 360 | unsigned long poll_considered; |
d386732b | 361 | /** @poll_invoked: Count how many requests blk_poll() polled. */ |
05229bee | 362 | unsigned long poll_invoked; |
d386732b | 363 | /** @poll_success: Count how many polled requests were completed. */ |
05229bee | 364 | unsigned long poll_success; |
9c1051aa OS |
365 | |
366 | #ifdef CONFIG_BLK_DEBUG_FS | |
d386732b AA |
367 | /** |
368 | * @debugfs_dir: debugfs directory for this hardware queue. Named | |
369 | * as cpu<cpu_number>. | |
370 | */ | |
9c1051aa | 371 | struct dentry *debugfs_dir; |
d386732b | 372 | /** @sched_debugfs_dir: debugfs directory for the scheduler. */ |
d332ce09 | 373 | struct dentry *sched_debugfs_dir; |
9c1051aa | 374 | #endif |
07319678 | 375 | |
2dd209f0 BVA |
376 | /** |
377 | * @hctx_list: if this hctx is not in use, this is an entry in | |
378 | * q->unused_hctx_list. | |
379 | */ | |
2f8f1336 ML |
380 | struct list_head hctx_list; |
381 | ||
d386732b AA |
382 | /** |
383 | * @srcu: Sleepable RCU. Use as lock when type of the hardware queue is | |
384 | * blocking (BLK_MQ_F_BLOCKING). Must be the last member - see also | |
385 | * blk_mq_hw_ctx_size(). | |
386 | */ | |
f36aaf8b | 387 | struct srcu_struct srcu[]; |
320ae51f JA |
388 | }; |
389 | ||
7a18312c | 390 | /** |
d386732b | 391 | * struct blk_mq_queue_map - Map software queues to hardware queues |
7a18312c BVA |
392 | * @mq_map: CPU ID to hardware queue index map. This is an array |
393 | * with nr_cpu_ids elements. Each element has a value in the range | |
394 | * [@queue_offset, @queue_offset + @nr_queues). | |
395 | * @nr_queues: Number of hardware queues to map CPU IDs onto. | |
396 | * @queue_offset: First hardware queue to map onto. Used by the PCIe NVMe | |
397 | * driver to map each hardware queue type (enum hctx_type) onto a distinct | |
398 | * set of hardware queues. | |
399 | */ | |
ed76e329 JA |
400 | struct blk_mq_queue_map { |
401 | unsigned int *mq_map; | |
402 | unsigned int nr_queues; | |
843477d4 | 403 | unsigned int queue_offset; |
ed76e329 JA |
404 | }; |
405 | ||
d386732b AA |
406 | /** |
407 | * enum hctx_type - Type of hardware queue | |
408 | * @HCTX_TYPE_DEFAULT: All I/O not otherwise accounted for. | |
409 | * @HCTX_TYPE_READ: Just for READ I/O. | |
410 | * @HCTX_TYPE_POLL: Polled I/O of any kind. | |
411 | * @HCTX_MAX_TYPES: Number of types of hctx. | |
412 | */ | |
e20ba6e1 | 413 | enum hctx_type { |
d386732b AA |
414 | HCTX_TYPE_DEFAULT, |
415 | HCTX_TYPE_READ, | |
416 | HCTX_TYPE_POLL, | |
e20ba6e1 CH |
417 | |
418 | HCTX_MAX_TYPES, | |
ed76e329 JA |
419 | }; |
420 | ||
7a18312c BVA |
421 | /** |
422 | * struct blk_mq_tag_set - tag set that can be shared between request queues | |
423 | * @map: One or more ctx -> hctx mappings. One map exists for each | |
424 | * hardware queue type (enum hctx_type) that the driver wishes | |
425 | * to support. There are no restrictions on maps being of the | |
426 | * same size, and it's perfectly legal to share maps between | |
427 | * types. | |
428 | * @nr_maps: Number of elements in the @map array. A number in the range | |
429 | * [1, HCTX_MAX_TYPES]. | |
430 | * @ops: Pointers to functions that implement block driver behavior. | |
431 | * @nr_hw_queues: Number of hardware queues supported by the block driver that | |
432 | * owns this data structure. | |
433 | * @queue_depth: Number of tags per hardware queue, reserved tags included. | |
434 | * @reserved_tags: Number of tags to set aside for BLK_MQ_REQ_RESERVED tag | |
435 | * allocations. | |
436 | * @cmd_size: Number of additional bytes to allocate per request. The block | |
437 | * driver owns these additional bytes. | |
438 | * @numa_node: NUMA node the storage adapter has been connected to. | |
439 | * @timeout: Request processing timeout in jiffies. | |
440 | * @flags: Zero or more BLK_MQ_F_* flags. | |
441 | * @driver_data: Pointer to data owned by the block driver that created this | |
442 | * tag set. | |
6a6223ec MCC |
443 | * @active_queues_shared_sbitmap: |
