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>
9 #include <linux/memblock.h> /* for max_pfn/max_low_pfn */
11 #include "blk-crypto-internal.h"
13 #include "blk-mq-sched.h"
17 /* Max future timer expiry for timeouts */
18 #define BLK_MAX_TIMEOUT (5 * HZ)
20 extern struct dentry *blk_debugfs_root;
22 struct blk_flush_queue {
23 unsigned int flush_pending_idx:1;
24 unsigned int flush_running_idx:1;
25 blk_status_t rq_status;
26 unsigned long flush_pending_since;
27 struct list_head flush_queue[2];
28 struct list_head flush_data_in_flight;
29 struct request *flush_rq;
31 spinlock_t mq_flush_lock;
34 extern struct kmem_cache *blk_requestq_cachep;
35 extern struct kobj_type blk_queue_ktype;
36 extern struct ida blk_queue_ida;
38 static inline struct blk_flush_queue *
39 blk_get_flush_queue(struct request_queue *q, struct blk_mq_ctx *ctx)
41 return blk_mq_map_queue(q, REQ_OP_FLUSH, ctx)->fq;
44 static inline void __blk_get_queue(struct request_queue *q)
46 kobject_get(&q->kobj);
49 bool is_flush_rq(struct request *req);
51 struct blk_flush_queue *blk_alloc_flush_queue(int node, int cmd_size,
53 void blk_free_flush_queue(struct blk_flush_queue *q);
55 void blk_freeze_queue(struct request_queue *q);
56 void __blk_mq_unfreeze_queue(struct request_queue *q, bool force_atomic);
57 void blk_queue_start_drain(struct request_queue *q);
58 int __bio_queue_enter(struct request_queue *q, struct bio *bio);
59 bool submit_bio_checks(struct bio *bio);
61 static inline bool blk_try_enter_queue(struct request_queue *q, bool pm)
64 if (!percpu_ref_tryget_live_rcu(&q->q_usage_counter))
68 * The code that increments the pm_only counter must ensure that the
69 * counter is globally visible before the queue is unfrozen.
71 if (blk_queue_pm_only(q) &&
72 (!pm || queue_rpm_status(q) == RPM_SUSPENDED))
85 static inline int bio_queue_enter(struct bio *bio)
87 struct request_queue *q = bdev_get_queue(bio->bi_bdev);
89 if (blk_try_enter_queue(q, false))
91 return __bio_queue_enter(q, bio);
94 #define BIO_INLINE_VECS 4
95 struct bio_vec *bvec_alloc(mempool_t *pool, unsigned short *nr_vecs,
97 void bvec_free(mempool_t *pool, struct bio_vec *bv, unsigned short nr_vecs);
99 static inline bool biovec_phys_mergeable(struct request_queue *q,
100 struct bio_vec *vec1, struct bio_vec *vec2)
102 unsigned long mask = queue_segment_boundary(q);
103 phys_addr_t addr1 = page_to_phys(vec1->bv_page) + vec1->bv_offset;
104 phys_addr_t addr2 = page_to_phys(vec2->bv_page) + vec2->bv_offset;
106 if (addr1 + vec1->bv_len != addr2)
108 if (xen_domain() && !xen_biovec_phys_mergeable(vec1, vec2->bv_page))
110 if ((addr1 | mask) != ((addr2 + vec2->bv_len - 1) | mask))
115 static inline bool __bvec_gap_to_prev(struct request_queue *q,
116 struct bio_vec *bprv, unsigned int offset)
118 return (offset & queue_virt_boundary(q)) ||
119 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
123 * Check if adding a bio_vec after bprv with offset would create a gap in
124 * the SG list. Most drivers don't care about this, but some do.
