include/linux/blkdev.h

   1 #ifndef _LINUX_BLKDEV_H
   2 #define _LINUX_BLKDEV_H
   3
   4 #ifdef CONFIG_BLOCK
   5
   6 #include <linux/sched.h>
   7 #include <linux/major.h>
   8 #include <linux/genhd.h>
   9 #include <linux/list.h>
  10 #include <linux/timer.h>
  11 #include <linux/workqueue.h>
  12 #include <linux/pagemap.h>
  13 #include <linux/backing-dev.h>
  14 #include <linux/wait.h>
  15 #include <linux/mempool.h>
  16 #include <linux/bio.h>
  17 #include <linux/module.h>
  18 #include <linux/stringify.h>
  19 #include <linux/bsg.h>
  20
  21 #include <asm/scatterlist.h>
  22
  23 struct scsi_ioctl_command;
  24
  25 struct request_queue;
  26 typedef struct request_queue request_queue_t __deprecated;
  27 struct elevator_queue;
  28 typedef struct elevator_queue elevator_t;
  29 struct request_pm_state;
  30 struct blk_trace;
  31 struct request;
  32 struct sg_io_hdr;
  33
  34 #define BLKDEV_MIN_RQ   4
  35 #define BLKDEV_MAX_RQ   128     /* Default maximum */
  36
  37 /*
  38  * This is the per-process anticipatory I/O scheduler state.
  39  */
  40 struct as_io_context {
  41         spinlock_t lock;
  42
  43         void (*dtor)(struct as_io_context *aic); /* destructor */
  44         void (*exit)(struct as_io_context *aic); /* called on task exit */
  45
  46         unsigned long state;
  47         atomic_t nr_queued; /* queued reads & sync writes */
  48         atomic_t nr_dispatched; /* number of requests gone to the drivers */
  49
  50         /* IO History tracking */
  51         /* Thinktime */
  52         unsigned long last_end_request;
  53         unsigned long ttime_total;
  54         unsigned long ttime_samples;
  55         unsigned long ttime_mean;
  56         /* Layout pattern */
  57         unsigned int seek_samples;
  58         sector_t last_request_pos;
  59         u64 seek_total;
  60         sector_t seek_mean;
  61 };
  62
  63 struct cfq_queue;
  64 struct cfq_io_context {
  65         struct rb_node rb_node;
  66         void *key;
  67
  68         struct cfq_queue *cfqq[2];
  69
  70         struct io_context *ioc;
  71
  72         unsigned long last_end_request;
  73         sector_t last_request_pos;
  74
  75         unsigned long ttime_total;
  76         unsigned long ttime_samples;
  77         unsigned long ttime_mean;
  78
  79         unsigned int seek_samples;
  80         u64 seek_total;
  81         sector_t seek_mean;
  82
  83         struct list_head queue_list;
  84
  85         void (*dtor)(struct io_context *); /* destructor */
  86         void (*exit)(struct io_context *); /* called on task exit */
  87 };
  88
  89 /*
  90  * This is the per-process I/O subsystem state.  It is refcounted and
  91  * kmalloc'ed. Currently all fields are modified in process io context
  92  * (apart from the atomic refcount), so require no locking.
  93  */
  94 struct io_context {
  95         atomic_t refcount;
  96         struct task_struct *task;
  97
  98         unsigned int ioprio_changed;
  99
 100         /*
 101          * For request batching
 102          */
 103         unsigned long last_waited; /* Time last woken after wait for request */
 104         int nr_batch_requests;     /* Number of requests left in the batch */
 105
 106         struct as_io_context *aic;
 107         struct rb_root cic_root;
 108         void *ioc_data;
 109 };
 110
 111 void put_io_context(struct io_context *ioc);
 112 void exit_io_context(void);
 113 struct io_context *get_io_context(gfp_t gfp_flags, int node);
 114 void copy_io_context(struct io_context **pdst, struct io_context **psrc);
 115 void swap_io_context(struct io_context **ioc1, struct io_context **ioc2);
 116
 117 struct request;
 118 typedef void (rq_end_io_fn)(struct request *, int);
 119
 120 struct request_list {
 121         int count[2];
 122         int starved[2];
 123         int elvpriv;
 124         mempool_t *rq_pool;
 125         wait_queue_head_t wait[2];
 126 };
 127
 128 /*
 129  * request command types
 130  */
 131 enum rq_cmd_type_bits {
 132         REQ_TYPE_FS             = 1,    /* fs request */
 133         REQ_TYPE_BLOCK_PC,              /* scsi command */
 134         REQ_TYPE_SENSE,                 /* sense request */
 135         REQ_TYPE_PM_SUSPEND,            /* suspend request */
 136         REQ_TYPE_PM_RESUME,             /* resume request */
 137         REQ_TYPE_PM_SHUTDOWN,           /* shutdown request */
 138         REQ_TYPE_FLUSH,                 /* flush request */
 139         REQ_TYPE_SPECIAL,               /* driver defined type */
 140         REQ_TYPE_LINUX_BLOCK,           /* generic block layer message */
 141         /*
 142          * for ATA/ATAPI devices. this really doesn't belong here, ide should
 143          * use REQ_TYPE_SPECIAL and use rq->cmd[0] with the range of driver
 144          * private REQ_LB opcodes to differentiate what type of request this is
 145          */
 146         REQ_TYPE_ATA_CMD,
 147         REQ_TYPE_ATA_TASK,
 148         REQ_TYPE_ATA_TASKFILE,
 149         REQ_TYPE_ATA_PC,
 150 };
 151
 152 /*
 153  * For request of type REQ_TYPE_LINUX_BLOCK, rq->cmd[0] is the opcode being
 154  * sent down (similar to how REQ_TYPE_BLOCK_PC means that ->cmd[] holds a
 155  * SCSI cdb.
