Merge tag 'nfsd-5.8' of git://linux-nfs.org/~bfields/linux
[linux-block.git] / block / blk.h
CommitLineData
b2441318 1/* SPDX-License-Identifier: GPL-2.0 */
8324aa91
JA
2#ifndef BLK_INTERNAL_H
3#define BLK_INTERNAL_H
4
a73f730d 5#include <linux/idr.h>
f70ced09 6#include <linux/blk-mq.h>
c6a564ff 7#include <linux/part_stat.h>
a892c8d5 8#include <linux/blk-crypto.h>
c39ae60d 9#include <xen/xen.h>
a892c8d5 10#include "blk-crypto-internal.h"
f70ced09 11#include "blk-mq.h"
c3e22192 12#include "blk-mq-sched.h"
a73f730d 13
0d2602ca
JA
14/* Max future timer expiry for timeouts */
15#define BLK_MAX_TIMEOUT (5 * HZ)
16
18fbda91
OS
17#ifdef CONFIG_DEBUG_FS
18extern struct dentry *blk_debugfs_root;
19#endif
20
7c94e1c1 21struct blk_flush_queue {
7c94e1c1
ML
22 unsigned int flush_pending_idx:1;
23 unsigned int flush_running_idx:1;
8d699663 24 blk_status_t rq_status;
7c94e1c1
ML
25 unsigned long flush_pending_since;
26 struct list_head flush_queue[2];
27 struct list_head flush_data_in_flight;
28 struct request *flush_rq;
0048b483
ML
29
30 /*
31 * flush_rq shares tag with this rq, both can't be active
32 * at the same time
33 */
34 struct request *orig_rq;
b3c6a599 35 struct lock_class_key key;
7c94e1c1
ML
36 spinlock_t mq_flush_lock;
37};
38
8324aa91
JA
39extern struct kmem_cache *blk_requestq_cachep;
40extern struct kobj_type blk_queue_ktype;
a73f730d 41extern struct ida blk_queue_ida;
8324aa91 42
f9afca4d
JA
43static inline struct blk_flush_queue *
44blk_get_flush_queue(struct request_queue *q, struct blk_mq_ctx *ctx)
7c94e1c1 45{
8ccdf4a3 46 return blk_mq_map_queue(q, REQ_OP_FLUSH, ctx)->fq;
7c94e1c1
ML
47}
48
09ac46c4
TH
49static inline void __blk_get_queue(struct request_queue *q)
50{
51 kobject_get(&q->kobj);
52}
53
8d699663
YY
54static inline bool
55is_flush_rq(struct request *req, struct blk_mq_hw_ctx *hctx)
56{
57 return hctx->fq->flush_rq == req;
58}
59
754a1572
GJ
60struct blk_flush_queue *blk_alloc_flush_queue(int node, int cmd_size,
61 gfp_t flags);
f70ced09 62void blk_free_flush_queue(struct blk_flush_queue *q);
f3552655 63
3ef28e83
DW
64void blk_freeze_queue(struct request_queue *q);
65
3dccdae5
CH
66static inline bool biovec_phys_mergeable(struct request_queue *q,
67 struct bio_vec *vec1, struct bio_vec *vec2)
6a9f5f24 68{
3dccdae5 69 unsigned long mask = queue_segment_boundary(q);
6e768461
CH
70 phys_addr_t addr1 = page_to_phys(vec1->bv_page) + vec1->bv_offset;
71 phys_addr_t addr2 = page_to_phys(vec2->bv_page) + vec2->bv_offset;
3dccdae5
CH
72
73 if (addr1 + vec1->bv_len != addr2)
6a9f5f24 74 return false;
0383ad43 75 if (xen_domain() && !xen_biovec_phys_mergeable(vec1, vec2->bv_page))
6a9f5f24 76 return false;
3dccdae5
CH
77 if ((addr1 | mask) != ((addr2 + vec2->bv_len - 1) | mask))
78 return false;
6a9f5f24
CH
79 return true;
80}
81
27ca1d4e
CH
82static inline bool __bvec_gap_to_prev(struct request_queue *q,
83 struct bio_vec *bprv, unsigned int offset)
84{
df376b2e 85 return (offset & queue_virt_boundary(q)) ||
27ca1d4e
CH
86 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
87}
88
89/*
90 * Check if adding a bio_vec after bprv with offset would create a gap in
91 * the SG list. Most drivers don't care about this, but some do.