444 | * number of active request queues per tag set. | |
32bc15af JG |
445 | * @__bitmap_tags: A shared tags sbitmap, used over all hctx's |
446 | * @__breserved_tags: | |
447 | * A shared reserved tags sbitmap, used over all hctx's | |
7a18312c BVA |
448 | * @tags: Tag sets. One tag set per hardware queue. Has @nr_hw_queues |
449 | * elements. | |
450 | * @tag_list_lock: Serializes tag_list accesses. | |
451 | * @tag_list: List of the request queues that use this tag set. See also | |
452 | * request_queue.tag_set_list. | |
453 | */ | |
24d2f903 | 454 | struct blk_mq_tag_set { |
ed76e329 | 455 | struct blk_mq_queue_map map[HCTX_MAX_TYPES]; |
7a18312c | 456 | unsigned int nr_maps; |
f8a5b122 | 457 | const struct blk_mq_ops *ops; |
7a18312c BVA |
458 | unsigned int nr_hw_queues; |
459 | unsigned int queue_depth; | |
320ae51f | 460 | unsigned int reserved_tags; |
7a18312c | 461 | unsigned int cmd_size; |
320ae51f JA |
462 | int numa_node; |
463 | unsigned int timeout; | |
7a18312c | 464 | unsigned int flags; |
24d2f903 | 465 | void *driver_data; |
f1b49fdc | 466 | atomic_t active_queues_shared_sbitmap; |
24d2f903 | 467 | |
32bc15af JG |
468 | struct sbitmap_queue __bitmap_tags; |
469 | struct sbitmap_queue __breserved_tags; | |
24d2f903 | 470 | struct blk_mq_tags **tags; |
0d2602ca JA |
471 | |
472 | struct mutex tag_list_lock; | |
473 | struct list_head tag_list; | |
320ae51f JA |
474 | }; |
475 | ||
d386732b AA |
476 | /** |
477 | * struct blk_mq_queue_data - Data about a request inserted in a queue | |
478 | * | |
479 | * @rq: Request pointer. | |
480 | * @last: If it is the last request in the queue. | |
481 | */ | |
74c45052 JA |
482 | struct blk_mq_queue_data { |
483 | struct request *rq; | |
74c45052 JA |
484 | bool last; |
485 | }; | |
486 | ||
7baa8572 | 487 | typedef bool (busy_iter_fn)(struct blk_mq_hw_ctx *, struct request *, void *, |
81481eb4 | 488 | bool); |
7baa8572 | 489 | typedef bool (busy_tag_iter_fn)(struct request *, void *, bool); |
05229bee | 490 | |
d386732b AA |
491 | /** |
492 | * struct blk_mq_ops - Callback functions that implements block driver | |
493 | * behaviour. | |
494 | */ | |
320ae51f | 495 | struct blk_mq_ops { |
d386732b AA |
496 | /** |
497 | * @queue_rq: Queue a new request from block IO. | |
320ae51f | 498 | */ |
0516c2f6 DW |
499 | blk_status_t (*queue_rq)(struct blk_mq_hw_ctx *, |
500 | const struct blk_mq_queue_data *); | |
320ae51f | 501 | |
d386732b AA |
502 | /** |
503 | * @commit_rqs: If a driver uses bd->last to judge when to submit | |
504 | * requests to hardware, it must define this function. In case of errors | |
505 | * that make us stop issuing further requests, this hook serves the | |
d666ba98 JA |
506 | * purpose of kicking the hardware (which the last request otherwise |
507 | * would have done). | |
508 | */ | |
0516c2f6 | 509 | void (*commit_rqs)(struct blk_mq_hw_ctx *); |
d666ba98 | 510 | |
d386732b AA |
511 | /** |
512 | * @get_budget: Reserve budget before queue request, once .queue_rq is | |
de148297 ML |
513 | * run, it is driver's responsibility to release the |
514 | * reserved budget. Also we have to handle failure case | |
515 | * of .get_budget for avoiding I/O deadlock. | |
516 | */ | |
2a5a24aa | 517 | int (*get_budget)(struct request_queue *); |
0516c2f6 | 518 | |
d386732b AA |
519 | /** |
520 | * @put_budget: Release the reserved budget. | |
521 | */ | |
2a5a24aa | 522 | void (*put_budget)(struct request_queue *, int); |
de148297 | 523 | |
85367040 ML |
524 | /** |
525 | * @set_rq_budget_token: store rq's budget token | |
d022d18c ML |
526 | */ |
527 | void (*set_rq_budget_token)(struct request *, int); | |
85367040 ML |
528 | /** |
529 | * @get_rq_budget_token: retrieve rq's budget token | |
d022d18c ML |
530 | */ |
531 | int (*get_rq_budget_token)(struct request *); | |
532 | ||
d386732b AA |
533 | /** |
534 | * @timeout: Called on request timeout. | |
320ae51f | 535 | */ |
0516c2f6 | 536 | enum blk_eh_timer_return (*timeout)(struct request *, bool); |
