include/linux/blkdev.h

   1 #ifndef _LINUX_BLKDEV_H
   2 #define _LINUX_BLKDEV_H
   3
   4 #include <linux/sched.h>
   5 #include <linux/major.h>
   6 #include <linux/genhd.h>
   7 #include <linux/list.h>
   8 #include <linux/timer.h>
   9 #include <linux/workqueue.h>
  10 #include <linux/pagemap.h>
  11 #include <linux/backing-dev.h>
  12 #include <linux/wait.h>
  13 #include <linux/mempool.h>
  14 #include <linux/bio.h>
  15 #include <linux/module.h>
  16 #include <linux/stringify.h>
  17
  18 #include <asm/scatterlist.h>
  19
  20 #ifdef CONFIG_LBD
  21 # include <asm/div64.h>
  22 # define sector_div(a, b) do_div(a, b)
  23 #else
  24 # define sector_div(n, b)( \
  25 { \
  26         int _res; \
  27         _res = (n) % (b); \
  28         (n) /= (b); \
  29         _res; \
  30 } \
  31 )
  32 #endif
  33
  34 #ifdef CONFIG_BLOCK
  35
  36 struct scsi_ioctl_command;
  37
  38 struct request_queue;
  39 typedef struct request_queue request_queue_t;
  40 struct elevator_queue;
  41 typedef struct elevator_queue elevator_t;
  42 struct request_pm_state;
  43 struct blk_trace;
  44
  45 #define BLKDEV_MIN_RQ   4
  46 #define BLKDEV_MAX_RQ   128     /* Default maximum */
  47
  48 /*
  49  * This is the per-process anticipatory I/O scheduler state.
  50  */
  51 struct as_io_context {
  52         spinlock_t lock;
  53
  54         void (*dtor)(struct as_io_context *aic); /* destructor */
  55         void (*exit)(struct as_io_context *aic); /* called on task exit */
  56
  57         unsigned long state;
  58         atomic_t nr_queued; /* queued reads & sync writes */
  59         atomic_t nr_dispatched; /* number of requests gone to the drivers */
  60
  61         /* IO History tracking */
  62         /* Thinktime */
  63         unsigned long last_end_request;
  64         unsigned long ttime_total;
  65         unsigned long ttime_samples;
  66         unsigned long ttime_mean;
  67         /* Layout pattern */
  68         unsigned int seek_samples;
  69         sector_t last_request_pos;
  70         u64 seek_total;
  71         sector_t seek_mean;
  72 };
  73
  74 struct cfq_queue;
  75 struct cfq_io_context {
  76         struct rb_node rb_node;
  77         void *key;
  78
  79         struct cfq_queue *cfqq[2];
  80
  81         struct io_context *ioc;
  82
  83         unsigned long last_end_request;
  84         sector_t last_request_pos;
  85
  86         unsigned long ttime_total;
  87         unsigned long ttime_samples;
  88         unsigned long ttime_mean;
  89
  90         unsigned int seek_samples;
  91         u64 seek_total;
  92         sector_t seek_mean;
  93
  94         struct list_head queue_list;
  95
  96         void (*dtor)(struct io_context *); /* destructor */
  97         void (*exit)(struct io_context *); /* called on task exit */
  98 };
  99
 100 /*
 101  * This is the per-process I/O subsystem state.  It is refcounted and
 102  * kmalloc'ed. Currently all fields are modified in process io context
 103  * (apart from the atomic refcount), so require no locking.
