include/block/block_int.h

   1 /*
   2  * QEMU System Emulator block driver
   3  *
   4  * Copyright (c) 2003 Fabrice Bellard
   5  *
   6  * Permission is hereby granted, free of charge, to any person obtaining a copy
   7  * of this software and associated documentation files (the "Software"), to deal
   8  * in the Software without restriction, including without limitation the rights
   9  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
  10  * copies of the Software, and to permit persons to whom the Software is
  11  * furnished to do so, subject to the following conditions:
  12  *
  13  * The above copyright notice and this permission notice shall be included in
  14  * all copies or substantial portions of the Software.
  15  *
  16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
  19  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  22  * THE SOFTWARE.
  23  */
  24 #ifndef BLOCK_INT_H
  25 #define BLOCK_INT_H
  26
  27 #include "block/accounting.h"
  28 #include "block/block.h"
  29 #include "block/throttle-groups.h"
  30 #include "qemu/option.h"
  31 #include "qemu/queue.h"
  32 #include "qemu/coroutine.h"
  33 #include "qemu/timer.h"
  34 #include "qapi-types.h"
  35 #include "qemu/hbitmap.h"
  36 #include "block/snapshot.h"
  37 #include "qemu/main-loop.h"
  38 #include "qemu/throttle.h"
  39
  40 #define BLOCK_FLAG_ENCRYPT          1
  41 #define BLOCK_FLAG_LAZY_REFCOUNTS   8
  42
  43 #define BLOCK_OPT_SIZE              "size"
  44 #define BLOCK_OPT_ENCRYPT           "encryption"
  45 #define BLOCK_OPT_COMPAT6           "compat6"
  46 #define BLOCK_OPT_HWVERSION         "hwversion"
  47 #define BLOCK_OPT_BACKING_FILE      "backing_file"
  48 #define BLOCK_OPT_BACKING_FMT       "backing_fmt"
  49 #define BLOCK_OPT_CLUSTER_SIZE      "cluster_size"
  50 #define BLOCK_OPT_TABLE_SIZE        "table_size"
  51 #define BLOCK_OPT_PREALLOC          "preallocation"
  52 #define BLOCK_OPT_SUBFMT            "subformat"
  53 #define BLOCK_OPT_COMPAT_LEVEL      "compat"
  54 #define BLOCK_OPT_LAZY_REFCOUNTS    "lazy_refcounts"
  55 #define BLOCK_OPT_ADAPTER_TYPE      "adapter_type"
  56 #define BLOCK_OPT_REDUNDANCY        "redundancy"
  57 #define BLOCK_OPT_NOCOW             "nocow"
  58 #define BLOCK_OPT_OBJECT_SIZE       "object_size"
  59 #define BLOCK_OPT_REFCOUNT_BITS     "refcount_bits"
  60
  61 #define BLOCK_PROBE_BUF_SIZE        512
  62
  63 enum BdrvTrackedRequestType {
  64     BDRV_TRACKED_READ,
  65     BDRV_TRACKED_WRITE,
  66     BDRV_TRACKED_FLUSH,
  67     BDRV_TRACKED_IOCTL,
  68     BDRV_TRACKED_DISCARD,
  69 };
  70
  71 typedef struct BdrvTrackedRequest {
  72     BlockDriverState *bs;
  73     int64_t offset;
  74     unsigned int bytes;
  75     enum BdrvTrackedRequestType type;
  76
  77     bool serialising;
  78     int64_t overlap_offset;
  79     unsigned int overlap_bytes;
  80
  81     QLIST_ENTRY(BdrvTrackedRequest) list;
  82     Coroutine *co; /* owner, used for deadlock detection */
  83     CoQueue wait_queue; /* coroutines blocked on this request */
  84
  85     struct BdrvTrackedRequest *waiting_for;
  86 } BdrvTrackedRequest;
  87
  88 struct BlockDriver {
  89     const char *format_name;
  90     int instance_size;
  91
  92     /* set to true if the BlockDriver is a block filter */
  93     bool is_filter;
  94     /* for snapshots block filter like Quorum can implement the
  95      * following recursive callback.
