include/linux/blkdev.h

   1 #ifndef _LINUX_BLKDEV_H
   2 #define _LINUX_BLKDEV_H
   3
   4 #include <linux/sched.h>
   5
   6 #ifdef CONFIG_BLOCK
   7
   8 #include <linux/major.h>
   9 #include <linux/genhd.h>
  10 #include <linux/list.h>
  11 #include <linux/llist.h>
  12 #include <linux/timer.h>
  13 #include <linux/workqueue.h>
  14 #include <linux/pagemap.h>
  15 #include <linux/backing-dev.h>
  16 #include <linux/wait.h>
  17 #include <linux/mempool.h>
  18 #include <linux/bio.h>
  19 #include <linux/stringify.h>
  20 #include <linux/gfp.h>
  21 #include <linux/bsg.h>
  22 #include <linux/smp.h>
  23 #include <linux/rcupdate.h>
  24
  25 #include <asm/scatterlist.h>
  26
  27 struct module;
  28 struct scsi_ioctl_command;
  29
  30 struct request_queue;
  31 struct elevator_queue;
  32 struct request_pm_state;
  33 struct blk_trace;
  34 struct request;
  35 struct sg_io_hdr;
  36 struct bsg_job;
  37 struct blkcg_gq;
  38
  39 #define BLKDEV_MIN_RQ   4
  40 #define BLKDEV_MAX_RQ   128     /* Default maximum */
  41
  42 /*
  43  * Maximum number of blkcg policies allowed to be registered concurrently.
  44  * Defined here to simplify include dependency.
  45  */
  46 #define BLKCG_MAX_POLS          2
  47
  48 struct request;
  49 typedef void (rq_end_io_fn)(struct request *, int);
  50
  51 #define BLK_RL_SYNCFULL         (1U << 0)
  52 #define BLK_RL_ASYNCFULL        (1U << 1)
  53
  54 struct request_list {
  55         struct request_queue    *q;     /* the queue this rl belongs to */
  56 #ifdef CONFIG_BLK_CGROUP
  57         struct blkcg_gq         *blkg;  /* blkg this request pool belongs to */
  58 #endif
  59         /*
  60          * count[], starved[], and wait[] are indexed by
  61          * BLK_RW_SYNC/BLK_RW_ASYNC
  62          */
  63         int                     count[2];
  64         int                     starved[2];
  65         mempool_t               *rq_pool;
  66         wait_queue_head_t       wait[2];
  67         unsigned int            flags;
  68 };
  69
  70 /*
  71  * request command types
  72  */
  73 enum rq_cmd_type_bits {
  74         REQ_TYPE_FS             = 1,    /* fs request */
  75         REQ_TYPE_BLOCK_PC,              /* scsi command */
  76         REQ_TYPE_SENSE,                 /* sense request */
  77         REQ_TYPE_PM_SUSPEND,            /* suspend request */
  78         REQ_TYPE_PM_RESUME,             /* resume request */
  79         REQ_TYPE_PM_SHUTDOWN,           /* shutdown request */
  80         REQ_TYPE_SPECIAL,               /* driver defined type */
  81         /*
  82          * for ATA/ATAPI devices. this really doesn't belong here, ide should
  83          * use REQ_TYPE_SPECIAL and use rq->cmd[0] with the range of driver
  84          * private REQ_LB opcodes to differentiate what type of request this is
  85          */
  86         REQ_TYPE_ATA_TASKFILE,
  87         REQ_TYPE_ATA_PC,
  88 };
  89
  90 #define BLK_MAX_CDB     16
  91
  92 /*
  93  * try to put the fields that are referenced together in the same cacheline.
  94  * if you modify this structure, be sure to check block/blk-core.c:blk_rq_init()
  95  * as well!
  96  */
  97 struct request {
  98         union {
  99                 struct list_head queuelist;
 100                 struct llist_node ll_list;
 101         };
 102         union {
 103                 struct call_single_data csd;
 104                 struct work_struct mq_flush_data;
 105         };
 106
 107         struct request_queue *q;
 108         struct blk_mq_ctx *mq_ctx;
 109
 110         u64 cmd_flags;
 111         enum rq_cmd_type_bits cmd_type;
 112         unsigned long atomic_flags;
 113
 114         int cpu;
 115
 116         /* the following two fields are internal, NEVER access directly */
 117         unsigned int __data_len;        /* total data len */
 118         sector_t __sector;              /* sector cursor */
 119
 120         struct bio *bio;
 121         struct bio *biotail;
 122
 123         struct hlist_node hash; /* merge hash */
 124         /*
 125          * The rb_node is only used inside the io scheduler, requests
 126          * are pruned when moved to the dispatch queue. So let the
 127          * completion_data share space with the rb_node.
 128          */
 129         union {
 130                 struct rb_node rb_node; /* sort/lookup */
 131                 void *completion_data;
 132         };
 133
 134         /*
 135          * Three pointers are available for the IO schedulers, if they need
 136          * more they have to dynamically allocate it.  Flush requests are
 137          * never put on the IO scheduler. So let the flush fields share
 138          * space with the elevator data.
 139          */
 140         union {
 141                 struct {
 142                         struct io_cq            *icq;
 143                         void                    *priv[2];
 144                 } elv;
 145
 146                 struct {
 147                         unsigned int            seq;
 148                         struct list_head        list;
 149                         rq_end_io_fn            *saved_end_io;
 150                 } flush;
 151         };
 152
 153         struct gendisk *rq_disk;
 154         struct hd_struct *part;
 155         unsigned long start_time;
 156 #ifdef CONFIG_BLK_CGROUP
 157         struct request_list *rl;                /* rl this rq is alloced from */
 158         unsigned long long start_time_ns;
 159         unsigned long long io_start_time_ns;    /* when passed to hardware */
 160 #endif
 161         /* Number of scatter-gather DMA addr+len pairs after
 162          * physical address coalescing is performed.
 163          */
 164         unsigned short nr_phys_segments;
 165 #if defined(CONFIG_BLK_DEV_INTEGRITY)
 166         unsigned short nr_integrity_segments;
 167 #endif
 168
 169         unsigned short ioprio;
 170
 171         void *special;          /* opaque pointer available for LLD use */
 172         char *buffer;           /* kaddr of the current segment if available */
 173
 174         int tag;
 175         int errors;
 176
 177         /*
 178          * when request is used as a packet command carrier
 179          */
 180         unsigned char __cmd[BLK_MAX_CDB];
 181         unsigned char *cmd;
 182         unsigned short cmd_len;
 183
 184         unsigned int extra_len; /* length of alignment and padding */
 185         unsigned int sense_len;
 186         unsigned int resid_len; /* residual count */
 187         void *sense;
 188
 189         unsigned long deadline;
 190         struct list_head timeout_list;
 191         unsigned int timeout;
 192         int retries;
 193
 194         /*
 195          * completion callback.
