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);
59 #define BIO_INLINE_VECS 4
60 struct bio_vec *bvec_alloc(mempool_t *pool, unsigned short *nr_vecs,
62 void bvec_free(mempool_t *pool, struct bio_vec *bv, unsigned short nr_vecs);
64 static inline bool biovec_phys_mergeable(struct request_queue *q,
65 struct bio_vec *vec1, struct bio_vec *vec2)
67 unsigned long mask = queue_segment_boundary(q);
68 phys_addr_t addr1 = page_to_phys(vec1->bv_page) + vec1->bv_offset;
69 phys_addr_t addr2 = page_to_phys(vec2->bv_page) + vec2->bv_offset;
71 if (addr1 + vec1->bv_len != addr2)
73 if (xen_domain() && !xen_biovec_phys_mergeable(vec1, vec2->bv_page))
75 if ((addr1 | mask) != ((addr2 + vec2->bv_len - 1) | mask))
80 static inline bool __bvec_gap_to_prev(struct request_queue *q,
81 struct bio_vec *bprv, unsigned int offset)
83 return (offset & queue_virt_boundary(q)) ||
84 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
88 * Check if adding a bio_vec after bprv with offset would create a gap in
89 * the SG list. Most drivers don't care about this, but some do.
91 static inline bool bvec_gap_to_prev(struct request_queue *q,
92 struct bio_vec *bprv, unsigned int offset)
94 if (!queue_virt_boundary(q))
96 return __bvec_gap_to_prev(q, bprv, offset);
99 static inline bool rq_mergeable(struct request *rq)
101 if (blk_rq_is_passthrough(rq))
104 if (req_op(rq) == REQ_OP_FLUSH)
107 if (req_op(rq) == REQ_OP_WRITE_ZEROES)
110 if (req_op(rq) == REQ_OP_ZONE_APPEND)
113 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
115 if (rq->rq_flags & RQF_NOMERGE_FLAGS)
122 * There are two different ways to handle DISCARD merges:
123 * 1) If max_discard_segments > 1, the driver treats every bio as a range and
124 * send the bios to controller together. The ranges don't need to be
126 * 2) Otherwise, the request will be normal read/write requests. The ranges
127 * need to be contiguous.
129 static inline bool blk_discard_mergable(struct request *req)
131 if (req_op(req) == REQ_OP_DISCARD &&
132 queue_max_discard_segments(req->q) > 1)
137 #ifdef CONFIG_BLK_DEV_INTEGRITY
138 void blk_flush_integrity(void);
139 bool __bio_integrity_endio(struct bio *);
140 void bio_integrity_free(struct bio *bio);
141 static inline bool bio_integrity_endio(struct bio *bio)
143 if (bio_integrity(bio))
144 return __bio_integrity_endio(bio);
148 bool blk_integrity_merge_rq(struct request_queue *, struct request *,
150 bool blk_integrity_merge_bio(struct request_queue *, struct request *,
153 static inline bool integrity_req_gap_back_merge(struct request *req,
156 struct bio_integrity_payload *bip = bio_integrity(req->bio);
157 struct bio_integrity_payload *bip_next = bio_integrity(next);
159 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
160 bip_next->bip_vec[0].bv_offset);
163 static inline bool integrity_req_gap_front_merge(struct request *req,
166 struct bio_integrity_payload *bip = bio_integrity(bio);
167 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
169 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
170 bip_next->bip_vec[0].bv_offset);
173 int blk_integrity_add(struct gendisk *disk);
174 void blk_integrity_del(struct gendisk *);
175 #else /* CONFIG_BLK_DEV_INTEGRITY */
176 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
177 struct request *r1, struct request *r2)
181 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
182 struct request *r, struct bio *b)
186 static inline bool integrity_req_gap_back_merge(struct request *req,
191 static inline bool integrity_req_gap_front_merge(struct request *req,
197 static inline void blk_flush_integrity(void)
200 static inline bool bio_integrity_endio(struct bio *bio)
204 static inline void bio_integrity_free(struct bio *bio)
207 static inline int blk_integrity_add(struct gendisk *disk)
211 static inline void blk_integrity_del(struct gendisk *disk)
214 #endif /* CONFIG_BLK_DEV_INTEGRITY */
216 unsigned long blk_rq_timeout(unsigned long timeout);
217 void blk_add_timer(struct request *req);
218 void blk_print_req_error(struct request *req, blk_status_t status);