 104  */
 105 struct io_context {
 106         atomic_t refcount;
 107         struct task_struct *task;
 108
 109         unsigned int ioprio_changed;
 110
 111         /*
 112          * For request batching
 113          */
 114         unsigned long last_waited; /* Time last woken after wait for request */
 115         int nr_batch_requests;     /* Number of requests left in the batch */
 116
 117         struct as_io_context *aic;
 118         struct rb_root cic_root;
 119         void *ioc_data;
 120 };
 121
 122 void put_io_context(struct io_context *ioc);
 123 void exit_io_context(void);
 124 struct io_context *get_io_context(gfp_t gfp_flags, int node);
 125 void copy_io_context(struct io_context **pdst, struct io_context **psrc);
 126 void swap_io_context(struct io_context **ioc1, struct io_context **ioc2);
 127
 128 struct request;
 129 typedef void (rq_end_io_fn)(struct request *, int);
 130
 131 struct request_list {
 132         int count[2];
 133         int starved[2];
 134         int elvpriv;
 135         mempool_t *rq_pool;
 136         wait_queue_head_t wait[2];
 137 };
 138
 139 /*
 140  * request command types
 141  */
 142 enum rq_cmd_type_bits {
 143         REQ_TYPE_FS             = 1,    /* fs request */
 144         REQ_TYPE_BLOCK_PC,              /* scsi command */
 145         REQ_TYPE_SENSE,                 /* sense request */
 146         REQ_TYPE_PM_SUSPEND,            /* suspend request */
 147         REQ_TYPE_PM_RESUME,             /* resume request */
 148         REQ_TYPE_PM_SHUTDOWN,           /* shutdown request */
 149         REQ_TYPE_FLUSH,                 /* flush request */
 150         REQ_TYPE_SPECIAL,               /* driver defined type */
 151         REQ_TYPE_LINUX_BLOCK,           /* generic block layer message */
 152         /*
 153          * for ATA/ATAPI devices. this really doesn't belong here, ide should
 154          * use REQ_TYPE_SPECIAL and use rq->cmd[0] with the range of driver
 155          * private REQ_LB opcodes to differentiate what type of request this is
 156          */
 157         REQ_TYPE_ATA_CMD,
 158         REQ_TYPE_ATA_TASK,
 159         REQ_TYPE_ATA_TASKFILE,
 160         REQ_TYPE_ATA_PC,
 161 };
 162
 163 /*
 164  * For request of type REQ_TYPE_LINUX_BLOCK, rq->cmd[0] is the opcode being
 165  * sent down (similar to how REQ_TYPE_BLOCK_PC means that ->cmd[] holds a
 166  * SCSI cdb.
 167  *
 168  * 0x00 -> 0x3f are driver private, to be used for whatever purpose they need,
 169  * typically to differentiate REQ_TYPE_SPECIAL requests.
 170  *
 171  */
 172 enum {
 173         /*
 174          * just examples for now
 175          */
 176         REQ_LB_OP_EJECT = 0x40,         /* eject request */
 177         REQ_LB_OP_FLUSH = 0x41,         /* flush device */
 178 };
 179
 180 /*
 181  * request type modified bits. first three bits match BIO_RW* bits, important
 182  */
 183 enum rq_flag_bits {
 184         __REQ_RW,               /* not set, read. set, write */
 185         __REQ_FAILFAST,         /* no low level driver retries */
 186         __REQ_SORTED,           /* elevator knows about this request */
 187         __REQ_SOFTBARRIER,      /* may not be passed by ioscheduler */
 188         __REQ_HARDBARRIER,      /* may not be passed by drive either */
 189         __REQ_FUA,              /* forced unit access */
 190         __REQ_NOMERGE,          /* don't touch this for merging */
 191         __REQ_STARTED,          /* drive already may have started this one */
 192         __REQ_DONTPREP,         /* don't call prep for this one */
 193         __REQ_QUEUED,           /* uses queueing */
 194         __REQ_ELVPRIV,          /* elevator private data attached */
 195         __REQ_FAILED,           /* set if the request failed */
 196         __REQ_QUIET,            /* don't worry about errors */
 197         __REQ_PREEMPT,          /* set for "ide_preempt" requests */
 198         __REQ_ORDERED_COLOR,    /* is before or after barrier */
 199         __REQ_RW_SYNC,          /* request is sync (O_DIRECT) */
 200         __REQ_ALLOCED,          /* request came from our alloc pool */
 201         __REQ_RW_META,          /* metadata io request */
 202         __REQ_NR_BITS,          /* stops here */
 203 };
 204
 205 #define REQ_RW          (1 << __REQ_RW)
 206 #define REQ_FAILFAST    (1 << __REQ_FAILFAST)
 207 #define REQ_SORTED      (1 << __REQ_SORTED)
 208 #define REQ_SOFTBARRIER (1 << __REQ_SOFTBARRIER)
 209 #define REQ_HARDBARRIER (1 << __REQ_HARDBARRIER)
 210 #define REQ_FUA         (1 << __REQ_FUA)
 211 #define REQ_NOMERGE     (1 << __REQ_NOMERGE)
 212 #define REQ_STARTED     (1 << __REQ_STARTED)
 213 #define REQ_DONTPREP    (1 << __REQ_DONTPREP)
 214 #define REQ_QUEUED      (1 << __REQ_QUEUED)
 215 #define REQ_ELVPRIV     (1 << __REQ_ELVPRIV)
 216 #define REQ_FAILED      (1 << __REQ_FAILED)
 217 #define REQ_QUIET       (1 << __REQ_QUIET)
 218 #define REQ_PREEMPT     (1 << __REQ_PREEMPT)
 219 #define REQ_ORDERED_COLOR       (1 << __REQ_ORDERED_COLOR)
 220 #define REQ_RW_SYNC     (1 << __REQ_RW_SYNC)
 221 #define REQ_ALLOCED     (1 << __REQ_ALLOCED)
 222 #define REQ_RW_META     (1 << __REQ_RW_META)
 223
 224 #define BLK_MAX_CDB     16
 225
 226 /*
 227  * try to put the fields that are referenced together in the same cacheline
 228  */
 229 struct request {
 230         struct list_head queuelist;
 231         struct list_head donelist;
 232
 233         request_queue_t *q;
 234
 235         unsigned int cmd_flags;
 236         enum rq_cmd_type_bits cmd_type;
 237
 238         /* Maintain bio traversal state for part by part I/O submission.
