xref: /kernel/linux/linux-5.10/block/blk.h (revision 8c2ecf20)
1/* SPDX-License-Identifier: GPL-2.0 */
2#ifndef BLK_INTERNAL_H
3#define BLK_INTERNAL_H
4
5#include <linux/idr.h>
6#include <linux/blk-mq.h>
7#include <linux/part_stat.h>
8#include <linux/blk-crypto.h>
9#include <xen/xen.h>
10#include "blk-crypto-internal.h"
11#include "blk-mq.h"
12#include "blk-mq-sched.h"
13
14/* Max future timer expiry for timeouts */
15#define BLK_MAX_TIMEOUT		(5 * HZ)
16
17extern struct dentry *blk_debugfs_root;
18
19struct blk_flush_queue {
20	unsigned int		flush_pending_idx:1;
21	unsigned int		flush_running_idx:1;
22	blk_status_t 		rq_status;
23	unsigned long		flush_pending_since;
24	struct list_head	flush_queue[2];
25	struct list_head	flush_data_in_flight;
26	struct request		*flush_rq;
27
28	spinlock_t		mq_flush_lock;
29};
30
31extern struct kmem_cache *blk_requestq_cachep;
32extern struct kobj_type blk_queue_ktype;
33extern struct ida blk_queue_ida;
34
35static inline struct blk_flush_queue *
36blk_get_flush_queue(struct request_queue *q, struct blk_mq_ctx *ctx)
37{
38	return blk_mq_map_queue(q, REQ_OP_FLUSH, ctx)->fq;
39}
40
41static inline void __blk_get_queue(struct request_queue *q)
42{
43	kobject_get(&q->kobj);
44}
45
46bool is_flush_rq(struct request *req);
47
48struct blk_flush_queue *blk_alloc_flush_queue(int node, int cmd_size,
49					      gfp_t flags);
50void blk_free_flush_queue(struct blk_flush_queue *q);
51
52void blk_freeze_queue(struct request_queue *q);
53
54static inline bool biovec_phys_mergeable(struct request_queue *q,
55		struct bio_vec *vec1, struct bio_vec *vec2)
56{
57	unsigned long mask = queue_segment_boundary(q);
58	phys_addr_t addr1 = page_to_phys(vec1->bv_page) + vec1->bv_offset;
59	phys_addr_t addr2 = page_to_phys(vec2->bv_page) + vec2->bv_offset;
60
61	if (addr1 + vec1->bv_len != addr2)
62		return false;
63	if (xen_domain() && !xen_biovec_phys_mergeable(vec1, vec2->bv_page))
64		return false;
65	if ((addr1 | mask) != ((addr2 + vec2->bv_len - 1) | mask))
66		return false;
67	return true;
68}
69
70static inline bool __bvec_gap_to_prev(struct request_queue *q,
71		struct bio_vec *bprv, unsigned int offset)
72{
73	return (offset & queue_virt_boundary(q)) ||
74		((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
75}
76
77/*
78 * Check if adding a bio_vec after bprv with offset would create a gap in
79 * the SG list. Most drivers don't care about this, but some do.
