1 /* SPDX-License-Identifier: GPL-2.0 */
5 #include <linux/blk-crypto.h>
6 #include <linux/memblock.h> /* for max_pfn/max_low_pfn */
8 #include "blk-crypto-internal.h"
12 /* Max future timer expiry for timeouts */
13 #define BLK_MAX_TIMEOUT (5 * HZ)
15 extern struct dentry *blk_debugfs_root;
17 struct blk_flush_queue {
18 unsigned int flush_pending_idx:1;
19 unsigned int flush_running_idx:1;
20 blk_status_t rq_status;
21 unsigned long flush_pending_since;
22 struct list_head flush_queue[2];
23 struct list_head flush_data_in_flight;
24 struct request *flush_rq;
26 spinlock_t mq_flush_lock;
29 extern struct kmem_cache *blk_requestq_cachep;
30 extern struct kobj_type blk_queue_ktype;
31 extern struct ida blk_queue_ida;
33 static inline void __blk_get_queue(struct request_queue *q)
35 kobject_get(&q->kobj);
38 bool is_flush_rq(struct request *req);
40 struct blk_flush_queue *blk_alloc_flush_queue(int node, int cmd_size,
42 void blk_free_flush_queue(struct blk_flush_queue *q);
44 void blk_freeze_queue(struct request_queue *q);
45 void __blk_mq_unfreeze_queue(struct request_queue *q, bool force_atomic);
46 void blk_queue_start_drain(struct request_queue *q);
47 int __bio_queue_enter(struct request_queue *q, struct bio *bio);
48 bool submit_bio_checks(struct bio *bio);
50 static inline bool blk_try_enter_queue(struct request_queue *q, bool pm)
53 if (!percpu_ref_tryget_live_rcu(&q->q_usage_counter))
57 * The code that increments the pm_only counter must ensure that the
58 * counter is globally visible before the queue is unfrozen.
60 if (blk_queue_pm_only(q) &&
61 (!pm || queue_rpm_status(q) == RPM_SUSPENDED))
74 static inline int bio_queue_enter(struct bio *bio)
76 struct request_queue *q = bdev_get_queue(bio->bi_bdev);
78 if (blk_try_enter_queue(q, false))
80 return __bio_queue_enter(q, bio);
83 #define BIO_INLINE_VECS 4
84 struct bio_vec *bvec_alloc(mempool_t *pool, unsigned short *nr_vecs,
86 void bvec_free(mempool_t *pool, struct bio_vec *bv, unsigned short nr_vecs);
88 static inline bool biovec_phys_mergeable(struct request_queue *q,
89 struct bio_vec *vec1, struct bio_vec *vec2)
91 unsigned long mask = queue_segment_boundary(q);
92 phys_addr_t addr1 = page_to_phys(vec1->bv_page) + vec1->bv_offset;
93 phys_addr_t addr2 = page_to_phys(vec2->bv_page) + vec2->bv_offset;
95 if (addr1 + vec1->bv_len != addr2)
97 if (xen_domain() && !xen_biovec_phys_mergeable(vec1, vec2->bv_page))
99 if ((addr1 | mask) != ((addr2 + vec2->bv_len - 1) | mask))
104 static inline bool __bvec_gap_to_prev(struct request_queue *q,
105 struct bio_vec *bprv, unsigned int offset)
107 return (offset & queue_virt_boundary(q)) ||
108 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
112 * Check if adding a bio_vec after bprv with offset would create a gap in
113 * the SG list. Most drivers don't care about this, but some do.
115 static inline bool bvec_gap_to_prev(struct request_queue *q,
116 struct bio_vec *bprv, unsigned int offset)
118 if (!queue_virt_boundary(q))
120 return __bvec_gap_to_prev(q, bprv, offset);
123 static inline bool rq_mergeable(struct request *rq)
125 if (blk_rq_is_passthrough(rq))
128 if (req_op(rq) == REQ_OP_FLUSH)
131 if (req_op(rq) == REQ_OP_WRITE_ZEROES)
134 if (req_op(rq) == REQ_OP_ZONE_APPEND)
137 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
139 if (rq->rq_flags & RQF_NOMERGE_FLAGS)
146 * There are two different ways to handle DISCARD merges:
147 * 1) If max_discard_segments > 1, the driver treats every bio as a range and
148 * send the bios to controller together. The ranges don't need to be
150 * 2) Otherwise, the request will be normal read/write requests. The ranges
151 * need to be contiguous.