 196          */
 197         rq_end_io_fn *end_io;
 198         void *end_io_data;
 199
 200         /* for bidi */
 201         struct request *next_rq;
 202 };
 203
 204 static inline unsigned short req_get_ioprio(struct request *req)
 205 {
 206         return req->ioprio;
 207 }
 208
 209 /*
 210  * State information carried for REQ_TYPE_PM_SUSPEND and REQ_TYPE_PM_RESUME
 211  * requests. Some step values could eventually be made generic.
 212  */
 213 struct request_pm_state
 214 {
 215         /* PM state machine step value, currently driver specific */
 216         int     pm_step;
 217         /* requested PM state value (S1, S2, S3, S4, ...) */
 218         u32     pm_state;
 219         void*   data;           /* for driver use */
 220 };
 221
 222 #include <linux/elevator.h>
 223
 224 struct blk_queue_ctx;
 225
 226 typedef void (request_fn_proc) (struct request_queue *q);
 227 typedef void (make_request_fn) (struct request_queue *q, struct bio *bio);
 228 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
 229 typedef void (unprep_rq_fn) (struct request_queue *, struct request *);
 230
 231 struct bio_vec;
 232 struct bvec_merge_data {
 233         struct block_device *bi_bdev;
 234         sector_t bi_sector;
 235         unsigned bi_size;
 236         unsigned long bi_rw;
 237 };
 238 typedef int (merge_bvec_fn) (struct request_queue *, struct bvec_merge_data *,
 239                              struct bio_vec *);
 240 typedef void (softirq_done_fn)(struct request *);
 241 typedef int (dma_drain_needed_fn)(struct request *);
 242 typedef int (lld_busy_fn) (struct request_queue *q);
 243 typedef int (bsg_job_fn) (struct bsg_job *);
 244
 245 enum blk_eh_timer_return {
 246         BLK_EH_NOT_HANDLED,
 247         BLK_EH_HANDLED,
 248         BLK_EH_RESET_TIMER,
 249 };
 250
 251 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
 252
 253 enum blk_queue_state {
 254         Queue_down,
 255         Queue_up,
 256 };
 257
 258 struct blk_queue_tag {
 259         struct request **tag_index;     /* map of busy tags */
 260         unsigned long *tag_map;         /* bit map of free/busy tags */
 261         int busy;                       /* current depth */
 262         int max_depth;                  /* what we will send to device */
 263         int real_max_depth;             /* what the array can hold */
 264         atomic_t refcnt;                /* map can be shared */
 265 };
 266
 267 #define BLK_SCSI_MAX_CMDS       (256)
 268 #define BLK_SCSI_CMD_PER_LONG   (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
 269
 270 struct queue_limits {
 271         unsigned long           bounce_pfn;
 272         unsigned long           seg_boundary_mask;
 273
 274         unsigned int            max_hw_sectors;
 275         unsigned int            max_sectors;
 276         unsigned int            max_segment_size;
 277         unsigned int            physical_block_size;
 278         unsigned int            alignment_offset;
 279         unsigned int            io_min;
 280         unsigned int            io_opt;
 281         unsigned int            max_discard_sectors;
 282         unsigned int            max_write_same_sectors;
 283         unsigned int            discard_granularity;
 284         unsigned int            discard_alignment;
 285
 286         unsigned short          logical_block_size;
 287         unsigned short          max_segments;
 288         unsigned short          max_integrity_segments;
 289
 290         unsigned char           misaligned;
 291         unsigned char           discard_misaligned;
 292         unsigned char           cluster;
 293         unsigned char           discard_zeroes_data;
 294 };
 295
 296 struct request_queue {
 297         /*
 298          * Together with queue_head for cacheline sharing
 299          */
 300         struct list_head        queue_head;
 301         struct request          *last_merge;
 302         struct elevator_queue   *elevator;
 303         int                     nr_rqs[2];      /* # allocated [a]sync rqs */
 304         int                     nr_rqs_elvpriv; /* # allocated rqs w/ elvpriv */
 305
 306         /*
 307          * If blkcg is not used, @q->root_rl serves all requests.  If blkcg
 308          * is used, root blkg allocates from @q->root_rl and all other
 309          * blkgs from their own blkg->rl.  Which one to use should be
 310          * determined using bio_request_list().
 311          */
 312         struct request_list     root_rl;
 313
 314         request_fn_proc         *request_fn;
 315         make_request_fn         *make_request_fn;
 316         prep_rq_fn              *prep_rq_fn;
 317         unprep_rq_fn            *unprep_rq_fn;
 318         merge_bvec_fn           *merge_bvec_fn;
 319         softirq_done_fn         *softirq_done_fn;
 320         rq_timed_out_fn         *rq_timed_out_fn;
 321         dma_drain_needed_fn     *dma_drain_needed;
 322         lld_busy_fn             *lld_busy_fn;
 323
 324         struct blk_mq_ops       *mq_ops;
 325
 326         unsigned int            *mq_map;
 327
 328         /* sw queues */
 329         struct blk_mq_ctx       *queue_ctx;
 330         unsigned int            nr_queues;
 331
 332         /* hw dispatch queues */
 333         struct blk_mq_hw_ctx    **queue_hw_ctx;
 334         unsigned int            nr_hw_queues;
 335
 336         /*
 337          * Dispatch queue sorting
 338          */
 339         sector_t                end_sector;
 340         struct request          *boundary_rq;
 341
 342         /*
 343          * Delayed queue handling
 344          */
 345         struct delayed_work     delay_work;
 346
 347         struct backing_dev_info backing_dev_info;
 348
 349         /*
 350          * The queue owner gets to use this for whatever they like.
 351          * ll_rw_blk doesn't touch it.
 352          */
 353         void                    *queuedata;
 354
 355         /*
 356          * various queue flags, see QUEUE_* below
 357          */
 358         unsigned long           queue_flags;
 359
 360         /*
 361          * ida allocated id for this queue.  Used to index queues from
 362          * ioctx.
 363          */
 364         int                     id;
 365
 366         /*
 367          * queue needs bounce pages for pages above this limit
 368          */
 369         gfp_t                   bounce_gfp;
 370
 371         /*
 372          * protects queue structures from reentrancy. ->__queue_lock should
 373          * _never_ be used directly, it is queue private. always use
 374          * ->queue_lock.
 375          */
 376         spinlock_t              __queue_lock;
 377         spinlock_t              *queue_lock;
 378
 379         /*
 380          * queue kobject
 381          */
 382         struct kobject kobj;
 383
 384         /*
 385          * mq queue kobject
 386          */
 387         struct kobject mq_kobj;
 388
 389 #ifdef CONFIG_PM_RUNTIME
 390         struct device           *dev;
 391         int                     rpm_status;
 392         unsigned int            nr_pending;
 393 #endif
 394
 395         /*
 396          * queue settings
 397          */
 398         unsigned long           nr_requests;    /* Max # of requests */
 399         unsigned int            nr_congestion_on;
 400         unsigned int            nr_congestion_off;
 401         unsigned int            nr_batching;
 402
 403         unsigned int            dma_drain_size;
 404         void                    *dma_drain_buffer;
 405         unsigned int            dma_pad_mask;
 406         unsigned int            dma_alignment;
 407
 408         struct blk_queue_tag    *queue_tags;
 409         struct list_head        tag_busy_list;
 410
 411         unsigned int            nr_sorted;
 412         unsigned int            in_flight[2];
 413         /*
 414          * Number of active block driver functions for which blk_drain_queue()
 415          * must wait. Must be incremented around functions that unlock the
 416          * queue_lock internally, e.g. scsi_request_fn().