 239          * hard_* are block layer internals, no driver should touch them!
 240          */
 241
 242         sector_t sector;                /* next sector to submit */
 243         sector_t hard_sector;           /* next sector to complete */
 244         unsigned long nr_sectors;       /* no. of sectors left to submit */
 245         unsigned long hard_nr_sectors;  /* no. of sectors left to complete */
 246         /* no. of sectors left to submit in the current segment */
 247         unsigned int current_nr_sectors;
 248
 249         /* no. of sectors left to complete in the current segment */
 250         unsigned int hard_cur_sectors;
 251
 252         struct bio *bio;
 253         struct bio *biotail;
 254
 255         struct hlist_node hash; /* merge hash */
 256         /*
 257          * The rb_node is only used inside the io scheduler, requests
 258          * are pruned when moved to the dispatch queue. So let the
 259          * completion_data share space with the rb_node.
 260          */
 261         union {
 262                 struct rb_node rb_node; /* sort/lookup */
 263                 void *completion_data;
 264         };
 265
 266         /*
 267          * two pointers are available for the IO schedulers, if they need
 268          * more they have to dynamically allocate it.
 269          */
 270         void *elevator_private;
 271         void *elevator_private2;
 272
 273         struct gendisk *rq_disk;
 274         unsigned long start_time;
 275
 276         /* Number of scatter-gather DMA addr+len pairs after
 277          * physical address coalescing is performed.
 278          */
 279         unsigned short nr_phys_segments;
 280
 281         /* Number of scatter-gather addr+len pairs after
 282          * physical and DMA remapping hardware coalescing is performed.
 283          * This is the number of scatter-gather entries the driver
 284          * will actually have to deal with after DMA mapping is done.
 285          */
 286         unsigned short nr_hw_segments;
 287
 288         unsigned short ioprio;
 289
 290         void *special;
 291         char *buffer;
 292
 293         int tag;
 294         int errors;
 295
 296         int ref_count;
 297
 298         /*
 299          * when request is used as a packet command carrier
 300          */
 301         unsigned int cmd_len;
 302         unsigned char cmd[BLK_MAX_CDB];
 303
 304         unsigned int data_len;
 305         unsigned int sense_len;
 306         void *data;
 307         void *sense;
 308
 309         unsigned int timeout;
 310         int retries;
 311
 312         /*
 313          * completion callback.
 314          */
 315         rq_end_io_fn *end_io;
 316         void *end_io_data;
 317 };
 318
 319 /*
 320  * State information carried for REQ_TYPE_PM_SUSPEND and REQ_TYPE_PM_RESUME
 321  * requests. Some step values could eventually be made generic.
 322  */
 323 struct request_pm_state
 324 {
 325         /* PM state machine step value, currently driver specific */
 326         int     pm_step;
 327         /* requested PM state value (S1, S2, S3, S4, ...) */
 328         u32     pm_state;
 329         void*   data;           /* for driver use */
 330 };
 331
 332 #include <linux/elevator.h>
 333
 334 typedef void (request_fn_proc) (request_queue_t *q);
 335 typedef int (make_request_fn) (request_queue_t *q, struct bio *bio);
 336 typedef int (prep_rq_fn) (request_queue_t *, struct request *);
 337 typedef void (unplug_fn) (request_queue_t *);
 338
 339 struct bio_vec;
 340 typedef int (merge_bvec_fn) (request_queue_t *, struct bio *, struct bio_vec *);
 341 typedef int (issue_flush_fn) (request_queue_t *, struct gendisk *, sector_t *);
 342 typedef void (prepare_flush_fn) (request_queue_t *, struct request *);
 343 typedef void (softirq_done_fn)(struct request *);
 344
 345 enum blk_queue_state {
 346         Queue_down,
 347         Queue_up,
 348 };
 349
 350 struct blk_queue_tag {
 351         struct request **tag_index;     /* map of busy tags */
 352         unsigned long *tag_map;         /* bit map of free/busy tags */
 353         struct list_head busy_list;     /* fifo list of busy tags */
 354         int busy;                       /* current depth */
 355         int max_depth;                  /* what we will send to device */
 356         int real_max_depth;             /* what the array can hold */
 357         atomic_t refcnt;                /* map can be shared */
 358 };
 359
 360 struct request_queue
 361 {
 362         /*
 363          * Together with queue_head for cacheline sharing
 364          */
 365         struct list_head        queue_head;
 366         struct request          *last_merge;
 367         elevator_t              *elevator;
 368
 369         /*
 370          * the queue request freelist, one for reads and one for writes
 371          */
 372         struct request_list     rq;
 373
 374         request_fn_proc         *request_fn;
 375         make_request_fn         *make_request_fn;
 376         prep_rq_fn              *prep_rq_fn;
 377         unplug_fn               *unplug_fn;
 378         merge_bvec_fn           *merge_bvec_fn;
 379         issue_flush_fn          *issue_flush_fn;
 380         prepare_flush_fn        *prepare_flush_fn;
 381         softirq_done_fn         *softirq_done_fn;
 382
 383         /*
 384          * Dispatch queue sorting
 385          */
 386         sector_t                end_sector;
 387         struct request          *boundary_rq;
 388
 389         /*
 390          * Auto-unplugging state
 391          */
 392         struct timer_list       unplug_timer;
 393         int                     unplug_thresh;  /* After this many requests */
 394         unsigned long           unplug_delay;   /* After this many jiffies */
 395         struct work_struct      unplug_work;
 396
 397         struct backing_dev_info backing_dev_info;
 398
 399         /*
 400          * The queue owner gets to use this for whatever they like.
