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 struct elevator_queue;
  27 typedef struct elevator_queue elevator_t;
  28 struct request_pm_state;
  29 struct blk_trace;
  30 struct request;
  31 struct sg_io_hdr;
  32
  33 #define BLKDEV_MIN_RQ   4
  34 #define BLKDEV_MAX_RQ   128     /* Default maximum */
  35
  36 int put_io_context(struct io_context *ioc);
  37 void exit_io_context(void);
  38 struct io_context *get_io_context(gfp_t gfp_flags, int node);
  39 struct io_context *alloc_io_context(gfp_t gfp_flags, int node);
  40 void copy_io_context(struct io_context **pdst, struct io_context **psrc);
  41
  42 struct request;
  43 typedef void (rq_end_io_fn)(struct request *, int);
  44
  45 struct request_list {
  46         int count[2];
  47         int starved[2];
  48         int elvpriv;
  49         mempool_t *rq_pool;
  50         wait_queue_head_t wait[2];
  51 };
  52
  53 /*
  54  * request command types
  55  */
  56 enum rq_cmd_type_bits {
  57         REQ_TYPE_FS             = 1,    /* fs request */
  58         REQ_TYPE_BLOCK_PC,              /* scsi command */
  59         REQ_TYPE_SENSE,                 /* sense request */
  60         REQ_TYPE_PM_SUSPEND,            /* suspend request */
  61         REQ_TYPE_PM_RESUME,             /* resume request */
  62         REQ_TYPE_PM_SHUTDOWN,           /* shutdown request */
  63         REQ_TYPE_FLUSH,                 /* flush request */
  64         REQ_TYPE_SPECIAL,               /* driver defined type */
  65         REQ_TYPE_LINUX_BLOCK,           /* generic block layer message */
  66         /*
  67          * for ATA/ATAPI devices. this really doesn't belong here, ide should
  68          * use REQ_TYPE_SPECIAL and use rq->cmd[0] with the range of driver
  69          * private REQ_LB opcodes to differentiate what type of request this is
  70          */
  71         REQ_TYPE_ATA_TASKFILE,
  72         REQ_TYPE_ATA_PC,
  73 };
  74
  75 /*
  76  * For request of type REQ_TYPE_LINUX_BLOCK, rq->cmd[0] is the opcode being
  77  * sent down (similar to how REQ_TYPE_BLOCK_PC means that ->cmd[] holds a
  78  * SCSI cdb.
  79  *
  80  * 0x00 -> 0x3f are driver private, to be used for whatever purpose they need,
  81  * typically to differentiate REQ_TYPE_SPECIAL requests.
  82  *
  83  */
  84 enum {
  85         /*
  86          * just examples for now
  87          */
  88         REQ_LB_OP_EJECT = 0x40,         /* eject request */
  89         REQ_LB_OP_FLUSH = 0x41,         /* flush device */
  90 };
  91
  92 /*
  93  * request type modified bits. first three bits match BIO_RW* bits, important
  94  */
  95 enum rq_flag_bits {
  96         __REQ_RW,               /* not set, read. set, write */
  97         __REQ_FAILFAST,         /* no low level driver retries */
  98         __REQ_SORTED,           /* elevator knows about this request */
  99         __REQ_SOFTBARRIER,      /* may not be passed by ioscheduler */
 100         __REQ_HARDBARRIER,      /* may not be passed by drive either */
 101         __REQ_FUA,              /* forced unit access */
 102         __REQ_NOMERGE,          /* don't touch this for merging */
 103         __REQ_STARTED,          /* drive already may have started this one */
 104         __REQ_DONTPREP,         /* don't call prep for this one */
 105         __REQ_QUEUED,           /* uses queueing */
 106         __REQ_ELVPRIV,          /* elevator private data attached */
 107         __REQ_FAILED,           /* set if the request failed */
 108         __REQ_QUIET,            /* don't worry about errors */
 109         __REQ_PREEMPT,          /* set for "ide_preempt" requests */
 110         __REQ_ORDERED_COLOR,    /* is before or after barrier */
 111         __REQ_RW_SYNC,          /* request is sync (O_DIRECT) */
 112         __REQ_ALLOCED,          /* request came from our alloc pool */
 113         __REQ_RW_META,          /* metadata io request */
 114         __REQ_COPY_USER,        /* contains copies of user pages */
 115         __REQ_INTEGRITY,        /* integrity metadata has been remapped */
 116         __REQ_NR_BITS,          /* stops here */
 117 };
 118
 119 #define REQ_RW          (1 << __REQ_RW)
 120 #define REQ_FAILFAST    (1 << __REQ_FAILFAST)
 121 #define REQ_SORTED      (1 << __REQ_SORTED)
 122 #define REQ_SOFTBARRIER (1 << __REQ_SOFTBARRIER)
 123 #define REQ_HARDBARRIER (1 << __REQ_HARDBARRIER)
 124 #define REQ_FUA         (1 << __REQ_FUA)
 125 #define REQ_NOMERGE     (1 << __REQ_NOMERGE)
 126 #define REQ_STARTED     (1 << __REQ_STARTED)
 127 #define REQ_DONTPREP    (1 << __REQ_DONTPREP)
 128 #define REQ_QUEUED      (1 << __REQ_QUEUED)
 129 #define REQ_ELVPRIV     (1 << __REQ_ELVPRIV)
 130 #define REQ_FAILED      (1 << __REQ_FAILED)
 131 #define REQ_QUIET       (1 << __REQ_QUIET)
 132 #define REQ_PREEMPT     (1 << __REQ_PREEMPT)
 133 #define REQ_ORDERED_COLOR       (1 << __REQ_ORDERED_COLOR)
 134 #define REQ_RW_SYNC     (1 << __REQ_RW_SYNC)
 135 #define REQ_ALLOCED     (1 << __REQ_ALLOCED)
 136 #define REQ_RW_META     (1 << __REQ_RW_META)
 137 #define REQ_COPY_USER   (1 << __REQ_COPY_USER)
 138 #define REQ_INTEGRITY   (1 << __REQ_INTEGRITY)
 139
 140 #define BLK_MAX_CDB     16
 141
 142 /*
 143  * try to put the fields that are referenced together in the same cacheline.