320ae51f | 537 | |
d386732b AA |
538 | /** |
539 | * @poll: Called to poll for completion of a specific tag. | |
05229bee | 540 | */ |
0516c2f6 | 541 | int (*poll)(struct blk_mq_hw_ctx *); |
05229bee | 542 | |
d386732b AA |
543 | /** |
544 | * @complete: Mark the request as complete. | |
545 | */ | |
0516c2f6 | 546 | void (*complete)(struct request *); |
30a91cb4 | 547 | |
d386732b AA |
548 | /** |
549 | * @init_hctx: Called when the block layer side of a hardware queue has | |
550 | * been set up, allowing the driver to allocate/init matching | |
551 | * structures. | |
320ae51f | 552 | */ |
0516c2f6 | 553 | int (*init_hctx)(struct blk_mq_hw_ctx *, void *, unsigned int); |
d386732b AA |
554 | /** |
555 | * @exit_hctx: Ditto for exit/teardown. | |
556 | */ | |
0516c2f6 | 557 | void (*exit_hctx)(struct blk_mq_hw_ctx *, unsigned int); |
e9b267d9 | 558 | |
d386732b AA |
559 | /** |
560 | * @init_request: Called for every command allocated by the block layer | |
561 | * to allow the driver to set up driver specific data. | |
f70ced09 ML |
562 | * |
563 | * Tag greater than or equal to queue_depth is for setting up | |
564 | * flush request. | |
e9b267d9 | 565 | */ |
0516c2f6 DW |
566 | int (*init_request)(struct blk_mq_tag_set *set, struct request *, |
567 | unsigned int, unsigned int); | |
d386732b AA |
568 | /** |
569 | * @exit_request: Ditto for exit/teardown. | |
570 | */ | |
0516c2f6 DW |
571 | void (*exit_request)(struct blk_mq_tag_set *set, struct request *, |
572 | unsigned int); | |
d386732b AA |
573 | |
574 | /** | |
575 | * @initialize_rq_fn: Called from inside blk_get_request(). | |
576 | */ | |
d280bab3 | 577 | void (*initialize_rq_fn)(struct request *rq); |
da695ba2 | 578 | |
d386732b AA |
579 | /** |
580 | * @cleanup_rq: Called before freeing one request which isn't completed | |
581 | * yet, and usually for freeing the driver private data. | |
226b4fc7 | 582 | */ |
0516c2f6 | 583 | void (*cleanup_rq)(struct request *); |
226b4fc7 | 584 | |
d386732b AA |
585 | /** |
586 | * @busy: If set, returns whether or not this queue currently is busy. | |
9ba20527 | 587 | */ |
0516c2f6 | 588 | bool (*busy)(struct request_queue *); |
9ba20527 | 589 | |
d386732b AA |
590 | /** |
591 | * @map_queues: This allows drivers specify their own queue mapping by | |
592 | * overriding the setup-time function that builds the mq_map. | |
593 | */ | |
0516c2f6 | 594 | int (*map_queues)(struct blk_mq_tag_set *set); |
2836ee4b BVA |
595 | |
596 | #ifdef CONFIG_BLK_DEBUG_FS | |
d386732b AA |
597 | /** |
598 | * @show_rq: Used by the debugfs implementation to show driver-specific | |
2836ee4b BVA |
599 | * information about a request. |
600 | */ | |
601 | void (*show_rq)(struct seq_file *m, struct request *rq); | |
602 | #endif | |
320ae51f JA |
603 | }; |
604 | ||
605 | enum { | |
320ae51f | 606 | BLK_MQ_F_SHOULD_MERGE = 1 << 0, |
51db1c37 | 607 | BLK_MQ_F_TAG_QUEUE_SHARED = 1 << 1, |
bf0beec0 ML |
608 | /* |
609 | * Set when this device requires underlying blk-mq device for | |
610 | * completing IO: | |
611 | */ | |
612 | BLK_MQ_F_STACKING = 1 << 2, | |
32bc15af | 613 | BLK_MQ_F_TAG_HCTX_SHARED = 1 << 3, |
1b792f2f | 614 | BLK_MQ_F_BLOCKING = 1 << 5, |
90b71980 | 615 | /* Do not allow an I/O scheduler to be configured. */ |
d3484991 | 616 | BLK_MQ_F_NO_SCHED = 1 << 6, |
90b71980 BVA |
617 | /* |
618 | * Select 'none' during queue registration in case of a single hwq | |
619 | * or shared hwqs instead of 'mq-deadline'. | |
620 | */ | |
621 | BLK_MQ_F_NO_SCHED_BY_DEFAULT = 1 << 7, | |
24391c0d SL |
622 | BLK_MQ_F_ALLOC_POLICY_START_BIT = 8, |
623 | BLK_MQ_F_ALLOC_POLICY_BITS = 1, | |
320ae51f | 624 | |
5d12f905 | 625 | BLK_MQ_S_STOPPED = 0, |
0d2602ca | 626 | BLK_MQ_S_TAG_ACTIVE = 1, |
bd166ef1 | 627 | BLK_MQ_S_SCHED_RESTART = 2, |
320ae51f | 628 | |
bf0beec0 ML |
629 | /* hw queue is inactive after all its CPUs become offline */ |
630 | BLK_MQ_S_INACTIVE = 3, | |
631 | ||
a4391c64 | 632 | BLK_MQ_MAX_DEPTH = 10240, |