 156  *
 157  * 0x00 -> 0x3f are driver private, to be used for whatever purpose they need,
 158  * typically to differentiate REQ_TYPE_SPECIAL requests.
 159  *
 160  */
 161 enum {
 162         /*
 163          * just examples for now
 164          */
 165         REQ_LB_OP_EJECT = 0x40,         /* eject request */
 166         REQ_LB_OP_FLUSH = 0x41,         /* flush device */
 167 };
 168
 169 /*
 170  * request type modified bits. first three bits match BIO_RW* bits, important
 171  */
 172 enum rq_flag_bits {
 173         __REQ_RW,               /* not set, read. set, write */
 174         __REQ_FAILFAST,         /* no low level driver retries */
 175         __REQ_SORTED,           /* elevator knows about this request */
 176         __REQ_SOFTBARRIER,      /* may not be passed by ioscheduler */
 177         __REQ_HARDBARRIER,      /* may not be passed by drive either */
 178         __REQ_FUA,              /* forced unit access */
 179         __REQ_NOMERGE,          /* don't touch this for merging */
 180         __REQ_STARTED,          /* drive already may have started this one */
 181         __REQ_DONTPREP,         /* don't call prep for this one */
 182         __REQ_QUEUED,           /* uses queueing */
 183         __REQ_ELVPRIV,          /* elevator private data attached */
 184         __REQ_FAILED,           /* set if the request failed */
 185         __REQ_QUIET,            /* don't worry about errors */
 186         __REQ_PREEMPT,          /* set for "ide_preempt" requests */
 187         __REQ_ORDERED_COLOR,    /* is before or after barrier */
 188         __REQ_RW_SYNC,          /* request is sync (O_DIRECT) */
 189         __REQ_ALLOCED,          /* request came from our alloc pool */
 190         __REQ_RW_META,          /* metadata io request */
 191         __REQ_NR_BITS,          /* stops here */
 192 };
 193
 194 #define REQ_RW          (1 << __REQ_RW)
 195 #define REQ_FAILFAST    (1 << __REQ_FAILFAST)
 196 #define REQ_SORTED      (1 << __REQ_SORTED)
 197 #define REQ_SOFTBARRIER (1 << __REQ_SOFTBARRIER)
 198 #define REQ_HARDBARRIER (1 << __REQ_HARDBARRIER)
 199 #define REQ_FUA         (1 << __REQ_FUA)
 200 #define REQ_NOMERGE     (1 << __REQ_NOMERGE)
 201 #define REQ_STARTED     (1 << __REQ_STARTED)
 202 #define REQ_DONTPREP    (1 << __REQ_DONTPREP)
 203 #define REQ_QUEUED      (1 << __REQ_QUEUED)
 204 #define REQ_ELVPRIV     (1 << __REQ_ELVPRIV)
 205 #define REQ_FAILED      (1 << __REQ_FAILED)
 206 #define REQ_QUIET       (1 << __REQ_QUIET)
 207 #define REQ_PREEMPT     (1 << __REQ_PREEMPT)
 208 #define REQ_ORDERED_COLOR       (1 << __REQ_ORDERED_COLOR)
 209 #define REQ_RW_SYNC     (1 << __REQ_RW_SYNC)
 210 #define REQ_ALLOCED     (1 << __REQ_ALLOCED)
 211 #define REQ_RW_META     (1 << __REQ_RW_META)
 212
 213 #define BLK_MAX_CDB     16
 214
 215 /*
 216  * try to put the fields that are referenced together in the same cacheline
 217  */
 218 struct request {
 219         struct list_head queuelist;
 220         struct list_head donelist;
 221
 222         struct request_queue *q;
 223
 224         unsigned int cmd_flags;
 225         enum rq_cmd_type_bits cmd_type;
 226
 227         /* Maintain bio traversal state for part by part I/O submission.