  96      * It's purpose is to recurse on the filter children while calling
  97      * bdrv_recurse_is_first_non_filter on them.
  98      * For a sample implementation look in the future Quorum block filter.
  99      */
 100     bool (*bdrv_recurse_is_first_non_filter)(BlockDriverState *bs,
 101                                              BlockDriverState *candidate);
 102
 103     int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename);
 104     int (*bdrv_probe_device)(const char *filename);
 105
 106     /* Any driver implementing this callback is expected to be able to handle
 107      * NULL file names in its .bdrv_open() implementation */
 108     void (*bdrv_parse_filename)(const char *filename, QDict *options, Error **errp);
 109     /* Drivers not implementing bdrv_parse_filename nor bdrv_open should have
 110      * this field set to true, except ones that are defined only by their
 111      * child's bs.
 112      * An example of the last type will be the quorum block driver.
 113      */
 114     bool bdrv_needs_filename;
 115
 116     /* Set if a driver can support backing files */
 117     bool supports_backing;
 118
 119     /* For handling image reopen for split or non-split files */
 120     int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state,
 121                                BlockReopenQueue *queue, Error **errp);
 122     void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state);
 123     void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state);
 124     void (*bdrv_join_options)(QDict *options, QDict *old_options);
 125
 126     int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags,
 127                      Error **errp);
 128     int (*bdrv_file_open)(BlockDriverState *bs, QDict *options, int flags,
 129                           Error **errp);
 130     void (*bdrv_close)(BlockDriverState *bs);
 131     int (*bdrv_create)(const char *filename, QemuOpts *opts, Error **errp);
 132     int (*bdrv_set_key)(BlockDriverState *bs, const char *key);
 133     int (*bdrv_make_empty)(BlockDriverState *bs);
 134
 135     void (*bdrv_refresh_filename)(BlockDriverState *bs, QDict *options);
 136
 137     /* aio */
 138     BlockAIOCB *(*bdrv_aio_readv)(BlockDriverState *bs,
 139         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
 140         BlockCompletionFunc *cb, void *opaque);
 141     BlockAIOCB *(*bdrv_aio_writev)(BlockDriverState *bs,
 142         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
 143         BlockCompletionFunc *cb, void *opaque);
 144     BlockAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs,
 145         BlockCompletionFunc *cb, void *opaque);
 146     BlockAIOCB *(*bdrv_aio_discard)(BlockDriverState *bs,
 147         int64_t sector_num, int nb_sectors,
 148         BlockCompletionFunc *cb, void *opaque);
 149
 150     int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs,
 151         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
 152     int coroutine_fn (*bdrv_co_preadv)(BlockDriverState *bs,
 153         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags);
 154     int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs,
 155         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
 156     int coroutine_fn (*bdrv_co_writev_flags)(BlockDriverState *bs,
 157         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, int flags);
 158     int coroutine_fn (*bdrv_co_pwritev)(BlockDriverState *bs,
 159         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags);
 160
 161     /*
 162      * Efficiently zero a region of the disk image.  Typically an image format
 163      * would use a compact metadata representation to implement this.  This
 164      * function pointer may be NULL or return -ENOSUP and .bdrv_co_writev()
 165      * will be called instead.
 166      */
 167     int coroutine_fn (*bdrv_co_write_zeroes)(BlockDriverState *bs,
 168         int64_t sector_num, int nb_sectors, BdrvRequestFlags flags);
 169     int coroutine_fn (*bdrv_co_discard)(BlockDriverState *bs,
 170         int64_t sector_num, int nb_sectors);
 171     int64_t coroutine_fn (*bdrv_co_get_block_status)(BlockDriverState *bs,
 172         int64_t sector_num, int nb_sectors, int *pnum,
 173         BlockDriverState **file);
 174
 175     /*
 176      * Invalidate any cached meta-data.