 144  * if you modify this structure, be sure to check block/blk-core.c:rq_init()
 145  * as well!
 146  */
 147 struct request {
 148         struct list_head queuelist;
 149         struct list_head donelist;
 150
 151         struct request_queue *q;
 152
 153         unsigned int cmd_flags;
 154         enum rq_cmd_type_bits cmd_type;
 155
 156         /* Maintain bio traversal state for part by part I/O submission.
 157          * hard_* are block layer internals, no driver should touch them!
 158          */
 159
 160         sector_t sector;                /* next sector to submit */
 161         sector_t hard_sector;           /* next sector to complete */
 162         unsigned long nr_sectors;       /* no. of sectors left to submit */
 163         unsigned long hard_nr_sectors;  /* no. of sectors left to complete */
 164         /* no. of sectors left to submit in the current segment */
 165         unsigned int current_nr_sectors;
 166
 167         /* no. of sectors left to complete in the current segment */
 168         unsigned int hard_cur_sectors;
 169
 170         struct bio *bio;
 171         struct bio *biotail;
 172
 173         struct hlist_node hash; /* merge hash */
 174         /*
 175          * The rb_node is only used inside the io scheduler, requests
 176          * are pruned when moved to the dispatch queue. So let the
 177          * completion_data share space with the rb_node.
 178          */
 179         union {
 180                 struct rb_node rb_node; /* sort/lookup */
 181                 void *completion_data;
 182         };
 183
 184         /*
 185          * two pointers are available for the IO schedulers, if they need
 186          * more they have to dynamically allocate it.
 187          */
 188         void *elevator_private;
 189         void *elevator_private2;
 190
 191         struct gendisk *rq_disk;
 192         unsigned long start_time;
 193
 194         /* Number of scatter-gather DMA addr+len pairs after
 195          * physical address coalescing is performed.
 196          */
 197         unsigned short nr_phys_segments;
 198
 199         /* Number of scatter-gather addr+len pairs after
 200          * physical and DMA remapping hardware coalescing is performed.
 201          * This is the number of scatter-gather entries the driver
 202          * will actually have to deal with after DMA mapping is done.
 203          */
 204         unsigned short nr_hw_segments;
 205
 206         unsigned short ioprio;
 207
 208         void *special;
 209         char *buffer;
 210
 211         int tag;
 212         int errors;
 213
 214         int ref_count;
 215
 216         /*
 217          * when request is used as a packet command carrier
 218          */
 219         unsigned short cmd_len;
 220         unsigned char __cmd[BLK_MAX_CDB];
 221         unsigned char *cmd;
 222
 223         unsigned int data_len;
 224         unsigned int extra_len; /* length of alignment and padding */
 225         unsigned int sense_len;
 226         void *data;
 227         void *sense;
 228
 229         unsigned int timeout;
 230         int retries;
 231
 232         /*
 233          * completion callback.
 234          */
 235         rq_end_io_fn *end_io;
 236         void *end_io_data;
 237
 238         /* for bidi */
 239         struct request *next_rq;
 240 };
 241
 242 /*
 243  * State information carried for REQ_TYPE_PM_SUSPEND and REQ_TYPE_PM_RESUME
 244  * requests. Some step values could eventually be made generic.
