1 /* SPDX-License-Identifier: GPL-2.0 */
6 #include <linux/blk-mq.h>
7 #include <linux/part_stat.h>
8 #include <linux/blk-crypto.h>
10 #include "blk-crypto-internal.h"
12 #include "blk-mq-sched.h"
14 /* Max future timer expiry for timeouts */
15 #define BLK_MAX_TIMEOUT (5 * HZ)
17 extern struct dentry *blk_debugfs_root;
19 struct 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;
28 struct lock_class_key key;
29 spinlock_t mq_flush_lock;
32 enum bio_merge_status {
38 extern struct kmem_cache *blk_requestq_cachep;
39 extern struct kobj_type blk_queue_ktype;
40 extern struct ida blk_queue_ida;
42 static inline struct blk_flush_queue *
43 blk_get_flush_queue(struct request_queue *q, struct blk_mq_ctx *ctx)
45 return blk_mq_map_queue(q, REQ_OP_FLUSH, ctx)->fq;
48 static inline void __blk_get_queue(struct request_queue *q)
50 kobject_get(&q->kobj);
54 is_flush_rq(struct request *req, struct blk_mq_hw_ctx *hctx)
56 return hctx->fq->flush_rq == req;
59 struct blk_flush_queue *blk_alloc_flush_queue(int node, int cmd_size,
61 void blk_free_flush_queue(struct blk_flush_queue *q);
63 void blk_freeze_queue(struct request_queue *q);
65 static inline bool biovec_phys_mergeable(struct request_queue *q,
66 struct bio_vec *vec1, struct bio_vec *vec2)
68 unsigned long mask = queue_segment_boundary(q);
69 phys_addr_t addr1 = page_to_phys(vec1->bv_page) + vec1->bv_offset;
70 phys_addr_t addr2 = page_to_phys(vec2->bv_page) + vec2->bv_offset;
72 if (addr1 + vec1->bv_len != addr2)
74 if (xen_domain() && !xen_biovec_phys_mergeable(vec1, vec2->bv_page))
76 if ((addr1 | mask) != ((addr2 + vec2->bv_len - 1) | mask))
81 static inline bool __bvec_gap_to_prev(struct request_queue *q,
82 struct bio_vec *bprv, unsigned int offset)
84 return (offset & queue_virt_boundary(q)) ||
85 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
89 * Check if adding a bio_vec after bprv with offset would create a gap in
90 * the SG list. Most drivers don't care about this, but some do.
92 static inline bool bvec_gap_to_prev(struct request_queue *q,
93 struct bio_vec *bprv, unsigned int offset)
95 if (!queue_virt_boundary(q))
97 return __bvec_gap_to_prev(q, bprv, offset);
100 static inline void blk_rq_bio_prep(struct request *rq, struct bio *bio,
101 unsigned int nr_segs)
103 rq->nr_phys_segments = nr_segs;
104 rq->__data_len = bio->bi_iter.bi_size;
105 rq->bio = rq->biotail = bio;
106 rq->ioprio = bio_prio(bio);
109 rq->rq_disk = bio->bi_disk;
112 #ifdef CONFIG_BLK_DEV_INTEGRITY
113 void blk_flush_integrity(void);
114 bool __bio_integrity_endio(struct bio *);
115 void bio_integrity_free(struct bio *bio);
116 static inline bool bio_integrity_endio(struct bio *bio)
118 if (bio_integrity(bio))
119 return __bio_integrity_endio(bio);
123 bool blk_integrity_merge_rq(struct request_queue *, struct request *,
126 static inline bool integrity_req_gap_back_merge(struct request *req,
129 struct bio_integrity_payload *bip = bio_integrity(req->bio);
130 struct bio_integrity_payload *bip_next = bio_integrity(next);
132 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
133 bip_next->bip_vec[0].bv_offset);
136 static inline bool integrity_req_gap_front_merge(struct request *req,
139 struct bio_integrity_payload *bip = bio_integrity(bio);
140 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
142 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
143 bip_next->bip_vec[0].bv_offset);
146 void blk_integrity_add(struct gendisk *);
147 void blk_integrity_del(struct gendisk *);
148 #else /* CONFIG_BLK_DEV_INTEGRITY */
149 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
150 struct request *r1, struct request *r2)
154 static inline bool integrity_req_gap_back_merge(struct request *req,
159 static inline bool integrity_req_gap_front_merge(struct request *req,
165 static inline void blk_flush_integrity(void)