 177      */
 178     void (*bdrv_invalidate_cache)(BlockDriverState *bs, Error **errp);
 179     int (*bdrv_inactivate)(BlockDriverState *bs);
 180
 181     /*
 182      * Flushes all data for all layers by calling bdrv_co_flush for underlying
 183      * layers, if needed. This function is needed for deterministic
 184      * synchronization of the flush finishing callback.
 185      */
 186     int coroutine_fn (*bdrv_co_flush)(BlockDriverState *bs);
 187
 188     /*
 189      * Flushes all data that was already written to the OS all the way down to
 190      * the disk (for example raw-posix calls fsync()).
 191      */
 192     int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs);
 193
 194     /*
 195      * Flushes all internal caches to the OS. The data may still sit in a
 196      * writeback cache of the host OS, but it will survive a crash of the qemu
 197      * process.
 198      */
 199     int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs);
 200
 201     const char *protocol_name;
 202     int (*bdrv_truncate)(BlockDriverState *bs, int64_t offset);
 203
 204     int64_t (*bdrv_getlength)(BlockDriverState *bs);
 205     bool has_variable_length;
 206     int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
 207
 208     int (*bdrv_write_compressed)(BlockDriverState *bs, int64_t sector_num,
 209                                  const uint8_t *buf, int nb_sectors);
 210
 211     int (*bdrv_snapshot_create)(BlockDriverState *bs,
 212                                 QEMUSnapshotInfo *sn_info);
 213     int (*bdrv_snapshot_goto)(BlockDriverState *bs,
 214                               const char *snapshot_id);
 215     int (*bdrv_snapshot_delete)(BlockDriverState *bs,
 216                                 const char *snapshot_id,
 217                                 const char *name,
 218                                 Error **errp);
 219     int (*bdrv_snapshot_list)(BlockDriverState *bs,
 220                               QEMUSnapshotInfo **psn_info);
 221     int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
 222                                   const char *snapshot_id,
 223                                   const char *name,
 224                                   Error **errp);
 225     int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
 226     ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs);
 227
 228     int (*bdrv_save_vmstate)(BlockDriverState *bs, QEMUIOVector *qiov,
 229                              int64_t pos);
 230     int (*bdrv_load_vmstate)(BlockDriverState *bs, uint8_t *buf,
 231                              int64_t pos, int size);
 232
 233     int (*bdrv_change_backing_file)(BlockDriverState *bs,
 234         const char *backing_file, const char *backing_fmt);
 235
 236     /* removable device specific */
 237     bool (*bdrv_is_inserted)(BlockDriverState *bs);
 238     int (*bdrv_media_changed)(BlockDriverState *bs);
 239     void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
 240     void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
 241
 242     /* to control generic scsi devices */
 243     BlockAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
 244         unsigned long int req, void *buf,
 245         BlockCompletionFunc *cb, void *opaque);
 246
 247     /* List of options for creating images, terminated by name == NULL */
 248     QemuOptsList *create_opts;
 249
 250     /*
 251      * Returns 0 for completed check, -errno for internal errors.
 252      * The check results are stored in result.
 253      */
 254     int (*bdrv_check)(BlockDriverState* bs, BdrvCheckResult *result,
 255         BdrvCheckMode fix);
 256
 257     int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts,
 258                               BlockDriverAmendStatusCB *status_cb,
 259                               void *cb_opaque);
 260
 261     void (*bdrv_debug_event)(BlockDriverState *bs, BlkdebugEvent event);
 262
 263     /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
 264     int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
 265         const char *tag);
 266     int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs,
 267         const char *tag);
 268     int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
 269     bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
 270
 271     void (*bdrv_refresh_limits)(BlockDriverState *bs, Error **errp);
 272
 273     /*
 274      * Returns 1 if newly created images are guaranteed to contain only
 275      * zeros, 0 otherwise.
