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 <xen/xen.h>
   8 #include "blk-mq.h"
   9 #include "blk-mq-sched.h"
  10
  11 /* Max future timer expiry for timeouts */
  12 #define BLK_MAX_TIMEOUT         (5 * HZ)
  13
  14 #ifdef CONFIG_DEBUG_FS
  15 extern struct dentry *blk_debugfs_root;
  16 #endif
  17
  18 struct blk_flush_queue {
  19         unsigned int            flush_queue_delayed:1;
  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         /*
  29          * flush_rq shares tag with this rq, both can't be active
  30          * at the same time
  31          */
  32         struct request          *orig_rq;
  33         spinlock_t              mq_flush_lock;
  34 };
  35
  36 extern struct kmem_cache *blk_requestq_cachep;
  37 extern struct kobj_type blk_queue_ktype;
  38 extern struct ida blk_queue_ida;
  39
  40 static inline struct blk_flush_queue *
  41 blk_get_flush_queue(struct request_queue *q, struct blk_mq_ctx *ctx)
  42 {
  43         return blk_mq_map_queue(q, REQ_OP_FLUSH, ctx)->fq;
  44 }
  45
  46 static inline void __blk_get_queue(struct request_queue *q)
  47 {
  48         kobject_get(&q->kobj);
  49 }
  50
  51 static inline bool
  52 is_flush_rq(struct request *req, struct blk_mq_hw_ctx *hctx)
  53 {
  54         return hctx->fq->flush_rq == req;
  55 }
  56
  57 struct blk_flush_queue *blk_alloc_flush_queue(struct request_queue *q,
  58                 int node, int cmd_size, gfp_t flags);
  59 void blk_free_flush_queue(struct blk_flush_queue *q);
  60
  61 void blk_freeze_queue(struct request_queue *q);
  62
  63 static inline void blk_queue_enter_live(struct request_queue *q)
  64 {
  65         /*
  66          * Given that running in generic_make_request() context
  67          * guarantees that a live reference against q_usage_counter has
  68          * been established, further references under that same context
  69          * need not check that the queue has been frozen (marked dead).
  70          */
  71         percpu_ref_get(&q->q_usage_counter);
  72 }
  73
  74 static inline bool biovec_phys_mergeable(struct request_queue *q,
  75                 struct bio_vec *vec1, struct bio_vec *vec2)
  76 {
  77         unsigned long mask = queue_segment_boundary(q);
  78         phys_addr_t addr1 = page_to_phys(vec1->bv_page) + vec1->bv_offset;
  79         phys_addr_t addr2 = page_to_phys(vec2->bv_page) + vec2->bv_offset;
  80
  81         if (addr1 + vec1->bv_len != addr2)
  82                 return false;
  83         if (xen_domain() && !xen_biovec_phys_mergeable(vec1, vec2->bv_page))
  84                 return false;
  85         if ((addr1 | mask) != ((addr2 + vec2->bv_len - 1) | mask))
  86                 return false;
  87         return true;
  88 }
  89
  90 static inline bool __bvec_gap_to_prev(struct request_queue *q,
  91                 struct bio_vec *bprv, unsigned int offset)
  92 {
  93         return (offset & queue_virt_boundary(q)) ||
  94                 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
  95 }
  96
  97 /*
  98  * Check if adding a bio_vec after bprv with offset would create a gap in
  99  * the SG list. Most drivers don't care about this, but some do.