80 */
81static inline bool bvec_gap_to_prev(struct request_queue *q,
82		struct bio_vec *bprv, unsigned int offset)
83{
84	if (!queue_virt_boundary(q))
85		return false;
86	return __bvec_gap_to_prev(q, bprv, offset);
87}
88
89static inline void blk_rq_bio_prep(struct request *rq, struct bio *bio,
90		unsigned int nr_segs)
91{
92	rq->nr_phys_segments = nr_segs;
93	rq->__data_len = bio->bi_iter.bi_size;
94	rq->bio = rq->biotail = bio;
95	rq->ioprio = bio_prio(bio);
96
97	if (bio->bi_disk)
98		rq->rq_disk = bio->bi_disk;
99}
100
101#ifdef CONFIG_BLK_DEV_INTEGRITY
102void blk_flush_integrity(void);
103bool __bio_integrity_endio(struct bio *);
104void bio_integrity_free(struct bio *bio);
105static inline bool bio_integrity_endio(struct bio *bio)
106{
107	if (bio_integrity(bio))
108		return __bio_integrity_endio(bio);
109	return true;
110}
111
112bool blk_integrity_merge_rq(struct request_queue *, struct request *,
113		struct request *);
114bool blk_integrity_merge_bio(struct request_queue *, struct request *,
115		struct bio *);
116
117static inline bool integrity_req_gap_back_merge(struct request *req,
118		struct bio *next)
119{
120	struct bio_integrity_payload *bip = bio_integrity(req->bio);
121	struct bio_integrity_payload *bip_next = bio_integrity(next);
122
123	return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
124				bip_next->bip_vec[0].bv_offset);
125}
126
127static inline bool integrity_req_gap_front_merge(struct request *req,
128		struct bio *bio)
129{
130	struct bio_integrity_payload *bip = bio_integrity(bio);
131	struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
132
133	return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
134				bip_next->bip_vec[0].bv_offset);
135}
136
137void blk_integrity_add(struct gendisk *);
138void blk_integrity_del(struct gendisk *);
139#else /* CONFIG_BLK_DEV_INTEGRITY */
140static inline bool blk_integrity_merge_rq(struct request_queue *rq,
141		struct request *r1, struct request *r2)
142{
143	return true;
144}
145static inline bool blk_integrity_merge_bio(struct request_queue *rq,
146		struct request *r, struct bio *b)
147{
148	return true;
149}
150static inline bool integrity_req_gap_back_merge(struct request *req,
151		struct bio *next)
152{
153	return false;
154}
155static inline bool integrity_req_gap_front_merge(struct request *req,
156		struct bio *bio)
157{
158	return false;
159}
160
161static inline void blk_flush_integrity(void)
162{
163}
164static inline bool bio_integrity_endio(struct bio *bio)
165{
166	return true;
167}
168static inline void bio_integrity_free(struct bio *bio)
169{
170}
171static inline void blk_integrity_add(struct gendisk *disk)
172{
173}
174static inline void blk_integrity_del(struct gendisk *disk)
175{
176}
177#endif /* CONFIG_BLK_DEV_INTEGRITY */
178
179unsigned long blk_rq_timeout(unsigned long timeout);
180void blk_add_timer(struct request *req);
181
182bool blk_attempt_plug_merge(struct request_queue *q, struct bio *bio,
183		unsigned int nr_segs, struct request **same_queue_rq);
184bool blk_bio_list_merge(struct request_queue *q, struct list_head *list,
185			struct bio *bio, unsigned int nr_segs);
186
187void blk_account_io_start(struct request *req);
188void blk_account_io_done(struct request *req, u64 now);
189
190/*
191 * Plug flush limits
192 */
193#define BLK_MAX_REQUEST_COUNT	32
194#define BLK_PLUG_FLUSH_SIZE	(128 * 1024)
195
196/*
197 * Internal elevator interface
198 */
199#define ELV_ON_HASH(rq) ((rq)->rq_flags & RQF_HASHED)
200
201void blk_insert_flush(struct request *rq);
202
203void elevator_init_mq(struct request_queue *q);
204int elevator_switch_mq(struct request_queue *q,
205			      struct elevator_type *new_e);
206void __elevator_exit(struct request_queue *, struct elevator_queue *);
207int elv_register_queue(struct request_queue *q, bool uevent);
208void elv_unregister_queue(struct request_queue *q);
209
210static inline void elevator_exit(struct request_queue *q,
211		struct elevator_queue *e)
212{
213	lockdep_assert_held(&q->sysfs_lock);
214
215	blk_mq_sched_free_requests(q);
216	__elevator_exit(q, e);
217}
218
219struct hd_struct *__disk_get_part(struct gendisk *disk, int partno);
220
