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 extern struct kmem_cache *blk_requestq_cachep;
33 extern struct kobj_type blk_queue_ktype;
34 extern struct ida blk_queue_ida;
36 static inline struct blk_flush_queue *
37 blk_get_flush_queue(struct request_queue *q, struct blk_mq_ctx *ctx)
39 return blk_mq_map_queue(q, REQ_OP_FLUSH, ctx)->fq;
42 static inline void __blk_get_queue(struct request_queue *q)
44 kobject_get(&q->kobj);
47 bool is_flush_rq(struct request *req);
49 struct blk_flush_queue *blk_alloc_flush_queue(int node, int cmd_size,
51 void blk_free_flush_queue(struct blk_flush_queue *q);
53 void blk_freeze_queue(struct request_queue *q);
55 static inline bool biovec_phys_mergeable(struct request_queue *q,
56 struct bio_vec *vec1, struct bio_vec *vec2)
58 unsigned long mask = queue_segment_boundary(q);
59 phys_addr_t addr1 = page_to_phys(vec1->bv_page) + vec1->bv_offset;
60 phys_addr_t addr2 = page_to_phys(vec2->bv_page) + vec2->bv_offset;
62 if (addr1 + vec1->bv_len != addr2)
64 if (xen_domain() && !xen_biovec_phys_mergeable(vec1, vec2->bv_page))
66 if ((addr1 | mask) != ((addr2 + vec2->bv_len - 1) | mask))
71 static inline bool __bvec_gap_to_prev(struct request_queue *q,
72 struct bio_vec *bprv, unsigned int offset)
74 return (offset & queue_virt_boundary(q)) ||
75 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
79 * Check if adding a bio_vec after bprv with offset would create a gap in
80 * the SG list. Most drivers don't care about this, but some do.
82 static inline bool bvec_gap_to_prev(struct request_queue *q,
83 struct bio_vec *bprv, unsigned int offset)
85 if (!queue_virt_boundary(q))
87 return __bvec_gap_to_prev(q, bprv, offset);
90 static inline void blk_rq_bio_prep(struct request *rq, struct bio *bio,
93 rq->nr_phys_segments = nr_segs;
94 rq->__data_len = bio->bi_iter.bi_size;
95 rq->bio = rq->biotail = bio;
96 rq->ioprio = bio_prio(bio);
99 rq->rq_disk = bio->bi_disk;
102 #ifdef CONFIG_BLK_DEV_INTEGRITY
103 void blk_flush_integrity(void);
104 bool __bio_integrity_endio(struct bio *);
105 void bio_integrity_free(struct bio *bio);
106 static inline bool bio_integrity_endio(struct bio *bio)
108 if (bio_integrity(bio))
109 return __bio_integrity_endio(bio);
113 bool blk_integrity_merge_rq(struct request_queue *, struct request *,
115 bool blk_integrity_merge_bio(struct request_queue *, struct request *,
118 static inline bool integrity_req_gap_back_merge(struct request *req,
121 struct bio_integrity_payload *bip = bio_integrity(req->bio);
122 struct bio_integrity_payload *bip_next = bio_integrity(next);
124 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
125 bip_next->bip_vec[0].bv_offset);
128 static inline bool integrity_req_gap_front_merge(struct request *req,
131 struct bio_integrity_payload *bip = bio_integrity(bio);
132 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
134 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
135 bip_next->bip_vec[0].bv_offset);
138 void blk_integrity_add(struct gendisk *);
139 void blk_integrity_del(struct gendisk *);
140 #else /* CONFIG_BLK_DEV_INTEGRITY */
141 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
142 struct request *r1, struct request *r2)
146 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
147 struct request *r, struct bio *b)
151 static inline bool integrity_req_gap_back_merge(struct request *req,
156 static inline bool integrity_req_gap_front_merge(struct request *req,
162 static inline void blk_flush_integrity(void)
165 static inline bool bio_integrity_endio(struct bio *bio)
169 static inline void bio_integrity_free(struct bio *bio)
172 static inline void blk_integrity_add(struct gendisk *disk)
175 static inline void blk_integrity_del(struct gendisk *disk)
178 #endif /* CONFIG_BLK_DEV_INTEGRITY */
180 unsigned long blk_rq_timeout(unsigned long timeout);
181 void blk_add_timer(struct request *req);
183 bool blk_attempt_plug_merge(struct request_queue *q, struct bio *bio,
184 unsigned int nr_segs, struct request **same_queue_rq);
185 bool blk_bio_list_merge(struct request_queue *q, struct list_head *list,
186 struct bio *bio, unsigned int nr_segs);
188 void blk_account_io_start(struct request *req);