153 static inline bool blk_discard_mergable(struct request *req)
155 if (req_op(req) == REQ_OP_DISCARD &&
156 queue_max_discard_segments(req->q) > 1)
161 #ifdef CONFIG_BLK_DEV_INTEGRITY
162 void blk_flush_integrity(void);
163 bool __bio_integrity_endio(struct bio *);
164 void bio_integrity_free(struct bio *bio);
165 static inline bool bio_integrity_endio(struct bio *bio)
167 if (bio_integrity(bio))
168 return __bio_integrity_endio(bio);
172 bool blk_integrity_merge_rq(struct request_queue *, struct request *,
174 bool blk_integrity_merge_bio(struct request_queue *, struct request *,
177 static inline bool integrity_req_gap_back_merge(struct request *req,
180 struct bio_integrity_payload *bip = bio_integrity(req->bio);
181 struct bio_integrity_payload *bip_next = bio_integrity(next);
183 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
184 bip_next->bip_vec[0].bv_offset);
187 static inline bool integrity_req_gap_front_merge(struct request *req,
190 struct bio_integrity_payload *bip = bio_integrity(bio);
191 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
193 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
194 bip_next->bip_vec[0].bv_offset);
197 int blk_integrity_add(struct gendisk *disk);
198 void blk_integrity_del(struct gendisk *);
199 #else /* CONFIG_BLK_DEV_INTEGRITY */
200 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
201 struct request *r1, struct request *r2)
205 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
206 struct request *r, struct bio *b)
210 static inline bool integrity_req_gap_back_merge(struct request *req,
215 static inline bool integrity_req_gap_front_merge(struct request *req,
221 static inline void blk_flush_integrity(void)
224 static inline bool bio_integrity_endio(struct bio *bio)
228 static inline void bio_integrity_free(struct bio *bio)
231 static inline int blk_integrity_add(struct gendisk *disk)
235 static inline void blk_integrity_del(struct gendisk *disk)
238 #endif /* CONFIG_BLK_DEV_INTEGRITY */
240 unsigned long blk_rq_timeout(unsigned long timeout);
241 void blk_add_timer(struct request *req);
242 const char *blk_status_to_str(blk_status_t status);
244 bool blk_attempt_plug_merge(struct request_queue *q, struct bio *bio,
245 unsigned int nr_segs);
246 bool blk_bio_list_merge(struct request_queue *q, struct list_head *list,
247 struct bio *bio, unsigned int nr_segs);
252 #define BLK_MAX_REQUEST_COUNT 32
253 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
256 * Internal elevator interface
258 #define ELV_ON_HASH(rq) ((rq)->rq_flags & RQF_HASHED)
260 void blk_insert_flush(struct request *rq);
262 int elevator_switch_mq(struct request_queue *q,
263 struct elevator_type *new_e);
264 void elevator_exit(struct request_queue *q);
265 int elv_register_queue(struct request_queue *q, bool uevent);
266 void elv_unregister_queue(struct request_queue *q);
268 ssize_t part_size_show(struct device *dev, struct device_attribute *attr,
270 ssize_t part_stat_show(struct device *dev, struct device_attribute *attr,
272 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
274 ssize_t part_fail_show(struct device *dev, struct device_attribute *attr,
276 ssize_t part_fail_store(struct device *dev, struct device_attribute *attr,
277 const char *buf, size_t count);
278 ssize_t part_timeout_show(struct device *, struct device_attribute *, char *);
279 ssize_t part_timeout_store(struct device *, struct device_attribute *,
280 const char *, size_t);
282 static inline bool blk_may_split(struct request_queue *q, struct bio *bio)
284 switch (bio_op(bio)) {
286 case REQ_OP_SECURE_ERASE:
287 case REQ_OP_WRITE_ZEROES:
288 case REQ_OP_WRITE_SAME:
289 return true; /* non-trivial splitting decisions */
295 * All drivers must accept single-segments bios that are <= PAGE_SIZE.
296 * This is a quick and dirty check that relies on the fact that
297 * bi_io_vec[0] is always valid if a bio has data. The check might
298 * lead to occasional false negatives when bios are cloned, but compared
299 * to the performance impact of cloned bios themselves the loop below
300 * doesn't matter anyway.
302 return q->limits.chunk_sectors || bio->bi_vcnt != 1 ||
303 bio->bi_io_vec->bv_len + bio->bi_io_vec->bv_offset > PAGE_SIZE;
306 void __blk_queue_split(struct request_queue *q, struct bio **bio,
307 unsigned int *nr_segs);
308 int ll_back_merge_fn(struct request *req, struct bio *bio,
309 unsigned int nr_segs);
310 bool blk_attempt_req_merge(struct request_queue *q, struct request *rq,
311 struct request *next);
312 unsigned int blk_recalc_rq_segments(struct request *rq);
313 void blk_rq_set_mixed_merge(struct request *rq);
314 bool blk_rq_merge_ok(struct request *rq, struct bio *bio);
315 enum elv_merge blk_try_merge(struct request *rq, struct bio *bio);
317 int blk_dev_init(void);
320 * Contribute to IO statistics IFF:
322 * a) it's attached to a gendisk, and
323 * b) the queue had IO stats enabled when this request was started
325 static inline bool blk_do_io_stat(struct request *rq)
327 return (rq->rq_flags & RQF_IO_STAT) && rq->q->disk;
330 void update_io_ticks(struct block_device *part, unsigned long now, bool end);
332 static inline void req_set_nomerge(struct request_queue *q, struct request *req)
334 req->cmd_flags |= REQ_NOMERGE;
335 if (req == q->last_merge)
336 q->last_merge = NULL;
340 * The max size one bio can handle is UINT_MAX becasue bvec_iter.bi_size
341 * is defined as 'unsigned int', meantime it has to aligned to with logical
342 * block size which is the minimum accepted unit by hardware.