 228          * hard_* are block layer internals, no driver should touch them!
 229          */
 230
 231         sector_t sector;                /* next sector to submit */
 232         sector_t hard_sector;           /* next sector to complete */
 233         unsigned long nr_sectors;       /* no. of sectors left to submit */
 234         unsigned long hard_nr_sectors;  /* no. of sectors left to complete */
 235         /* no. of sectors left to submit in the current segment */
 236         unsigned int current_nr_sectors;
 237
 238         /* no. of sectors left to complete in the current segment */
 239         unsigned int hard_cur_sectors;
 240
 241         struct bio *bio;
 242         struct bio *biotail;
 243
 244         struct hlist_node hash; /* merge hash */
 245         /*
 246          * The rb_node is only used inside the io scheduler, requests
 247          * are pruned when moved to the dispatch queue. So let the
 248          * completion_data share space with the rb_node.
 249          */
 250         union {
 251                 struct rb_node rb_node; /* sort/lookup */
 252                 void *completion_data;
 253         };
 254
 255         /*
 256          * two pointers are available for the IO schedulers, if they need
 257          * more they have to dynamically allocate it.
 258          */
 259         void *elevator_private;
 260         void *elevator_private2;
 261
 262         struct gendisk *rq_disk;
 263         unsigned long start_time;
 264
 265         /* Number of scatter-gather DMA addr+len pairs after
 266          * physical address coalescing is performed.
 267          */
 268         unsigned short nr_phys_segments;
 269
 270         /* Number of scatter-gather addr+len pairs after
 271          * physical and DMA remapping hardware coalescing is performed.
 272          * This is the number of scatter-gather entries the driver
 273          * will actually have to deal with after DMA mapping is done.
 274          */
 275         unsigned short nr_hw_segments;
 276
 277         unsigned short ioprio;
 278
 279         void *special;
 280         char *buffer;
 281
 282         int tag;
 283         int errors;
 284
 285         int ref_count;
 286
 287         /*
 288          * when request is used as a packet command carrier
 289          */
 290         unsigned int cmd_len;
 291         unsigned char cmd[BLK_MAX_CDB];
 292
 293         unsigned int data_len;
 294         unsigned int sense_len;
 295         void *data;
 296         void *sense;
 297
 298         unsigned int timeout;
 299         int retries;
 300
 301         /*
 302          * completion callback.
 303          */
 304         rq_end_io_fn *end_io;
 305         void *end_io_data;
 306
 307         /* for bidi */
 308         struct request *next_rq;
 309 };
 310
 311 /*
 312  * State information carried for REQ_TYPE_PM_SUSPEND and REQ_TYPE_PM_RESUME
 313  * requests. Some step values could eventually be made generic.
 314  */
 315 struct request_pm_state
 316 {
 317         /* PM state machine step value, currently driver specific */
 318         int     pm_step;
 319         /* requested PM state value (S1, S2, S3, S4, ...) */
 320         u32     pm_state;
 321         void*   data;           /* for driver use */
 322 };
 323
 324 #include <linux/elevator.h>
 325
 326 typedef void (request_fn_proc) (struct request_queue *q);
 327 typedef int (make_request_fn) (struct request_queue *q, struct bio *bio);
 328 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
 329 typedef void (unplug_fn) (struct request_queue *);
 330
 331 struct bio_vec;
 332 typedef int (merge_bvec_fn) (struct request_queue *, struct bio *, struct bio_vec *);
 333 typedef void (prepare_flush_fn) (struct request_queue *, struct request *);
 334 typedef void (softirq_done_fn)(struct request *);
 335
 336 enum blk_queue_state {
 337         Queue_down,
 338         Queue_up,
 339 };
 340
 341 struct blk_queue_tag {
 342         struct request **tag_index;     /* map of busy tags */
 343         unsigned long *tag_map;         /* bit map of free/busy tags */
 344         int busy;                       /* current depth */
 345         int max_depth;                  /* what we will send to device */
 346         int real_max_depth;             /* what the array can hold */
 347         atomic_t refcnt;                /* map can be shared */
 348 };
 349
 350 struct request_queue
 351 {
 352         /*
 353          * Together with queue_head for cacheline sharing
 354          */
 355         struct list_head        queue_head;
 356         struct request          *last_merge;
 357         elevator_t              *elevator;
 358
 359         /*
 360          * the queue request freelist, one for reads and one for writes
 361          */
 362         struct request_list     rq;
 363
 364         request_fn_proc         *request_fn;
 365         make_request_fn         *make_request_fn;
 366         prep_rq_fn              *prep_rq_fn;
 367         unplug_fn               *unplug_fn;
 368         merge_bvec_fn           *merge_bvec_fn;
 369         prepare_flush_fn        *prepare_flush_fn;
 370         softirq_done_fn         *softirq_done_fn;
 371
 372         /*
 373          * Dispatch queue sorting
 374          */
 375         sector_t                end_sector;
 376         struct request          *boundary_rq;
 377
 378         /*
 379          * Auto-unplugging state
 380          */
 381         struct timer_list       unplug_timer;
 382         int                     unplug_thresh;  /* After this many requests */
 383         unsigned long           unplug_delay;   /* After this many jiffies */
 384         struct work_struct      unplug_work;
 385
 386         struct backing_dev_info backing_dev_info;
 387
 388         /*
 389          * The queue owner gets to use this for whatever they like.