506e931f JA |
633 | |
634 | BLK_MQ_CPU_WORK_BATCH = 8, | |
320ae51f | 635 | }; |
24391c0d SL |
636 | #define BLK_MQ_FLAG_TO_ALLOC_POLICY(flags) \ |
637 | ((flags >> BLK_MQ_F_ALLOC_POLICY_START_BIT) & \ | |
638 | ((1 << BLK_MQ_F_ALLOC_POLICY_BITS) - 1)) | |
639 | #define BLK_ALLOC_POLICY_TO_MQ_FLAG(policy) \ | |
640 | ((policy & ((1 << BLK_MQ_F_ALLOC_POLICY_BITS) - 1)) \ | |
641 | << BLK_MQ_F_ALLOC_POLICY_START_BIT) | |
320ae51f | 642 | |
4dcc4874 CH |
643 | struct gendisk *__blk_mq_alloc_disk(struct blk_mq_tag_set *set, void *queuedata, |
644 | struct lock_class_key *lkclass); | |
b461dfc4 CH |
645 | #define blk_mq_alloc_disk(set, queuedata) \ |
646 | ({ \ | |
647 | static struct lock_class_key __key; \ | |
b461dfc4 | 648 | \ |
4dcc4874 | 649 | __blk_mq_alloc_disk(set, queuedata, &__key); \ |
b461dfc4 | 650 | }) |
24d2f903 | 651 | struct request_queue *blk_mq_init_queue(struct blk_mq_tag_set *); |
26a9750a CH |
652 | int blk_mq_init_allocated_queue(struct blk_mq_tag_set *set, |
653 | struct request_queue *q); | |
b21d5b30 | 654 | void blk_mq_unregister_dev(struct device *, struct request_queue *); |
320ae51f | 655 | |
24d2f903 | 656 | int blk_mq_alloc_tag_set(struct blk_mq_tag_set *set); |
cdb14e0f CH |
657 | int blk_mq_alloc_sq_tag_set(struct blk_mq_tag_set *set, |
658 | const struct blk_mq_ops *ops, unsigned int queue_depth, | |
659 | unsigned int set_flags); | |
24d2f903 CH |
660 | void blk_mq_free_tag_set(struct blk_mq_tag_set *set); |
661 | ||
320ae51f JA |
662 | void blk_mq_flush_plug_list(struct blk_plug *plug, bool from_schedule); |
663 | ||
320ae51f | 664 | void blk_mq_free_request(struct request *rq); |
6f3b0e8b | 665 | |
3c94d83c | 666 | bool blk_mq_queue_inflight(struct request_queue *q); |
ae879912 | 667 | |
6f3b0e8b | 668 | enum { |
9a95e4ef BVA |
669 | /* return when out of requests */ |
670 | BLK_MQ_REQ_NOWAIT = (__force blk_mq_req_flags_t)(1 << 0), | |
671 | /* allocate from reserved pool */ | |
672 | BLK_MQ_REQ_RESERVED = (__force blk_mq_req_flags_t)(1 << 1), | |
0854bcdc BVA |
673 | /* set RQF_PM */ |
674 | BLK_MQ_REQ_PM = (__force blk_mq_req_flags_t)(1 << 2), | |
6f3b0e8b CH |
675 | }; |
676 | ||
cd6ce148 | 677 | struct request *blk_mq_alloc_request(struct request_queue *q, unsigned int op, |
9a95e4ef | 678 | blk_mq_req_flags_t flags); |
cd6ce148 | 679 | struct request *blk_mq_alloc_request_hctx(struct request_queue *q, |
9a95e4ef BVA |
680 | unsigned int op, blk_mq_req_flags_t flags, |
681 | unsigned int hctx_idx); | |
0e62f51f | 682 | struct request *blk_mq_tag_to_rq(struct blk_mq_tags *tags, unsigned int tag); |
320ae51f | 683 | |
205fb5f5 BVA |
684 | enum { |
685 | BLK_MQ_UNIQUE_TAG_BITS = 16, | |
686 | BLK_MQ_UNIQUE_TAG_MASK = (1 << BLK_MQ_UNIQUE_TAG_BITS) - 1, | |
687 | }; | |
688 | ||
689 | u32 blk_mq_unique_tag(struct request *rq); | |
690 | ||
691 | static inline u16 blk_mq_unique_tag_to_hwq(u32 unique_tag) | |
692 | { | |
693 | return unique_tag >> BLK_MQ_UNIQUE_TAG_BITS; | |
694 | } | |
695 | ||
696 | static inline u16 blk_mq_unique_tag_to_tag(u32 unique_tag) | |
697 | { | |
698 | return unique_tag & BLK_MQ_UNIQUE_TAG_MASK; | |
699 | } | |
700 | ||
27a46989 PB |
701 | /** |
702 | * blk_mq_rq_state() - read the current MQ_RQ_* state of a request | |
703 | * @rq: target request. | |
704 | */ | |
705 | static inline enum mq_rq_state blk_mq_rq_state(struct request *rq) | |
706 | { | |
707 | return READ_ONCE(rq->state); | |
708 | } | |
709 | ||
710 | static inline int blk_mq_request_started(struct request *rq) | |
711 | { | |
712 | return blk_mq_rq_state(rq) != MQ_RQ_IDLE; | |
713 | } | |
714 | ||
715 | static inline int blk_mq_request_completed(struct request *rq) | |
716 | { | |
717 | return blk_mq_rq_state(rq) == MQ_RQ_COMPLETE; | |
718 | } | |
320ae51f | 719 | |
83fba8c8 CL |
720 | /* |
721 | * | |
722 | * Set the state to complete when completing a request from inside ->queue_rq. | |
723 | * This is used by drivers that want to ensure special complete actions that | |