 401          * ll_rw_blk doesn't touch it.
 402          */
 403         void                    *queuedata;
 404
 405         /*
 406          * queue needs bounce pages for pages above this limit
 407          */
 408         unsigned long           bounce_pfn;
 409         gfp_t                   bounce_gfp;
 410
 411         /*
 412          * various queue flags, see QUEUE_* below
 413          */
 414         unsigned long           queue_flags;
 415
 416         /*
 417          * protects queue structures from reentrancy. ->__queue_lock should
 418          * _never_ be used directly, it is queue private. always use
 419          * ->queue_lock.
 420          */
 421         spinlock_t              __queue_lock;
 422         spinlock_t              *queue_lock;
 423
 424         /*
 425          * queue kobject
 426          */
 427         struct kobject kobj;
 428
 429         /*
 430          * queue settings
 431          */
 432         unsigned long           nr_requests;    /* Max # of requests */
 433         unsigned int            nr_congestion_on;
 434         unsigned int            nr_congestion_off;
 435         unsigned int            nr_batching;
 436
 437         unsigned int            max_sectors;
 438         unsigned int            max_hw_sectors;
 439         unsigned short          max_phys_segments;
 440         unsigned short          max_hw_segments;
 441         unsigned short          hardsect_size;
 442         unsigned int            max_segment_size;
 443
 444         unsigned long           seg_boundary_mask;
 445         unsigned int            dma_alignment;
 446
 447         struct blk_queue_tag    *queue_tags;
 448
 449         unsigned int            nr_sorted;
 450         unsigned int            in_flight;
 451
 452         /*
 453          * sg stuff
 454          */
 455         unsigned int            sg_timeout;
 456         unsigned int            sg_reserved_size;
 457         int                     node;
 458 #ifdef CONFIG_BLK_DEV_IO_TRACE
 459         struct blk_trace        *blk_trace;
 460 #endif
 461         /*
 462          * reserved for flush operations
 463          */
 464         unsigned int            ordered, next_ordered, ordseq;
 465         int                     orderr, ordcolor;
 466         struct request          pre_flush_rq, bar_rq, post_flush_rq;
 467         struct request          *orig_bar_rq;
 468         unsigned int            bi_size;
 469
 470         struct mutex            sysfs_lock;
 471 };
 472
 473 #define QUEUE_FLAG_CLUSTER      0       /* cluster several segments into 1 */
 474 #define QUEUE_FLAG_QUEUED       1       /* uses generic tag queueing */
 475 #define QUEUE_FLAG_STOPPED      2       /* queue is stopped */
 476 #define QUEUE_FLAG_READFULL     3       /* write queue has been filled */
 477 #define QUEUE_FLAG_WRITEFULL    4       /* read queue has been filled */
 478 #define QUEUE_FLAG_DEAD         5       /* queue being torn down */
 479 #define QUEUE_FLAG_REENTER      6       /* Re-entrancy avoidance */
 480 #define QUEUE_FLAG_PLUGGED      7       /* queue is plugged */
 481 #define QUEUE_FLAG_ELVSWITCH    8       /* don't use elevator, just do FIFO */
 482
 483 enum {
 484         /*
 485          * Hardbarrier is supported with one of the following methods.