 390          * ll_rw_blk doesn't touch it.
 391          */
 392         void                    *queuedata;
 393
 394         /*
 395          * queue needs bounce pages for pages above this limit
 396          */
 397         unsigned long           bounce_pfn;
 398         gfp_t                   bounce_gfp;
 399
 400         /*
 401          * various queue flags, see QUEUE_* below
 402          */
 403         unsigned long           queue_flags;
 404
 405         /*
 406          * protects queue structures from reentrancy. ->__queue_lock should
 407          * _never_ be used directly, it is queue private. always use
 408          * ->queue_lock.
 409          */
 410         spinlock_t              __queue_lock;
 411         spinlock_t              *queue_lock;
 412
 413         /*
 414          * queue kobject
 415          */
 416         struct kobject kobj;
 417
 418         /*
 419          * queue settings
 420          */
 421         unsigned long           nr_requests;    /* Max # of requests */
 422         unsigned int            nr_congestion_on;
 423         unsigned int            nr_congestion_off;
 424         unsigned int            nr_batching;
 425
 426         unsigned int            max_sectors;
 427         unsigned int            max_hw_sectors;
 428         unsigned short          max_phys_segments;
 429         unsigned short          max_hw_segments;
 430         unsigned short          hardsect_size;
 431         unsigned int            max_segment_size;
 432
 433         unsigned long           seg_boundary_mask;
 434         unsigned int            dma_alignment;
 435
 436         struct blk_queue_tag    *queue_tags;
 437         struct list_head        tag_busy_list;
 438
 439         unsigned int            nr_sorted;
 440         unsigned int            in_flight;
 441
 442         /*
 443          * sg stuff
 444          */
 445         unsigned int            sg_timeout;
 446         unsigned int            sg_reserved_size;
 447         int                     node;
 448 #ifdef CONFIG_BLK_DEV_IO_TRACE
 449         struct blk_trace        *blk_trace;
 450 #endif
 451         /*
 452          * reserved for flush operations
 453          */
 454         unsigned int            ordered, next_ordered, ordseq;
 455         int                     orderr, ordcolor;
 456         struct request          pre_flush_rq, bar_rq, post_flush_rq;
 457         struct request          *orig_bar_rq;
 458
 459         struct mutex            sysfs_lock;
 460
 461 #if defined(CONFIG_BLK_DEV_BSG)
 462         struct bsg_class_device bsg_dev;
 463 #endif
 464 };
 465
 466 #define QUEUE_FLAG_CLUSTER      0       /* cluster several segments into 1 */
 467 #define QUEUE_FLAG_QUEUED       1       /* uses generic tag queueing */
 468 #define QUEUE_FLAG_STOPPED      2       /* queue is stopped */
 469 #define QUEUE_FLAG_READFULL     3       /* read queue has been filled */
 470 #define QUEUE_FLAG_WRITEFULL    4       /* write queue has been filled */
 471 #define QUEUE_FLAG_DEAD         5       /* queue being torn down */
 472 #define QUEUE_FLAG_REENTER      6       /* Re-entrancy avoidance */
 473 #define QUEUE_FLAG_PLUGGED      7       /* queue is plugged */
 474 #define QUEUE_FLAG_ELVSWITCH    8       /* don't use elevator, just do FIFO */
 475 #define QUEUE_FLAG_BIDI         9       /* queue supports bidi requests */
 476
 477 enum {
 478         /*
 479          * Hardbarrier is supported with one of the following methods.