 245  */
 246 struct request_pm_state
 247 {
 248         /* PM state machine step value, currently driver specific */
 249         int     pm_step;
 250         /* requested PM state value (S1, S2, S3, S4, ...) */
 251         u32     pm_state;
 252         void*   data;           /* for driver use */
 253 };
 254
 255 #include <linux/elevator.h>
 256
 257 typedef void (request_fn_proc) (struct request_queue *q);
 258 typedef int (make_request_fn) (struct request_queue *q, struct bio *bio);
 259 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
 260 typedef void (unplug_fn) (struct request_queue *);
 261
 262 struct bio_vec;
 263 typedef int (merge_bvec_fn) (struct request_queue *, struct bio *, struct bio_vec *);
 264 typedef void (prepare_flush_fn) (struct request_queue *, struct request *);
 265 typedef void (softirq_done_fn)(struct request *);
 266 typedef int (dma_drain_needed_fn)(struct request *);
 267
 268 enum blk_queue_state {
 269         Queue_down,
 270         Queue_up,
 271 };
 272
 273 struct blk_queue_tag {
 274         struct request **tag_index;     /* map of busy tags */
 275         unsigned long *tag_map;         /* bit map of free/busy tags */
 276         int busy;                       /* current depth */
 277         int max_depth;                  /* what we will send to device */
 278         int real_max_depth;             /* what the array can hold */
 279         atomic_t refcnt;                /* map can be shared */
 280 };
 281
 282 struct request_queue
 283 {
 284         /*
 285          * Together with queue_head for cacheline sharing
 286          */
 287         struct list_head        queue_head;
 288         struct request          *last_merge;
 289         elevator_t              *elevator;
 290
 291         /*
 292          * the queue request freelist, one for reads and one for writes
 293          */
 294         struct request_list     rq;
 295
 296         request_fn_proc         *request_fn;
 297         make_request_fn         *make_request_fn;
 298         prep_rq_fn              *prep_rq_fn;
 299         unplug_fn               *unplug_fn;
 300         merge_bvec_fn           *merge_bvec_fn;
 301         prepare_flush_fn        *prepare_flush_fn;
 302         softirq_done_fn         *softirq_done_fn;
 303         dma_drain_needed_fn     *dma_drain_needed;
 304
 305         /*
 306          * Dispatch queue sorting
 307          */
 308         sector_t                end_sector;
 309         struct request          *boundary_rq;
 310
 311         /*
 312          * Auto-unplugging state
 313          */
 314         struct timer_list       unplug_timer;
 315         int                     unplug_thresh;  /* After this many requests */
 316         unsigned long           unplug_delay;   /* After this many jiffies */
 317         struct work_struct      unplug_work;
 318
 319         struct backing_dev_info backing_dev_info;
 320
 321         /*
 322          * The queue owner gets to use this for whatever they like.
 323          * ll_rw_blk doesn't touch it.
 324          */
 325         void                    *queuedata;
 326
 327         /*
 328          * queue needs bounce pages for pages above this limit
 329          */
 330         unsigned long           bounce_pfn;
 331         gfp_t                   bounce_gfp;
 332
 333         /*
 334          * various queue flags, see QUEUE_* below
 335          */
 336         unsigned long           queue_flags;
 337
 338         /*
 339          * protects queue structures from reentrancy. ->__queue_lock should
 340          * _never_ be used directly, it is queue private. always use
 341          * ->queue_lock.
 342          */
 343         spinlock_t              __queue_lock;
 344         spinlock_t              *queue_lock;
 345
 346         /*
 347          * queue kobject
 348          */
 349         struct kobject kobj;
 350
 351         /*
 352          * queue settings
 353          */
 354         unsigned long           nr_requests;    /* Max # of requests */
 355         unsigned int            nr_congestion_on;
 356         unsigned int            nr_congestion_off;
 357         unsigned int            nr_batching;
 358
 359         unsigned int            max_sectors;
 360         unsigned int            max_hw_sectors;
 361         unsigned short          max_phys_segments;
 362         unsigned short          max_hw_segments;
 363         unsigned short          hardsect_size;
 364         unsigned int            max_segment_size;
 365
 366         unsigned long           seg_boundary_mask;
 367         void                    *dma_drain_buffer;
 368         unsigned int            dma_drain_size;
 369         unsigned int            dma_pad_mask;
 370         unsigned int            dma_alignment;
 371
 372         struct blk_queue_tag    *queue_tags;
 373         struct list_head        tag_busy_list;
 374
 375         unsigned int            nr_sorted;
 376         unsigned int            in_flight;
 377
 378         /*
 379          * sg stuff
 380          */
 381         unsigned int            sg_timeout;
 382         unsigned int            sg_reserved_size;
 383         int                     node;
 384 #ifdef CONFIG_BLK_DEV_IO_TRACE
 385         struct blk_trace        *blk_trace;
 386 #endif
 387         /*
 388          * reserved for flush operations
 389          */
 390         unsigned int            ordered, next_ordered, ordseq;
 391         int                     orderr, ordcolor;
 392         struct request          pre_flush_rq, bar_rq, post_flush_rq;
 393         struct request          *orig_bar_rq;
 394
 395         struct mutex            sysfs_lock;
 396
 397 #if defined(CONFIG_BLK_DEV_BSG)
 398         struct bsg_class_device bsg_dev;
 399 #endif
 400 };
 401