168 static inline bool bio_integrity_endio(struct bio *bio)
172 static inline void bio_integrity_free(struct bio *bio)
175 static inline void blk_integrity_add(struct gendisk *disk)
178 static inline void blk_integrity_del(struct gendisk *disk)
181 #endif /* CONFIG_BLK_DEV_INTEGRITY */
183 unsigned long blk_rq_timeout(unsigned long timeout);
184 void blk_add_timer(struct request *req);
186 enum bio_merge_status bio_attempt_front_merge(struct request *req,
188 unsigned int nr_segs);
189 enum bio_merge_status bio_attempt_back_merge(struct request *req,
191 unsigned int nr_segs);
192 enum bio_merge_status bio_attempt_discard_merge(struct request_queue *q,
195 bool blk_attempt_plug_merge(struct request_queue *q, struct bio *bio,
196 unsigned int nr_segs, struct request **same_queue_rq);
197 bool blk_bio_list_merge(struct request_queue *q, struct list_head *list,
198 struct bio *bio, unsigned int nr_segs);
200 void blk_account_io_start(struct request *req);
201 void blk_account_io_done(struct request *req, u64 now);
204 * Internal elevator interface
206 #define ELV_ON_HASH(rq) ((rq)->rq_flags & RQF_HASHED)
208 void blk_insert_flush(struct request *rq);
210 void elevator_init_mq(struct request_queue *q);
211 int elevator_switch_mq(struct request_queue *q,
212 struct elevator_type *new_e);
213 void __elevator_exit(struct request_queue *, struct elevator_queue *);
214 int elv_register_queue(struct request_queue *q, bool uevent);
215 void elv_unregister_queue(struct request_queue *q);
217 static inline void elevator_exit(struct request_queue *q,
218 struct elevator_queue *e)
220 lockdep_assert_held(&q->sysfs_lock);
222 blk_mq_sched_free_requests(q);
223 __elevator_exit(q, e);
226 struct hd_struct *__disk_get_part(struct gendisk *disk, int partno);
228 ssize_t part_size_show(struct device *dev, struct device_attribute *attr,
230 ssize_t part_stat_show(struct device *dev, struct device_attribute *attr,
232 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
234 ssize_t part_fail_show(struct device *dev, struct device_attribute *attr,
236 ssize_t part_fail_store(struct device *dev, struct device_attribute *attr,
237 const char *buf, size_t count);
238 ssize_t part_timeout_show(struct device *, struct device_attribute *, char *);
239 ssize_t part_timeout_store(struct device *, struct device_attribute *,
240 const char *, size_t);
242 void __blk_queue_split(struct bio **bio, unsigned int *nr_segs);
243 int ll_back_merge_fn(struct request *req, struct bio *bio,
244 unsigned int nr_segs);
245 int ll_front_merge_fn(struct request *req, struct bio *bio,
246 unsigned int nr_segs);
247 struct request *attempt_back_merge(struct request_queue *q, struct request *rq);
248 struct request *attempt_front_merge(struct request_queue *q, struct request *rq);
249 int blk_attempt_req_merge(struct request_queue *q, struct request *rq,
250 struct request *next);
251 unsigned int blk_recalc_rq_segments(struct request *rq);
252 void blk_rq_set_mixed_merge(struct request *rq);
253 bool blk_rq_merge_ok(struct request *rq, struct bio *bio);
254 enum elv_merge blk_try_merge(struct request *rq, struct bio *bio);
256 int blk_dev_init(void);
259 * Contribute to IO statistics IFF:
261 * a) it's attached to a gendisk, and
262 * b) the queue had IO stats enabled when this request was started
264 static inline bool blk_do_io_stat(struct request *rq)
266 return rq->rq_disk && (rq->rq_flags & RQF_IO_STAT);
269 static inline void req_set_nomerge(struct request_queue *q, struct request *req)
271 req->cmd_flags |= REQ_NOMERGE;
272 if (req == q->last_merge)
273 q->last_merge = NULL;
277 * The max size one bio can handle is UINT_MAX becasue bvec_iter.bi_size
278 * is defined as 'unsigned int', meantime it has to aligned to with logical
279 * block size which is the minimum accepted unit by hardware.