 100  */
 101 static inline bool bvec_gap_to_prev(struct request_queue *q,
 102                 struct bio_vec *bprv, unsigned int offset)
 103 {
 104         if (!queue_virt_boundary(q))
 105                 return false;
 106         return __bvec_gap_to_prev(q, bprv, offset);
 107 }
 108
 109 static inline void blk_rq_bio_prep(struct request *rq, struct bio *bio,
 110                 unsigned int nr_segs)
 111 {
 112         rq->nr_phys_segments = nr_segs;
 113         rq->__data_len = bio->bi_iter.bi_size;
 114         rq->bio = rq->biotail = bio;
 115         rq->ioprio = bio_prio(bio);
 116
 117         if (bio->bi_disk)
 118                 rq->rq_disk = bio->bi_disk;
 119 }
 120
 121 #ifdef CONFIG_BLK_DEV_INTEGRITY
 122 void blk_flush_integrity(void);
 123 bool __bio_integrity_endio(struct bio *);
 124 void bio_integrity_free(struct bio *bio);
 125 static inline bool bio_integrity_endio(struct bio *bio)
 126 {
 127         if (bio_integrity(bio))
 128                 return __bio_integrity_endio(bio);
 129         return true;
 130 }
 131
 132 static inline bool integrity_req_gap_back_merge(struct request *req,
 133                 struct bio *next)
 134 {
 135         struct bio_integrity_payload *bip = bio_integrity(req->bio);
 136         struct bio_integrity_payload *bip_next = bio_integrity(next);
 137
 138         return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
 139                                 bip_next->bip_vec[0].bv_offset);
 140 }
 141
 142 static inline bool integrity_req_gap_front_merge(struct request *req,
 143                 struct bio *bio)
 144 {
 145         struct bio_integrity_payload *bip = bio_integrity(bio);
 146         struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
 147
 148         return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
 149                                 bip_next->bip_vec[0].bv_offset);
 150 }
 151 #else /* CONFIG_BLK_DEV_INTEGRITY */
 152 static inline bool integrity_req_gap_back_merge(struct request *req,
 153                 struct bio *next)
 154 {
 155         return false;
 156 }
 157 static inline bool integrity_req_gap_front_merge(struct request *req,
 158                 struct bio *bio)
 159 {
 160         return false;
 161 }
 162
 163 static inline void blk_flush_integrity(void)
 164 {
 165 }
 166 static inline bool bio_integrity_endio(struct bio *bio)
 167 {
 168         return true;
 169 }
 170 static inline void bio_integrity_free(struct bio *bio)
 171 {
 172 }
 173 #endif /* CONFIG_BLK_DEV_INTEGRITY */
 174
 175 unsigned long blk_rq_timeout(unsigned long timeout);
 176 void blk_add_timer(struct request *req);
 177
 178 bool bio_attempt_front_merge(struct request *req, struct bio *bio,
 179                 unsigned int nr_segs);
 180 bool bio_attempt_back_merge(struct request *req, struct bio *bio,
 181                 unsigned int nr_segs);
 182 bool bio_attempt_discard_merge(struct request_queue *q, struct request *req,
 183                 struct bio *bio);
 184 bool blk_attempt_plug_merge(struct request_queue *q, struct bio *bio,
 185                 unsigned int nr_segs, struct request **same_queue_rq);
 186
 187 void blk_account_io_start(struct request *req, bool new_io);
 188 void blk_account_io_completion(struct request *req, unsigned int bytes);
 189 void blk_account_io_done(struct request *req, u64 now);
 190
 191 /*
 192  * Internal elevator interface
 193  */
 194 #define ELV_ON_HASH(rq) ((rq)->rq_flags & RQF_HASHED)
 195
 196 void blk_insert_flush(struct request *rq);
 197
 198 void elevator_init_mq(struct request_queue *q);
 199 int elevator_switch_mq(struct request_queue *q,
 200                               struct elevator_type *new_e);
 201 void __elevator_exit(struct request_queue *, struct elevator_queue *);
 202 int elv_register_queue(struct request_queue *q, bool uevent);
 203 void elv_unregister_queue(struct request_queue *q);
 204
 205 static inline void elevator_exit(struct request_queue *q,
 206                 struct elevator_queue *e)
 207 {
 208         lockdep_assert_held(&q->sysfs_lock);
 209
 210         blk_mq_sched_free_requests(q);
 211         __elevator_exit(q, e);
 212 }
 213
 214 struct hd_struct *__disk_get_part(struct gendisk *disk, int partno);
 215
 216 #ifdef CONFIG_FAIL_IO_TIMEOUT
 217 int blk_should_fake_timeout(struct request_queue *);
 218 ssize_t part_timeout_show(struct device *, struct device_attribute *, char *);
 219 ssize_t part_timeout_store(struct device *, struct device_attribute *,
 220                                 const char *, size_t);
 221 #else
 222 static inline int blk_should_fake_timeout(struct request_queue *q)
 223 {
 224         return 0;
 225 }
 226 #endif
 227
 228 void __blk_queue_split(struct request_queue *q, struct bio **bio,
 229                 unsigned int *nr_segs);
 230 int ll_back_merge_fn(struct request *req, struct bio *bio,
 231                 unsigned int nr_segs);
 232 int ll_front_merge_fn(struct request *req,  struct bio *bio,
 233                 unsigned int nr_segs);
 234 struct request *attempt_back_merge(struct request_queue *q, struct request *rq);
 235 struct request *attempt_front_merge(struct request_queue *q, struct request *rq);
 236 int blk_attempt_req_merge(struct request_queue *q, struct request *rq,
 237                                 struct request *next);
 238 unsigned int blk_recalc_rq_segments(struct request *rq);
 239 void blk_rq_set_mixed_merge(struct request *rq);
 240 bool blk_rq_merge_ok(struct request *rq, struct bio *bio);
 241 enum elv_merge blk_try_merge(struct request *rq, struct bio *bio);
 242
 243 int blk_dev_init(void);
 244
 245 /*
 246  * Contribute to IO statistics IFF:
 247  *
 248  *      a) it's attached to a gendisk, and
 249  *      b) the queue had IO stats enabled when this request was started
 250  */
 251 static inline bool blk_do_io_stat(struct request *rq)
 252 {
 253         return rq->rq_disk && (rq->rq_flags & RQF_IO_STAT);
 254 }
 255
 256 static inline void req_set_nomerge(struct request_queue *q, struct request *req)
 257 {
 258         req->cmd_flags |= REQ_NOMERGE;
 259         if (req == q->last_merge)
 260                 q->last_merge = NULL;
 261 }
 262
 263 /*
 264  * The max size one bio can handle is UINT_MAX becasue bvec_iter.bi_size
 265  * is defined as 'unsigned int', meantime it has to aligned to with logical
 266  * block size which is the minimum accepted unit by hardware.