 486          *
 487          * NONE         : hardbarrier unsupported
 488          * DRAIN        : ordering by draining is enough
 489          * DRAIN_FLUSH  : ordering by draining w/ pre and post flushes
 490          * DRAIN_FUA    : ordering by draining w/ pre flush and FUA write
 491          * TAG          : ordering by tag is enough
 492          * TAG_FLUSH    : ordering by tag w/ pre and post flushes
 493          * TAG_FUA      : ordering by tag w/ pre flush and FUA write
 494          */
 495         QUEUE_ORDERED_NONE      = 0x00,
 496         QUEUE_ORDERED_DRAIN     = 0x01,
 497         QUEUE_ORDERED_TAG       = 0x02,
 498
 499         QUEUE_ORDERED_PREFLUSH  = 0x10,
 500         QUEUE_ORDERED_POSTFLUSH = 0x20,
 501         QUEUE_ORDERED_FUA       = 0x40,
 502
 503         QUEUE_ORDERED_DRAIN_FLUSH = QUEUE_ORDERED_DRAIN |
 504                         QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_POSTFLUSH,
 505         QUEUE_ORDERED_DRAIN_FUA = QUEUE_ORDERED_DRAIN |
 506                         QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_FUA,
 507         QUEUE_ORDERED_TAG_FLUSH = QUEUE_ORDERED_TAG |
 508                         QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_POSTFLUSH,
 509         QUEUE_ORDERED_TAG_FUA   = QUEUE_ORDERED_TAG |
 510                         QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_FUA,
 511
 512         /*
 513          * Ordered operation sequence
 514          */
 515         QUEUE_ORDSEQ_STARTED    = 0x01, /* flushing in progress */
 516         QUEUE_ORDSEQ_DRAIN      = 0x02, /* waiting for the queue to be drained */
 517         QUEUE_ORDSEQ_PREFLUSH   = 0x04, /* pre-flushing in progress */
 518         QUEUE_ORDSEQ_BAR        = 0x08, /* original barrier req in progress */
 519         QUEUE_ORDSEQ_POSTFLUSH  = 0x10, /* post-flushing in progress */
 520         QUEUE_ORDSEQ_DONE       = 0x20,
 521 };
 522
 523 #define blk_queue_plugged(q)    test_bit(QUEUE_FLAG_PLUGGED, &(q)->queue_flags)
 524 #define blk_queue_tagged(q)     test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
 525 #define blk_queue_stopped(q)    test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
 526 #define blk_queue_flushing(q)   ((q)->ordseq)
 527
 528 #define blk_fs_request(rq)      ((rq)->cmd_type == REQ_TYPE_FS)
 529 #define blk_pc_request(rq)      ((rq)->cmd_type == REQ_TYPE_BLOCK_PC)
 530 #define blk_special_request(rq) ((rq)->cmd_type == REQ_TYPE_SPECIAL)
 531 #define blk_sense_request(rq)   ((rq)->cmd_type == REQ_TYPE_SENSE)
 532
 533 #define blk_noretry_request(rq) ((rq)->cmd_flags & REQ_FAILFAST)
 534 #define blk_rq_started(rq)      ((rq)->cmd_flags & REQ_STARTED)
 535
 536 #define blk_account_rq(rq)      (blk_rq_started(rq) && blk_fs_request(rq))
 537
 538 #define blk_pm_suspend_request(rq)      ((rq)->cmd_type == REQ_TYPE_PM_SUSPEND)
 539 #define blk_pm_resume_request(rq)       ((rq)->cmd_type == REQ_TYPE_PM_RESUME)
 540 #define blk_pm_request(rq)      \
 541         (blk_pm_suspend_request(rq) || blk_pm_resume_request(rq))
 542
 543 #define blk_sorted_rq(rq)       ((rq)->cmd_flags & REQ_SORTED)
 544 #define blk_barrier_rq(rq)      ((rq)->cmd_flags & REQ_HARDBARRIER)
 545 #define blk_fua_rq(rq)          ((rq)->cmd_flags & REQ_FUA)
 546
 547 #define list_entry_rq(ptr)      list_entry((ptr), struct request, queuelist)
 548
 549 #define rq_data_dir(rq)         ((rq)->cmd_flags & 1)
 550
 551 /*
 552  * We regard a request as sync, if it's a READ or a SYNC write.