92 */
93static inline bool bvec_gap_to_prev(struct request_queue *q,
94 struct bio_vec *bprv, unsigned int offset)
95{
96 if (!queue_virt_boundary(q))
97 return false;
98 return __bvec_gap_to_prev(q, bprv, offset);
99}
100
1aa0a133
CH
101static inline void blk_rq_bio_prep(struct request *rq, struct bio *bio,
102 unsigned int nr_segs)
103{
104 rq->nr_phys_segments = nr_segs;
105 rq->__data_len = bio->bi_iter.bi_size;
106 rq->bio = rq->biotail = bio;
107 rq->ioprio = bio_prio(bio);
108
109 if (bio->bi_disk)
110 rq->rq_disk = bio->bi_disk;
111}
112
5a48fc14
DW
113#ifdef CONFIG_BLK_DEV_INTEGRITY
114void blk_flush_integrity(void);
7c20f116 115bool __bio_integrity_endio(struct bio *);
ece841ab 116void bio_integrity_free(struct bio *bio);
7c20f116
CH
117static inline bool bio_integrity_endio(struct bio *bio)
118{
119 if (bio_integrity(bio))
120 return __bio_integrity_endio(bio);
121 return true;
122}
43b729bf
CH
123
124static inline bool integrity_req_gap_back_merge(struct request *req,
125 struct bio *next)
126{
127 struct bio_integrity_payload *bip = bio_integrity(req->bio);
128 struct bio_integrity_payload *bip_next = bio_integrity(next);
129
130 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
131 bip_next->bip_vec[0].bv_offset);
132}
133
134static inline bool integrity_req_gap_front_merge(struct request *req,
135 struct bio *bio)
136{
137 struct bio_integrity_payload *bip = bio_integrity(bio);
138 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
139
140 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
141 bip_next->bip_vec[0].bv_offset);
142}
581e2600
CH
143
144void blk_integrity_add(struct gendisk *);
145void blk_integrity_del(struct gendisk *);
43b729bf
CH
146#else /* CONFIG_BLK_DEV_INTEGRITY */
147static inline bool integrity_req_gap_back_merge(struct request *req,
148 struct bio *next)
149{
150 return false;
151}
152static inline bool integrity_req_gap_front_merge(struct request *req,
153 struct bio *bio)
154{
155 return false;
156}
157
5a48fc14
DW
158static inline void blk_flush_integrity(void)
159{
160}
7c20f116
CH
161static inline bool bio_integrity_endio(struct bio *bio)
162{
163 return true;
164}
ece841ab
JT
165static inline void bio_integrity_free(struct bio *bio)
166{
167}
581e2600
CH
168static inline void blk_integrity_add(struct gendisk *disk)
169{
170}
171static inline void blk_integrity_del(struct gendisk *disk)
172{
173}
43b729bf 174#endif /* CONFIG_BLK_DEV_INTEGRITY */
8324aa91 175
0d2602ca 176unsigned long blk_rq_timeout(unsigned long timeout);
87ee7b11 177void blk_add_timer(struct request *req);
320ae51f 178
14ccb66b
CH
179bool bio_attempt_front_merge(struct request *req, struct bio *bio,
180 unsigned int nr_segs);
181bool bio_attempt_back_merge(struct request *req, struct bio *bio,
182 unsigned int nr_segs);
1e739730
CH
183bool bio_attempt_discard_merge(struct request_queue *q, struct request *req,
184 struct bio *bio);
320ae51f 185bool blk_attempt_plug_merge(struct request_queue *q, struct bio *bio,
14ccb66b 186 unsigned int nr_segs, struct request **same_queue_rq);
320ae51f 187
b5af37ab 188void blk_account_io_start(struct request *req);
522a7775 189void blk_account_io_done(struct request *req, u64 now);
320ae51f 190
158dbda0
TH
191/*
192 * Internal elevator interface
193 */
e8064021 194#define ELV_ON_HASH(rq) ((rq)->rq_flags & RQF_HASHED)
158dbda0 195