220 bool blk_attempt_plug_merge(struct request_queue *q, struct bio *bio,
221 unsigned int nr_segs, bool *same_queue_rq);
222 bool blk_bio_list_merge(struct request_queue *q, struct list_head *list,
223 struct bio *bio, unsigned int nr_segs);
225 void __blk_account_io_start(struct request *req);
226 void __blk_account_io_done(struct request *req, u64 now);
231 #define BLK_MAX_REQUEST_COUNT 32
232 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
235 * Internal elevator interface
237 #define ELV_ON_HASH(rq) ((rq)->rq_flags & RQF_HASHED)
239 bool blk_insert_flush(struct request *rq);
241 int elevator_switch_mq(struct request_queue *q,
242 struct elevator_type *new_e);
243 void __elevator_exit(struct request_queue *, struct elevator_queue *);
244 int elv_register_queue(struct request_queue *q, bool uevent);
245 void elv_unregister_queue(struct request_queue *q);
247 static inline void elevator_exit(struct request_queue *q,
248 struct elevator_queue *e)
250 lockdep_assert_held(&q->sysfs_lock);
252 blk_mq_sched_free_rqs(q);
253 __elevator_exit(q, e);
256 ssize_t part_size_show(struct device *dev, struct device_attribute *attr,
258 ssize_t part_stat_show(struct device *dev, struct device_attribute *attr,
260 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
262 ssize_t part_fail_show(struct device *dev, struct device_attribute *attr,
264 ssize_t part_fail_store(struct device *dev, struct device_attribute *attr,
265 const char *buf, size_t count);
266 ssize_t part_timeout_show(struct device *, struct device_attribute *, char *);
267 ssize_t part_timeout_store(struct device *, struct device_attribute *,
268 const char *, size_t);
270 static inline bool blk_may_split(struct request_queue *q, struct bio *bio)
272 switch (bio_op(bio)) {
274 case REQ_OP_SECURE_ERASE:
275 case REQ_OP_WRITE_ZEROES:
276 case REQ_OP_WRITE_SAME:
277 return true; /* non-trivial splitting decisions */
283 * All drivers must accept single-segments bios that are <= PAGE_SIZE.
284 * This is a quick and dirty check that relies on the fact that
285 * bi_io_vec[0] is always valid if a bio has data. The check might
286 * lead to occasional false negatives when bios are cloned, but compared
287 * to the performance impact of cloned bios themselves the loop below
288 * doesn't matter anyway.
290 return q->limits.chunk_sectors || bio->bi_vcnt != 1 ||
291 bio->bi_io_vec->bv_len + bio->bi_io_vec->bv_offset > PAGE_SIZE;
294 void __blk_queue_split(struct request_queue *q, struct bio **bio,
295 unsigned int *nr_segs);
296 int ll_back_merge_fn(struct request *req, struct bio *bio,
297 unsigned int nr_segs);
298 bool blk_attempt_req_merge(struct request_queue *q, struct request *rq,
299 struct request *next);
300 unsigned int blk_recalc_rq_segments(struct request *rq);
301 void blk_rq_set_mixed_merge(struct request *rq);
302 bool blk_rq_merge_ok(struct request *rq, struct bio *bio);
303 enum elv_merge blk_try_merge(struct request *rq, struct bio *bio);
305 int blk_dev_init(void);
308 * Contribute to IO statistics IFF:
310 * a) it's attached to a gendisk, and
311 * b) the queue had IO stats enabled when this request was started
313 static inline bool blk_do_io_stat(struct request *rq)
315 return (rq->rq_flags & RQF_IO_STAT) && rq->rq_disk;
318 static inline void blk_account_io_done(struct request *req, u64 now)
321 * Account IO completion. flush_rq isn't accounted as a
322 * normal IO on queueing nor completion. Accounting the
323 * containing request is enough.
325 if (blk_do_io_stat(req) && req->part &&
326 !(req->rq_flags & RQF_FLUSH_SEQ))
327 __blk_account_io_done(req, now);
330 static inline void blk_account_io_start(struct request *req)
332 if (blk_do_io_stat(req))
333 __blk_account_io_start(req);
336 static inline void req_set_nomerge(struct request_queue *q, struct request *req)
338 req->cmd_flags |= REQ_NOMERGE;
339 if (req == q->last_merge)
340 q->last_merge = NULL;
344 * The max size one bio can handle is UINT_MAX becasue bvec_iter.bi_size
345 * is defined as 'unsigned int', meantime it has to aligned to with logical
346 * block size which is the minimum accepted unit by hardware.