281 static inline unsigned int bio_allowed_max_sectors(struct request_queue *q)
283 return round_down(UINT_MAX, queue_logical_block_size(q)) >> 9;
287 * The max bio size which is aligned to q->limits.discard_granularity. This
288 * is a hint to split large discard bio in generic block layer, then if device
289 * driver needs to split the discard bio into smaller ones, their bi_size can
290 * be very probably and easily aligned to discard_granularity of the device's
293 static inline unsigned int bio_aligned_discard_max_sectors(
294 struct request_queue *q)
296 return round_down(UINT_MAX, q->limits.discard_granularity) >>
301 * Internal io_context interface
303 void get_io_context(struct io_context *ioc);
304 struct io_cq *ioc_lookup_icq(struct io_context *ioc, struct request_queue *q);
305 struct io_cq *ioc_create_icq(struct io_context *ioc, struct request_queue *q,
307 void ioc_clear_queue(struct request_queue *q);
309 int create_task_io_context(struct task_struct *task, gfp_t gfp_mask, int node);
312 * Internal throttling interface
314 #ifdef CONFIG_BLK_DEV_THROTTLING
315 extern int blk_throtl_init(struct request_queue *q);
316 extern void blk_throtl_exit(struct request_queue *q);
317 extern void blk_throtl_register_queue(struct request_queue *q);
318 bool blk_throtl_bio(struct bio *bio);
319 #else /* CONFIG_BLK_DEV_THROTTLING */
320 static inline int blk_throtl_init(struct request_queue *q) { return 0; }
321 static inline void blk_throtl_exit(struct request_queue *q) { }
322 static inline void blk_throtl_register_queue(struct request_queue *q) { }
323 static inline bool blk_throtl_bio(struct bio *bio) { return false; }
324 #endif /* CONFIG_BLK_DEV_THROTTLING */
325 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW
326 extern ssize_t blk_throtl_sample_time_show(struct request_queue *q, char *page);
327 extern ssize_t blk_throtl_sample_time_store(struct request_queue *q,
328 const char *page, size_t count);
329 extern void blk_throtl_bio_endio(struct bio *bio);
330 extern void blk_throtl_stat_add(struct request *rq, u64 time);
332 static inline void blk_throtl_bio_endio(struct bio *bio) { }
333 static inline void blk_throtl_stat_add(struct request *rq, u64 time) { }
337 extern int init_emergency_isa_pool(void);
338 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
340 static inline int init_emergency_isa_pool(void)
344 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
347 #endif /* CONFIG_BOUNCE */
349 #ifdef CONFIG_BLK_CGROUP_IOLATENCY
350 extern int blk_iolatency_init(struct request_queue *q);
352 static inline int blk_iolatency_init(struct request_queue *q) { return 0; }
355 struct bio *blk_next_bio(struct bio *bio, unsigned int nr_pages, gfp_t gfp);
357 #ifdef CONFIG_BLK_DEV_ZONED
358 void blk_queue_free_zone_bitmaps(struct request_queue *q);
360 static inline void blk_queue_free_zone_bitmaps(struct request_queue *q) {}
363 struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector);
365 int blk_alloc_devt(struct hd_struct *part, dev_t *devt);
366 void blk_free_devt(dev_t devt);
367 void blk_invalidate_devt(dev_t devt);
368 char *disk_name(struct gendisk *hd, int partno, char *buf);
369 #define ADDPART_FLAG_NONE 0
370 #define ADDPART_FLAG_RAID 1
371 #define ADDPART_FLAG_WHOLEDISK 2
372 void delete_partition(struct hd_struct *part);
373 int bdev_add_partition(struct block_device *bdev, int partno,
374 sector_t start, sector_t length);
375 int bdev_del_partition(struct block_device *bdev, int partno);
376 int bdev_resize_partition(struct block_device *bdev, int partno,
377 sector_t start, sector_t length);
378 int disk_expand_part_tbl(struct gendisk *disk, int target);
379 int hd_ref_init(struct hd_struct *part);
381 /* no need to get/put refcount of part0 */
382 static inline int hd_struct_try_get(struct hd_struct *part)
385 return percpu_ref_tryget_live(&part->ref);
389 static inline void hd_struct_put(struct hd_struct *part)
392 percpu_ref_put(&part->ref);
395 static inline void hd_free_part(struct hd_struct *part)
397 free_percpu(part->dkstats);
399 percpu_ref_exit(&part->ref);
403 * Any access of part->nr_sects which is not protected by partition
404 * bd_mutex or gendisk bdev bd_mutex, should be done using this
407 * Code written along the lines of i_size_read() and i_size_write().
408 * CONFIG_PREEMPTION case optimizes the case of UP kernel with preemption
411 static inline sector_t part_nr_sects_read(struct hd_struct *part)
413 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
417 seq = read_seqcount_begin(&part->nr_sects_seq);
418 nr_sects = part->nr_sects;
419 } while (read_seqcount_retry(&part->nr_sects_seq, seq));
421 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
425 nr_sects = part->nr_sects;
429 return part->nr_sects;
434 * Should be called with mutex lock held (typically bd_mutex) of partition
435 * to provide mutual exlusion among writers otherwise seqcount might be
436 * left in wrong state leaving the readers spinning infinitely.
438 static inline void part_nr_sects_write(struct hd_struct *part, sector_t size)
440 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
442 write_seqcount_begin(&part->nr_sects_seq);
443 part->nr_sects = size;
444 write_seqcount_end(&part->nr_sects_seq);
446 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
448 part->nr_sects = size;
451 part->nr_sects = size;
455 int bio_add_hw_page(struct request_queue *q, struct bio *bio,
456 struct page *page, unsigned int len, unsigned int offset,
457 unsigned int max_sectors, bool *same_page);
459 #endif /* BLK_INTERNAL_H */