block/blk.h

   1 /* SPDX-License-Identifier: GPL-2.0 */
   2 #ifndef BLK_INTERNAL_H
   3 #define BLK_INTERNAL_H
   4
   5 #include <linux/idr.h>
   6 #include <linux/blk-mq.h>
   7 #include <linux/part_stat.h>
   8 #include <linux/blk-crypto.h>
   9 #include <xen/xen.h>
  10 #include "blk-crypto-internal.h"
  11 #include "blk-mq.h"
  12 #include "blk-mq-sched.h"
  13
  14 /* Max future timer expiry for timeouts */
  15 #define BLK_MAX_TIMEOUT         (5 * HZ)
  16
  17 extern struct dentry *blk_debugfs_root;
  18
  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;
  27
  28         struct lock_class_key   key;
  29         spinlock_t              mq_flush_lock;
  30 };
  31
  32 extern struct kmem_cache *blk_requestq_cachep;
  33 extern struct kobj_type blk_queue_ktype;
  34 extern struct ida blk_queue_ida;
  35
  36 static inline struct blk_flush_queue *
  37 blk_get_flush_queue(struct request_queue *q, struct blk_mq_ctx *ctx)
  38 {
  39         return blk_mq_map_queue(q, REQ_OP_FLUSH, ctx)->fq;
  40 }
  41
  42 static inline void __blk_get_queue(struct request_queue *q)
  43 {
  44         kobject_get(&q->kobj);
  45 }
  46
  47 static inline bool
  48 is_flush_rq(struct request *req, struct blk_mq_hw_ctx *hctx)
  49 {
  50         return hctx->fq->flush_rq == req;
  51 }
  52
  53 struct blk_flush_queue *blk_alloc_flush_queue(int node, int cmd_size,
  54                                               gfp_t flags);
  55 void blk_free_flush_queue(struct blk_flush_queue *q);
  56
  57 void blk_freeze_queue(struct request_queue *q);
  58
  59 static inline bool biovec_phys_mergeable(struct request_queue *q,
  60                 struct bio_vec *vec1, struct bio_vec *vec2)
  61 {
  62         unsigned long mask = queue_segment_boundary(q);
  63         phys_addr_t addr1 = page_to_phys(vec1->bv_page) + vec1->bv_offset;
  64         phys_addr_t addr2 = page_to_phys(vec2->bv_page) + vec2->bv_offset;
  65
  66         if (addr1 + vec1->bv_len != addr2)
  67                 return false;
  68         if (xen_domain() && !xen_biovec_phys_mergeable(vec1, vec2->bv_page))
  69                 return false;
  70         if ((addr1 | mask) != ((addr2 + vec2->bv_len - 1) | mask))
  71                 return false;
  72         return true;
  73 }
  74
  75 static inline bool __bvec_gap_to_prev(struct request_queue *q,
  76                 struct bio_vec *bprv, unsigned int offset)
  77 {
  78         return (offset & queue_virt_boundary(q)) ||
  79                 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
  80 }
  81
  82 /*
  83  * Check if adding a bio_vec after bprv with offset would create a gap in
  84  * the SG list. Most drivers don't care about this, but some do.