ae1b1539 196void blk_insert_flush(struct request *rq);
dd831006 197
954b4a5c 198void elevator_init_mq(struct request_queue *q);
d48ece20
JW
199int elevator_switch_mq(struct request_queue *q,
200 struct elevator_type *new_e);
c3e22192 201void __elevator_exit(struct request_queue *, struct elevator_queue *);
cecf5d87 202int elv_register_queue(struct request_queue *q, bool uevent);
83d016ac
BVA
203void elv_unregister_queue(struct request_queue *q);
204
c3e22192
ML
205static inline void elevator_exit(struct request_queue *q,
206 struct elevator_queue *e)
207{
284b94be
ML
208 lockdep_assert_held(&q->sysfs_lock);
209
c3e22192
ML
210 blk_mq_sched_free_requests(q);
211 __elevator_exit(q, e);
212}
213
807d4af2
CH
214struct hd_struct *__disk_get_part(struct gendisk *disk, int partno);
215
3ad5cee5
CH
216ssize_t part_size_show(struct device *dev, struct device_attribute *attr,
217 char *buf);
218ssize_t part_stat_show(struct device *dev, struct device_attribute *attr,
219 char *buf);
220ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
221 char *buf);
222ssize_t part_fail_show(struct device *dev, struct device_attribute *attr,
223 char *buf);
224ssize_t part_fail_store(struct device *dev, struct device_attribute *attr,
225 const char *buf, size_t count);
226
581d4e28
JA
227#ifdef CONFIG_FAIL_IO_TIMEOUT
228int blk_should_fake_timeout(struct request_queue *);
229ssize_t part_timeout_show(struct device *, struct device_attribute *, char *);
230ssize_t part_timeout_store(struct device *, struct device_attribute *,
231 const char *, size_t);
232#else
233static inline int blk_should_fake_timeout(struct request_queue *q)
234{
235 return 0;
236}
237#endif
238
14ccb66b
CH
239void __blk_queue_split(struct request_queue *q, struct bio **bio,
240 unsigned int *nr_segs);
241int ll_back_merge_fn(struct request *req, struct bio *bio,
242 unsigned int nr_segs);
243int ll_front_merge_fn(struct request *req, struct bio *bio,
244 unsigned int nr_segs);
b973cb7e
JA
245struct request *attempt_back_merge(struct request_queue *q, struct request *rq);
246struct request *attempt_front_merge(struct request_queue *q, struct request *rq);
5e84ea3a
JA
247int blk_attempt_req_merge(struct request_queue *q, struct request *rq,
248 struct request *next);
e9cd19c0 249unsigned int blk_recalc_rq_segments(struct request *rq);
80a761fd 250void blk_rq_set_mixed_merge(struct request *rq);
050c8ea8 251bool blk_rq_merge_ok(struct request *rq, struct bio *bio);
34fe7c05 252enum elv_merge blk_try_merge(struct request *rq, struct bio *bio);
d6d48196 253
ff88972c
AB
254int blk_dev_init(void);
255
c2553b58
JA
256/*
257 * Contribute to IO statistics IFF:
258 *
259 * a) it's attached to a gendisk, and
48d9b0d4 260 * b) the queue had IO stats enabled when this request was started
c2553b58 261 */
599d067d 262static inline bool blk_do_io_stat(struct request *rq)
fb8ec18c 263{
48d9b0d4 264 return rq->rq_disk && (rq->rq_flags & RQF_IO_STAT);
fb8ec18c
JA
265}
266
6cf7677f
CH
267static inline void req_set_nomerge(struct request_queue *q, struct request *req)
268{
269 req->cmd_flags |= REQ_NOMERGE;
270 if (req == q->last_merge)
271 q->last_merge = NULL;
272}
273
1adfc5e4
ML
274/*
275 * The max size one bio can handle is UINT_MAX becasue bvec_iter.bi_size
276 * is defined as 'unsigned int', meantime it has to aligned to with logical
277 * block size which is the minimum accepted unit by hardware.