 276      */
 277     int (*bdrv_has_zero_init)(BlockDriverState *bs);
 278
 279     /* Remove fd handlers, timers, and other event loop callbacks so the event
 280      * loop is no longer in use.  Called with no in-flight requests and in
 281      * depth-first traversal order with parents before child nodes.
 282      */
 283     void (*bdrv_detach_aio_context)(BlockDriverState *bs);
 284
 285     /* Add fd handlers, timers, and other event loop callbacks so I/O requests
 286      * can be processed again.  Called with no in-flight requests and in
 287      * depth-first traversal order with child nodes before parent nodes.
 288      */
 289     void (*bdrv_attach_aio_context)(BlockDriverState *bs,
 290                                     AioContext *new_context);
 291
 292     /* io queue for linux-aio */
 293     void (*bdrv_io_plug)(BlockDriverState *bs);
 294     void (*bdrv_io_unplug)(BlockDriverState *bs);
 295
 296     /**
 297      * Try to get @bs's logical and physical block size.
 298      * On success, store them in @bsz and return zero.
 299      * On failure, return negative errno.
 300      */
 301     int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz);
 302     /**
 303      * Try to get @bs's geometry (cyls, heads, sectors)
 304      * On success, store them in @geo and return 0.
 305      * On failure return -errno.
 306      * Only drivers that want to override guest geometry implement this
 307      * callback; see hd_geometry_guess().
 308      */
 309     int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo);
 310
 311     /**
 312      * Drain and stop any internal sources of requests in the driver, and
 313      * remain so until next I/O callback (e.g. bdrv_co_writev) is called.
 314      */
 315     void (*bdrv_drain)(BlockDriverState *bs);
 316
 317     QLIST_ENTRY(BlockDriver) list;
 318 };
 319
 320 typedef struct BlockLimits {
 321     /* maximum number of sectors that can be discarded at once */
 322     int max_discard;
 323
 324     /* optimal alignment for discard requests in sectors */
 325     int64_t discard_alignment;
 326
 327     /* maximum number of sectors that can zeroized at once */
 328     int max_write_zeroes;
 329
 330     /* optimal alignment for write zeroes requests in sectors */
 331     int64_t write_zeroes_alignment;
 332
 333     /* optimal transfer length in sectors */
 334     int opt_transfer_length;
 335
 336     /* maximal transfer length in sectors */
 337     int max_transfer_length;
 338
 339     /* memory alignment so that no bounce buffer is needed */
 340     size_t min_mem_alignment;
 341
 342     /* memory alignment for bounce buffer */
 343     size_t opt_mem_alignment;
 344
 345     /* maximum number of iovec elements */
 346     int max_iov;
 347 } BlockLimits;
 348
 349 typedef struct BdrvOpBlocker BdrvOpBlocker;
 350
 351 typedef struct BdrvAioNotifier {
 352     void (*attached_aio_context)(AioContext *new_context, void *opaque);
 353     void (*detach_aio_context)(void *opaque);
 354
 355     void *opaque;
 356
 357     QLIST_ENTRY(BdrvAioNotifier) list;
 358 } BdrvAioNotifier;
 359
 360 struct BdrvChildRole {
 361     void (*inherit_options)(int *child_flags, QDict *child_options,
 362                             int parent_flags, QDict *parent_options);
 363 };
 364
 365 extern const BdrvChildRole child_file;
 366 extern const BdrvChildRole child_format;
 367
 368 struct BdrvChild {
 369     BlockDriverState *bs;
 370     char *name;
 371     const BdrvChildRole *role;
 372     QLIST_ENTRY(BdrvChild) next;
 373     QLIST_ENTRY(BdrvChild) next_parent;
 374 };
 375
 376 /*
 377  * Note: the function bdrv_append() copies and swaps contents of
 378  * BlockDriverStates, so if you add new fields to this struct, please
 379  * inspect bdrv_append() to determine if the new fields need to be
 380  * copied as well.