 417          */
 418         unsigned int            request_fn_active;
 419
 420         unsigned int            rq_timeout;
 421         struct timer_list       timeout;
 422         struct list_head        timeout_list;
 423
 424         struct list_head        icq_list;
 425 #ifdef CONFIG_BLK_CGROUP
 426         DECLARE_BITMAP          (blkcg_pols, BLKCG_MAX_POLS);
 427         struct blkcg_gq         *root_blkg;
 428         struct list_head        blkg_list;
 429 #endif
 430
 431         struct queue_limits     limits;
 432
 433         /*
 434          * sg stuff
 435          */
 436         unsigned int            sg_timeout;
 437         unsigned int            sg_reserved_size;
 438         int                     node;
 439 #ifdef CONFIG_BLK_DEV_IO_TRACE
 440         struct blk_trace        *blk_trace;
 441 #endif
 442         /*
 443          * for flush operations
 444          */
 445         unsigned int            flush_flags;
 446         unsigned int            flush_not_queueable:1;
 447         unsigned int            flush_queue_delayed:1;
 448         unsigned int            flush_pending_idx:1;
 449         unsigned int            flush_running_idx:1;
 450         unsigned long           flush_pending_since;
 451         struct list_head        flush_queue[2];
 452         struct list_head        flush_data_in_flight;
 453         union {
 454                 struct request  flush_rq;
 455                 struct {
 456                         spinlock_t mq_flush_lock;
 457                         struct work_struct mq_flush_work;
 458                 };
 459         };
 460
 461         struct mutex            sysfs_lock;
 462
 463         int                     bypass_depth;
 464
 465 #if defined(CONFIG_BLK_DEV_BSG)
 466         bsg_job_fn              *bsg_job_fn;
 467         int                     bsg_job_size;
 468         struct bsg_class_device bsg_dev;
 469 #endif
 470
 471 #ifdef CONFIG_BLK_DEV_THROTTLING
 472         /* Throttle data */
 473         struct throtl_data *td;
 474 #endif
 475         struct rcu_head         rcu_head;
 476         wait_queue_head_t       mq_freeze_wq;
 477         struct percpu_counter   mq_usage_counter;
 478         struct list_head        all_q_node;
 479 };
 480
 481 #define QUEUE_FLAG_QUEUED       1       /* uses generic tag queueing */
 482 #define QUEUE_FLAG_STOPPED      2       /* queue is stopped */
 483 #define QUEUE_FLAG_SYNCFULL     3       /* read queue has been filled */
 484 #define QUEUE_FLAG_ASYNCFULL    4       /* write queue has been filled */
 485 #define QUEUE_FLAG_DYING        5       /* queue being torn down */
 486 #define QUEUE_FLAG_BYPASS       6       /* act as dumb FIFO queue */
 487 #define QUEUE_FLAG_BIDI         7       /* queue supports bidi requests */
 488 #define QUEUE_FLAG_NOMERGES     8       /* disable merge attempts */
 489 #define QUEUE_FLAG_SAME_COMP    9       /* complete on same CPU-group */
 490 #define QUEUE_FLAG_FAIL_IO     10       /* fake timeout */
 491 #define QUEUE_FLAG_STACKABLE   11       /* supports request stacking */
 492 #define QUEUE_FLAG_NONROT      12       /* non-rotational device (SSD) */
 493 #define QUEUE_FLAG_VIRT        QUEUE_FLAG_NONROT /* paravirt device */
 494 #define QUEUE_FLAG_IO_STAT     13       /* do IO stats */
 495 #define QUEUE_FLAG_DISCARD     14       /* supports DISCARD */
 496 #define QUEUE_FLAG_NOXMERGES   15       /* No extended merges */
 497 #define QUEUE_FLAG_ADD_RANDOM  16       /* Contributes to random pool */
 498 #define QUEUE_FLAG_SECDISCARD  17       /* supports SECDISCARD */
 499 #define QUEUE_FLAG_SAME_FORCE  18       /* force complete on same CPU */
 500 #define QUEUE_FLAG_DEAD        19       /* queue tear-down finished */
 501 #define QUEUE_FLAG_INIT_DONE   20       /* queue is initialized */
 502
 503 #define QUEUE_FLAG_DEFAULT      ((1 << QUEUE_FLAG_IO_STAT) |            \
 504                                  (1 << QUEUE_FLAG_STACKABLE)    |       \
 505                                  (1 << QUEUE_FLAG_SAME_COMP)    |       \
 506                                  (1 << QUEUE_FLAG_ADD_RANDOM))
 507
 508 #define QUEUE_FLAG_MQ_DEFAULT   ((1 << QUEUE_FLAG_IO_STAT) |            \
 509                                  (1 << QUEUE_FLAG_SAME_COMP))
 510
 511 static inline void queue_lockdep_assert_held(struct request_queue *q)
 512 {
 513         if (q->queue_lock)
 514                 lockdep_assert_held(q->queue_lock);
 515 }
 516
 517 static inline void queue_flag_set_unlocked(unsigned int flag,
 518                                            struct request_queue *q)
 519 {
 520         __set_bit(flag, &q->queue_flags);
 521 }
 522
 523 static inline int queue_flag_test_and_clear(unsigned int flag,
 524                                             struct request_queue *q)
 525 {
 526         queue_lockdep_assert_held(q);
 527
 528         if (test_bit(flag, &q->queue_flags)) {
 529                 __clear_bit(flag, &q->queue_flags);
 530                 return 1;
 531         }
 532
 533         return 0;
 534 }
 535
 536 static inline int queue_flag_test_and_set(unsigned int flag,
 537                                           struct request_queue *q)
 538 {
 539         queue_lockdep_assert_held(q);
 540
 541         if (!test_bit(flag, &q->queue_flags)) {
 542                 __set_bit(flag, &q->queue_flags);
 543                 return 0;
 544         }
 545
 546         return 1;
 547 }
 548
 549 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
 550 {
 551         queue_lockdep_assert_held(q);
 552         __set_bit(flag, &q->queue_flags);
 553 }
 554
 555 static inline void queue_flag_clear_unlocked(unsigned int flag,
 556                                              struct request_queue *q)
 557 {
 558         __clear_bit(flag, &q->queue_flags);
 559 }
 560
 561 static inline int queue_in_flight(struct request_queue *q)
 562 {
 563         return q->in_flight[0] + q->in_flight[1];
 564 }
 565
 566 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
 567 {
 568         queue_lockdep_assert_held(q);
 569         __clear_bit(flag, &q->queue_flags);
 570 }
 571
 572 #define blk_queue_tagged(q)     test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
 573 #define blk_queue_stopped(q)    test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
 574 #define blk_queue_dying(q)      test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
 575 #define blk_queue_dead(q)       test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
 576 #define blk_queue_bypass(q)     test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
 577 #define blk_queue_init_done(q)  test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
 578 #define blk_queue_nomerges(q)   test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
 579 #define blk_queue_noxmerges(q)  \
 580         test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
 581 #define blk_queue_nonrot(q)     test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
 582 #define blk_queue_io_stat(q)    test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
 583 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
 584 #define blk_queue_stackable(q)  \
 585         test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
 586 #define blk_queue_discard(q)    test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
 587 #define blk_queue_secdiscard(q) (blk_queue_discard(q) && \
 588         test_bit(QUEUE_FLAG_SECDISCARD, &(q)->queue_flags))
 589
 590 #define blk_noretry_request(rq) \
 591         ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
 592                              REQ_FAILFAST_DRIVER))
 593