724 | * need access to the request are called on failure, e.g. by nvme for | |
725 | * multipathing. | |
726 | */ | |
727 | static inline void blk_mq_set_request_complete(struct request *rq) | |
728 | { | |
729 | WRITE_ONCE(rq->state, MQ_RQ_COMPLETE); | |
730 | } | |
731 | ||
e2490073 | 732 | void blk_mq_start_request(struct request *rq); |
2a842aca CH |
733 | void blk_mq_end_request(struct request *rq, blk_status_t error); |
734 | void __blk_mq_end_request(struct request *rq, blk_status_t error); | |
320ae51f | 735 | |
2b053aca | 736 | void blk_mq_requeue_request(struct request *rq, bool kick_requeue_list); |
6fca6a61 | 737 | void blk_mq_kick_requeue_list(struct request_queue *q); |
2849450a | 738 | void blk_mq_delay_kick_requeue_list(struct request_queue *q, unsigned long msecs); |
15f73f5b | 739 | void blk_mq_complete_request(struct request *rq); |
40d09b53 | 740 | bool blk_mq_complete_request_remote(struct request *rq); |
fd001443 | 741 | bool blk_mq_queue_stopped(struct request_queue *q); |
320ae51f JA |
742 | void blk_mq_stop_hw_queue(struct blk_mq_hw_ctx *hctx); |
743 | void blk_mq_start_hw_queue(struct blk_mq_hw_ctx *hctx); | |
280d45f6 | 744 | void blk_mq_stop_hw_queues(struct request_queue *q); |
2f268556 | 745 | void blk_mq_start_hw_queues(struct request_queue *q); |
ae911c5e | 746 | void blk_mq_start_stopped_hw_queue(struct blk_mq_hw_ctx *hctx, bool async); |
1b4a3258 | 747 | void blk_mq_start_stopped_hw_queues(struct request_queue *q, bool async); |
97e01209 | 748 | void blk_mq_quiesce_queue(struct request_queue *q); |
e4e73913 | 749 | void blk_mq_unquiesce_queue(struct request_queue *q); |
7587a5ae | 750 | void blk_mq_delay_run_hw_queue(struct blk_mq_hw_ctx *hctx, unsigned long msecs); |
626fb735 | 751 | void blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx, bool async); |
b94ec296 | 752 | void blk_mq_run_hw_queues(struct request_queue *q, bool async); |
b9151e7b | 753 | void blk_mq_delay_run_hw_queues(struct request_queue *q, unsigned long msecs); |
e0489487 SG |
754 | void blk_mq_tagset_busy_iter(struct blk_mq_tag_set *tagset, |
755 | busy_tag_iter_fn *fn, void *priv); | |
f9934a80 | 756 | void blk_mq_tagset_wait_completed_request(struct blk_mq_tag_set *tagset); |
c761d96b | 757 | void blk_mq_freeze_queue(struct request_queue *q); |
b4c6a028 | 758 | void blk_mq_unfreeze_queue(struct request_queue *q); |
1671d522 | 759 | void blk_freeze_queue_start(struct request_queue *q); |
6bae363e | 760 | void blk_mq_freeze_queue_wait(struct request_queue *q); |
f91328c4 KB |
761 | int blk_mq_freeze_queue_wait_timeout(struct request_queue *q, |
762 | unsigned long timeout); | |
320ae51f | 763 | |
ed76e329 | 764 | int blk_mq_map_queues(struct blk_mq_queue_map *qmap); |
868f2f0b KB |
765 | void blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set, int nr_hw_queues); |
766 | ||
852ec809 | 767 | void blk_mq_quiesce_queue_nowait(struct request_queue *q); |
4f084b41 | 768 | |
9cf2bab6 JA |
769 | unsigned int blk_mq_rq_cpu(struct request *rq); |
770 | ||
15f73f5b CH |
771 | bool __blk_should_fake_timeout(struct request_queue *q); |
772 | static inline bool blk_should_fake_timeout(struct request_queue *q) | |
773 | { | |
774 | if (IS_ENABLED(CONFIG_FAIL_IO_TIMEOUT) && | |
775 | test_bit(QUEUE_FLAG_FAIL_IO, &q->queue_flags)) | |
776 | return __blk_should_fake_timeout(q); | |
777 | return false; | |
778 | } | |
779 | ||
d386732b AA |
780 | /** |
781 | * blk_mq_rq_from_pdu - cast a PDU to a request | |
782 | * @pdu: the PDU (Protocol Data Unit) to be casted | |
783 | * | |
784 | * Return: request | |
785 | * | |
320ae51f | 786 | * Driver command data is immediately after the request. So subtract request |
d386732b | 787 | * size to get back to the original request. |
320ae51f JA |
788 | */ |
789 | static inline struct request *blk_mq_rq_from_pdu(void *pdu) | |
790 | { | |
791 | return pdu - sizeof(struct request); | |
792 | } | |
d386732b AA |
793 | |
794 | /** | |
795 | * blk_mq_rq_to_pdu - cast a request to a PDU | |