126 static inline bool bvec_gap_to_prev(struct request_queue *q,
127 struct bio_vec *bprv, unsigned int offset)
129 if (!queue_virt_boundary(q))
131 return __bvec_gap_to_prev(q, bprv, offset);
134 static inline bool rq_mergeable(struct request *rq)
136 if (blk_rq_is_passthrough(rq))
139 if (req_op(rq) == REQ_OP_FLUSH)
142 if (req_op(rq) == REQ_OP_WRITE_ZEROES)
145 if (req_op(rq) == REQ_OP_ZONE_APPEND)
148 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
150 if (rq->rq_flags & RQF_NOMERGE_FLAGS)
157 * There are two different ways to handle DISCARD merges:
158 * 1) If max_discard_segments > 1, the driver treats every bio as a range and
159 * send the bios to controller together. The ranges don't need to be
161 * 2) Otherwise, the request will be normal read/write requests. The ranges
162 * need to be contiguous.
164 static inline bool blk_discard_mergable(struct request *req)
166 if (req_op(req) == REQ_OP_DISCARD &&
167 queue_max_discard_segments(req->q) > 1)
172 #ifdef CONFIG_BLK_DEV_INTEGRITY
173 void blk_flush_integrity(void);
174 bool __bio_integrity_endio(struct bio *);
175 void bio_integrity_free(struct bio *bio);
176 static inline bool bio_integrity_endio(struct bio *bio)
178 if (bio_integrity(bio))
179 return __bio_integrity_endio(bio);
183 bool blk_integrity_merge_rq(struct request_queue *, struct request *,
185 bool blk_integrity_merge_bio(struct request_queue *, struct request *,
188 static inline bool integrity_req_gap_back_merge(struct request *req,
191 struct bio_integrity_payload *bip = bio_integrity(req->bio);
192 struct bio_integrity_payload *bip_next = bio_integrity(next);
194 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
195 bip_next->bip_vec[0].bv_offset);
198 static inline bool integrity_req_gap_front_merge(struct request *req,
201 struct bio_integrity_payload *bip = bio_integrity(bio);
202 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
204 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
205 bip_next->bip_vec[0].bv_offset);
208 int blk_integrity_add(struct gendisk *disk);
209 void blk_integrity_del(struct gendisk *);
210 #else /* CONFIG_BLK_DEV_INTEGRITY */
211 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
212 struct request *r1, struct request *r2)
216 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
217 struct request *r, struct bio *b)
221 static inline bool integrity_req_gap_back_merge(struct request *req,
226 static inline bool integrity_req_gap_front_merge(struct request *req,
232 static inline void blk_flush_integrity(void)
235 static inline bool bio_integrity_endio(struct bio *bio)
239 static inline void bio_integrity_free(struct bio *bio)
242 static inline int blk_integrity_add(struct gendisk *disk)
246 static inline void blk_integrity_del(struct gendisk *disk)
249 #endif /* CONFIG_BLK_DEV_INTEGRITY */
251 unsigned long blk_rq_timeout(unsigned long timeout);
252 void blk_add_timer(struct request *req);
253 void blk_print_req_error(struct request *req, blk_status_t status);
255 bool blk_attempt_plug_merge(struct request_queue *q, struct bio *bio,
256 unsigned int nr_segs, bool *same_queue_rq);
257 bool blk_bio_list_merge(struct request_queue *q, struct list_head *list,
258 struct bio *bio, unsigned int nr_segs);
260 void __blk_account_io_start(struct request *req);
261 void __blk_account_io_done(struct request *req, u64 now);
266 #define BLK_MAX_REQUEST_COUNT 32
267 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
270 * Internal elevator interface
272 #define ELV_ON_HASH(rq) ((rq)->rq_flags & RQF_HASHED)
274 void blk_insert_flush(struct request *rq);
276 int elevator_switch_mq(struct request_queue *q,
277 struct elevator_type *new_e);
278 void __elevator_exit(struct request_queue *, struct elevator_queue *);
279 int elv_register_queue(struct request_queue *q, bool uevent);
280 void elv_unregister_queue(struct request_queue *q);
282 static inline void elevator_exit(struct request_queue *q,
283 struct elevator_queue *e)
285 lockdep_assert_held(&q->sysfs_lock);
287 blk_mq_sched_free_rqs(q);
288 __elevator_exit(q, e);
291 ssize_t part_size_show(struct device *dev, struct device_attribute *attr,
293 ssize_t part_stat_show(struct device *dev, struct device_attribute *attr,
295 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
297 ssize_t part_fail_show(struct device *dev, struct device_attribute *attr,
299 ssize_t part_fail_store(struct device *dev, struct device_attribute *attr,
300 const char *buf, size_t count);
301 ssize_t part_timeout_show(struct device *, struct device_attribute *, char *);
302 ssize_t part_timeout_store(struct device *, struct device_attribute *,
303 const char *, size_t);
305 static inline bool blk_may_split(struct request_queue *q, struct bio *bio)
307 switch (bio_op(bio)) {
309 case REQ_OP_SECURE_ERASE:
310 case REQ_OP_WRITE_ZEROES:
311 case REQ_OP_WRITE_SAME:
312 return true; /* non-trivial splitting decisions */
318 * All drivers must accept single-segments bios that are <= PAGE_SIZE.