 594 #define blk_account_rq(rq) \
 595         (((rq)->cmd_flags & REQ_STARTED) && \
 596          ((rq)->cmd_type == REQ_TYPE_FS))
 597
 598 #define blk_pm_request(rq)      \
 599         ((rq)->cmd_type == REQ_TYPE_PM_SUSPEND || \
 600          (rq)->cmd_type == REQ_TYPE_PM_RESUME)
 601
 602 #define blk_rq_cpu_valid(rq)    ((rq)->cpu != -1)
 603 #define blk_bidi_rq(rq)         ((rq)->next_rq != NULL)
 604 /* rq->queuelist of dequeued request must be list_empty() */
 605 #define blk_queued_rq(rq)       (!list_empty(&(rq)->queuelist))
 606
 607 #define list_entry_rq(ptr)      list_entry((ptr), struct request, queuelist)
 608
 609 #define rq_data_dir(rq)         (((rq)->cmd_flags & 1) != 0)
 610
 611 static inline unsigned int blk_queue_cluster(struct request_queue *q)
 612 {
 613         return q->limits.cluster;
 614 }
 615
 616 /*
 617  * We regard a request as sync, if either a read or a sync write
 618  */
 619 static inline bool rw_is_sync(unsigned int rw_flags)
 620 {
 621         return !(rw_flags & REQ_WRITE) || (rw_flags & REQ_SYNC);
 622 }
 623
 624 static inline bool rq_is_sync(struct request *rq)
 625 {
 626         return rw_is_sync(rq->cmd_flags);
 627 }
 628
 629 static inline bool blk_rl_full(struct request_list *rl, bool sync)
 630 {
 631         unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
 632
 633         return rl->flags & flag;
 634 }
 635
 636 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
 637 {
 638         unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
 639
 640         rl->flags |= flag;
 641 }
 642
 643 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
 644 {
 645         unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
 646
 647         rl->flags &= ~flag;
 648 }
 649
 650 static inline bool rq_mergeable(struct request *rq)
 651 {
 652         if (rq->cmd_type != REQ_TYPE_FS)
 653                 return false;
 654
 655         if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
 656                 return false;
 657
 658         return true;
 659 }
 660
 661 static inline bool blk_check_merge_flags(unsigned int flags1,
 662                                          unsigned int flags2)
 663 {
 664         if ((flags1 & REQ_DISCARD) != (flags2 & REQ_DISCARD))
 665                 return false;
 666
 667         if ((flags1 & REQ_SECURE) != (flags2 & REQ_SECURE))
 668                 return false;
 669
 670         if ((flags1 & REQ_WRITE_SAME) != (flags2 & REQ_WRITE_SAME))
 671                 return false;
 672
 673         return true;
 674 }
 675
 676 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
 677 {
 678         if (bio_data(a) == bio_data(b))
 679                 return true;
 680
 681         return false;
 682 }
 683
 684 /*
 685  * q->prep_rq_fn return values
 686  */
 687 #define BLKPREP_OK              0       /* serve it */
 688 #define BLKPREP_KILL            1       /* fatal error, kill */
 689 #define BLKPREP_DEFER           2       /* leave on queue */
 690
 691 extern unsigned long blk_max_low_pfn, blk_max_pfn;
 692
 693 /*
 694  * standard bounce addresses:
 695  *
 696  * BLK_BOUNCE_HIGH      : bounce all highmem pages
 697  * BLK_BOUNCE_ANY       : don't bounce anything
 698  * BLK_BOUNCE_ISA       : bounce pages above ISA DMA boundary
 699  */
 700
 701 #if BITS_PER_LONG == 32
 702 #define BLK_BOUNCE_HIGH         ((u64)blk_max_low_pfn << PAGE_SHIFT)
 703 #else
 704 #define BLK_BOUNCE_HIGH         -1ULL
 705 #endif
 706 #define BLK_BOUNCE_ANY          (-1ULL)
 707 #define BLK_BOUNCE_ISA          (DMA_BIT_MASK(24))
 708
 709 /*
 710  * default timeout for SG_IO if none specified
 711  */
 712 #define BLK_DEFAULT_SG_TIMEOUT  (60 * HZ)
 713 #define BLK_MIN_SG_TIMEOUT      (7 * HZ)
 714
 715 #ifdef CONFIG_BOUNCE
 716 extern int init_emergency_isa_pool(void);
 717 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
 718 #else
 719 static inline int init_emergency_isa_pool(void)
 720 {
 721         return 0;
 722 }
 723 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
 724 {
 725 }
 726 #endif /* CONFIG_MMU */
 727
 728 struct rq_map_data {
 729         struct page **pages;
 730         int page_order;
 731         int nr_entries;
 732         unsigned long offset;
 733         int null_mapped;
 734         int from_user;
 735 };
 736
 737 struct req_iterator {
 738         int i;
 739         struct bio *bio;
 740 };
 741
 742 /* This should not be used directly - use rq_for_each_segment */
 743 #define for_each_bio(_bio)              \
 744         for (; _bio; _bio = _bio->bi_next)
 745 #define __rq_for_each_bio(_bio, rq)     \
 746         if ((rq->bio))                  \
 747                 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
 748
 749 #define rq_for_each_segment(bvl, _rq, _iter)                    \
 750         __rq_for_each_bio(_iter.bio, _rq)                       \
 751                 bio_for_each_segment(bvl, _iter.bio, _iter.i)
 752
 753 #define rq_iter_last(rq, _iter)                                 \
 754                 (_iter.bio->bi_next == NULL && _iter.i == _iter.bio->bi_vcnt-1)
 755
 756 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
 757 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
 758 #endif
 759 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
 760 extern void rq_flush_dcache_pages(struct request *rq);
 761 #else
 762 static inline void rq_flush_dcache_pages(struct request *rq)
 763 {
 764 }
 765 #endif
 766
 767 extern int blk_register_queue(struct gendisk *disk);
 768 extern void blk_unregister_queue(struct gendisk *disk);
 769 extern void generic_make_request(struct bio *bio);
 770 extern void blk_rq_init(struct request_queue *q, struct request *rq);
 771 extern void blk_put_request(struct request *);
 772 extern void __blk_put_request(struct request_queue *, struct request *);
 773 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
 774 extern struct request *blk_make_request(struct request_queue *, struct bio *,
 775                                         gfp_t);
 776 extern void blk_requeue_request(struct request_queue *, struct request *);
 777 extern void blk_add_request_payload(struct request *rq, struct page *page,
 778                 unsigned int len);
 779 extern int blk_rq_check_limits(struct request_queue *q, struct request *rq);
 780 extern int blk_lld_busy(struct request_queue *q);
 781 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
 782                              struct bio_set *bs, gfp_t gfp_mask,
 783                              int (*bio_ctr)(struct bio *, struct bio *, void *),
 784                              void *data);
 785 extern void blk_rq_unprep_clone(struct request *rq);
 786 extern int blk_insert_cloned_request(struct request_queue *q,
 787                                      struct request *rq);
 788 extern void blk_delay_queue(struct request_queue *, unsigned long);
 789 extern void blk_recount_segments(struct request_queue *, struct bio *);
 790 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
 791 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
 792                               unsigned int, void __user *);
 793 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
 794                           unsigned int, void __user *);
 795 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
 796                          struct scsi_ioctl_command __user *);
 797
 798 extern void blk_queue_bio(struct request_queue *q, struct bio *bio);
 799
 800 /*
 801  * A queue has just exitted congestion.  Note this in the global counter of
 802  * congested queues, and wake up anyone who was waiting for requests to be
 803  * put back.