796 | * @rq: the request to be casted | |
797 | * | |
798 | * Return: pointer to the PDU | |
799 | * | |
800 | * Driver command data is immediately after the request. So add request to get | |
801 | * the PDU. | |
802 | */ | |
320ae51f JA |
803 | static inline void *blk_mq_rq_to_pdu(struct request *rq) |
804 | { | |
2963e3f7 | 805 | return rq + 1; |
320ae51f JA |
806 | } |
807 | ||
320ae51f | 808 | #define queue_for_each_hw_ctx(q, hctx, i) \ |
0d0b7d42 JA |
809 | for ((i) = 0; (i) < (q)->nr_hw_queues && \ |
810 | ({ hctx = (q)->queue_hw_ctx[i]; 1; }); (i)++) | |
320ae51f | 811 | |
320ae51f | 812 | #define hctx_for_each_ctx(hctx, ctx, i) \ |
0d0b7d42 JA |
813 | for ((i) = 0; (i) < (hctx)->nr_ctx && \ |
814 | ({ ctx = (hctx)->ctxs[(i)]; 1; }); (i)++) | |
320ae51f | 815 | |
7b7ab780 SG |
816 | static inline blk_qc_t request_to_qc_t(struct blk_mq_hw_ctx *hctx, |
817 | struct request *rq) | |
818 | { | |
819 | if (rq->tag != -1) | |
820 | return rq->tag | (hctx->queue_num << BLK_QC_T_SHIFT); | |
821 | ||
822 | return rq->internal_tag | (hctx->queue_num << BLK_QC_T_SHIFT) | | |
823 | BLK_QC_T_INTERNAL; | |
824 | } | |
825 | ||
226b4fc7 ML |
826 | static inline void blk_mq_cleanup_rq(struct request *rq) |
827 | { | |
828 | if (rq->q->mq_ops->cleanup_rq) | |
829 | rq->q->mq_ops->cleanup_rq(rq); | |
830 | } | |
831 | ||
53ffabfd CK |
832 | static inline void blk_rq_bio_prep(struct request *rq, struct bio *bio, |
833 | unsigned int nr_segs) | |
834 | { | |
835 | rq->nr_phys_segments = nr_segs; | |
836 | rq->__data_len = bio->bi_iter.bi_size; | |
837 | rq->bio = rq->biotail = bio; | |
838 | rq->ioprio = bio_prio(bio); | |
839 | ||
309dca30 CH |
840 | if (bio->bi_bdev) |
841 | rq->rq_disk = bio->bi_bdev->bd_disk; | |
53ffabfd CK |
842 | } |
843 | ||
c62b37d9 | 844 | blk_qc_t blk_mq_submit_bio(struct bio *bio); |
fb01a293 ML |
845 | void blk_mq_hctx_set_fq_lock_class(struct blk_mq_hw_ctx *hctx, |
846 | struct lock_class_key *key); | |
8cf7961d | 847 | |
24b83deb CH |
848 | static inline bool rq_is_sync(struct request *rq) |
849 | { | |
850 | return op_is_sync(rq->cmd_flags); | |
851 | } | |
852 | ||
853 | void blk_rq_init(struct request_queue *q, struct request *rq); | |
854 | void blk_put_request(struct request *rq); | |
855 | struct request *blk_get_request(struct request_queue *q, unsigned int op, | |
856 | blk_mq_req_flags_t flags); | |
857 | int blk_rq_prep_clone(struct request *rq, struct request *rq_src, | |
858 | struct bio_set *bs, gfp_t gfp_mask, | |
859 | int (*bio_ctr)(struct bio *, struct bio *, void *), void *data); | |
860 | void blk_rq_unprep_clone(struct request *rq); | |
861 | blk_status_t blk_insert_cloned_request(struct request_queue *q, | |
862 | struct request *rq); | |
863 | ||
864 | struct rq_map_data { | |
865 | struct page **pages; | |
866 | int page_order; | |
867 | int nr_entries; | |
868 | unsigned long offset; | |
869 | int null_mapped; | |
870 | int from_user; | |
871 | }; | |
872 | ||
873 | int blk_rq_map_user(struct request_queue *, struct request *, | |
874 | struct rq_map_data *, void __user *, unsigned long, gfp_t); | |
875 | int blk_rq_map_user_iov(struct request_queue *, struct request *, | |
876 | struct rq_map_data *, const struct iov_iter *, gfp_t); | |
877 | int blk_rq_unmap_user(struct bio *); | |
878 | int blk_rq_map_kern(struct request_queue *, struct request *, void *, | |
879 | unsigned int, gfp_t); | |
880 | int blk_rq_append_bio(struct request *rq, struct bio *bio); | |
881 | void blk_execute_rq_nowait(struct gendisk *, struct request *, int, | |
882 | rq_end_io_fn *); | |
883 | blk_status_t blk_execute_rq(struct gendisk *bd_disk, struct request *rq, | |
884 | int at_head); | |
885 | ||
886 | struct req_iterator { | |
887 | struct bvec_iter iter; | |
888 | struct bio *bio; | |
889 | }; | |
890 | ||
891 | #define __rq_for_each_bio(_bio, rq) \ | |
892 | if ((rq->bio)) \ | |
893 | for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next) | |
894 | ||
895 | #define rq_for_each_segment(bvl, _rq, _iter) \ | |