221ssize_t part_size_show(struct device *dev, struct device_attribute *attr,
222		char *buf);
223ssize_t part_stat_show(struct device *dev, struct device_attribute *attr,
224		char *buf);
225ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
226		char *buf);
227ssize_t part_fail_show(struct device *dev, struct device_attribute *attr,
228		char *buf);
229ssize_t part_fail_store(struct device *dev, struct device_attribute *attr,
230		const char *buf, size_t count);
231ssize_t part_timeout_show(struct device *, struct device_attribute *, char *);
232ssize_t part_timeout_store(struct device *, struct device_attribute *,
233				const char *, size_t);
234
235void __blk_queue_split(struct bio **bio, unsigned int *nr_segs);
236int ll_back_merge_fn(struct request *req, struct bio *bio,
237		unsigned int nr_segs);
238bool blk_attempt_req_merge(struct request_queue *q, struct request *rq,
239				struct request *next);
240unsigned int blk_recalc_rq_segments(struct request *rq);
241void blk_rq_set_mixed_merge(struct request *rq);
242bool blk_rq_merge_ok(struct request *rq, struct bio *bio);
243enum elv_merge blk_try_merge(struct request *rq, struct bio *bio);
244
245int blk_dev_init(void);
246
247/*
248 * Contribute to IO statistics IFF:
249 *
250 *	a) it's attached to a gendisk, and
251 *	b) the queue had IO stats enabled when this request was started
252 */
253static inline bool blk_do_io_stat(struct request *rq)
254{
255	return rq->rq_disk && (rq->rq_flags & RQF_IO_STAT);
256}
257
258static inline void req_set_nomerge(struct request_queue *q, struct request *req)
259{
260	req->cmd_flags |= REQ_NOMERGE;
261	if (req == q->last_merge)
262		q->last_merge = NULL;
263}
264
265/*
266 * The max size one bio can handle is UINT_MAX becasue bvec_iter.bi_size
267 * is defined as 'unsigned int', meantime it has to aligned to with logical
268 * block size which is the minimum accepted unit by hardware.
269 */
270static inline unsigned int bio_allowed_max_sectors(struct request_queue *q)
271{
272	return round_down(UINT_MAX, queue_logical_block_size(q)) >> 9;
273}
274
275/*
276 * The max bio size which is aligned to q->limits.discard_granularity. This
277 * is a hint to split large discard bio in generic block layer, then if device
278 * driver needs to split the discard bio into smaller ones, their bi_size can
279 * be very probably and easily aligned to discard_granularity of the device's
280 * queue.
281 */
282static inline unsigned int bio_aligned_discard_max_sectors(
283					struct request_queue *q)
284{
285	return round_down(UINT_MAX, q->limits.discard_granularity) >>
286			SECTOR_SHIFT;
287}
288
289/*
290 * Internal io_context interface
291 */
292void get_io_context(struct io_context *ioc);
293struct io_cq *ioc_lookup_icq(struct io_context *ioc, struct request_queue *q);
294struct io_cq *ioc_create_icq(struct io_context *ioc, struct request_queue *q,
295			     gfp_t gfp_mask);
296void ioc_clear_queue(struct request_queue *q);
297
298int create_task_io_context(struct task_struct *task, gfp_t gfp_mask, int node);
299
300/*
301 * Internal throttling interface
302 */
303#ifdef CONFIG_BLK_DEV_THROTTLING
304extern int blk_throtl_init(struct request_queue *q);
305extern void blk_throtl_exit(struct request_queue *q);
306extern void blk_throtl_register_queue(struct request_queue *q);
307extern void blk_throtl_charge_bio_split(struct bio *bio);
308bool blk_throtl_bio(struct bio *bio);
309#else /* CONFIG_BLK_DEV_THROTTLING */
310static inline int blk_throtl_init(struct request_queue *q) { return 0; }
311static inline void blk_throtl_exit(struct request_queue *q) { }
312static inline void blk_throtl_register_queue(struct request_queue *q) { }
313static inline void blk_throtl_charge_bio_split(struct bio *bio) { }
314static inline bool blk_throtl_bio(struct bio *bio) { return false; }
315#endif /* CONFIG_BLK_DEV_THROTTLING */
316#ifdef CONFIG_BLK_DEV_THROTTLING_LOW
317extern ssize_t blk_throtl_sample_time_show(struct request_queue *q, char *page);
318extern ssize_t blk_throtl_sample_time_store(struct request_queue *q,
319	const char *page, size_t count);
320extern void blk_throtl_bio_endio(struct bio *bio);