 480          *
 481          * NONE         : hardbarrier unsupported
 482          * DRAIN        : ordering by draining is enough
 483          * DRAIN_FLUSH  : ordering by draining w/ pre and post flushes
 484          * DRAIN_FUA    : ordering by draining w/ pre flush and FUA write
 485          * TAG          : ordering by tag is enough
 486          * TAG_FLUSH    : ordering by tag w/ pre and post flushes
 487          * TAG_FUA      : ordering by tag w/ pre flush and FUA write
 488          */
 489         QUEUE_ORDERED_NONE      = 0x00,
 490         QUEUE_ORDERED_DRAIN     = 0x01,
 491         QUEUE_ORDERED_TAG       = 0x02,
 492
 493         QUEUE_ORDERED_PREFLUSH  = 0x10,
 494         QUEUE_ORDERED_POSTFLUSH = 0x20,
 495         QUEUE_ORDERED_FUA       = 0x40,
 496
 497         QUEUE_ORDERED_DRAIN_FLUSH = QUEUE_ORDERED_DRAIN |
 498                         QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_POSTFLUSH,
 499         QUEUE_ORDERED_DRAIN_FUA = QUEUE_ORDERED_DRAIN |
 500                         QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_FUA,
 501         QUEUE_ORDERED_TAG_FLUSH = QUEUE_ORDERED_TAG |
 502                         QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_POSTFLUSH,
 503         QUEUE_ORDERED_TAG_FUA   = QUEUE_ORDERED_TAG |
 504                         QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_FUA,
 505
 506         /*
 507          * Ordered operation sequence
 508          */
 509         QUEUE_ORDSEQ_STARTED    = 0x01, /* flushing in progress */
 510         QUEUE_ORDSEQ_DRAIN      = 0x02, /* waiting for the queue to be drained */
 511         QUEUE_ORDSEQ_PREFLUSH   = 0x04, /* pre-flushing in progress */
 512         QUEUE_ORDSEQ_BAR        = 0x08, /* original barrier req in progress */
 513         QUEUE_ORDSEQ_POSTFLUSH  = 0x10, /* post-flushing in progress */
 514         QUEUE_ORDSEQ_DONE       = 0x20,
 515 };
 516
 517 #define blk_queue_plugged(q)    test_bit(QUEUE_FLAG_PLUGGED, &(q)->queue_flags)
 518 #define blk_queue_tagged(q)     test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
 519 #define blk_queue_stopped(q)    test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
 520 #define blk_queue_flushing(q)   ((q)->ordseq)
 521
 522 #define blk_fs_request(rq)      ((rq)->cmd_type == REQ_TYPE_FS)
 523 #define blk_pc_request(rq)      ((rq)->cmd_type == REQ_TYPE_BLOCK_PC)
 524 #define blk_special_request(rq) ((rq)->cmd_type == REQ_TYPE_SPECIAL)
 525 #define blk_sense_request(rq)   ((rq)->cmd_type == REQ_TYPE_SENSE)
 526
 527 #define blk_noretry_request(rq) ((rq)->cmd_flags & REQ_FAILFAST)
 528 #define blk_rq_started(rq)      ((rq)->cmd_flags & REQ_STARTED)
 529
 530 #define blk_account_rq(rq)      (blk_rq_started(rq) && blk_fs_request(rq))
 531
 532 #define blk_pm_suspend_request(rq)      ((rq)->cmd_type == REQ_TYPE_PM_SUSPEND)
 533 #define blk_pm_resume_request(rq)       ((rq)->cmd_type == REQ_TYPE_PM_RESUME)
 534 #define blk_pm_request(rq)      \
 535         (blk_pm_suspend_request(rq) || blk_pm_resume_request(rq))
 536
 537 #define blk_sorted_rq(rq)       ((rq)->cmd_flags & REQ_SORTED)
 538 #define blk_barrier_rq(rq)      ((rq)->cmd_flags & REQ_HARDBARRIER)
 539 #define blk_fua_rq(rq)          ((rq)->cmd_flags & REQ_FUA)
 540 #define blk_bidi_rq(rq)         ((rq)->next_rq != NULL)
 541 #define blk_empty_barrier(rq)   (blk_barrier_rq(rq) && blk_fs_request(rq) && !(rq)->hard_nr_sectors)
 542
 543 #define list_entry_rq(ptr)      list_entry((ptr), struct request, queuelist)
 544
 545 #define rq_data_dir(rq)         ((rq)->cmd_flags & 1)
 546
 547 /*
 548  * We regard a request as sync, if it's a READ or a SYNC write.