 553  */
 554 #define rq_is_sync(rq)          (rq_data_dir((rq)) == READ || (rq)->cmd_flags & REQ_RW_SYNC)
 555 #define rq_is_meta(rq)          ((rq)->cmd_flags & REQ_RW_META)
 556
 557 static inline int blk_queue_full(struct request_queue *q, int rw)
 558 {
 559         if (rw == READ)
 560                 return test_bit(QUEUE_FLAG_READFULL, &q->queue_flags);
 561         return test_bit(QUEUE_FLAG_WRITEFULL, &q->queue_flags);
 562 }
 563
 564 static inline void blk_set_queue_full(struct request_queue *q, int rw)
 565 {
 566         if (rw == READ)
 567                 set_bit(QUEUE_FLAG_READFULL, &q->queue_flags);
 568         else
 569                 set_bit(QUEUE_FLAG_WRITEFULL, &q->queue_flags);
 570 }
 571
 572 static inline void blk_clear_queue_full(struct request_queue *q, int rw)
 573 {
 574         if (rw == READ)
 575                 clear_bit(QUEUE_FLAG_READFULL, &q->queue_flags);
 576         else
 577                 clear_bit(QUEUE_FLAG_WRITEFULL, &q->queue_flags);
 578 }
 579
 580
 581 /*
 582  * mergeable request must not have _NOMERGE or _BARRIER bit set, nor may
 583  * it already be started by driver.
 584  */
 585 #define RQ_NOMERGE_FLAGS        \
 586         (REQ_NOMERGE | REQ_STARTED | REQ_HARDBARRIER | REQ_SOFTBARRIER)
 587 #define rq_mergeable(rq)        \
 588         (!((rq)->cmd_flags & RQ_NOMERGE_FLAGS) && blk_fs_request((rq)))
 589
 590 /*
 591  * q->prep_rq_fn return values
 592  */
 593 #define BLKPREP_OK              0       /* serve it */
 594 #define BLKPREP_KILL            1       /* fatal error, kill */
 595 #define BLKPREP_DEFER           2       /* leave on queue */
 596
 597 extern unsigned long blk_max_low_pfn, blk_max_pfn;
 598
 599 /*
 600  * standard bounce addresses:
 601  *
 602  * BLK_BOUNCE_HIGH      : bounce all highmem pages
 603  * BLK_BOUNCE_ANY       : don't bounce anything
 604  * BLK_BOUNCE_ISA       : bounce pages above ISA DMA boundary
 605  */
 606 #define BLK_BOUNCE_HIGH         ((u64)blk_max_low_pfn << PAGE_SHIFT)
 607 #define BLK_BOUNCE_ANY          ((u64)blk_max_pfn << PAGE_SHIFT)
 608 #define BLK_BOUNCE_ISA          (ISA_DMA_THRESHOLD)
 609
 610 #ifdef CONFIG_MMU
 611 extern int init_emergency_isa_pool(void);
 612 extern void blk_queue_bounce(request_queue_t *q, struct bio **bio);
 613 #else
 614 static inline int init_emergency_isa_pool(void)
 615 {
 616         return 0;
 617 }
 618 static inline void blk_queue_bounce(request_queue_t *q, struct bio **bio)
 619 {
 620 }
 621 #endif /* CONFIG_MMU */
 622
 623 #define rq_for_each_bio(_bio, rq)       \
 624         if ((rq->bio))                  \
 625                 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
 626
 627 extern int blk_register_queue(struct gendisk *disk);
 628 extern void blk_unregister_queue(struct gendisk *disk);
 629 extern void register_disk(struct gendisk *dev);
 630 extern void generic_make_request(struct bio *bio);
 631 extern void blk_put_request(struct request *);
 632 extern void __blk_put_request(request_queue_t *, struct request *);
 633 extern void blk_end_sync_rq(struct request *rq, int error);
 634 extern struct request *blk_get_request(request_queue_t *, int, gfp_t);
 635 extern void blk_insert_request(request_queue_t *, struct request *, int, void *);
 636 extern void blk_requeue_request(request_queue_t *, struct request *);
 637 extern void blk_plug_device(request_queue_t *);
 638 extern int blk_remove_plug(request_queue_t *);
 639 extern void blk_recount_segments(request_queue_t *, struct bio *);
 640 extern int scsi_cmd_ioctl(struct file *, struct gendisk *, unsigned int, void __user *);
 641 extern int sg_scsi_ioctl(struct file *, struct request_queue *,
 642                 struct gendisk *, struct scsi_ioctl_command __user *);
 643
 644 /*
 645  * Temporary export, until SCSI gets fixed up.
 646  */
 647 extern int ll_back_merge_fn(request_queue_t *, struct request *, struct bio *);
 648
 649 /*
 650  * A queue has just exitted congestion.  Note this in the global counter of
 651  * congested queues, and wake up anyone who was waiting for requests to be
 652  * put back.
 653  */
 654 static inline void blk_clear_queue_congested(request_queue_t *q, int rw)
 655 {
 656         clear_bdi_congested(&q->backing_dev_info, rw);
 657 }
 658
 659 /*
 660  * A queue has just entered congestion.  Flag that in the queue's VM-visible
 661  * state flags and increment the global gounter of congested queues.