 402 #define QUEUE_FLAG_CLUSTER      0       /* cluster several segments into 1 */
 403 #define QUEUE_FLAG_QUEUED       1       /* uses generic tag queueing */
 404 #define QUEUE_FLAG_STOPPED      2       /* queue is stopped */
 405 #define QUEUE_FLAG_READFULL     3       /* read queue has been filled */
 406 #define QUEUE_FLAG_WRITEFULL    4       /* write queue has been filled */
 407 #define QUEUE_FLAG_DEAD         5       /* queue being torn down */
 408 #define QUEUE_FLAG_REENTER      6       /* Re-entrancy avoidance */
 409 #define QUEUE_FLAG_PLUGGED      7       /* queue is plugged */
 410 #define QUEUE_FLAG_ELVSWITCH    8       /* don't use elevator, just do FIFO */
 411 #define QUEUE_FLAG_BIDI         9       /* queue supports bidi requests */
 412 #define QUEUE_FLAG_NOMERGES    10       /* disable merge attempts */
 413
 414 static inline int queue_is_locked(struct request_queue *q)
 415 {
 416 #ifdef CONFIG_SMP
 417         spinlock_t *lock = q->queue_lock;
 418         return lock && spin_is_locked(lock);
 419 #else
 420         return 1;
 421 #endif
 422 }
 423
 424 static inline void queue_flag_set_unlocked(unsigned int flag,
 425                                            struct request_queue *q)
 426 {
 427         __set_bit(flag, &q->queue_flags);
 428 }
 429
 430 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
 431 {
 432         WARN_ON_ONCE(!queue_is_locked(q));
 433         __set_bit(flag, &q->queue_flags);
 434 }
 435
 436 static inline void queue_flag_clear_unlocked(unsigned int flag,
 437                                              struct request_queue *q)
 438 {
 439         __clear_bit(flag, &q->queue_flags);
 440 }
 441
 442 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
 443 {
 444         WARN_ON_ONCE(!queue_is_locked(q));
 445         __clear_bit(flag, &q->queue_flags);
 446 }
 447
 448 enum {
 449         /*
 450          * Hardbarrier is supported with one of the following methods.
 451          *
 452          * NONE         : hardbarrier unsupported
 453          * DRAIN        : ordering by draining is enough
 454          * DRAIN_FLUSH  : ordering by draining w/ pre and post flushes
 455          * DRAIN_FUA    : ordering by draining w/ pre flush and FUA write
 456          * TAG          : ordering by tag is enough
 457          * TAG_FLUSH    : ordering by tag w/ pre and post flushes
 458          * TAG_FUA      : ordering by tag w/ pre flush and FUA write
 459          */
 460         QUEUE_ORDERED_NONE      = 0x00,
 461         QUEUE_ORDERED_DRAIN     = 0x01,
 462         QUEUE_ORDERED_TAG       = 0x02,
 463
 464         QUEUE_ORDERED_PREFLUSH  = 0x10,
 465         QUEUE_ORDERED_POSTFLUSH = 0x20,
 466         QUEUE_ORDERED_FUA       = 0x40,
 467
 468         QUEUE_ORDERED_DRAIN_FLUSH = QUEUE_ORDERED_DRAIN |
 469                         QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_POSTFLUSH,
 470         QUEUE_ORDERED_DRAIN_FUA = QUEUE_ORDERED_DRAIN |
 471                         QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_FUA,
 472         QUEUE_ORDERED_TAG_FLUSH = QUEUE_ORDERED_TAG |
 473                         QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_POSTFLUSH,
 474         QUEUE_ORDERED_TAG_FUA   = QUEUE_ORDERED_TAG |
 475                         QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_FUA,
 476
 477         /*
 478          * Ordered operation sequence
 479          */
 480         QUEUE_ORDSEQ_STARTED    = 0x01, /* flushing in progress */
 481         QUEUE_ORDSEQ_DRAIN      = 0x02, /* waiting for the queue to be drained */
 482         QUEUE_ORDSEQ_PREFLUSH   = 0x04, /* pre-flushing in progress */
 483         QUEUE_ORDSEQ_BAR        = 0x08, /* original barrier req in progress */
 484         QUEUE_ORDSEQ_POSTFLUSH  = 0x10, /* post-flushing in progress */
 485         QUEUE_ORDSEQ_DONE       = 0x20,
 486 };
 487
 488 #define blk_queue_plugged(q)    test_bit(QUEUE_FLAG_PLUGGED, &(q)->queue_flags)
 489 #define blk_queue_tagged(q)     test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
 490 #define blk_queue_stopped(q)    test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
 491 #define blk_queue_nomerges(q)   test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
 492 #define blk_queue_flushing(q)   ((q)->ordseq)
 493
 494 #define blk_fs_request(rq)      ((rq)->cmd_type == REQ_TYPE_FS)
 495 #define blk_pc_request(rq)      ((rq)->cmd_type == REQ_TYPE_BLOCK_PC)
 496 #define blk_special_request(rq) ((rq)->cmd_type == REQ_TYPE_SPECIAL)
 497 #define blk_sense_request(rq)   ((rq)->cmd_type == REQ_TYPE_SENSE)
 498
 499 #define blk_noretry_request(rq) ((rq)->cmd_flags & REQ_FAILFAST)
 500 #define blk_rq_started(rq)      ((rq)->cmd_flags & REQ_STARTED)
 501
 502 #define blk_account_rq(rq)      (blk_rq_started(rq) && blk_fs_request(rq))
 503
 504 #define blk_pm_suspend_request(rq)      ((rq)->cmd_type == REQ_TYPE_PM_SUSPEND)
 505 #define blk_pm_resume_request(rq)       ((rq)->cmd_type == REQ_TYPE_PM_RESUME)
 506 #define blk_pm_request(rq)      \
 507         (blk_pm_suspend_request(rq) || blk_pm_resume_request(rq))
 508
 509 #define blk_sorted_rq(rq)       ((rq)->cmd_flags & REQ_SORTED)
 510 #define blk_barrier_rq(rq)      ((rq)->cmd_flags & REQ_HARDBARRIER)
 511 #define blk_fua_rq(rq)          ((rq)->cmd_flags & REQ_FUA)
 512 #define blk_bidi_rq(rq)         ((rq)->next_rq != NULL)
 513 #define blk_empty_barrier(rq)   (blk_barrier_rq(rq) && blk_fs_request(rq) && !(rq)->hard_nr_sectors)
 514 /* rq->queuelist of dequeued request must be list_empty() */
 515 #define blk_queued_rq(rq)       (!list_empty(&(rq)->queuelist))
 516
 517 #define list_entry_rq(ptr)      list_entry((ptr), struct request, queuelist)
 518
 519 #define rq_data_dir(rq)         ((rq)->cmd_flags & 1)
 520
 521 /*
 522  * We regard a request as sync, if it's a READ or a SYNC write.