 549  */
 550 #define rq_is_sync(rq)          (rq_data_dir((rq)) == READ || (rq)->cmd_flags & REQ_RW_SYNC)
 551 #define rq_is_meta(rq)          ((rq)->cmd_flags & REQ_RW_META)
 552
 553 static inline int blk_queue_full(struct request_queue *q, int rw)
 554 {
 555         if (rw == READ)
 556                 return test_bit(QUEUE_FLAG_READFULL, &q->queue_flags);
 557         return test_bit(QUEUE_FLAG_WRITEFULL, &q->queue_flags);
 558 }
 559
 560 static inline void blk_set_queue_full(struct request_queue *q, int rw)
 561 {
 562         if (rw == READ)
 563                 set_bit(QUEUE_FLAG_READFULL, &q->queue_flags);
 564         else
 565                 set_bit(QUEUE_FLAG_WRITEFULL, &q->queue_flags);
 566 }
 567
 568 static inline void blk_clear_queue_full(struct request_queue *q, int rw)
 569 {
 570         if (rw == READ)
 571                 clear_bit(QUEUE_FLAG_READFULL, &q->queue_flags);
 572         else
 573                 clear_bit(QUEUE_FLAG_WRITEFULL, &q->queue_flags);
 574 }
 575
 576
 577 /*
 578  * mergeable request must not have _NOMERGE or _BARRIER bit set, nor may
 579  * it already be started by driver.
 580  */
 581 #define RQ_NOMERGE_FLAGS        \
 582         (REQ_NOMERGE | REQ_STARTED | REQ_HARDBARRIER | REQ_SOFTBARRIER)
 583 #define rq_mergeable(rq)        \
 584         (!((rq)->cmd_flags & RQ_NOMERGE_FLAGS) && blk_fs_request((rq)))
 585
 586 /*
 587  * q->prep_rq_fn return values
 588  */
 589 #define BLKPREP_OK              0       /* serve it */
 590 #define BLKPREP_KILL            1       /* fatal error, kill */
 591 #define BLKPREP_DEFER           2       /* leave on queue */
 592
 593 extern unsigned long blk_max_low_pfn, blk_max_pfn;
 594
 595 /*
 596  * standard bounce addresses:
 597  *
 598  * BLK_BOUNCE_HIGH      : bounce all highmem pages
 599  * BLK_BOUNCE_ANY       : don't bounce anything
 600  * BLK_BOUNCE_ISA       : bounce pages above ISA DMA boundary
 601  */
 602 #define BLK_BOUNCE_HIGH         ((u64)blk_max_low_pfn << PAGE_SHIFT)
 603 #define BLK_BOUNCE_ANY          ((u64)blk_max_pfn << PAGE_SHIFT)
 604 #define BLK_BOUNCE_ISA          (ISA_DMA_THRESHOLD)
 605
 606 /*
 607  * default timeout for SG_IO if none specified
 608  */
 609 #define BLK_DEFAULT_SG_TIMEOUT  (60 * HZ)
 610
 611 #ifdef CONFIG_BOUNCE
 612 extern int init_emergency_isa_pool(void);
 613 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
 614 #else
 615 static inline int init_emergency_isa_pool(void)
 616 {
 617         return 0;
 618 }
 619 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
 620 {
 621 }
 622 #endif /* CONFIG_MMU */
 623
 624 struct req_iterator {
 625         int i;
 626         struct bio *bio;
 627 };
 628
 629 /* This should not be used directly - use rq_for_each_segment */
 630 #define __rq_for_each_bio(_bio, rq)     \
 631         if ((rq->bio))                  \
 632                 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
 633
 634 #define rq_for_each_segment(bvl, _rq, _iter)                    \
 635         __rq_for_each_bio(_iter.bio, _rq)                       \
 636                 bio_for_each_segment(bvl, _iter.bio, _iter.i)
 637
 638 #define rq_iter_last(rq, _iter)                                 \
 639                 (_iter.bio->bi_next == NULL && _iter.i == _iter.bio->bi_vcnt-1)
 640
 641 extern int blk_register_queue(struct gendisk *disk);
 642 extern void blk_unregister_queue(struct gendisk *disk);
 643 extern void register_disk(struct gendisk *dev);
 644 extern void generic_make_request(struct bio *bio);
 645 extern void blk_put_request(struct request *);
 646 extern void __blk_put_request(struct request_queue *, struct request *);
 647 extern void blk_end_sync_rq(struct request *rq, int error);
 648 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
 649 extern void blk_insert_request(struct request_queue *, struct request *, int, void *);
 650 extern void blk_requeue_request(struct request_queue *, struct request *);
 651 extern void blk_plug_device(struct request_queue *);
 652 extern int blk_remove_plug(struct request_queue *);
 653 extern void blk_recount_segments(struct request_queue *, struct bio *);
 654 extern int scsi_cmd_ioctl(struct file *, struct request_queue *,
 655                           struct gendisk *, unsigned int, void __user *);
 656 extern int sg_scsi_ioctl(struct file *, struct request_queue *,
 657                 struct gendisk *, struct scsi_ioctl_command __user *);
 658
 659 /*
 660  * Temporary export, until SCSI gets fixed up.