 381  */
 382 struct BlockDriverState {
 383     int64_t total_sectors; /* if we are reading a disk image, give its
 384                               size in sectors */
 385     int read_only; /* if true, the media is read only */
 386     int open_flags; /* flags used to open the file, re-used for re-open */
 387     int encrypted; /* if true, the media is encrypted */
 388     int valid_key; /* if true, a valid encryption key has been set */
 389     int sg;        /* if true, the device is a /dev/sg* */
 390     int copy_on_read; /* if true, copy read backing sectors into image
 391                          note this is a reference count */
 392     bool probed;
 393
 394     BlockDriver *drv; /* NULL means no media */
 395     void *opaque;
 396
 397     BlockBackend *blk;          /* owning backend, if any */
 398
 399     AioContext *aio_context; /* event loop used for fd handlers, timers, etc */
 400     /* long-running tasks intended to always use the same AioContext as this
 401      * BDS may register themselves in this list to be notified of changes
 402      * regarding this BDS's context */
 403     QLIST_HEAD(, BdrvAioNotifier) aio_notifiers;
 404
 405     char filename[PATH_MAX];
 406     char backing_file[PATH_MAX]; /* if non zero, the image is a diff of
 407                                     this file image */
 408     char backing_format[16]; /* if non-zero and backing_file exists */
 409
 410     QDict *full_open_options;
 411     char exact_filename[PATH_MAX];
 412
 413     BdrvChild *backing;
 414     BdrvChild *file;
 415
 416     /* Callback before write request is processed */
 417     NotifierWithReturnList before_write_notifiers;
 418
 419     /* number of in-flight serialising requests */
 420     unsigned int serialising_in_flight;
 421
 422     /* I/O throttling.
 423      * throttle_state tells us if this BDS has I/O limits configured.
 424      * io_limits_disabled tells us if they are currently being enforced */
 425     CoQueue      throttled_reqs[2];
 426     unsigned int io_limits_disabled;
 427
 428     /* The following fields are protected by the ThrottleGroup lock.
 429      * See the ThrottleGroup documentation for details. */
 430     ThrottleState *throttle_state;
 431     ThrottleTimers throttle_timers;
 432     unsigned       pending_reqs[2];
 433     QLIST_ENTRY(BlockDriverState) round_robin;
 434
 435     /* Offset after the highest byte written to */
 436     uint64_t wr_highest_offset;
 437
 438     /* I/O Limits */
 439     BlockLimits bl;
 440
 441     /* Whether produces zeros when read beyond eof */
 442     bool zero_beyond_eof;
 443
 444     /* Alignment requirement for offset/length of I/O requests */
 445     unsigned int request_alignment;
 446     /* Flags honored during pwrite (so far: BDRV_REQ_FUA) */
 447     unsigned int supported_write_flags;
 448     /* Flags honored during write_zeroes (so far: BDRV_REQ_FUA,
 449      * BDRV_REQ_MAY_UNMAP) */
 450     unsigned int supported_zero_flags;
 451
 452     /* the following member gives a name to every node on the bs graph. */
 453     char node_name[32];
 454     /* element of the list of named nodes building the graph */
 455     QTAILQ_ENTRY(BlockDriverState) node_list;
 456     /* element of the list of all BlockDriverStates (all_bdrv_states) */
 457     QTAILQ_ENTRY(BlockDriverState) bs_list;
 458     /* element of the list of monitor-owned BDS */
 459     QTAILQ_ENTRY(BlockDriverState) monitor_list;
 460     QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps;
 461     int refcnt;
 462
 463     QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
 464
 465     /* operation blockers */
 466     QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX];
 467