 804  */
 805 static inline void blk_clear_queue_congested(struct request_queue *q, int sync)
 806 {
 807         clear_bdi_congested(&q->backing_dev_info, sync);
 808 }
 809
 810 /*
 811  * A queue has just entered congestion.  Flag that in the queue's VM-visible
 812  * state flags and increment the global gounter of congested queues.
 813  */
 814 static inline void blk_set_queue_congested(struct request_queue *q, int sync)
 815 {
 816         set_bdi_congested(&q->backing_dev_info, sync);
 817 }
 818
 819 extern void blk_start_queue(struct request_queue *q);
 820 extern void blk_stop_queue(struct request_queue *q);
 821 extern void blk_sync_queue(struct request_queue *q);
 822 extern void __blk_stop_queue(struct request_queue *q);
 823 extern void __blk_run_queue(struct request_queue *q);
 824 extern void blk_run_queue(struct request_queue *);
 825 extern void blk_run_queue_async(struct request_queue *q);
 826 extern int blk_rq_map_user(struct request_queue *, struct request *,
 827                            struct rq_map_data *, void __user *, unsigned long,
 828                            gfp_t);
 829 extern int blk_rq_unmap_user(struct bio *);
 830 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
 831 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
 832                                struct rq_map_data *, struct sg_iovec *, int,
 833                                unsigned int, gfp_t);
 834 extern int blk_execute_rq(struct request_queue *, struct gendisk *,
 835                           struct request *, int);
 836 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
 837                                   struct request *, int, rq_end_io_fn *);
 838
 839 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
 840 {
 841         return bdev->bd_disk->queue;
 842 }
 843
 844 /*
 845  * blk_rq_pos()                 : the current sector
 846  * blk_rq_bytes()               : bytes left in the entire request
 847  * blk_rq_cur_bytes()           : bytes left in the current segment
 848  * blk_rq_err_bytes()           : bytes left till the next error boundary
 849  * blk_rq_sectors()             : sectors left in the entire request
 850  * blk_rq_cur_sectors()         : sectors left in the current segment
 851  */
 852 static inline sector_t blk_rq_pos(const struct request *rq)
 853 {
 854         return rq->__sector;
 855 }
 856
 857 static inline unsigned int blk_rq_bytes(const struct request *rq)
 858 {
 859         return rq->__data_len;
 860 }
 861
 862 static inline int blk_rq_cur_bytes(const struct request *rq)
 863 {
 864         return rq->bio ? bio_cur_bytes(rq->bio) : 0;
 865 }
 866
 867 extern unsigned int blk_rq_err_bytes(const struct request *rq);
 868
 869 static inline unsigned int blk_rq_sectors(const struct request *rq)
 870 {
 871         return blk_rq_bytes(rq) >> 9;
 872 }
 873
 874 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
 875 {
 876         return blk_rq_cur_bytes(rq) >> 9;
 877 }
 878
 879 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
 880                                                      unsigned int cmd_flags)
 881 {
 882         if (unlikely(cmd_flags & REQ_DISCARD))
 883                 return min(q->limits.max_discard_sectors, UINT_MAX >> 9);
 884
 885         if (unlikely(cmd_flags & REQ_WRITE_SAME))
 886                 return q->limits.max_write_same_sectors;
 887
 888         return q->limits.max_sectors;
 889 }
 890
 891 static inline unsigned int blk_rq_get_max_sectors(struct request *rq)
 892 {
 893         struct request_queue *q = rq->q;
 894
 895         if (unlikely(rq->cmd_type == REQ_TYPE_BLOCK_PC))
 896                 return q->limits.max_hw_sectors;
 897
 898         return blk_queue_get_max_sectors(q, rq->cmd_flags);
 899 }
 900
 901 static inline unsigned int blk_rq_count_bios(struct request *rq)
 902 {
 903         unsigned int nr_bios = 0;
 904         struct bio *bio;
 905
 906         __rq_for_each_bio(bio, rq)
 907                 nr_bios++;
 908
 909         return nr_bios;
 910 }
 911
 912 /*
 913  * Request issue related functions.
 914  */
 915 extern struct request *blk_peek_request(struct request_queue *q);
 916 extern void blk_start_request(struct request *rq);
 917 extern struct request *blk_fetch_request(struct request_queue *q);
 918
 919 /*
 920  * Request completion related functions.
 921  *
 922  * blk_update_request() completes given number of bytes and updates
 923  * the request without completing it.
 924  *
 925  * blk_end_request() and friends.  __blk_end_request() must be called
 926  * with the request queue spinlock acquired.
 927  *
 928  * Several drivers define their own end_request and call
 929  * blk_end_request() for parts of the original function.
 930  * This prevents code duplication in drivers.