278 */
279static inline unsigned int bio_allowed_max_sectors(struct request_queue *q)
280{
281 return round_down(UINT_MAX, queue_logical_block_size(q)) >> 9;
282}
283
f2dbd76a
TH
284/*
285 * Internal io_context interface
286 */
287void get_io_context(struct io_context *ioc);
47fdd4ca 288struct io_cq *ioc_lookup_icq(struct io_context *ioc, struct request_queue *q);
24acfc34
TH
289struct io_cq *ioc_create_icq(struct io_context *ioc, struct request_queue *q,
290 gfp_t gfp_mask);
7e5a8794 291void ioc_clear_queue(struct request_queue *q);
f2dbd76a 292
24acfc34 293int create_task_io_context(struct task_struct *task, gfp_t gfp_mask, int node);
f2dbd76a 294
f2dbd76a
TH
295/*
296 * Internal throttling interface
297 */
bc9fcbf9 298#ifdef CONFIG_BLK_DEV_THROTTLING
bc9fcbf9
TH
299extern int blk_throtl_init(struct request_queue *q);
300extern void blk_throtl_exit(struct request_queue *q);
d61fcfa4 301extern void blk_throtl_register_queue(struct request_queue *q);
bc9fcbf9 302#else /* CONFIG_BLK_DEV_THROTTLING */
bc9fcbf9
TH
303static inline int blk_throtl_init(struct request_queue *q) { return 0; }
304static inline void blk_throtl_exit(struct request_queue *q) { }
d61fcfa4 305static inline void blk_throtl_register_queue(struct request_queue *q) { }
bc9fcbf9 306#endif /* CONFIG_BLK_DEV_THROTTLING */
297e3d85
SL
307#ifdef CONFIG_BLK_DEV_THROTTLING_LOW
308extern ssize_t blk_throtl_sample_time_show(struct request_queue *q, char *page);
309extern ssize_t blk_throtl_sample_time_store(struct request_queue *q,
310 const char *page, size_t count);
9e234eea 311extern void blk_throtl_bio_endio(struct bio *bio);
b9147dd1 312extern void blk_throtl_stat_add(struct request *rq, u64 time);
9e234eea
SL
313#else
314static inline void blk_throtl_bio_endio(struct bio *bio) { }
b9147dd1 315static inline void blk_throtl_stat_add(struct request *rq, u64 time) { }
297e3d85 316#endif
bc9fcbf9 317
3bce016a
CH
318#ifdef CONFIG_BOUNCE
319extern int init_emergency_isa_pool(void);
320extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
321#else
322static inline int init_emergency_isa_pool(void)
323{
324 return 0;
325}
326static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
327{
328}
329#endif /* CONFIG_BOUNCE */
330
d7067512
JB
331#ifdef CONFIG_BLK_CGROUP_IOLATENCY
332extern int blk_iolatency_init(struct request_queue *q);
333#else
334static inline int blk_iolatency_init(struct request_queue *q) { return 0; }
335#endif
336
a2d6b3a2
DLM
337struct bio *blk_next_bio(struct bio *bio, unsigned int nr_pages, gfp_t gfp);
338
bf505456
DLM
339#ifdef CONFIG_BLK_DEV_ZONED
340void blk_queue_free_zone_bitmaps(struct request_queue *q);
341#else
342static inline void blk_queue_free_zone_bitmaps(struct request_queue *q) {}
343#endif
344
581e2600
CH
345struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector);
346
347int blk_alloc_devt(struct hd_struct *part, dev_t *devt);
348void blk_free_devt(dev_t devt);
349void blk_invalidate_devt(dev_t devt);
350char *disk_name(struct gendisk *hd, int partno, char *buf);
351#define ADDPART_FLAG_NONE 0
352#define ADDPART_FLAG_RAID 1