344 static inline unsigned int bio_allowed_max_sectors(struct request_queue *q)
346 return round_down(UINT_MAX, queue_logical_block_size(q)) >> 9;
350 * The max bio size which is aligned to q->limits.discard_granularity. This
351 * is a hint to split large discard bio in generic block layer, then if device
352 * driver needs to split the discard bio into smaller ones, their bi_size can
353 * be very probably and easily aligned to discard_granularity of the device's
356 static inline unsigned int bio_aligned_discard_max_sectors(
357 struct request_queue *q)
359 return round_down(UINT_MAX, q->limits.discard_granularity) >>
364 * Internal io_context interface
366 struct io_cq *ioc_find_get_icq(struct request_queue *q);
367 struct io_cq *ioc_lookup_icq(struct request_queue *q);
368 void ioc_clear_queue(struct request_queue *q);
370 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW
371 extern ssize_t blk_throtl_sample_time_show(struct request_queue *q, char *page);
372 extern ssize_t blk_throtl_sample_time_store(struct request_queue *q,
373 const char *page, size_t count);
374 extern void blk_throtl_bio_endio(struct bio *bio);
375 extern void blk_throtl_stat_add(struct request *rq, u64 time);
377 static inline void blk_throtl_bio_endio(struct bio *bio) { }
378 static inline void blk_throtl_stat_add(struct request *rq, u64 time) { }
381 void __blk_queue_bounce(struct request_queue *q, struct bio **bio);
383 static inline bool blk_queue_may_bounce(struct request_queue *q)
385 return IS_ENABLED(CONFIG_BOUNCE) &&
386 q->limits.bounce == BLK_BOUNCE_HIGH &&
387 max_low_pfn >= max_pfn;
390 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
392 if (unlikely(blk_queue_may_bounce(q) && bio_has_data(*bio)))
393 __blk_queue_bounce(q, bio);
396 #ifdef CONFIG_BLK_CGROUP_IOLATENCY
397 extern int blk_iolatency_init(struct request_queue *q);
399 static inline int blk_iolatency_init(struct request_queue *q) { return 0; }
402 struct bio *blk_next_bio(struct bio *bio, unsigned int nr_pages, gfp_t gfp);
404 #ifdef CONFIG_BLK_DEV_ZONED
405 void blk_queue_free_zone_bitmaps(struct request_queue *q);
406 void blk_queue_clear_zone_settings(struct request_queue *q);
408 static inline void blk_queue_free_zone_bitmaps(struct request_queue *q) {}
409 static inline void blk_queue_clear_zone_settings(struct request_queue *q) {}
412 int blk_alloc_ext_minor(void);
413 void blk_free_ext_minor(unsigned int minor);
414 #define ADDPART_FLAG_NONE 0
415 #define ADDPART_FLAG_RAID 1
416 #define ADDPART_FLAG_WHOLEDISK 2
417 int bdev_add_partition(struct gendisk *disk, int partno, sector_t start,
419 int bdev_del_partition(struct gendisk *disk, int partno);
420 int bdev_resize_partition(struct gendisk *disk, int partno, sector_t start,
423 int bio_add_hw_page(struct request_queue *q, struct bio *bio,
424 struct page *page, unsigned int len, unsigned int offset,
425 unsigned int max_sectors, bool *same_page);
427 struct request_queue *blk_alloc_queue(int node_id);
428 int disk_scan_partitions(struct gendisk *disk, fmode_t mode);
430 int disk_alloc_events(struct gendisk *disk);
431 void disk_add_events(struct gendisk *disk);
432 void disk_del_events(struct gendisk *disk);
433 void disk_release_events(struct gendisk *disk);
434 extern struct device_attribute dev_attr_events;
435 extern struct device_attribute dev_attr_events_async;
436 extern struct device_attribute dev_attr_events_poll_msecs;
438 static inline void bio_clear_polled(struct bio *bio)
440 /* can't support alloc cache if we turn off polling */
441 bio_clear_flag(bio, BIO_PERCPU_CACHE);
442 bio->bi_opf &= ~REQ_POLLED;
445 long blkdev_ioctl(struct file *file, unsigned cmd, unsigned long arg);
446 long compat_blkdev_ioctl(struct file *file, unsigned cmd, unsigned long arg);
448 extern const struct address_space_operations def_blk_aops;
450 int disk_register_independent_access_ranges(struct gendisk *disk,
451 struct blk_independent_access_ranges *new_iars);
452 void disk_unregister_independent_access_ranges(struct gendisk *disk);
454 #ifdef CONFIG_FAIL_MAKE_REQUEST
455 bool should_fail_request(struct block_device *part, unsigned int bytes);
456 #else /* CONFIG_FAIL_MAKE_REQUEST */
457 static inline bool should_fail_request(struct block_device *part,
462 #endif /* CONFIG_FAIL_MAKE_REQUEST */
464 #endif /* BLK_INTERNAL_H */