 468     /* long-running background operation */
 469     BlockJob *job;
 470
 471     /* The node that this node inherited default options from (and a reopen on
 472      * which can affect this node by changing these defaults). This is always a
 473      * parent node of this node. */
 474     BlockDriverState *inherits_from;
 475     QLIST_HEAD(, BdrvChild) children;
 476     QLIST_HEAD(, BdrvChild) parents;
 477
 478     QDict *options;
 479     QDict *explicit_options;
 480     BlockdevDetectZeroesOptions detect_zeroes;
 481
 482     /* The error object in use for blocking operations on backing_hd */
 483     Error *backing_blocker;
 484
 485     /* threshold limit for writes, in bytes. "High water mark". */
 486     uint64_t write_threshold_offset;
 487     NotifierWithReturn write_threshold_notifier;
 488
 489     /* counters for nested bdrv_io_plug and bdrv_io_unplugged_begin */
 490     unsigned io_plugged;
 491     unsigned io_plug_disabled;
 492
 493     int quiesce_counter;
 494 };
 495
 496 struct BlockBackendRootState {
 497     int open_flags;
 498     bool read_only;
 499     BlockdevDetectZeroesOptions detect_zeroes;
 500
 501     char *throttle_group;
 502     ThrottleState *throttle_state;
 503 };
 504
 505 static inline BlockDriverState *backing_bs(BlockDriverState *bs)
 506 {
 507     return bs->backing ? bs->backing->bs : NULL;
 508 }
 509
 510
 511 /* Essential block drivers which must always be statically linked into qemu, and
 512  * which therefore can be accessed without using bdrv_find_format() */
 513 extern BlockDriver bdrv_file;
 514 extern BlockDriver bdrv_raw;
 515 extern BlockDriver bdrv_qcow2;
 516
 517 /**
 518  * bdrv_setup_io_funcs:
 519  *
 520  * Prepare a #BlockDriver for I/O request processing by populating
 521  * unimplemented coroutine and AIO interfaces with generic wrapper functions
 522  * that fall back to implemented interfaces.
 523  */
 524 void bdrv_setup_io_funcs(BlockDriver *bdrv);
 525
 526 int coroutine_fn bdrv_co_preadv(BlockDriverState *bs,
 527     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
 528     BdrvRequestFlags flags);
 529 int coroutine_fn bdrv_co_pwritev(BlockDriverState *bs,
 530     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
 531     BdrvRequestFlags flags);
 532
 533 int get_tmp_filename(char *filename, int size);
 534 BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size,
 535                             const char *filename);
 536
 537 void bdrv_set_io_limits(BlockDriverState *bs,
 538                         ThrottleConfig *cfg);
 539
 540
 541 /**
 542  * bdrv_add_before_write_notifier:
 543  *
 544  * Register a callback that is invoked before write requests are processed but
 545  * after any throttling or waiting for overlapping requests.
 546  */
 547 void bdrv_add_before_write_notifier(BlockDriverState *bs,
 548                                     NotifierWithReturn *notifier);
 549
 550 /**
 551  * bdrv_detach_aio_context:
 552  *
 553  * May be called from .bdrv_detach_aio_context() to detach children from the
 554  * current #AioContext.  This is only needed by block drivers that manage their
 555  * own children.  Both ->file and ->backing are automatically handled and
 556  * block drivers should not call this function on them explicitly.
 557  */
 558 void bdrv_detach_aio_context(BlockDriverState *bs);
 559
 560 /**
 561  * bdrv_attach_aio_context:
 562  *
 563  * May be called from .bdrv_attach_aio_context() to attach children to the new
 564  * #AioContext.  This is only needed by block drivers that manage their own
 565  * children.  Both ->file and ->backing are automatically handled and block
 566  * drivers should not call this function on them explicitly.