 931  */
 932 extern bool blk_update_request(struct request *rq, int error,
 933                                unsigned int nr_bytes);
 934 extern bool blk_end_request(struct request *rq, int error,
 935                             unsigned int nr_bytes);
 936 extern void blk_end_request_all(struct request *rq, int error);
 937 extern bool blk_end_request_cur(struct request *rq, int error);
 938 extern bool blk_end_request_err(struct request *rq, int error);
 939 extern bool __blk_end_request(struct request *rq, int error,
 940                               unsigned int nr_bytes);
 941 extern void __blk_end_request_all(struct request *rq, int error);
 942 extern bool __blk_end_request_cur(struct request *rq, int error);
 943 extern bool __blk_end_request_err(struct request *rq, int error);
 944
 945 extern void blk_complete_request(struct request *);
 946 extern void __blk_complete_request(struct request *);
 947 extern void blk_abort_request(struct request *);
 948 extern void blk_unprep_request(struct request *);
 949
 950 /*
 951  * Access functions for manipulating queue properties
 952  */
 953 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
 954                                         spinlock_t *lock, int node_id);
 955 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
 956 extern struct request_queue *blk_init_allocated_queue(struct request_queue *,
 957                                                       request_fn_proc *, spinlock_t *);
 958 extern void blk_cleanup_queue(struct request_queue *);
 959 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
 960 extern void blk_queue_bounce_limit(struct request_queue *, u64);
 961 extern void blk_limits_max_hw_sectors(struct queue_limits *, unsigned int);
 962 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
 963 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
 964 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
 965 extern void blk_queue_max_discard_sectors(struct request_queue *q,
 966                 unsigned int max_discard_sectors);
 967 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
 968                 unsigned int max_write_same_sectors);
 969 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
 970 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
 971 extern void blk_queue_alignment_offset(struct request_queue *q,
 972                                        unsigned int alignment);
 973 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
 974 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
 975 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
 976 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
 977 extern void blk_set_default_limits(struct queue_limits *lim);
 978 extern void blk_set_stacking_limits(struct queue_limits *lim);
 979 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
 980                             sector_t offset);
 981 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
 982                             sector_t offset);
 983 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
 984                               sector_t offset);
 985 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
 986 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
 987 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
 988 extern int blk_queue_dma_drain(struct request_queue *q,
 989                                dma_drain_needed_fn *dma_drain_needed,
 990                                void *buf, unsigned int size);
 991 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
 992 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
 993 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
 994 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
 995 extern void blk_queue_merge_bvec(struct request_queue *, merge_bvec_fn *);
 996 extern void blk_queue_dma_alignment(struct request_queue *, int);
 997 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
 998 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
 999 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1000 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1001 extern void blk_queue_flush(struct request_queue *q, unsigned int flush);
1002 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1003 extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
1004
1005 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1006 extern int blk_bio_map_sg(struct request_queue *q, struct bio *bio,
1007                           struct scatterlist *sglist);
1008 extern void blk_dump_rq_flags(struct request *, char *);
1009 extern long nr_blockdev_pages(void);
1010
1011 bool __must_check blk_get_queue(struct request_queue *);
1012 struct request_queue *blk_alloc_queue(gfp_t);
1013 struct request_queue *blk_alloc_queue_node(gfp_t, int);
1014 extern void blk_put_queue(struct request_queue *);
1015
1016 /*
1017  * block layer runtime pm functions
1018  */
1019 #ifdef CONFIG_PM_RUNTIME
1020 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1021 extern int blk_pre_runtime_suspend(struct request_queue *q);
1022 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1023 extern void blk_pre_runtime_resume(struct request_queue *q);
1024 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1025 #else
1026 static inline void blk_pm_runtime_init(struct request_queue *q,
1027         struct device *dev) {}
1028 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1029 {
1030         return -ENOSYS;
1031 }
1032 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1033 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1034 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1035 #endif
1036
1037 /*
1038  * blk_plug permits building a queue of related requests by holding the I/O
1039  * fragments for a short period. This allows merging of sequential requests
1040  * into single larger request. As the requests are moved from a per-task list to
1041  * the device's request_queue in a batch, this results in improved scalability
1042  * as the lock contention for request_queue lock is reduced.
1043  *
1044  * It is ok not to disable preemption when adding the request to the plug list
1045  * or when attempting a merge, because blk_schedule_flush_list() will only flush
1046  * the plug list when the task sleeps by itself. For details, please see
1047  * schedule() where blk_schedule_flush_plug() is called.
1048  */
1049 struct blk_plug {
1050         unsigned long magic; /* detect uninitialized use-cases */
1051         struct list_head list; /* requests */
1052         struct list_head mq_list; /* blk-mq requests */
1053         struct list_head cb_list; /* md requires an unplug callback */
1054 };
1055 #define BLK_MAX_REQUEST_COUNT 16
1056
1057 struct blk_plug_cb;
1058 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1059 struct blk_plug_cb {
1060         struct list_head list;
1061         blk_plug_cb_fn callback;
1062         void *data;
1063 };
1064 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1065                                              void *data, int size);
1066 extern void blk_start_plug(struct blk_plug *);
1067 extern void blk_finish_plug(struct blk_plug *);
1068 extern void blk_flush_plug_list(struct blk_plug *, bool);
1069
1070 static inline void blk_flush_plug(struct task_struct *tsk)
1071 {
1072         struct blk_plug *plug = tsk->plug;
1073
1074         if (plug)
1075                 blk_flush_plug_list(plug, false);
1076 }
1077
1078 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1079 {
1080         struct blk_plug *plug = tsk->plug;
1081
1082         if (plug)
1083                 blk_flush_plug_list(plug, true);
1084 }
1085
1086 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1087 {
1088         struct blk_plug *plug = tsk->plug;
1089
1090         return plug &&
1091                 (!list_empty(&plug->list) ||
1092                  !list_empty(&plug->mq_list) ||
1093                  !list_empty(&plug->cb_list));
1094 }
1095
1096 /*
1097  * tag stuff
1098  */
1099 #define blk_rq_tagged(rq)               ((rq)->cmd_flags & REQ_QUEUED)
1100 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1101 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1102 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1103 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *);
1104 extern void blk_queue_free_tags(struct request_queue *);
1105 extern int blk_queue_resize_tags(struct request_queue *, int);
1106 extern void blk_queue_invalidate_tags(struct request_queue *);
1107 extern struct blk_queue_tag *blk_init_tags(int);
1108 extern void blk_free_tags(struct blk_queue_tag *);