353#define ADDPART_FLAG_WHOLEDISK 2
cddae808 354void delete_partition(struct gendisk *disk, struct hd_struct *part);
fa9156ae
CH
355int bdev_add_partition(struct block_device *bdev, int partno,
356 sector_t start, sector_t length);
357int bdev_del_partition(struct block_device *bdev, int partno);
358int bdev_resize_partition(struct block_device *bdev, int partno,
359 sector_t start, sector_t length);
581e2600 360int disk_expand_part_tbl(struct gendisk *disk, int target);
8da2892e 361int hd_ref_init(struct hd_struct *part);
581e2600 362
27eb3af9 363/* no need to get/put refcount of part0 */
581e2600
CH
364static inline int hd_struct_try_get(struct hd_struct *part)
365{
27eb3af9
ML
366 if (part->partno)
367 return percpu_ref_tryget_live(&part->ref);
368 return 1;
581e2600
CH
369}
370
371static inline void hd_struct_put(struct hd_struct *part)
372{
27eb3af9
ML
373 if (part->partno)
374 percpu_ref_put(&part->ref);
581e2600
CH
375}
376
581e2600
CH
377static inline void hd_free_part(struct hd_struct *part)
378{
58d4f14f 379 free_percpu(part->dkstats);
581e2600
CH
380 kfree(part->info);
381 percpu_ref_exit(&part->ref);
382}
383
384/*
385 * Any access of part->nr_sects which is not protected by partition
386 * bd_mutex or gendisk bdev bd_mutex, should be done using this
387 * accessor function.
388 *
389 * Code written along the lines of i_size_read() and i_size_write().
390 * CONFIG_PREEMPTION case optimizes the case of UP kernel with preemption
391 * on.
392 */
393static inline sector_t part_nr_sects_read(struct hd_struct *part)
394{
395#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
396 sector_t nr_sects;
397 unsigned seq;
398 do {
399 seq = read_seqcount_begin(&part->nr_sects_seq);
400 nr_sects = part->nr_sects;
401 } while (read_seqcount_retry(&part->nr_sects_seq, seq));
402 return nr_sects;
403#elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
404 sector_t nr_sects;
405
406 preempt_disable();
407 nr_sects = part->nr_sects;
408 preempt_enable();
409 return nr_sects;
410#else
411 return part->nr_sects;
412#endif
413}
414
415/*
416 * Should be called with mutex lock held (typically bd_mutex) of partition
417 * to provide mutual exlusion among writers otherwise seqcount might be
418 * left in wrong state leaving the readers spinning infinitely.
419 */
420static inline void part_nr_sects_write(struct hd_struct *part, sector_t size)
421{
422#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
423 write_seqcount_begin(&part->nr_sects_seq);
424 part->nr_sects = size;
425 write_seqcount_end(&part->nr_sects_seq);
426#elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
427 preempt_disable();
428 part->nr_sects = size;
429 preempt_enable();
430#else
431 part->nr_sects = size;
432#endif
433}
434
3d745ea5
CH
435struct request_queue *__blk_alloc_queue(int node_id);
436
e4581105 437int bio_add_hw_page(struct request_queue *q, struct bio *bio,
130879f1 438 struct page *page, unsigned int len, unsigned int offset,
e4581105 439 unsigned int max_sectors, bool *same_page);
130879f1 440
bc9fcbf9 441#endif /* BLK_INTERNAL_H */