 567  */
 568 void bdrv_attach_aio_context(BlockDriverState *bs,
 569                              AioContext *new_context);
 570
 571 /**
 572  * bdrv_add_aio_context_notifier:
 573  *
 574  * If a long-running job intends to be always run in the same AioContext as a
 575  * certain BDS, it may use this function to be notified of changes regarding the
 576  * association of the BDS to an AioContext.
 577  *
 578  * attached_aio_context() is called after the target BDS has been attached to a
 579  * new AioContext; detach_aio_context() is called before the target BDS is being
 580  * detached from its old AioContext.
 581  */
 582 void bdrv_add_aio_context_notifier(BlockDriverState *bs,
 583         void (*attached_aio_context)(AioContext *new_context, void *opaque),
 584         void (*detach_aio_context)(void *opaque), void *opaque);
 585
 586 /**
 587  * bdrv_remove_aio_context_notifier:
 588  *
 589  * Unsubscribe of change notifications regarding the BDS's AioContext. The
 590  * parameters given here have to be the same as those given to
 591  * bdrv_add_aio_context_notifier().
 592  */
 593 void bdrv_remove_aio_context_notifier(BlockDriverState *bs,
 594                                       void (*aio_context_attached)(AioContext *,
 595                                                                    void *),
 596                                       void (*aio_context_detached)(void *),
 597                                       void *opaque);
 598
 599 #ifdef _WIN32
 600 int is_windows_drive(const char *filename);
 601 #endif
 602
 603 /**
 604  * stream_start:
 605  * @bs: Block device to operate on.
 606  * @base: Block device that will become the new base, or %NULL to
 607  * flatten the whole backing file chain onto @bs.
 608  * @base_id: The file name that will be written to @bs as the new
 609  * backing file if the job completes.  Ignored if @base is %NULL.
 610  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
 611  * @on_error: The action to take upon error.
 612  * @cb: Completion function for the job.
 613  * @opaque: Opaque pointer value passed to @cb.
 614  * @errp: Error object.
 615  *
 616  * Start a streaming operation on @bs.  Clusters that are unallocated
 617  * in @bs, but allocated in any image between @base and @bs (both
 618  * exclusive) will be written to @bs.  At the end of a successful
 619  * streaming job, the backing file of @bs will be changed to
 620  * @base_id in the written image and to @base in the live BlockDriverState.
 621  */
 622 void stream_start(BlockDriverState *bs, BlockDriverState *base,
 623                   const char *base_id, int64_t speed, BlockdevOnError on_error,
 624                   BlockCompletionFunc *cb,
 625                   void *opaque, Error **errp);
 626
 627 /**
 628  * commit_start:
 629  * @bs: Active block device.
 630  * @top: Top block device to be committed.
 631  * @base: Block device that will be written into, and become the new top.
 632  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
 633  * @on_error: The action to take upon error.
 634  * @cb: Completion function for the job.
 635  * @opaque: Opaque pointer value passed to @cb.
 636  * @backing_file_str: String to use as the backing file in @top's overlay
 637  * @errp: Error object.
 638  *
 639  */
 640 void commit_start(BlockDriverState *bs, BlockDriverState *base,
 641                  BlockDriverState *top, int64_t speed,
 642                  BlockdevOnError on_error, BlockCompletionFunc *cb,
 643                  void *opaque, const char *backing_file_str, Error **errp);
 644 /**
 645  * commit_active_start:
 646  * @bs: Active block device to be committed.
 647  * @base: Block device that will be written into, and become the new top.
 648  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
 649  * @on_error: The action to take upon error.
 650  * @cb: Completion function for the job.
 651  * @opaque: Opaque pointer value passed to @cb.
 652  * @errp: Error object.
 653  *
 654  */
 655 void commit_active_start(BlockDriverState *bs, BlockDriverState *base,
 656                          int64_t speed,
 657                          BlockdevOnError on_error,
 658                          BlockCompletionFunc *cb,
 659                          void *opaque, Error **errp);
 660 /*
 661  * mirror_start:
 662  * @bs: Block device to operate on.