1109
1110 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1111                                                 int tag)
1112 {
1113         if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1114                 return NULL;
1115         return bqt->tag_index[tag];
1116 }
1117
1118 #define BLKDEV_DISCARD_SECURE  0x01    /* secure discard */
1119
1120 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1121 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1122                 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1123 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1124                 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1125 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1126                         sector_t nr_sects, gfp_t gfp_mask);
1127 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1128                 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1129 {
1130         return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9),
1131                                     nr_blocks << (sb->s_blocksize_bits - 9),
1132                                     gfp_mask, flags);
1133 }
1134 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1135                 sector_t nr_blocks, gfp_t gfp_mask)
1136 {
1137         return blkdev_issue_zeroout(sb->s_bdev,
1138                                     block << (sb->s_blocksize_bits - 9),
1139                                     nr_blocks << (sb->s_blocksize_bits - 9),
1140                                     gfp_mask);
1141 }
1142
1143 extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm);
1144
1145 enum blk_default_limits {
1146         BLK_MAX_SEGMENTS        = 128,
1147         BLK_SAFE_MAX_SECTORS    = 255,
1148         BLK_DEF_MAX_SECTORS     = 1024,
1149         BLK_MAX_SEGMENT_SIZE    = 65536,
1150         BLK_SEG_BOUNDARY_MASK   = 0xFFFFFFFFUL,
1151 };
1152
1153 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1154
1155 static inline unsigned long queue_bounce_pfn(struct request_queue *q)
1156 {
1157         return q->limits.bounce_pfn;
1158 }
1159
1160 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1161 {
1162         return q->limits.seg_boundary_mask;
1163 }
1164
1165 static inline unsigned int queue_max_sectors(struct request_queue *q)
1166 {
1167         return q->limits.max_sectors;
1168 }
1169
1170 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1171 {
1172         return q->limits.max_hw_sectors;
1173 }
1174
1175 static inline unsigned short queue_max_segments(struct request_queue *q)
1176 {
1177         return q->limits.max_segments;
1178 }
1179
1180 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1181 {
1182         return q->limits.max_segment_size;
1183 }
1184
1185 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1186 {
1187         int retval = 512;
1188
1189         if (q && q->limits.logical_block_size)
1190                 retval = q->limits.logical_block_size;
1191
1192         return retval;
1193 }
1194
1195 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1196 {
1197         return queue_logical_block_size(bdev_get_queue(bdev));
1198 }
1199
1200 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1201 {
1202         return q->limits.physical_block_size;
1203 }
1204
1205 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1206 {
1207         return queue_physical_block_size(bdev_get_queue(bdev));
1208 }
1209
1210 static inline unsigned int queue_io_min(struct request_queue *q)
1211 {
1212         return q->limits.io_min;
1213 }
1214
1215 static inline int bdev_io_min(struct block_device *bdev)
1216 {
1217         return queue_io_min(bdev_get_queue(bdev));
1218 }
1219
1220 static inline unsigned int queue_io_opt(struct request_queue *q)
1221 {
1222         return q->limits.io_opt;
1223 }
1224
1225 static inline int bdev_io_opt(struct block_device *bdev)
1226 {
1227         return queue_io_opt(bdev_get_queue(bdev));
1228 }
1229
1230 static inline int queue_alignment_offset(struct request_queue *q)
1231 {
1232         if (q->limits.misaligned)
1233                 return -1;
1234
1235         return q->limits.alignment_offset;
1236 }
1237
1238 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1239 {
1240         unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1241         unsigned int alignment = (sector << 9) & (granularity - 1);
1242
1243         return (granularity + lim->alignment_offset - alignment)
1244                 & (granularity - 1);
1245 }
1246
1247 static inline int bdev_alignment_offset(struct block_device *bdev)
1248 {
1249         struct request_queue *q = bdev_get_queue(bdev);
1250
1251         if (q->limits.misaligned)
1252                 return -1;
1253
1254         if (bdev != bdev->bd_contains)
1255                 return bdev->bd_part->alignment_offset;
1256
1257         return q->limits.alignment_offset;
1258 }
1259
1260 static inline int queue_discard_alignment(struct request_queue *q)
1261 {
1262         if (q->limits.discard_misaligned)
1263                 return -1;
1264
1265         return q->limits.discard_alignment;
1266 }
1267
1268 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1269 {
1270         unsigned int alignment, granularity, offset;
1271
1272         if (!lim->max_discard_sectors)
1273                 return 0;
1274
1275         /* Why are these in bytes, not sectors? */
1276         alignment = lim->discard_alignment >> 9;
1277         granularity = lim->discard_granularity >> 9;
1278         if (!granularity)
1279                 return 0;
1280
1281         /* Offset of the partition start in 'granularity' sectors */
1282         offset = sector_div(sector, granularity);
1283
1284         /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1285         offset = (granularity + alignment - offset) % granularity;
1286
1287         /* Turn it back into bytes, gaah */
1288         return offset << 9;
1289 }
1290
1291 static inline int bdev_discard_alignment(struct block_device *bdev)
1292 {
1293         struct request_queue *q = bdev_get_queue(bdev);
1294
1295         if (bdev != bdev->bd_contains)
1296                 return bdev->bd_part->discard_alignment;
1297
1298         return q->limits.discard_alignment;
1299 }
1300
1301 static inline unsigned int queue_discard_zeroes_data(struct request_queue *q)
1302 {
1303         if (q->limits.max_discard_sectors && q->limits.discard_zeroes_data == 1)
1304                 return 1;
1305
1306         return 0;
1307 }
1308
1309 static inline unsigned int bdev_discard_zeroes_data(struct block_device *bdev)
1310 {
1311         return queue_discard_zeroes_data(bdev_get_queue(bdev));
1312 }
1313
1314 static inline unsigned int bdev_write_same(struct block_device *bdev)
1315 {
1316         struct request_queue *q = bdev_get_queue(bdev);
1317
1318         if (q)
1319                 return q->limits.max_write_same_sectors;
1320
1321         return 0;
1322 }
1323
1324 static inline int queue_dma_alignment(struct request_queue *q)
1325 {
1326         return q ? q->dma_alignment : 511;
1327 }
1328
1329 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1330                                  unsigned int len)
1331 {
1332         unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1333         return !(addr & alignment) && !(len & alignment);
1334 }
1335
1336 /* assumes size > 256 */
1337 static inline unsigned int blksize_bits(unsigned int size)
1338 {
1339         unsigned int bits = 8;
1340         do {
1341                 bits++;
1342                 size >>= 1;
1343         } while (size > 256);
1344         return bits;
1345 }
1346
1347 static inline unsigned int block_size(struct block_device *bdev)
1348 {
1349         return bdev->bd_block_size;
1350 }
1351
1352 static inline bool queue_flush_queueable(struct request_queue *q)
1353 {
1354         return !q->flush_not_queueable;
1355 }
1356
1357 typedef struct {struct page *v;} Sector;
1358
1359 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1360
1361 static inline void put_dev_sector(Sector p)
1362 {
1363         page_cache_release(p.v);
1364 }
1365
1366 struct work_struct;
1367 int kblockd_schedule_work(struct request_queue *q, struct work_struct *work);
1368 int kblockd_schedule_delayed_work(struct request_queue *q, struct delayed_work *dwork, unsigned long delay);
1369
1370 #ifdef CONFIG_BLK_CGROUP
1371 /*
1372  * This should not be using sched_clock(). A real patch is in progress
1373  * to fix this up, until that is in place we need to disable preemption
1374  * around sched_clock() in this function and set_io_start_time_ns().