189 void blk_account_io_done(struct request *req, u64 now);
194 #define BLK_MAX_REQUEST_COUNT 32
195 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
198 * Internal elevator interface
200 #define ELV_ON_HASH(rq) ((rq)->rq_flags & RQF_HASHED)
202 void blk_insert_flush(struct request *rq);
204 void elevator_init_mq(struct request_queue *q);
205 int elevator_switch_mq(struct request_queue *q,
206 struct elevator_type *new_e);
207 void __elevator_exit(struct request_queue *, struct elevator_queue *);
208 int elv_register_queue(struct request_queue *q, bool uevent);
209 void elv_unregister_queue(struct request_queue *q);
211 static inline void elevator_exit(struct request_queue *q,
212 struct elevator_queue *e)
214 lockdep_assert_held(&q->sysfs_lock);
216 blk_mq_sched_free_requests(q);
217 __elevator_exit(q, e);
220 struct hd_struct *__disk_get_part(struct gendisk *disk, int partno);
222 ssize_t part_size_show(struct device *dev, struct device_attribute *attr,
224 ssize_t part_stat_show(struct device *dev, struct device_attribute *attr,
226 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
228 ssize_t part_fail_show(struct device *dev, struct device_attribute *attr,
230 ssize_t part_fail_store(struct device *dev, struct device_attribute *attr,
231 const char *buf, size_t count);
232 ssize_t part_timeout_show(struct device *, struct device_attribute *, char *);
233 ssize_t part_timeout_store(struct device *, struct device_attribute *,
234 const char *, size_t);
236 void __blk_queue_split(struct bio **bio, unsigned int *nr_segs);
237 int ll_back_merge_fn(struct request *req, struct bio *bio,
238 unsigned int nr_segs);
239 int blk_attempt_req_merge(struct request_queue *q, struct request *rq,
240 struct request *next);
241 unsigned int blk_recalc_rq_segments(struct request *rq);
242 void blk_rq_set_mixed_merge(struct request *rq);
243 bool blk_rq_merge_ok(struct request *rq, struct bio *bio);
244 enum elv_merge blk_try_merge(struct request *rq, struct bio *bio);
246 int blk_dev_init(void);
249 * Contribute to IO statistics IFF:
251 * a) it's attached to a gendisk, and
252 * b) the queue had IO stats enabled when this request was started
254 static inline bool blk_do_io_stat(struct request *rq)
256 return rq->rq_disk && (rq->rq_flags & RQF_IO_STAT);
259 static inline void req_set_nomerge(struct request_queue *q, struct request *req)
261 req->cmd_flags |= REQ_NOMERGE;
262 if (req == q->last_merge)
263 q->last_merge = NULL;
267 * The max size one bio can handle is UINT_MAX becasue bvec_iter.bi_size
268 * is defined as 'unsigned int', meantime it has to aligned to with logical
269 * block size which is the minimum accepted unit by hardware.
271 static inline unsigned int bio_allowed_max_sectors(struct request_queue *q)
273 return round_down(UINT_MAX, queue_logical_block_size(q)) >> 9;
277 * The max bio size which is aligned to q->limits.discard_granularity. This
278 * is a hint to split large discard bio in generic block layer, then if device
279 * driver needs to split the discard bio into smaller ones, their bi_size can
280 * be very probably and easily aligned to discard_granularity of the device's
283 static inline unsigned int bio_aligned_discard_max_sectors(
284 struct request_queue *q)
286 return round_down(UINT_MAX, q->limits.discard_granularity) >>
291 * Internal io_context interface
293 void get_io_context(struct io_context *ioc);
294 struct io_cq *ioc_lookup_icq(struct io_context *ioc, struct request_queue *q);
295 struct io_cq *ioc_create_icq(struct io_context *ioc, struct request_queue *q,
297 void ioc_clear_queue(struct request_queue *q);
299 int create_task_io_context(struct task_struct *task, gfp_t gfp_mask, int node);
302 * Internal throttling interface
304 #ifdef CONFIG_BLK_DEV_THROTTLING
305 extern int blk_throtl_init(struct request_queue *q);
306 extern void blk_throtl_exit(struct request_queue *q);
307 extern void blk_throtl_register_queue(struct request_queue *q);
308 extern void blk_throtl_charge_bio_split(struct bio *bio);
309 bool blk_throtl_bio(struct bio *bio);
310 #else /* CONFIG_BLK_DEV_THROTTLING */
311 static inline int blk_throtl_init(struct request_queue *q) { return 0; }