 523  */
 524 #define rq_is_sync(rq)          (rq_data_dir((rq)) == READ || (rq)->cmd_flags & REQ_RW_SYNC)
 525 #define rq_is_meta(rq)          ((rq)->cmd_flags & REQ_RW_META)
 526
 527 static inline int blk_queue_full(struct request_queue *q, int rw)
 528 {
 529         if (rw == READ)
 530                 return test_bit(QUEUE_FLAG_READFULL, &q->queue_flags);
 531         return test_bit(QUEUE_FLAG_WRITEFULL, &q->queue_flags);
 532 }
 533
 534 static inline void blk_set_queue_full(struct request_queue *q, int rw)
 535 {
 536         if (rw == READ)
 537                 queue_flag_set(QUEUE_FLAG_READFULL, q);
 538         else
 539                 queue_flag_set(QUEUE_FLAG_WRITEFULL, q);
 540 }
 541
 542 static inline void blk_clear_queue_full(struct request_queue *q, int rw)
 543 {
 544         if (rw == READ)
 545                 queue_flag_clear(QUEUE_FLAG_READFULL, q);
 546         else
 547                 queue_flag_clear(QUEUE_FLAG_WRITEFULL, q);
 548 }
 549
 550
 551 /*
 552  * mergeable request must not have _NOMERGE or _BARRIER bit set, nor may
 553  * it already be started by driver.
 554  */
 555 #define RQ_NOMERGE_FLAGS        \
 556         (REQ_NOMERGE | REQ_STARTED | REQ_HARDBARRIER | REQ_SOFTBARRIER)
 557 #define rq_mergeable(rq)        \
 558         (!((rq)->cmd_flags & RQ_NOMERGE_FLAGS) && blk_fs_request((rq)))
 559
 560 /*
 561  * q->prep_rq_fn return values
 562  */
 563 #define BLKPREP_OK              0       /* serve it */
 564 #define BLKPREP_KILL            1       /* fatal error, kill */
 565 #define BLKPREP_DEFER           2       /* leave on queue */
 566
 567 extern unsigned long blk_max_low_pfn, blk_max_pfn;
 568
 569 /*
 570  * standard bounce addresses:
 571  *
 572  * BLK_BOUNCE_HIGH      : bounce all highmem pages
 573  * BLK_BOUNCE_ANY       : don't bounce anything
 574  * BLK_BOUNCE_ISA       : bounce pages above ISA DMA boundary
 575  */
 576
 577 #if BITS_PER_LONG == 32
 578 #define BLK_BOUNCE_HIGH         ((u64)blk_max_low_pfn << PAGE_SHIFT)
 579 #else
 580 #define BLK_BOUNCE_HIGH         -1ULL
 581 #endif
 582 #define BLK_BOUNCE_ANY          (-1ULL)
 583 #define BLK_BOUNCE_ISA          (ISA_DMA_THRESHOLD)
 584
 585 /*
 586  * default timeout for SG_IO if none specified
 587  */
 588 #define BLK_DEFAULT_SG_TIMEOUT  (60 * HZ)
 589
 590 #ifdef CONFIG_BOUNCE
 591 extern int init_emergency_isa_pool(void);
 592 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
 593 #else
 594 static inline int init_emergency_isa_pool(void)
 595 {
 596         return 0;
 597 }
 598 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
 599 {
 600 }
 601 #endif /* CONFIG_MMU */
 602
 603 struct req_iterator {
 604         int i;
 605         struct bio *bio;
 606 };
 607
 608 /* This should not be used directly - use rq_for_each_segment */
 609 #define __rq_for_each_bio(_bio, rq)     \
 610         if ((rq->bio))                  \
 611                 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
 612
 613 #define rq_for_each_segment(bvl, _rq, _iter)                    \
 614         __rq_for_each_bio(_iter.bio, _rq)                       \
 615                 bio_for_each_segment(bvl, _iter.bio, _iter.i)
 616
 617 #define rq_iter_last(rq, _iter)                                 \
 618                 (_iter.bio->bi_next == NULL && _iter.i == _iter.bio->bi_vcnt-1)
 619
 620 extern int blk_register_queue(struct gendisk *disk);
 621 extern void blk_unregister_queue(struct gendisk *disk);
 622 extern void register_disk(struct gendisk *dev);
 623 extern void generic_make_request(struct bio *bio);
 624 extern void blk_rq_init(struct request_queue *q, struct request *rq);
 625 extern void blk_put_request(struct request *);
 626 extern void __blk_put_request(struct request_queue *, struct request *);
 627 extern void blk_end_sync_rq(struct request *rq, int error);
 628 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
 629 extern void blk_insert_request(struct request_queue *, struct request *, int, void *);
 630 extern void blk_requeue_request(struct request_queue *, struct request *);
 631 extern void blk_plug_device(struct request_queue *);
 632 extern int blk_remove_plug(struct request_queue *);