  85  */
  86 static inline bool bvec_gap_to_prev(struct request_queue *q,
  87                 struct bio_vec *bprv, unsigned int offset)
  88 {
  89         if (!queue_virt_boundary(q))
  90                 return false;
  91         return __bvec_gap_to_prev(q, bprv, offset);
  92 }
  93
  94 static inline void blk_rq_bio_prep(struct request *rq, struct bio *bio,
  95                 unsigned int nr_segs)
  96 {
  97         rq->nr_phys_segments = nr_segs;
  98         rq->__data_len = bio->bi_iter.bi_size;
  99         rq->bio = rq->biotail = bio;
 100         rq->ioprio = bio_prio(bio);
 101
 102         if (bio->bi_disk)
 103                 rq->rq_disk = bio->bi_disk;
 104 }
 105
 106 #ifdef CONFIG_BLK_DEV_INTEGRITY
 107 void blk_flush_integrity(void);
 108 bool __bio_integrity_endio(struct bio *);
 109 void bio_integrity_free(struct bio *bio);
 110 static inline bool bio_integrity_endio(struct bio *bio)
 111 {
 112         if (bio_integrity(bio))
 113                 return __bio_integrity_endio(bio);
 114         return true;
 115 }
 116
 117 static inline bool integrity_req_gap_back_merge(struct request *req,
 118                 struct bio *next)
 119 {
 120         struct bio_integrity_payload *bip = bio_integrity(req->bio);
 121         struct bio_integrity_payload *bip_next = bio_integrity(next);
 122
 123         return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
 124                                 bip_next->bip_vec[0].bv_offset);
 125 }
 126
 127 static inline bool integrity_req_gap_front_merge(struct request *req,
 128                 struct bio *bio)
 129 {
 130         struct bio_integrity_payload *bip = bio_integrity(bio);
 131         struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
 132
 133         return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
 134                                 bip_next->bip_vec[0].bv_offset);
 135 }
 136
 137 void blk_integrity_add(struct gendisk *);
 138 void blk_integrity_del(struct gendisk *);
 139 #else /* CONFIG_BLK_DEV_INTEGRITY */
 140 static inline bool integrity_req_gap_back_merge(struct request *req,
 141                 struct bio *next)
 142 {
 143         return false;
 144 }
 145 static inline bool integrity_req_gap_front_merge(struct request *req,
 146                 struct bio *bio)
 147 {
 148         return false;
 149 }
 150
 151 static inline void blk_flush_integrity(void)
 152 {
 153 }
 154 static inline bool bio_integrity_endio(struct bio *bio)
 155 {
 156         return true;
 157 }
 158 static inline void bio_integrity_free(struct bio *bio)
 159 {
 160 }
 161 static inline void blk_integrity_add(struct gendisk *disk)
 162 {
 163 }
 164 static inline void blk_integrity_del(struct gendisk *disk)
 165 {
 166 }
 167 #endif /* CONFIG_BLK_DEV_INTEGRITY */
 168
 169 unsigned long blk_rq_timeout(unsigned long timeout);
 170 void blk_add_timer(struct request *req);
 171
 172 bool bio_attempt_front_merge(struct request *req, struct bio *bio,
 173                 unsigned int nr_segs);
 174 bool bio_attempt_back_merge(struct request *req, struct bio *bio,
 175                 unsigned int nr_segs);
 176 bool bio_attempt_discard_merge(struct request_queue *q, struct request *req,
 177                 struct bio *bio);
 178 bool blk_attempt_plug_merge(struct request_queue *q, struct bio *bio,
 179                 unsigned int nr_segs, struct request **same_queue_rq);
 180
 181 void blk_account_io_start(struct request *req);
 182 void blk_account_io_done(struct request *req, u64 now);
 183
 184 /*
 185  * Internal elevator interface
 186  */
 187 #define ELV_ON_HASH(rq) ((rq)->rq_flags & RQF_HASHED)
 188
 189 void blk_insert_flush(struct request *rq);
 190
 191 void elevator_init_mq(struct request_queue *q);
 192 int elevator_switch_mq(struct request_queue *q,
 193                               struct elevator_type *new_e);
 194 void __elevator_exit(struct request_queue *, struct elevator_queue *);
 195 int elv_register_queue(struct request_queue *q, bool uevent);
 196 void elv_unregister_queue(struct request_queue *q);
 197
 198 static inline void elevator_exit(struct request_queue *q,
 199                 struct elevator_queue *e)
 200 {
 201         lockdep_assert_held(&q->sysfs_lock);
 202
 203         blk_mq_sched_free_requests(q);
 204         __elevator_exit(q, e);
 205 }
 206
 207 struct hd_struct *__disk_get_part(struct gendisk *disk, int partno);
 208
 209 ssize_t part_size_show(struct device *dev, struct device_attribute *attr,
 210                 char *buf);
 211 ssize_t part_stat_show(struct device *dev, struct device_attribute *attr,
 212                 char *buf);