1375  */
1376 static inline void set_start_time_ns(struct request *req)
1377 {
1378         preempt_disable();
1379         req->start_time_ns = sched_clock();
1380         preempt_enable();
1381 }
1382
1383 static inline void set_io_start_time_ns(struct request *req)
1384 {
1385         preempt_disable();
1386         req->io_start_time_ns = sched_clock();
1387         preempt_enable();
1388 }
1389
1390 static inline uint64_t rq_start_time_ns(struct request *req)
1391 {
1392         return req->start_time_ns;
1393 }
1394
1395 static inline uint64_t rq_io_start_time_ns(struct request *req)
1396 {
1397         return req->io_start_time_ns;
1398 }
1399 #else
1400 static inline void set_start_time_ns(struct request *req) {}
1401 static inline void set_io_start_time_ns(struct request *req) {}
1402 static inline uint64_t rq_start_time_ns(struct request *req)
1403 {
1404         return 0;
1405 }
1406 static inline uint64_t rq_io_start_time_ns(struct request *req)
1407 {
1408         return 0;
1409 }
1410 #endif
1411
1412 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1413         MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1414 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1415         MODULE_ALIAS("block-major-" __stringify(major) "-*")
1416
1417 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1418
1419 #define INTEGRITY_FLAG_READ     2       /* verify data integrity on read */
1420 #define INTEGRITY_FLAG_WRITE    4       /* generate data integrity on write */
1421
1422 struct blk_integrity_exchg {
1423         void                    *prot_buf;
1424         void                    *data_buf;
1425         sector_t                sector;
1426         unsigned int            data_size;
1427         unsigned short          sector_size;
1428         const char              *disk_name;
1429 };
1430
1431 typedef void (integrity_gen_fn) (struct blk_integrity_exchg *);
1432 typedef int (integrity_vrfy_fn) (struct blk_integrity_exchg *);
1433 typedef void (integrity_set_tag_fn) (void *, void *, unsigned int);
1434 typedef void (integrity_get_tag_fn) (void *, void *, unsigned int);
1435
1436 struct blk_integrity {
1437         integrity_gen_fn        *generate_fn;
1438         integrity_vrfy_fn       *verify_fn;
1439         integrity_set_tag_fn    *set_tag_fn;
1440         integrity_get_tag_fn    *get_tag_fn;
1441
1442         unsigned short          flags;
1443         unsigned short          tuple_size;
1444         unsigned short          sector_size;
1445         unsigned short          tag_size;
1446
1447         const char              *name;
1448
1449         struct kobject          kobj;
1450 };
1451
1452 extern bool blk_integrity_is_initialized(struct gendisk *);
1453 extern int blk_integrity_register(struct gendisk *, struct blk_integrity *);
1454 extern void blk_integrity_unregister(struct gendisk *);
1455 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1456 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1457                                    struct scatterlist *);
1458 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1459 extern int blk_integrity_merge_rq(struct request_queue *, struct request *,
1460                                   struct request *);
1461 extern int blk_integrity_merge_bio(struct request_queue *, struct request *,
1462                                    struct bio *);
1463
1464 static inline
1465 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1466 {
1467         return bdev->bd_disk->integrity;
1468 }
1469
1470 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1471 {
1472         return disk->integrity;
1473 }
1474
1475 static inline int blk_integrity_rq(struct request *rq)
1476 {
1477         if (rq->bio == NULL)
1478                 return 0;
1479
1480         return bio_integrity(rq->bio);
1481 }
1482
1483 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1484                                                     unsigned int segs)
1485 {
1486         q->limits.max_integrity_segments = segs;
1487 }
1488
1489 static inline unsigned short
1490 queue_max_integrity_segments(struct request_queue *q)
1491 {
1492         return q->limits.max_integrity_segments;
1493 }
1494
1495 #else /* CONFIG_BLK_DEV_INTEGRITY */
1496
1497 struct bio;
1498 struct block_device;
1499 struct gendisk;
1500 struct blk_integrity;
1501
1502 static inline int blk_integrity_rq(struct request *rq)
1503 {
1504         return 0;
1505 }
1506 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1507                                             struct bio *b)
1508 {
1509         return 0;
1510 }
1511 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1512                                           struct bio *b,
1513                                           struct scatterlist *s)
1514 {
1515         return 0;
1516 }
1517 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1518 {
1519         return 0;
1520 }
1521 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1522 {
1523         return NULL;
1524 }
1525 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1526 {
1527         return 0;
1528 }
1529 static inline int blk_integrity_register(struct gendisk *d,
1530                                          struct blk_integrity *b)
1531 {
1532         return 0;
1533 }
1534 static inline void blk_integrity_unregister(struct gendisk *d)
1535 {
1536 }
1537 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1538                                                     unsigned int segs)
1539 {
1540 }
1541 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1542 {
1543         return 0;
1544 }
1545 static inline int blk_integrity_merge_rq(struct request_queue *rq,
1546                                          struct request *r1,
1547                                          struct request *r2)
1548 {
1549         return 0;
1550 }
1551 static inline int blk_integrity_merge_bio(struct request_queue *rq,
1552                                           struct request *r,
1553                                           struct bio *b)
1554 {
1555         return 0;
1556 }
1557 static inline bool blk_integrity_is_initialized(struct gendisk *g)
1558 {
1559         return 0;
1560 }
1561
1562 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1563
1564 struct block_device_operations {
1565         int (*open) (struct block_device *, fmode_t);
1566         void (*release) (struct gendisk *, fmode_t);
1567         int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1568         int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1569         int (*direct_access) (struct block_device *, sector_t,
1570                                                 void **, unsigned long *);
1571         unsigned int (*check_events) (struct gendisk *disk,
1572                                       unsigned int clearing);
1573         /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1574         int (*media_changed) (struct gendisk *);
1575         void (*unlock_native_capacity) (struct gendisk *);
1576         int (*revalidate_disk) (struct gendisk *);
1577         int (*getgeo)(struct block_device *, struct hd_geometry *);
1578         /* this callback is with swap_lock and sometimes page table lock held */
1579         void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1580         struct module *owner;
1581 };
1582
1583 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1584                                  unsigned long);
1585 #else /* CONFIG_BLOCK */
1586 /*
1587  * stubs for when the block layer is configured out
1588  */
1589 #define buffer_heads_over_limit 0
1590
1591 static inline long nr_blockdev_pages(void)
1592 {
1593         return 0;
1594 }
1595
1596 struct blk_plug {
1597 };
1598
1599 static inline void blk_start_plug(struct blk_plug *plug)
1600 {
1601 }
1602
1603 static inline void blk_finish_plug(struct blk_plug *plug)
1604 {
1605 }
1606
1607 static inline void blk_flush_plug(struct task_struct *task)
1608 {
1609 }
1610
1611 static inline void blk_schedule_flush_plug(struct task_struct *task)
1612 {
1613 }
1614
1615
1616 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1617 {
1618         return false;
1619 }
1620
1621 #endif /* CONFIG_BLOCK */
1622
1623 #endif