 633 extern void blk_recount_segments(struct request_queue *, struct bio *);
 634 extern int scsi_cmd_ioctl(struct file *, struct request_queue *,
 635                           struct gendisk *, unsigned int, void __user *);
 636 extern int sg_scsi_ioctl(struct file *, struct request_queue *,
 637                 struct gendisk *, struct scsi_ioctl_command __user *);
 638
 639 /*
 640  * Temporary export, until SCSI gets fixed up.
 641  */
 642 extern int blk_rq_append_bio(struct request_queue *q, struct request *rq,
 643                              struct bio *bio);
 644
 645 /*
 646  * A queue has just exitted congestion.  Note this in the global counter of
 647  * congested queues, and wake up anyone who was waiting for requests to be
 648  * put back.
 649  */
 650 static inline void blk_clear_queue_congested(struct request_queue *q, int rw)
 651 {
 652         clear_bdi_congested(&q->backing_dev_info, rw);
 653 }
 654
 655 /*
 656  * A queue has just entered congestion.  Flag that in the queue's VM-visible
 657  * state flags and increment the global gounter of congested queues.
 658  */
 659 static inline void blk_set_queue_congested(struct request_queue *q, int rw)
 660 {
 661         set_bdi_congested(&q->backing_dev_info, rw);
 662 }
 663
 664 extern void blk_start_queue(struct request_queue *q);
 665 extern void blk_stop_queue(struct request_queue *q);
 666 extern void blk_sync_queue(struct request_queue *q);
 667 extern void __blk_stop_queue(struct request_queue *q);
 668 extern void __blk_run_queue(struct request_queue *);
 669 extern void blk_run_queue(struct request_queue *);
 670 extern void blk_start_queueing(struct request_queue *);
 671 extern int blk_rq_map_user(struct request_queue *, struct request *, void __user *, unsigned long);
 672 extern int blk_rq_unmap_user(struct bio *);
 673 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
 674 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
 675                                struct sg_iovec *, int, unsigned int);
 676 extern int blk_execute_rq(struct request_queue *, struct gendisk *,
 677                           struct request *, int);
 678 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
 679                                   struct request *, int, rq_end_io_fn *);
 680 extern int blk_verify_command(unsigned char *, int);
 681 extern void blk_unplug(struct request_queue *q);
 682
 683 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
 684 {
 685         return bdev->bd_disk->queue;
 686 }
 687
 688 static inline void blk_run_backing_dev(struct backing_dev_info *bdi,
 689                                        struct page *page)
 690 {
 691         if (bdi && bdi->unplug_io_fn)
 692                 bdi->unplug_io_fn(bdi, page);
 693 }
 694
 695 static inline void blk_run_address_space(struct address_space *mapping)
 696 {
 697         if (mapping)
 698                 blk_run_backing_dev(mapping->backing_dev_info, NULL);
 699 }
 700
 701 /*
 702  * blk_end_request() and friends.
 703  * __blk_end_request() and end_request() must be called with
 704  * the request queue spinlock acquired.
 705  *
 706  * Several drivers define their own end_request and call
 707  * blk_end_request() for parts of the original function.
 708  * This prevents code duplication in drivers.
 709  */
 710 extern int blk_end_request(struct request *rq, int error,
 711                                 unsigned int nr_bytes);
 712 extern int __blk_end_request(struct request *rq, int error,
 713                                 unsigned int nr_bytes);
 714 extern int blk_end_bidi_request(struct request *rq, int error,
 715                                 unsigned int nr_bytes, unsigned int bidi_bytes);
 716 extern void end_request(struct request *, int);
 717 extern void end_queued_request(struct request *, int);
 718 extern void end_dequeued_request(struct request *, int);
 719 extern int blk_end_request_callback(struct request *rq, int error,
 720                                 unsigned int nr_bytes,
 721                                 int (drv_callback)(struct request *));
 722 extern void blk_complete_request(struct request *);
 723
 724 /*
 725  * blk_end_request() takes bytes instead of sectors as a complete size.