 662  */
 663 static inline void blk_set_queue_congested(request_queue_t *q, int rw)
 664 {
 665         set_bdi_congested(&q->backing_dev_info, rw);
 666 }
 667
 668 extern void blk_start_queue(request_queue_t *q);
 669 extern void blk_stop_queue(request_queue_t *q);
 670 extern void blk_sync_queue(struct request_queue *q);
 671 extern void __blk_stop_queue(request_queue_t *q);
 672 extern void blk_run_queue(request_queue_t *);
 673 extern void blk_start_queueing(request_queue_t *);
 674 extern int blk_rq_map_user(request_queue_t *, struct request *, void __user *, unsigned long);
 675 extern int blk_rq_unmap_user(struct bio *);
 676 extern int blk_rq_map_kern(request_queue_t *, struct request *, void *, unsigned int, gfp_t);
 677 extern int blk_rq_map_user_iov(request_queue_t *, struct request *,
 678                                struct sg_iovec *, int, unsigned int);
 679 extern int blk_execute_rq(request_queue_t *, struct gendisk *,
 680                           struct request *, int);
 681 extern void blk_execute_rq_nowait(request_queue_t *, struct gendisk *,
 682                                   struct request *, int, rq_end_io_fn *);
 683
 684 static inline request_queue_t *bdev_get_queue(struct block_device *bdev)
 685 {
 686         return bdev->bd_disk->queue;
 687 }
 688
 689 static inline void blk_run_backing_dev(struct backing_dev_info *bdi,
 690                                        struct page *page)
 691 {
 692         if (bdi && bdi->unplug_io_fn)
 693                 bdi->unplug_io_fn(bdi, page);
 694 }
 695
 696 static inline void blk_run_address_space(struct address_space *mapping)
 697 {
 698         if (mapping)
 699                 blk_run_backing_dev(mapping->backing_dev_info, NULL);
 700 }
 701
 702 /*
 703  * end_request() and friends. Must be called with the request queue spinlock
 704  * acquired. All functions called within end_request() _must_be_ atomic.
 705  *
 706  * Several drivers define their own end_request and call
 707  * end_that_request_first() and end_that_request_last()
 708  * for parts of the original function. This prevents
 709  * code duplication in drivers.
 710  */
 711 extern int end_that_request_first(struct request *, int, int);
 712 extern int end_that_request_chunk(struct request *, int, int);
 713 extern void end_that_request_last(struct request *, int);
 714 extern void end_request(struct request *req, int uptodate);
 715 extern void blk_complete_request(struct request *);
 716
 717 /*
 718  * end_that_request_first/chunk() takes an uptodate argument. we account
 719  * any value <= as an io error. 0 means -EIO for compatability reasons,
 720  * any other < 0 value is the direct error type. An uptodate value of
 721  * 1 indicates successful io completion
 722  */
 723 #define end_io_error(uptodate)  (unlikely((uptodate) <= 0))
 724
 725 static inline void blkdev_dequeue_request(struct request *req)
 726 {
 727         elv_dequeue_request(req->q, req);
 728 }
 729
 730 /*
 731  * Access functions for manipulating queue properties
 732  */
 733 extern request_queue_t *blk_init_queue_node(request_fn_proc *rfn,
 734                                         spinlock_t *lock, int node_id);
 735 extern request_queue_t *blk_init_queue(request_fn_proc *, spinlock_t *);
 736 extern void blk_cleanup_queue(request_queue_t *);
 737 extern void blk_queue_make_request(request_queue_t *, make_request_fn *);
 738 extern void blk_queue_bounce_limit(request_queue_t *, u64);
 739 extern void blk_queue_max_sectors(request_queue_t *, unsigned int);
 740 extern void blk_queue_max_phys_segments(request_queue_t *, unsigned short);
 741 extern void blk_queue_max_hw_segments(request_queue_t *, unsigned short);
 742 extern void blk_queue_max_segment_size(request_queue_t *, unsigned int);
 743 extern void blk_queue_hardsect_size(request_queue_t *, unsigned short);
 744 extern void blk_queue_stack_limits(request_queue_t *t, request_queue_t *b);
 745 extern void blk_queue_segment_boundary(request_queue_t *, unsigned long);
 746 extern void blk_queue_prep_rq(request_queue_t *, prep_rq_fn *pfn);
 747 extern void blk_queue_merge_bvec(request_queue_t *, merge_bvec_fn *);
 748 extern void blk_queue_dma_alignment(request_queue_t *, int);
 749 extern void blk_queue_softirq_done(request_queue_t *, softirq_done_fn *);
 750 extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
 751 extern int blk_queue_ordered(request_queue_t *, unsigned, prepare_flush_fn *);
 752 extern void blk_queue_issue_flush_fn(request_queue_t *, issue_flush_fn *);