319 * This is a quick and dirty check that relies on the fact that
320 * bi_io_vec[0] is always valid if a bio has data. The check might
321 * lead to occasional false negatives when bios are cloned, but compared
322 * to the performance impact of cloned bios themselves the loop below
323 * doesn't matter anyway.
325 return q->limits.chunk_sectors || bio->bi_vcnt != 1 ||
326 bio->bi_io_vec->bv_len + bio->bi_io_vec->bv_offset > PAGE_SIZE;
329 void __blk_queue_split(struct request_queue *q, struct bio **bio,
330 unsigned int *nr_segs);
331 int ll_back_merge_fn(struct request *req, struct bio *bio,
332 unsigned int nr_segs);
333 bool blk_attempt_req_merge(struct request_queue *q, struct request *rq,
334 struct request *next);
335 unsigned int blk_recalc_rq_segments(struct request *rq);
336 void blk_rq_set_mixed_merge(struct request *rq);
337 bool blk_rq_merge_ok(struct request *rq, struct bio *bio);
338 enum elv_merge blk_try_merge(struct request *rq, struct bio *bio);
340 int blk_dev_init(void);
343 * Contribute to IO statistics IFF:
345 * a) it's attached to a gendisk, and
346 * b) the queue had IO stats enabled when this request was started
348 static inline bool blk_do_io_stat(struct request *rq)
350 return (rq->rq_flags & RQF_IO_STAT) && rq->rq_disk;
353 static inline void blk_account_io_done(struct request *req, u64 now)
356 * Account IO completion. flush_rq isn't accounted as a
357 * normal IO on queueing nor completion. Accounting the
358 * containing request is enough.
360 if (blk_do_io_stat(req) && req->part &&
361 !(req->rq_flags & RQF_FLUSH_SEQ))
362 __blk_account_io_done(req, now);
365 static inline void blk_account_io_start(struct request *req)
367 if (blk_do_io_stat(req))
368 __blk_account_io_start(req);
371 static inline void req_set_nomerge(struct request_queue *q, struct request *req)
373 req->cmd_flags |= REQ_NOMERGE;
374 if (req == q->last_merge)
375 q->last_merge = NULL;
379 * The max size one bio can handle is UINT_MAX becasue bvec_iter.bi_size
380 * is defined as 'unsigned int', meantime it has to aligned to with logical
381 * block size which is the minimum accepted unit by hardware.