312 static inline void blk_throtl_exit(struct request_queue *q) { }
313 static inline void blk_throtl_register_queue(struct request_queue *q) { }
314 static inline void blk_throtl_charge_bio_split(struct bio *bio) { }
315 static inline bool blk_throtl_bio(struct bio *bio) { return false; }
316 #endif /* CONFIG_BLK_DEV_THROTTLING */
317 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW
318 extern ssize_t blk_throtl_sample_time_show(struct request_queue *q, char *page);
319 extern ssize_t blk_throtl_sample_time_store(struct request_queue *q,
320 const char *page, size_t count);
321 extern void blk_throtl_bio_endio(struct bio *bio);
322 extern void blk_throtl_stat_add(struct request *rq, u64 time);
324 static inline void blk_throtl_bio_endio(struct bio *bio) { }
325 static inline void blk_throtl_stat_add(struct request *rq, u64 time) { }
329 extern int init_emergency_isa_pool(void);
330 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
332 static inline int init_emergency_isa_pool(void)
336 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
339 #endif /* CONFIG_BOUNCE */
341 #ifdef CONFIG_BLK_CGROUP_IOLATENCY
342 extern int blk_iolatency_init(struct request_queue *q);
344 static inline int blk_iolatency_init(struct request_queue *q) { return 0; }
347 struct bio *blk_next_bio(struct bio *bio, unsigned int nr_pages, gfp_t gfp);
349 #ifdef CONFIG_BLK_DEV_ZONED
350 void blk_queue_free_zone_bitmaps(struct request_queue *q);
352 static inline void blk_queue_free_zone_bitmaps(struct request_queue *q) {}
355 struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector);
357 int blk_alloc_devt(struct hd_struct *part, dev_t *devt);
358 void blk_free_devt(dev_t devt);
359 void blk_invalidate_devt(dev_t devt);
360 char *disk_name(struct gendisk *hd, int partno, char *buf);
361 #define ADDPART_FLAG_NONE 0
362 #define ADDPART_FLAG_RAID 1
363 #define ADDPART_FLAG_WHOLEDISK 2
364 void delete_partition(struct hd_struct *part);
365 int bdev_add_partition(struct block_device *bdev, int partno,
366 sector_t start, sector_t length);
367 int bdev_del_partition(struct block_device *bdev, int partno);
368 int bdev_resize_partition(struct block_device *bdev, int partno,
369 sector_t start, sector_t length);
370 int disk_expand_part_tbl(struct gendisk *disk, int target);
371 int hd_ref_init(struct hd_struct *part);
373 /* no need to get/put refcount of part0 */
374 static inline int hd_struct_try_get(struct hd_struct *part)
377 return percpu_ref_tryget_live(&part->ref);
381 static inline void hd_struct_put(struct hd_struct *part)
384 percpu_ref_put(&part->ref);
387 static inline void hd_free_part(struct hd_struct *part)
389 free_percpu(part->dkstats);
391 percpu_ref_exit(&part->ref);
395 * Any access of part->nr_sects which is not protected by partition
396 * bd_mutex or gendisk bdev bd_mutex, should be done using this
399 * Code written along the lines of i_size_read() and i_size_write().
400 * CONFIG_PREEMPTION case optimizes the case of UP kernel with preemption
403 static inline sector_t part_nr_sects_read(struct hd_struct *part)
405 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
409 seq = read_seqcount_begin(&part->nr_sects_seq);
410 nr_sects = part->nr_sects;
411 } while (read_seqcount_retry(&part->nr_sects_seq, seq));
413 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
417 nr_sects = part->nr_sects;
421 return part->nr_sects;
426 * Should be called with mutex lock held (typically bd_mutex) of partition
427 * to provide mutual exlusion among writers otherwise seqcount might be
428 * left in wrong state leaving the readers spinning infinitely.
430 static inline void part_nr_sects_write(struct hd_struct *part, sector_t size)
432 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
434 write_seqcount_begin(&part->nr_sects_seq);
435 part->nr_sects = size;
436 write_seqcount_end(&part->nr_sects_seq);
438 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
440 part->nr_sects = size;
443 part->nr_sects = size;
447 int bio_add_hw_page(struct request_queue *q, struct bio *bio,
448 struct page *page, unsigned int len, unsigned int offset,
449 unsigned int max_sectors, bool *same_page);
451 #endif /* BLK_INTERNAL_H */