 267  */
 268 static inline unsigned int bio_allowed_max_sectors(struct request_queue *q)
 269 {
 270         return round_down(UINT_MAX, queue_logical_block_size(q)) >> 9;
 271 }
 272
 273 /*
 274  * Internal io_context interface
 275  */
 276 void get_io_context(struct io_context *ioc);
 277 struct io_cq *ioc_lookup_icq(struct io_context *ioc, struct request_queue *q);
 278 struct io_cq *ioc_create_icq(struct io_context *ioc, struct request_queue *q,
 279                              gfp_t gfp_mask);
 280 void ioc_clear_queue(struct request_queue *q);
 281
 282 int create_task_io_context(struct task_struct *task, gfp_t gfp_mask, int node);
 283
 284 /**
 285  * create_io_context - try to create task->io_context
 286  * @gfp_mask: allocation mask
 287  * @node: allocation node
 288  *
 289  * If %current->io_context is %NULL, allocate a new io_context and install
 290  * it.  Returns the current %current->io_context which may be %NULL if
 291  * allocation failed.
 292  *
 293  * Note that this function can't be called with IRQ disabled because
 294  * task_lock which protects %current->io_context is IRQ-unsafe.
 295  */
 296 static inline struct io_context *create_io_context(gfp_t gfp_mask, int node)
 297 {
 298         WARN_ON_ONCE(irqs_disabled());
 299         if (unlikely(!current->io_context))
 300                 create_task_io_context(current, gfp_mask, node);
 301         return current->io_context;
 302 }
 303
 304 /*
 305  * Internal throttling interface
 306  */
 307 #ifdef CONFIG_BLK_DEV_THROTTLING
 308 extern void blk_throtl_drain(struct request_queue *q);
 309 extern int blk_throtl_init(struct request_queue *q);
 310 extern void blk_throtl_exit(struct request_queue *q);
 311 extern void blk_throtl_register_queue(struct request_queue *q);
 312 #else /* CONFIG_BLK_DEV_THROTTLING */
 313 static inline void blk_throtl_drain(struct request_queue *q) { }
 314 static inline int blk_throtl_init(struct request_queue *q) { return 0; }
 315 static inline void blk_throtl_exit(struct request_queue *q) { }
 316 static inline void blk_throtl_register_queue(struct request_queue *q) { }
 317 #endif /* CONFIG_BLK_DEV_THROTTLING */
 318 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW
 319 extern ssize_t blk_throtl_sample_time_show(struct request_queue *q, char *page);
 320 extern ssize_t blk_throtl_sample_time_store(struct request_queue *q,
 321         const char *page, size_t count);
 322 extern void blk_throtl_bio_endio(struct bio *bio);
 323 extern void blk_throtl_stat_add(struct request *rq, u64 time);
 324 #else
 325 static inline void blk_throtl_bio_endio(struct bio *bio) { }
 326 static inline void blk_throtl_stat_add(struct request *rq, u64 time) { }
 327 #endif
 328
 329 #ifdef CONFIG_BOUNCE
 330 extern int init_emergency_isa_pool(void);
 331 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
 332 #else
 333 static inline int init_emergency_isa_pool(void)
 334 {
 335         return 0;
 336 }
 337 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
 338 {
 339 }
 340 #endif /* CONFIG_BOUNCE */
 341
 342 #ifdef CONFIG_BLK_CGROUP_IOLATENCY
 343 extern int blk_iolatency_init(struct request_queue *q);
 344 #else
 345 static inline int blk_iolatency_init(struct request_queue *q) { return 0; }
 346 #endif
 347
 348 struct bio *blk_next_bio(struct bio *bio, unsigned int nr_pages, gfp_t gfp);
 349
 350 #ifdef CONFIG_BLK_DEV_ZONED
 351 void blk_queue_free_zone_bitmaps(struct request_queue *q);
 352 #else
 353 static inline void blk_queue_free_zone_bitmaps(struct request_queue *q) {}
 354 #endif
 355
 356 #endif /* BLK_INTERNAL_H */