348 static inline unsigned int bio_allowed_max_sectors(struct request_queue *q)
350 return round_down(UINT_MAX, queue_logical_block_size(q)) >> 9;
354 * The max bio size which is aligned to q->limits.discard_granularity. This
355 * is a hint to split large discard bio in generic block layer, then if device
356 * driver needs to split the discard bio into smaller ones, their bi_size can
357 * be very probably and easily aligned to discard_granularity of the device's
360 static inline unsigned int bio_aligned_discard_max_sectors(
361 struct request_queue *q)
363 return round_down(UINT_MAX, q->limits.discard_granularity) >>
368 * Internal io_context interface
370 void get_io_context(struct io_context *ioc);
371 struct io_cq *ioc_lookup_icq(struct io_context *ioc, struct request_queue *q);
372 struct io_cq *ioc_create_icq(struct io_context *ioc, struct request_queue *q,
374 void ioc_clear_queue(struct request_queue *q);
376 int create_task_io_context(struct task_struct *task, gfp_t gfp_mask, int node);
378 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW
379 extern ssize_t blk_throtl_sample_time_show(struct request_queue *q, char *page);
380 extern ssize_t blk_throtl_sample_time_store(struct request_queue *q,
381 const char *page, size_t count);
382 extern void blk_throtl_bio_endio(struct bio *bio);
383 extern void blk_throtl_stat_add(struct request *rq, u64 time);
385 static inline void blk_throtl_bio_endio(struct bio *bio) { }
386 static inline void blk_throtl_stat_add(struct request *rq, u64 time) { }
389 void __blk_queue_bounce(struct request_queue *q, struct bio **bio);
391 static inline bool blk_queue_may_bounce(struct request_queue *q)
393 return IS_ENABLED(CONFIG_BOUNCE) &&
394 q->limits.bounce == BLK_BOUNCE_HIGH &&
395 max_low_pfn >= max_pfn;
398 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
400 if (unlikely(blk_queue_may_bounce(q) && bio_has_data(*bio)))
401 __blk_queue_bounce(q, bio);
404 #ifdef CONFIG_BLK_CGROUP_IOLATENCY
405 extern int blk_iolatency_init(struct request_queue *q);
407 static inline int blk_iolatency_init(struct request_queue *q) { return 0; }
410 struct bio *blk_next_bio(struct bio *bio, unsigned int nr_pages, gfp_t gfp);
412 #ifdef CONFIG_BLK_DEV_ZONED
413 void blk_queue_free_zone_bitmaps(struct request_queue *q);
414 void blk_queue_clear_zone_settings(struct request_queue *q);
416 static inline void blk_queue_free_zone_bitmaps(struct request_queue *q) {}
417 static inline void blk_queue_clear_zone_settings(struct request_queue *q) {}
420 int blk_alloc_ext_minor(void);
421 void blk_free_ext_minor(unsigned int minor);
422 #define ADDPART_FLAG_NONE 0
423 #define ADDPART_FLAG_RAID 1
424 #define ADDPART_FLAG_WHOLEDISK 2
425 int bdev_add_partition(struct gendisk *disk, int partno, sector_t start,
427 int bdev_del_partition(struct gendisk *disk, int partno);
428 int bdev_resize_partition(struct gendisk *disk, int partno, sector_t start,
431 int bio_add_hw_page(struct request_queue *q, struct bio *bio,
432 struct page *page, unsigned int len, unsigned int offset,
433 unsigned int max_sectors, bool *same_page);
435 struct request_queue *blk_alloc_queue(int node_id);
437 int disk_alloc_events(struct gendisk *disk);
438 void disk_add_events(struct gendisk *disk);
439 void disk_del_events(struct gendisk *disk);
440 void disk_release_events(struct gendisk *disk);
441 extern struct device_attribute dev_attr_events;
442 extern struct device_attribute dev_attr_events_async;
443 extern struct device_attribute dev_attr_events_poll_msecs;
445 static inline void bio_clear_polled(struct bio *bio)
447 /* can't support alloc cache if we turn off polling */
448 bio_clear_flag(bio, BIO_PERCPU_CACHE);
449 bio->bi_opf &= ~REQ_POLLED;
452 long blkdev_ioctl(struct file *file, unsigned cmd, unsigned long arg);
453 long compat_blkdev_ioctl(struct file *file, unsigned cmd, unsigned long arg);
455 extern const struct address_space_operations def_blk_aops;
457 #endif /* BLK_INTERNAL_H */