896 | __rq_for_each_bio(_iter.bio, _rq) \ | |
897 | bio_for_each_segment(bvl, _iter.bio, _iter.iter) | |
898 | ||
899 | #define rq_for_each_bvec(bvl, _rq, _iter) \ | |
900 | __rq_for_each_bio(_iter.bio, _rq) \ | |
901 | bio_for_each_bvec(bvl, _iter.bio, _iter.iter) | |
902 | ||
903 | #define rq_iter_last(bvec, _iter) \ | |
904 | (_iter.bio->bi_next == NULL && \ | |
905 | bio_iter_last(bvec, _iter.iter)) | |
906 | ||
907 | /* | |
908 | * blk_rq_pos() : the current sector | |
909 | * blk_rq_bytes() : bytes left in the entire request | |
910 | * blk_rq_cur_bytes() : bytes left in the current segment | |
911 | * blk_rq_err_bytes() : bytes left till the next error boundary | |
912 | * blk_rq_sectors() : sectors left in the entire request | |
913 | * blk_rq_cur_sectors() : sectors left in the current segment | |
914 | * blk_rq_stats_sectors() : sectors of the entire request used for stats | |
915 | */ | |
916 | static inline sector_t blk_rq_pos(const struct request *rq) | |
917 | { | |
918 | return rq->__sector; | |
919 | } | |
920 | ||
921 | static inline unsigned int blk_rq_bytes(const struct request *rq) | |
922 | { | |
923 | return rq->__data_len; | |
924 | } | |
925 | ||
926 | static inline int blk_rq_cur_bytes(const struct request *rq) | |
927 | { | |
928 | return rq->bio ? bio_cur_bytes(rq->bio) : 0; | |
929 | } | |
930 | ||
931 | unsigned int blk_rq_err_bytes(const struct request *rq); | |
932 | ||
933 | static inline unsigned int blk_rq_sectors(const struct request *rq) | |
934 | { | |
935 | return blk_rq_bytes(rq) >> SECTOR_SHIFT; | |
936 | } | |
937 | ||
938 | static inline unsigned int blk_rq_cur_sectors(const struct request *rq) | |
939 | { | |
940 | return blk_rq_cur_bytes(rq) >> SECTOR_SHIFT; | |
941 | } | |
942 | ||
943 | static inline unsigned int blk_rq_stats_sectors(const struct request *rq) | |
944 | { | |
945 | return rq->stats_sectors; | |
946 | } | |
947 | ||
948 | /* | |
949 | * Some commands like WRITE SAME have a payload or data transfer size which | |
950 | * is different from the size of the request. Any driver that supports such | |
951 | * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to | |
952 | * calculate the data transfer size. | |
953 | */ | |
954 | static inline unsigned int blk_rq_payload_bytes(struct request *rq) | |
955 | { | |
956 | if (rq->rq_flags & RQF_SPECIAL_PAYLOAD) | |
957 | return rq->special_vec.bv_len; | |
958 | return blk_rq_bytes(rq); | |
959 | } | |
960 | ||
961 | /* | |
962 | * Return the first full biovec in the request. The caller needs to check that | |
963 | * there are any bvecs before calling this helper. | |
964 | */ | |
965 | static inline struct bio_vec req_bvec(struct request *rq) | |
966 | { | |
967 | if (rq->rq_flags & RQF_SPECIAL_PAYLOAD) | |
968 | return rq->special_vec; | |
969 | return mp_bvec_iter_bvec(rq->bio->bi_io_vec, rq->bio->bi_iter); | |
970 | } | |
971 | ||
972 | static inline unsigned int blk_rq_count_bios(struct request *rq) | |
973 | { | |
974 | unsigned int nr_bios = 0; | |
975 | struct bio *bio; | |
976 | ||
977 | __rq_for_each_bio(bio, rq) | |
978 | nr_bios++; | |
979 | ||
980 | return nr_bios; | |
981 | } | |
982 | ||
983 | void blk_steal_bios(struct bio_list *list, struct request *rq); | |
984 | ||
985 | /* | |
986 | * Request completion related functions. | |
987 | * | |
988 | * blk_update_request() completes given number of bytes and updates | |
989 | * the request without completing it. | |
990 | */ | |
991 | bool blk_update_request(struct request *rq, blk_status_t error, | |
992 | unsigned int nr_bytes); | |
993 | void blk_abort_request(struct request *); | |
994 | ||
995 | /* | |
996 | * Number of physical segments as sent to the device. | |
997 | * | |
998 | * Normally this is the number of discontiguous data segments sent by the | |
999 | * submitter. But for data-less command like discard we might have no | |
1000 | * actual data segments submitted, but the driver might have to add it's | |