321extern void blk_throtl_stat_add(struct request *rq, u64 time);
322#else
323static inline void blk_throtl_bio_endio(struct bio *bio) { }
324static inline void blk_throtl_stat_add(struct request *rq, u64 time) { }
325#endif
326
327#ifdef CONFIG_BOUNCE
328extern int init_emergency_isa_pool(void);
329extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
330#else
331static inline int init_emergency_isa_pool(void)
332{
333	return 0;
334}
335static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
336{
337}
338#endif /* CONFIG_BOUNCE */
339
340#ifdef CONFIG_BLK_CGROUP_IOLATENCY
341extern int blk_iolatency_init(struct request_queue *q);
342#else
343static inline int blk_iolatency_init(struct request_queue *q) { return 0; }
344#endif
345
346struct bio *blk_next_bio(struct bio *bio, unsigned int nr_pages, gfp_t gfp);
347
348#ifdef CONFIG_BLK_DEV_ZONED
349void blk_queue_free_zone_bitmaps(struct request_queue *q);
350#else
351static inline void blk_queue_free_zone_bitmaps(struct request_queue *q) {}
352#endif
353
354struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector);
355
356int blk_alloc_devt(struct hd_struct *part, dev_t *devt);
357void blk_free_devt(dev_t devt);
358void blk_invalidate_devt(dev_t devt);
359char *disk_name(struct gendisk *hd, int partno, char *buf);
360#define ADDPART_FLAG_NONE	0
361#define ADDPART_FLAG_RAID	1
362#define ADDPART_FLAG_WHOLEDISK	2
363void delete_partition(struct hd_struct *part);
364int bdev_add_partition(struct block_device *bdev, int partno,
365		sector_t start, sector_t length);
366int bdev_del_partition(struct block_device *bdev, int partno);
367int bdev_resize_partition(struct block_device *bdev, int partno,
368		sector_t start, sector_t length);
369int disk_expand_part_tbl(struct gendisk *disk, int target);
370int hd_ref_init(struct hd_struct *part);
371
372/* no need to get/put refcount of part0 */
373static inline int hd_struct_try_get(struct hd_struct *part)
374{
375	if (part->partno)
376		return percpu_ref_tryget_live(&part->ref);
377	return 1;
378}
379
380static inline void hd_struct_put(struct hd_struct *part)
381{
382	if (part->partno)
383		percpu_ref_put(&part->ref);
384}
385
386static inline void hd_free_part(struct hd_struct *part)
387{
388	free_percpu(part->dkstats);
389	kfree(part->info);
390	percpu_ref_exit(&part->ref);
391}
392
393/*
394 * Any access of part->nr_sects which is not protected by partition
395 * bd_mutex or gendisk bdev bd_mutex, should be done using this
396 * accessor function.
397 *
398 * Code written along the lines of i_size_read() and i_size_write().
399 * CONFIG_PREEMPTION case optimizes the case of UP kernel with preemption
400 * on.
401 */
402static inline sector_t part_nr_sects_read(struct hd_struct *part)
403{
404#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
405	sector_t nr_sects;
406	unsigned seq;
407	do {
408		seq = read_seqcount_begin(&part->nr_sects_seq);
409		nr_sects = part->nr_sects;
410	} while (read_seqcount_retry(&part->nr_sects_seq, seq));
411	return nr_sects;
412#elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
413	sector_t nr_sects;
414
415	preempt_disable();
416	nr_sects = part->nr_sects;
417	preempt_enable();
418	return nr_sects;
419#else
420	return part->nr_sects;
421#endif
422}
423
424/*
425 * Should be called with mutex lock held (typically bd_mutex) of partition
426 * to provide mutual exlusion among writers otherwise seqcount might be
427 * left in wrong state leaving the readers spinning infinitely.
428 */
429static inline void part_nr_sects_write(struct hd_struct *part, sector_t size)
430{
431#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
432	preempt_disable();
433	write_seqcount_begin(&part->nr_sects_seq);
434	part->nr_sects = size;
435	write_seqcount_end(&part->nr_sects_seq);
436	preempt_enable();
437#elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
438	preempt_disable();
439	part->nr_sects = size;
440	preempt_enable();
441#else
442	part->nr_sects = size;
443#endif
444}
445
446int bio_add_hw_page(struct request_queue *q, struct bio *bio,
447		struct page *page, unsigned int len, unsigned int offset,
448		unsigned int max_sectors, bool *same_page);
449
450#endif /* BLK_INTERNAL_H */
451