 661  */
 662 extern int blk_rq_append_bio(struct request_queue *q, struct request *rq,
 663                              struct bio *bio);
 664
 665 /*
 666  * A queue has just exitted congestion.  Note this in the global counter of
 667  * congested queues, and wake up anyone who was waiting for requests to be
 668  * put back.
 669  */
 670 static inline void blk_clear_queue_congested(struct request_queue *q, int rw)
 671 {
 672         clear_bdi_congested(&q->backing_dev_info, rw);
 673 }
 674
 675 /*
 676  * A queue has just entered congestion.  Flag that in the queue's VM-visible
 677  * state flags and increment the global gounter of congested queues.
 678  */
 679 static inline void blk_set_queue_congested(struct request_queue *q, int rw)
 680 {
 681         set_bdi_congested(&q->backing_dev_info, rw);
 682 }
 683
 684 extern void blk_start_queue(struct request_queue *q);
 685 extern void blk_stop_queue(struct request_queue *q);
 686 extern void blk_sync_queue(struct request_queue *q);
 687 extern void __blk_stop_queue(struct request_queue *q);
 688 extern void blk_run_queue(struct request_queue *);
 689 extern void blk_start_queueing(struct request_queue *);
 690 extern int blk_rq_map_user(struct request_queue *, struct request *, void __user *, unsigned long);
 691 extern int blk_rq_unmap_user(struct bio *);
 692 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
 693 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
 694                                struct sg_iovec *, int, unsigned int);
 695 extern int blk_execute_rq(struct request_queue *, struct gendisk *,
 696                           struct request *, int);
 697 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
 698                                   struct request *, int, rq_end_io_fn *);
 699 extern int blk_verify_command(unsigned char *, int);
 700 extern void blk_unplug(struct request_queue *q);
 701
 702 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
 703 {
 704         return bdev->bd_disk->queue;
 705 }
 706
 707 static inline void blk_run_backing_dev(struct backing_dev_info *bdi,
 708                                        struct page *page)
 709 {
 710         if (bdi && bdi->unplug_io_fn)
 711                 bdi->unplug_io_fn(bdi, page);
 712 }
 713
 714 static inline void blk_run_address_space(struct address_space *mapping)
 715 {
 716         if (mapping)
 717                 blk_run_backing_dev(mapping->backing_dev_info, NULL);
 718 }
 719
 720 /*
 721  * end_request() and friends. Must be called with the request queue spinlock
 722  * acquired. All functions called within end_request() _must_be_ atomic.
 723  *
 724  * Several drivers define their own end_request and call
 725  * end_that_request_first() and end_that_request_last()
 726  * for parts of the original function. This prevents
 727  * code duplication in drivers.
 728  */
 729 extern int end_that_request_first(struct request *, int, int);
 730 extern int end_that_request_chunk(struct request *, int, int);
 731 extern void end_that_request_last(struct request *, int);
 732 extern void end_request(struct request *, int);
 733 extern void end_queued_request(struct request *, int);
 734 extern void end_dequeued_request(struct request *, int);
 735 extern void blk_complete_request(struct request *);
 736
 737 /*
 738  * end_that_request_first/chunk() takes an uptodate argument. we account
 739  * any value <= as an io error. 0 means -EIO for compatability reasons,
 740  * any other < 0 value is the direct error type. An uptodate value of
 741  * 1 indicates successful io completion
 742  */
 743 #define end_io_error(uptodate)  (unlikely((uptodate) <= 0))
 744
 745 static inline void blkdev_dequeue_request(struct request *req)
 746 {
 747         elv_dequeue_request(req->q, req);
 748 }
 749
 750 /*
 751  * Access functions for manipulating queue properties
 752  */
 753 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
 754                                         spinlock_t *lock, int node_id);
 755 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
 756 extern void blk_cleanup_queue(struct request_queue *);
 757 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
 758 extern void blk_queue_bounce_limit(struct request_queue *, u64);
 759 extern void blk_queue_max_sectors(struct request_queue *, unsigned int);
 760 extern void blk_queue_max_phys_segments(struct request_queue *, unsigned short);
 761 extern void blk_queue_max_hw_segments(struct request_queue *, unsigned short);
 762 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
 763 extern void blk_queue_hardsect_size(struct request_queue *, unsigned short);
 764 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
 765 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