1001 | * own special payload. In that case we still return 1 here so that this | |
1002 | * special payload will be mapped. | |
1003 | */ | |
1004 | static inline unsigned short blk_rq_nr_phys_segments(struct request *rq) | |
1005 | { | |
1006 | if (rq->rq_flags & RQF_SPECIAL_PAYLOAD) | |
1007 | return 1; | |
1008 | return rq->nr_phys_segments; | |
1009 | } | |
1010 | ||
1011 | /* | |
1012 | * Number of discard segments (or ranges) the driver needs to fill in. | |
1013 | * Each discard bio merged into a request is counted as one segment. | |
1014 | */ | |
1015 | static inline unsigned short blk_rq_nr_discard_segments(struct request *rq) | |
1016 | { | |
1017 | return max_t(unsigned short, rq->nr_phys_segments, 1); | |
1018 | } | |
1019 | ||
1020 | int __blk_rq_map_sg(struct request_queue *q, struct request *rq, | |
1021 | struct scatterlist *sglist, struct scatterlist **last_sg); | |
1022 | static inline int blk_rq_map_sg(struct request_queue *q, struct request *rq, | |
1023 | struct scatterlist *sglist) | |
1024 | { | |
1025 | struct scatterlist *last_sg = NULL; | |
1026 | ||
1027 | return __blk_rq_map_sg(q, rq, sglist, &last_sg); | |
1028 | } | |
1029 | void blk_dump_rq_flags(struct request *, char *); | |
1030 | ||
1031 | #ifdef CONFIG_BLK_DEV_ZONED | |
1032 | static inline unsigned int blk_rq_zone_no(struct request *rq) | |
1033 | { | |
1034 | return blk_queue_zone_no(rq->q, blk_rq_pos(rq)); | |
1035 | } | |
1036 | ||
1037 | static inline unsigned int blk_rq_zone_is_seq(struct request *rq) | |
1038 | { | |
1039 | return blk_queue_zone_is_seq(rq->q, blk_rq_pos(rq)); | |
1040 | } | |
1041 | ||
1042 | bool blk_req_needs_zone_write_lock(struct request *rq); | |
1043 | bool blk_req_zone_write_trylock(struct request *rq); | |
1044 | void __blk_req_zone_write_lock(struct request *rq); | |
1045 | void __blk_req_zone_write_unlock(struct request *rq); | |
1046 | ||
1047 | static inline void blk_req_zone_write_lock(struct request *rq) | |
1048 | { | |
1049 | if (blk_req_needs_zone_write_lock(rq)) | |
1050 | __blk_req_zone_write_lock(rq); | |
1051 | } | |
1052 | ||
1053 | static inline void blk_req_zone_write_unlock(struct request *rq) | |
1054 | { | |
1055 | if (rq->rq_flags & RQF_ZONE_WRITE_LOCKED) | |
1056 | __blk_req_zone_write_unlock(rq); | |
1057 | } | |
1058 | ||
1059 | static inline bool blk_req_zone_is_write_locked(struct request *rq) | |
1060 | { | |
1061 | return rq->q->seq_zones_wlock && | |
1062 | test_bit(blk_rq_zone_no(rq), rq->q->seq_zones_wlock); | |
1063 | } | |
1064 | ||
1065 | static inline bool blk_req_can_dispatch_to_zone(struct request *rq) | |
1066 | { | |
1067 | if (!blk_req_needs_zone_write_lock(rq)) | |
1068 | return true; | |
1069 | return !blk_req_zone_is_write_locked(rq); | |
1070 | } | |
1071 | #else /* CONFIG_BLK_DEV_ZONED */ | |
1072 | static inline bool blk_req_needs_zone_write_lock(struct request *rq) | |
1073 | { | |
1074 | return false; | |
1075 | } | |
1076 | ||
1077 | static inline void blk_req_zone_write_lock(struct request *rq) | |
1078 | { | |
1079 | } | |
1080 | ||
1081 | static inline void blk_req_zone_write_unlock(struct request *rq) | |
1082 | { | |
1083 | } | |
1084 | static inline bool blk_req_zone_is_write_locked(struct request *rq) | |
1085 | { | |
1086 | return false; | |
1087 | } | |
1088 | ||
1089 | static inline bool blk_req_can_dispatch_to_zone(struct request *rq) | |
1090 | { | |
1091 | return true; | |
1092 | } | |
1093 | #endif /* CONFIG_BLK_DEV_ZONED */ | |
1094 | ||
1095 | #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE | |
1096 | # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform" | |
320ae51f | 1097 | #endif |
24b83deb CH |
1098 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE |
1099 | void rq_flush_dcache_pages(struct request *rq); | |
1100 | #else | |
1101 | static inline void rq_flush_dcache_pages(struct request *rq) | |
1102 | { | |
1103 | } | |
1104 | #endif /* ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE */ | |
1105 | #endif /* BLK_MQ_H */ |