 753 extern int blk_do_ordered(request_queue_t *, struct request **);
 754 extern unsigned blk_ordered_cur_seq(request_queue_t *);
 755 extern unsigned blk_ordered_req_seq(struct request *);
 756 extern void blk_ordered_complete_seq(request_queue_t *, unsigned, int);
 757
 758 extern int blk_rq_map_sg(request_queue_t *, struct request *, struct scatterlist *);
 759 extern void blk_dump_rq_flags(struct request *, char *);
 760 extern void generic_unplug_device(request_queue_t *);
 761 extern void __generic_unplug_device(request_queue_t *);
 762 extern long nr_blockdev_pages(void);
 763
 764 int blk_get_queue(request_queue_t *);
 765 request_queue_t *blk_alloc_queue(gfp_t);
 766 request_queue_t *blk_alloc_queue_node(gfp_t, int);
 767 extern void blk_put_queue(request_queue_t *);
 768
 769 /*
 770  * tag stuff
 771  */
 772 #define blk_queue_tag_depth(q)          ((q)->queue_tags->busy)
 773 #define blk_queue_tag_queue(q)          ((q)->queue_tags->busy < (q)->queue_tags->max_depth)
 774 #define blk_rq_tagged(rq)               ((rq)->cmd_flags & REQ_QUEUED)
 775 extern int blk_queue_start_tag(request_queue_t *, struct request *);
 776 extern struct request *blk_queue_find_tag(request_queue_t *, int);
 777 extern void blk_queue_end_tag(request_queue_t *, struct request *);
 778 extern int blk_queue_init_tags(request_queue_t *, int, struct blk_queue_tag *);
 779 extern void blk_queue_free_tags(request_queue_t *);
 780 extern int blk_queue_resize_tags(request_queue_t *, int);
 781 extern void blk_queue_invalidate_tags(request_queue_t *);
 782 extern struct blk_queue_tag *blk_init_tags(int);
 783 extern void blk_free_tags(struct blk_queue_tag *);
 784
 785 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
 786                                                 int tag)
 787 {
 788         if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
 789                 return NULL;
 790         return bqt->tag_index[tag];
 791 }
 792
 793 extern void blk_rq_bio_prep(request_queue_t *, struct request *, struct bio *);
 794 extern int blkdev_issue_flush(struct block_device *, sector_t *);
 795
 796 #define MAX_PHYS_SEGMENTS 128
 797 #define MAX_HW_SEGMENTS 128
 798 #define SAFE_MAX_SECTORS 255
 799 #define BLK_DEF_MAX_SECTORS 1024
 800
 801 #define MAX_SEGMENT_SIZE        65536
 802
 803 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
 804
 805 static inline int queue_hardsect_size(request_queue_t *q)
 806 {
 807         int retval = 512;
 808
 809         if (q && q->hardsect_size)
 810                 retval = q->hardsect_size;
 811
 812         return retval;
 813 }
 814
 815 static inline int bdev_hardsect_size(struct block_device *bdev)
 816 {
 817         return queue_hardsect_size(bdev_get_queue(bdev));
 818 }
 819
 820 static inline int queue_dma_alignment(request_queue_t *q)
 821 {
 822         int retval = 511;
 823
 824         if (q && q->dma_alignment)
 825                 retval = q->dma_alignment;
 826
 827         return retval;
 828 }
 829
 830 /* assumes size > 256 */
 831 static inline unsigned int blksize_bits(unsigned int size)
 832 {
 833         unsigned int bits = 8;
 834         do {
 835                 bits++;
 836                 size >>= 1;
 837         } while (size > 256);
 838         return bits;
 839 }
 840
 841 static inline unsigned int block_size(struct block_device *bdev)
 842 {
 843         return bdev->bd_block_size;
 844 }
 845
 846 typedef struct {struct page *v;} Sector;
 847
 848 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
 849
 850 static inline void put_dev_sector(Sector p)
 851 {
 852         page_cache_release(p.v);
 853 }
 854
 855 struct work_struct;
 856 int kblockd_schedule_work(struct work_struct *work);
 857 void kblockd_flush(void);
 858
 859 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
 860         MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
 861 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
 862         MODULE_ALIAS("block-major-" __stringify(major) "-*")
 863
 864
 865 #else /* CONFIG_BLOCK */
 866 /*
 867  * stubs for when the block layer is configured out
 868  */
 869 #define buffer_heads_over_limit 0
 870
 871 static inline long blk_congestion_wait(int rw, long timeout)
 872 {
 873         return io_schedule_timeout(timeout);
 874 }
 875
 876 static inline long nr_blockdev_pages(void)
 877 {
 878         return 0;
 879 }
 880
 881 static inline void exit_io_context(void)
 882 {
 883 }
 884
 885 #endif /* CONFIG_BLOCK */
 886
 887 #endif