 213 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
 214                 char *buf);
 215 ssize_t part_fail_show(struct device *dev, struct device_attribute *attr,
 216                 char *buf);
 217 ssize_t part_fail_store(struct device *dev, struct device_attribute *attr,
 218                 const char *buf, size_t count);
 219 ssize_t part_timeout_show(struct device *, struct device_attribute *, char *);
 220 ssize_t part_timeout_store(struct device *, struct device_attribute *,
 221                                 const char *, size_t);
 222
 223 void __blk_queue_split(struct bio **bio, unsigned int *nr_segs);
 224 int ll_back_merge_fn(struct request *req, struct bio *bio,
 225                 unsigned int nr_segs);
 226 int ll_front_merge_fn(struct request *req,  struct bio *bio,
 227                 unsigned int nr_segs);
 228 struct request *attempt_back_merge(struct request_queue *q, struct request *rq);
 229 struct request *attempt_front_merge(struct request_queue *q, struct request *rq);
 230 int blk_attempt_req_merge(struct request_queue *q, struct request *rq,
 231                                 struct request *next);
 232 unsigned int blk_recalc_rq_segments(struct request *rq);
 233 void blk_rq_set_mixed_merge(struct request *rq);
 234 bool blk_rq_merge_ok(struct request *rq, struct bio *bio);
 235 enum elv_merge blk_try_merge(struct request *rq, struct bio *bio);
 236
 237 int blk_dev_init(void);
 238
 239 /*
 240  * Contribute to IO statistics IFF:
 241  *
 242  *      a) it's attached to a gendisk, and
 243  *      b) the queue had IO stats enabled when this request was started
 244  */
 245 static inline bool blk_do_io_stat(struct request *rq)
 246 {
 247         return rq->rq_disk && (rq->rq_flags & RQF_IO_STAT);
 248 }
 249
 250 static inline void req_set_nomerge(struct request_queue *q, struct request *req)
 251 {
 252         req->cmd_flags |= REQ_NOMERGE;
 253         if (req == q->last_merge)
 254                 q->last_merge = NULL;
 255 }
 256
 257 /*
 258  * The max size one bio can handle is UINT_MAX becasue bvec_iter.bi_size
 259  * is defined as 'unsigned int', meantime it has to aligned to with logical
 260  * block size which is the minimum accepted unit by hardware.
 261  */
 262 static inline unsigned int bio_allowed_max_sectors(struct request_queue *q)
 263 {
 264         return round_down(UINT_MAX, queue_logical_block_size(q)) >> 9;
 265 }
 266
 267 /*
 268  * The max bio size which is aligned to q->limits.discard_granularity. This
 269  * is a hint to split large discard bio in generic block layer, then if device
 270  * driver needs to split the discard bio into smaller ones, their bi_size can
 271  * be very probably and easily aligned to discard_granularity of the device's
 272  * queue.
 273  */
 274 static inline unsigned int bio_aligned_discard_max_sectors(
 275                                         struct request_queue *q)
 276 {
 277         return round_down(UINT_MAX, q->limits.discard_granularity) >>
 278                         SECTOR_SHIFT;
 279 }
 280
 281 /*
 282  * Internal io_context interface
 283  */
 284 void get_io_context(struct io_context *ioc);
 285 struct io_cq *ioc_lookup_icq(struct io_context *ioc, struct request_queue *q);
 286 struct io_cq *ioc_create_icq(struct io_context *ioc, struct request_queue *q,
 287                              gfp_t gfp_mask);
 288 void ioc_clear_queue(struct request_queue *q);
 289
 290 int create_task_io_context(struct task_struct *task, gfp_t gfp_mask, int node);
 291
 292 /*
 293  * Internal throttling interface
 294  */
 295 #ifdef CONFIG_BLK_DEV_THROTTLING
 296 extern int blk_throtl_init(struct request_queue *q);
 297 extern void blk_throtl_exit(struct request_queue *q);
 298 extern void blk_throtl_register_queue(struct request_queue *q);
 299 bool blk_throtl_bio(struct bio *bio);
 300 #else /* CONFIG_BLK_DEV_THROTTLING */
 301 static inline int blk_throtl_init(struct request_queue *q) { return 0; }
 302 static inline void blk_throtl_exit(struct request_queue *q) { }
 303 static inline void blk_throtl_register_queue(struct request_queue *q) { }
 304 static inline bool blk_throtl_bio(struct bio *bio) { return false; }
 305 #endif /* CONFIG_BLK_DEV_THROTTLING */
 306 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW
 307 extern ssize_t blk_throtl_sample_time_show(struct request_queue *q, char *page);
 308 extern ssize_t blk_throtl_sample_time_store(struct request_queue *q,
 309         const char *page, size_t count);
 310 extern void blk_throtl_bio_endio(struct bio *bio);
 311 extern void blk_throtl_stat_add(struct request *rq, u64 time);
 312 #else