383 static inline unsigned int bio_allowed_max_sectors(struct request_queue *q)
385 return round_down(UINT_MAX, queue_logical_block_size(q)) >> 9;
389 * The max bio size which is aligned to q->limits.discard_granularity. This
390 * is a hint to split large discard bio in generic block layer, then if device
391 * driver needs to split the discard bio into smaller ones, their bi_size can
392 * be very probably and easily aligned to discard_granularity of the device's
395 static inline unsigned int bio_aligned_discard_max_sectors(
396 struct request_queue *q)
398 return round_down(UINT_MAX, q->limits.discard_granularity) >>
403 * Internal io_context interface
405 void get_io_context(struct io_context *ioc);
406 struct io_cq *ioc_lookup_icq(struct io_context *ioc, struct request_queue *q);
407 struct io_cq *ioc_create_icq(struct io_context *ioc, struct request_queue *q,
409 void ioc_clear_queue(struct request_queue *q);
411 int create_task_io_context(struct task_struct *task, gfp_t gfp_mask, int node);
413 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW
414 extern ssize_t blk_throtl_sample_time_show(struct request_queue *q, char *page);
415 extern ssize_t blk_throtl_sample_time_store(struct request_queue *q,
416 const char *page, size_t count);
417 extern void blk_throtl_bio_endio(struct bio *bio);
418 extern void blk_throtl_stat_add(struct request *rq, u64 time);
420 static inline void blk_throtl_bio_endio(struct bio *bio) { }
421 static inline void blk_throtl_stat_add(struct request *rq, u64 time) { }
424 void __blk_queue_bounce(struct request_queue *q, struct bio **bio);
426 static inline bool blk_queue_may_bounce(struct request_queue *q)
428 return IS_ENABLED(CONFIG_BOUNCE) &&
429 q->limits.bounce == BLK_BOUNCE_HIGH &&
430 max_low_pfn >= max_pfn;
433 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
435 if (unlikely(blk_queue_may_bounce(q) && bio_has_data(*bio)))
436 __blk_queue_bounce(q, bio);
439 #ifdef CONFIG_BLK_CGROUP_IOLATENCY
440 extern int blk_iolatency_init(struct request_queue *q);
442 static inline int blk_iolatency_init(struct request_queue *q) { return 0; }
445 struct bio *blk_next_bio(struct bio *bio, unsigned int nr_pages, gfp_t gfp);
447 #ifdef CONFIG_BLK_DEV_ZONED
448 void blk_queue_free_zone_bitmaps(struct request_queue *q);
449 void blk_queue_clear_zone_settings(struct request_queue *q);
451 static inline void blk_queue_free_zone_bitmaps(struct request_queue *q) {}
452 static inline void blk_queue_clear_zone_settings(struct request_queue *q) {}
455 int blk_alloc_ext_minor(void);
456 void blk_free_ext_minor(unsigned int minor);
457 #define ADDPART_FLAG_NONE 0
458 #define ADDPART_FLAG_RAID 1
459 #define ADDPART_FLAG_WHOLEDISK 2
460 int bdev_add_partition(struct gendisk *disk, int partno, sector_t start,
462 int bdev_del_partition(struct gendisk *disk, int partno);
463 int bdev_resize_partition(struct gendisk *disk, int partno, sector_t start,
466 int bio_add_hw_page(struct request_queue *q, struct bio *bio,
467 struct page *page, unsigned int len, unsigned int offset,
468 unsigned int max_sectors, bool *same_page);
470 struct request_queue *blk_alloc_queue(int node_id);
472 int disk_alloc_events(struct gendisk *disk);
473 void disk_add_events(struct gendisk *disk);
474 void disk_del_events(struct gendisk *disk);
475 void disk_release_events(struct gendisk *disk);
476 extern struct device_attribute dev_attr_events;
477 extern struct device_attribute dev_attr_events_async;
478 extern struct device_attribute dev_attr_events_poll_msecs;
480 static inline void bio_clear_polled(struct bio *bio)
482 /* can't support alloc cache if we turn off polling */
483 bio_clear_flag(bio, BIO_PERCPU_CACHE);
484 bio->bi_opf &= ~REQ_POLLED;
487 long blkdev_ioctl(struct file *file, unsigned cmd, unsigned long arg);
488 long compat_blkdev_ioctl(struct file *file, unsigned cmd, unsigned long arg);
490 extern const struct address_space_operations def_blk_aops;
492 int disk_register_independent_access_ranges(struct gendisk *disk,
493 struct blk_independent_access_ranges *new_iars);
494 void disk_unregister_independent_access_ranges(struct gendisk *disk);
496 #endif /* BLK_INTERNAL_H */