 766 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
 767 extern void blk_queue_merge_bvec(struct request_queue *, merge_bvec_fn *);
 768 extern void blk_queue_dma_alignment(struct request_queue *, int);
 769 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
 770 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
 771 extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
 772 extern int blk_queue_ordered(struct request_queue *, unsigned, prepare_flush_fn *);
 773 extern int blk_do_ordered(struct request_queue *, struct request **);
 774 extern unsigned blk_ordered_cur_seq(struct request_queue *);
 775 extern unsigned blk_ordered_req_seq(struct request *);
 776 extern void blk_ordered_complete_seq(struct request_queue *, unsigned, int);
 777
 778 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
 779 extern void blk_dump_rq_flags(struct request *, char *);
 780 extern void generic_unplug_device(struct request_queue *);
 781 extern void __generic_unplug_device(struct request_queue *);
 782 extern long nr_blockdev_pages(void);
 783
 784 int blk_get_queue(struct request_queue *);
 785 struct request_queue *blk_alloc_queue(gfp_t);
 786 struct request_queue *blk_alloc_queue_node(gfp_t, int);
 787 extern void blk_put_queue(struct request_queue *);
 788
 789 /*
 790  * tag stuff
 791  */
 792 #define blk_queue_tag_depth(q)          ((q)->queue_tags->busy)
 793 #define blk_queue_tag_queue(q)          ((q)->queue_tags->busy < (q)->queue_tags->max_depth)
 794 #define blk_rq_tagged(rq)               ((rq)->cmd_flags & REQ_QUEUED)
 795 extern int blk_queue_start_tag(struct request_queue *, struct request *);
 796 extern struct request *blk_queue_find_tag(struct request_queue *, int);
 797 extern void blk_queue_end_tag(struct request_queue *, struct request *);
 798 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *);
 799 extern void blk_queue_free_tags(struct request_queue *);
 800 extern int blk_queue_resize_tags(struct request_queue *, int);
 801 extern void blk_queue_invalidate_tags(struct request_queue *);
 802 extern struct blk_queue_tag *blk_init_tags(int);
 803 extern void blk_free_tags(struct blk_queue_tag *);
 804
 805 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
 806                                                 int tag)
 807 {
 808         if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
 809                 return NULL;
 810         return bqt->tag_index[tag];
 811 }
 812
 813 extern int blkdev_issue_flush(struct block_device *, sector_t *);
 814
 815 #define MAX_PHYS_SEGMENTS 128
 816 #define MAX_HW_SEGMENTS 128
 817 #define SAFE_MAX_SECTORS 255
 818 #define BLK_DEF_MAX_SECTORS 1024
 819
 820 #define MAX_SEGMENT_SIZE        65536
 821
 822 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
 823
 824 static inline int queue_hardsect_size(struct request_queue *q)
 825 {
 826         int retval = 512;
 827
 828         if (q && q->hardsect_size)
 829                 retval = q->hardsect_size;
 830
 831         return retval;
 832 }
 833
 834 static inline int bdev_hardsect_size(struct block_device *bdev)
 835 {
 836         return queue_hardsect_size(bdev_get_queue(bdev));
 837 }
 838
 839 static inline int queue_dma_alignment(struct request_queue *q)
 840 {
 841         int retval = 511;
 842
 843         if (q && q->dma_alignment)
 844                 retval = q->dma_alignment;
 845
 846         return retval;
 847 }
 848
 849 /* assumes size > 256 */
 850 static inline unsigned int blksize_bits(unsigned int size)
 851 {
 852         unsigned int bits = 8;
 853         do {
 854                 bits++;
 855                 size >>= 1;
 856         } while (size > 256);
 857         return bits;
 858 }
 859
 860 static inline unsigned int block_size(struct block_device *bdev)
 861 {
 862         return bdev->bd_block_size;
 863 }
 864
 865 typedef struct {struct page *v;} Sector;
 866
 867 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
 868
 869 static inline void put_dev_sector(Sector p)
 870 {
 871         page_cache_release(p.v);
 872 }
 873
 874 struct work_struct;
 875 int kblockd_schedule_work(struct work_struct *work);
 876 void kblockd_flush_work(struct work_struct *work);
 877
 878 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
 879         MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
 880 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
 881         MODULE_ALIAS("block-major-" __stringify(major) "-*")
 882
 883
 884 #else /* CONFIG_BLOCK */
 885 /*
 886  * stubs for when the block layer is configured out
 887  */
 888 #define buffer_heads_over_limit 0
 889
 890 static inline long nr_blockdev_pages(void)
 891 {
 892         return 0;
 893 }
 894
 895 static inline void exit_io_context(void)
 896 {
 897 }
 898
 899 #endif /* CONFIG_BLOCK */
 900
 901 #endif