 313 static inline void blk_throtl_bio_endio(struct bio *bio) { }
 314 static inline void blk_throtl_stat_add(struct request *rq, u64 time) { }
 315 #endif
 316
 317 #ifdef CONFIG_BOUNCE
 318 extern int init_emergency_isa_pool(void);
 319 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
 320 #else
 321 static inline int init_emergency_isa_pool(void)
 322 {
 323         return 0;
 324 }
 325 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
 326 {
 327 }
 328 #endif /* CONFIG_BOUNCE */
 329
 330 #ifdef CONFIG_BLK_CGROUP_IOLATENCY
 331 extern int blk_iolatency_init(struct request_queue *q);
 332 #else
 333 static inline int blk_iolatency_init(struct request_queue *q) { return 0; }
 334 #endif
 335
 336 struct bio *blk_next_bio(struct bio *bio, unsigned int nr_pages, gfp_t gfp);
 337
 338 #ifdef CONFIG_BLK_DEV_ZONED
 339 void blk_queue_free_zone_bitmaps(struct request_queue *q);
 340 #else
 341 static inline void blk_queue_free_zone_bitmaps(struct request_queue *q) {}
 342 #endif
 343
 344 struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector);
 345
 346 int blk_alloc_devt(struct hd_struct *part, dev_t *devt);
 347 void blk_free_devt(dev_t devt);
 348 void blk_invalidate_devt(dev_t devt);
 349 char *disk_name(struct gendisk *hd, int partno, char *buf);
 350 #define ADDPART_FLAG_NONE       0
 351 #define ADDPART_FLAG_RAID       1
 352 #define ADDPART_FLAG_WHOLEDISK  2
 353 void delete_partition(struct gendisk *disk, struct hd_struct *part);
 354 int bdev_add_partition(struct block_device *bdev, int partno,
 355                 sector_t start, sector_t length);
 356 int bdev_del_partition(struct block_device *bdev, int partno);
 357 int bdev_resize_partition(struct block_device *bdev, int partno,
 358                 sector_t start, sector_t length);
 359 int disk_expand_part_tbl(struct gendisk *disk, int target);
 360 int hd_ref_init(struct hd_struct *part);
 361
 362 /* no need to get/put refcount of part0 */
 363 static inline int hd_struct_try_get(struct hd_struct *part)
 364 {
 365         if (part->partno)
 366                 return percpu_ref_tryget_live(&part->ref);
 367         return 1;
 368 }
 369
 370 static inline void hd_struct_put(struct hd_struct *part)
 371 {
 372         if (part->partno)
 373                 percpu_ref_put(&part->ref);
 374 }
 375
 376 static inline void hd_free_part(struct hd_struct *part)
 377 {
 378         free_percpu(part->dkstats);
 379         kfree(part->info);
 380         percpu_ref_exit(&part->ref);
 381 }
 382
 383 /*
 384  * Any access of part->nr_sects which is not protected by partition
 385  * bd_mutex or gendisk bdev bd_mutex, should be done using this
 386  * accessor function.
 387  *
 388  * Code written along the lines of i_size_read() and i_size_write().
 389  * CONFIG_PREEMPTION case optimizes the case of UP kernel with preemption
 390  * on.
 391  */
 392 static inline sector_t part_nr_sects_read(struct hd_struct *part)
 393 {
 394 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
 395         sector_t nr_sects;
 396         unsigned seq;
 397         do {
 398                 seq = read_seqcount_begin(&part->nr_sects_seq);
 399                 nr_sects = part->nr_sects;
 400         } while (read_seqcount_retry(&part->nr_sects_seq, seq));
 401         return nr_sects;
 402 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
 403         sector_t nr_sects;
 404
 405         preempt_disable();
 406         nr_sects = part->nr_sects;
 407         preempt_enable();
 408         return nr_sects;
 409 #else
 410         return part->nr_sects;
 411 #endif
 412 }
 413
 414 /*
 415  * Should be called with mutex lock held (typically bd_mutex) of partition
 416  * to provide mutual exlusion among writers otherwise seqcount might be
 417  * left in wrong state leaving the readers spinning infinitely.
 418  */
 419 static inline void part_nr_sects_write(struct hd_struct *part, sector_t size)
 420 {
 421 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
 422         preempt_disable();
 423         write_seqcount_begin(&part->nr_sects_seq);
 424         part->nr_sects = size;
 425         write_seqcount_end(&part->nr_sects_seq);
 426         preempt_enable();
 427 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
 428         preempt_disable();
 429         part->nr_sects = size;
 430         preempt_enable();
 431 #else
 432         part->nr_sects = size;
 433 #endif
 434 }
 435
 436 int bio_add_hw_page(struct request_queue *q, struct bio *bio,
 437                 struct page *page, unsigned int len, unsigned int offset,
 438                 unsigned int max_sectors, bool *same_page);
 439
 440 #endif /* BLK_INTERNAL_H */