 726  * blk_rq_bytes() returns bytes left to complete in the entire request.
 727  * blk_rq_cur_bytes() returns bytes left to complete in the current segment.
 728  */
 729 extern unsigned int blk_rq_bytes(struct request *rq);
 730 extern unsigned int blk_rq_cur_bytes(struct request *rq);
 731
 732 static inline void blkdev_dequeue_request(struct request *req)
 733 {
 734         elv_dequeue_request(req->q, req);
 735 }
 736
 737 /*
 738  * Access functions for manipulating queue properties
 739  */
 740 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
 741                                         spinlock_t *lock, int node_id);
 742 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
 743 extern void blk_cleanup_queue(struct request_queue *);
 744 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
 745 extern void blk_queue_bounce_limit(struct request_queue *, u64);
 746 extern void blk_queue_max_sectors(struct request_queue *, unsigned int);
 747 extern void blk_queue_max_phys_segments(struct request_queue *, unsigned short);
 748 extern void blk_queue_max_hw_segments(struct request_queue *, unsigned short);
 749 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
 750 extern void blk_queue_hardsect_size(struct request_queue *, unsigned short);
 751 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
 752 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
 753 extern int blk_queue_dma_drain(struct request_queue *q,
 754                                dma_drain_needed_fn *dma_drain_needed,
 755                                void *buf, unsigned int size);
 756 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
 757 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
 758 extern void blk_queue_merge_bvec(struct request_queue *, merge_bvec_fn *);
 759 extern void blk_queue_dma_alignment(struct request_queue *, int);
 760 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
 761 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
 762 extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
 763 extern int blk_queue_ordered(struct request_queue *, unsigned, prepare_flush_fn *);
 764 extern int blk_do_ordered(struct request_queue *, struct request **);
 765 extern unsigned blk_ordered_cur_seq(struct request_queue *);
 766 extern unsigned blk_ordered_req_seq(struct request *);
 767 extern void blk_ordered_complete_seq(struct request_queue *, unsigned, int);
 768
 769 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
 770 extern void blk_dump_rq_flags(struct request *, char *);
 771 extern void generic_unplug_device(struct request_queue *);
 772 extern void __generic_unplug_device(struct request_queue *);
 773 extern long nr_blockdev_pages(void);
 774
 775 int blk_get_queue(struct request_queue *);
 776 struct request_queue *blk_alloc_queue(gfp_t);
 777 struct request_queue *blk_alloc_queue_node(gfp_t, int);
 778 extern void blk_put_queue(struct request_queue *);
 779
 780 /*
 781  * tag stuff
 782  */
 783 #define blk_queue_tag_depth(q)          ((q)->queue_tags->busy)
 784 #define blk_queue_tag_queue(q)          ((q)->queue_tags->busy < (q)->queue_tags->max_depth)
 785 #define blk_rq_tagged(rq)               ((rq)->cmd_flags & REQ_QUEUED)
 786 extern int blk_queue_start_tag(struct request_queue *, struct request *);
 787 extern struct request *blk_queue_find_tag(struct request_queue *, int);
 788 extern void blk_queue_end_tag(struct request_queue *, struct request *);
 789 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *);
 790 extern void blk_queue_free_tags(struct request_queue *);
 791 extern int blk_queue_resize_tags(struct request_queue *, int);
 792 extern void blk_queue_invalidate_tags(struct request_queue *);
 793 extern struct blk_queue_tag *blk_init_tags(int);
 794 extern void blk_free_tags(struct blk_queue_tag *);
 795
 796 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
 797                                                 int tag)
 798 {
 799         if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
 800                 return NULL;
 801         return bqt->tag_index[tag];
 802 }
 803
 804 extern int blkdev_issue_flush(struct block_device *, sector_t *);
 805
 806 #define MAX_PHYS_SEGMENTS 128
 807 #define MAX_HW_SEGMENTS 128
 808 #define SAFE_MAX_SECTORS 255
 809 #define BLK_DEF_MAX_SECTORS 1024
 810
 811 #define MAX_SEGMENT_SIZE        65536
 812
 813 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
 814
 815 static inline int queue_hardsect_size(struct request_queue *q)
 816 {
 817         int retval = 512;
 818
 819         if (q && q->hardsect_size)
 820                 retval = q->hardsect_size;
 821
 822         return retval;
 823 }
 824
 825 static inline int bdev_hardsect_size(struct block_device *bdev)
 826 {
 827         return queue_hardsect_size(bdev_get_queue(bdev));
 828 }
 829
 830 static inline int queue_dma_alignment(struct request_queue *q)
 831 {
 832         return q ? q->dma_alignment : 511;
 833 }
 834
 835 /* assumes size > 256 */
 836 static inline unsigned int blksize_bits(unsigned int size)
 837 {