 663  * @target: Block device to write to.
 664  * @replaces: Block graph node name to replace once the mirror is done. Can
 665  *            only be used when full mirroring is selected.
 666  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
 667  * @granularity: The chosen granularity for the dirty bitmap.
 668  * @buf_size: The amount of data that can be in flight at one time.
 669  * @mode: Whether to collapse all images in the chain to the target.
 670  * @on_source_error: The action to take upon error reading from the source.
 671  * @on_target_error: The action to take upon error writing to the target.
 672  * @unmap: Whether to unmap target where source sectors only contain zeroes.
 673  * @cb: Completion function for the job.
 674  * @opaque: Opaque pointer value passed to @cb.
 675  * @errp: Error object.
 676  *
 677  * Start a mirroring operation on @bs.  Clusters that are allocated
 678  * in @bs will be written to @bs until the job is cancelled or
 679  * manually completed.  At the end of a successful mirroring job,
 680  * @bs will be switched to read from @target.
 681  */
 682 void mirror_start(BlockDriverState *bs, BlockDriverState *target,
 683                   const char *replaces,
 684                   int64_t speed, uint32_t granularity, int64_t buf_size,
 685                   MirrorSyncMode mode, BlockdevOnError on_source_error,
 686                   BlockdevOnError on_target_error,
 687                   bool unmap,
 688                   BlockCompletionFunc *cb,
 689                   void *opaque, Error **errp);
 690
 691 /*
 692  * backup_start:
 693  * @bs: Block device to operate on.
 694  * @target: Block device to write to.
 695  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
 696  * @sync_mode: What parts of the disk image should be copied to the destination.
 697  * @sync_bitmap: The dirty bitmap if sync_mode is MIRROR_SYNC_MODE_INCREMENTAL.
 698  * @on_source_error: The action to take upon error reading from the source.
 699  * @on_target_error: The action to take upon error writing to the target.
 700  * @cb: Completion function for the job.
 701  * @opaque: Opaque pointer value passed to @cb.
 702  * @txn: Transaction that this job is part of (may be NULL).
 703  *
 704  * Start a backup operation on @bs.  Clusters in @bs are written to @target
 705  * until the job is cancelled or manually completed.
 706  */
 707 void backup_start(BlockDriverState *bs, BlockDriverState *target,
 708                   int64_t speed, MirrorSyncMode sync_mode,
 709                   BdrvDirtyBitmap *sync_bitmap,
 710                   BlockdevOnError on_source_error,
 711                   BlockdevOnError on_target_error,
 712                   BlockCompletionFunc *cb, void *opaque,
 713                   BlockJobTxn *txn, Error **errp);
 714
 715 void hmp_drive_add_node(Monitor *mon, const char *optstr);
 716
 717 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
 718                                   const char *child_name,
 719                                   const BdrvChildRole *child_role);
 720 void bdrv_root_unref_child(BdrvChild *child);
 721
 722 void bdrv_no_throttling_begin(BlockDriverState *bs);
 723 void bdrv_no_throttling_end(BlockDriverState *bs);
 724
 725 void blk_dev_change_media_cb(BlockBackend *blk, bool load);
 726 bool blk_dev_has_removable_media(BlockBackend *blk);
 727 bool blk_dev_has_tray(BlockBackend *blk);
 728 void blk_dev_eject_request(BlockBackend *blk, bool force);
 729 bool blk_dev_is_tray_open(BlockBackend *blk);
 730 bool blk_dev_is_medium_locked(BlockBackend *blk);
 731 void blk_dev_resize_cb(BlockBackend *blk);
 732
 733 void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector, int nr_sectors);
 734 bool bdrv_requests_pending(BlockDriverState *bs);
 735
 736 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap **out);
 737 void bdrv_undo_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap *in);
 738
 739 void blockdev_close_all_bdrv_states(void);
 740
 741 #endif /* BLOCK_INT_H */