 838         unsigned int bits = 8;
 839         do {
 840                 bits++;
 841                 size >>= 1;
 842         } while (size > 256);
 843         return bits;
 844 }
 845
 846 static inline unsigned int block_size(struct block_device *bdev)
 847 {
 848         return bdev->bd_block_size;
 849 }
 850
 851 typedef struct {struct page *v;} Sector;
 852
 853 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
 854
 855 static inline void put_dev_sector(Sector p)
 856 {
 857         page_cache_release(p.v);
 858 }
 859
 860 struct work_struct;
 861 int kblockd_schedule_work(struct work_struct *work);
 862 void kblockd_flush_work(struct work_struct *work);
 863
 864 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
 865         MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
 866 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
 867         MODULE_ALIAS("block-major-" __stringify(major) "-*")
 868
 869
 870 #if defined(CONFIG_BLK_DEV_INTEGRITY)
 871
 872 #define INTEGRITY_FLAG_READ     1       /* verify data integrity on read */
 873 #define INTEGRITY_FLAG_WRITE    2       /* generate data integrity on write */
 874
 875 struct blk_integrity_exchg {
 876         void                    *prot_buf;
 877         void                    *data_buf;
 878         sector_t                sector;
 879         unsigned int            data_size;
 880         unsigned short          sector_size;
 881         const char              *disk_name;
 882 };
 883
 884 typedef void (integrity_gen_fn) (struct blk_integrity_exchg *);
 885 typedef int (integrity_vrfy_fn) (struct blk_integrity_exchg *);
 886 typedef void (integrity_set_tag_fn) (void *, void *, unsigned int);
 887 typedef void (integrity_get_tag_fn) (void *, void *, unsigned int);
 888
 889 struct blk_integrity {
 890         integrity_gen_fn        *generate_fn;
 891         integrity_vrfy_fn       *verify_fn;
 892         integrity_set_tag_fn    *set_tag_fn;
 893         integrity_get_tag_fn    *get_tag_fn;
 894
 895         unsigned short          flags;
 896         unsigned short          tuple_size;
 897         unsigned short          sector_size;
 898         unsigned short          tag_size;
 899
 900         const char              *name;
 901
 902         struct kobject          kobj;
 903 };
 904
 905 extern int blk_integrity_register(struct gendisk *, struct blk_integrity *);
 906 extern void blk_integrity_unregister(struct gendisk *);
 907 extern int blk_integrity_compare(struct block_device *, struct block_device *);
 908 extern int blk_rq_map_integrity_sg(struct request *, struct scatterlist *);
 909 extern int blk_rq_count_integrity_sg(struct request *);
 910
 911 static inline unsigned short blk_integrity_tuple_size(struct blk_integrity *bi)
 912 {
 913         if (bi)
 914                 return bi->tuple_size;
 915
 916         return 0;
 917 }
 918
 919 static inline struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
 920 {
 921         return bdev->bd_disk->integrity;
 922 }
 923
 924 static inline unsigned int bdev_get_tag_size(struct block_device *bdev)
 925 {
 926         struct blk_integrity *bi = bdev_get_integrity(bdev);
 927
 928         if (bi)
 929                 return bi->tag_size;
 930
 931         return 0;
 932 }
 933
 934 static inline int bdev_integrity_enabled(struct block_device *bdev, int rw)
 935 {
 936         struct blk_integrity *bi = bdev_get_integrity(bdev);
 937
 938         if (bi == NULL)
 939                 return 0;
 940
 941         if (rw == READ && bi->verify_fn != NULL &&
 942             test_bit(INTEGRITY_FLAG_READ, &bi->flags))
 943                 return 1;
 944
 945         if (rw == WRITE && bi->generate_fn != NULL &&
 946             test_bit(INTEGRITY_FLAG_WRITE, &bi->flags))
 947                 return 1;
 948
 949         return 0;
 950 }
 951
 952 static inline int blk_integrity_rq(struct request *rq)
 953 {
 954         BUG_ON(rq->bio == NULL);
 955
 956         return bio_integrity(rq->bio);
 957 }
 958
 959 #else /* CONFIG_BLK_DEV_INTEGRITY */
 960
 961 #define blk_integrity_rq(rq)                    (0)
 962 #define blk_rq_count_integrity_sg(a)            (0)
 963 #define blk_rq_map_integrity_sg(a, b)           (0)
 964 #define bdev_get_integrity(a)                   (0)
 965 #define bdev_get_tag_size(a)                    (0)
 966 #define blk_integrity_compare(a, b)             (0)
 967 #define blk_integrity_register(a, b)            (0)
 968 #define blk_integrity_unregister(a)             do { } while (0);
 969
 970 #endif /* CONFIG_BLK_DEV_INTEGRITY */
 971
 972
 973 #else /* CONFIG_BLOCK */
 974 /*
 975  * stubs for when the block layer is configured out
 976  */
 977 #define buffer_heads_over_limit 0
 978
 979 static inline long nr_blockdev_pages(void)
 980 {
 981         return 0;
 982 }
 983
 984 static inline void exit_io_context(void)
 985 {
 986 }
 987
 988 struct io_context;
 989 static inline int put_io_context(struct io_context *ioc)
 990 {
 991         return 1;
 992 }
 993
 994
 995 #endif /* CONFIG_BLOCK */
 996
 997 #endif