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 "qemu/option.h"
  30 #include "qemu/queue.h"
  31 #include "qemu/coroutine.h"
  32 #include "qemu/timer.h"
  33 #include "qapi-types.h"
  34 #include "qemu/hbitmap.h"
  35 #include "block/snapshot.h"
  36 #include "qemu/main-loop.h"
  37 #include "qemu/throttle.h"
  38
  39 #define BLOCK_FLAG_ENCRYPT          1
  40 #define BLOCK_FLAG_LAZY_REFCOUNTS   8
  41
  42 #define BLOCK_OPT_SIZE              "size"
  43 #define BLOCK_OPT_ENCRYPT           "encryption"
  44 #define BLOCK_OPT_COMPAT6           "compat6"
  45 #define BLOCK_OPT_HWVERSION         "hwversion"
  46 #define BLOCK_OPT_BACKING_FILE      "backing_file"
  47 #define BLOCK_OPT_BACKING_FMT       "backing_fmt"
  48 #define BLOCK_OPT_CLUSTER_SIZE      "cluster_size"
  49 #define BLOCK_OPT_TABLE_SIZE        "table_size"
  50 #define BLOCK_OPT_PREALLOC          "preallocation"
  51 #define BLOCK_OPT_SUBFMT            "subformat"
  52 #define BLOCK_OPT_COMPAT_LEVEL      "compat"
  53 #define BLOCK_OPT_LAZY_REFCOUNTS    "lazy_refcounts"
  54 #define BLOCK_OPT_ADAPTER_TYPE      "adapter_type"
  55 #define BLOCK_OPT_REDUNDANCY        "redundancy"
  56 #define BLOCK_OPT_NOCOW             "nocow"
  57 #define BLOCK_OPT_OBJECT_SIZE       "object_size"
  58 #define BLOCK_OPT_REFCOUNT_BITS     "refcount_bits"
  59
  60 #define BLOCK_PROBE_BUF_SIZE        512
  61
  62 enum BdrvTrackedRequestType {
  63     BDRV_TRACKED_READ,
  64     BDRV_TRACKED_WRITE,
  65     BDRV_TRACKED_FLUSH,
  66     BDRV_TRACKED_IOCTL,
  67     BDRV_TRACKED_DISCARD,
  68 };
  69
  70 typedef struct BdrvTrackedRequest {
  71     BlockDriverState *bs;
  72     int64_t offset;
  73     unsigned int bytes;
  74     enum BdrvTrackedRequestType type;
  75
  76     bool serialising;
  77     int64_t overlap_offset;
  78     unsigned int overlap_bytes;
  79
  80     QLIST_ENTRY(BdrvTrackedRequest) list;
  81     Coroutine *co; /* owner, used for deadlock detection */
  82     CoQueue wait_queue; /* coroutines blocked on this request */
  83
  84     struct BdrvTrackedRequest *waiting_for;
  85 } BdrvTrackedRequest;
  86
  87 struct BlockDriver {
  88     const char *format_name;
  89     int instance_size;
  90
  91     /* set to true if the BlockDriver is a block filter */
  92     bool is_filter;
  93     /* for snapshots block filter like Quorum can implement the
  94      * following recursive callback.
  95      * It's purpose is to recurse on the filter children while calling
  96      * bdrv_recurse_is_first_non_filter on them.
  97      * For a sample implementation look in the future Quorum block filter.
  98      */
  99     bool (*bdrv_recurse_is_first_non_filter)(BlockDriverState *bs,
 100                                              BlockDriverState *candidate);
 101
 102     int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename);
 103     int (*bdrv_probe_device)(const char *filename);
 104
 105     /* Any driver implementing this callback is expected to be able to handle
 106      * NULL file names in its .bdrv_open() implementation */
 107     void (*bdrv_parse_filename)(const char *filename, QDict *options, Error **errp);
 108     /* Drivers not implementing bdrv_parse_filename nor bdrv_open should have
 109      * this field set to true, except ones that are defined only by their
 110      * child's bs.
 111      * An example of the last type will be the quorum block driver.
 112      */
 113     bool bdrv_needs_filename;
 114
 115     /* Set if a driver can support backing files */
 116     bool supports_backing;
 117
 118     /* For handling image reopen for split or non-split files */
 119     int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state,
 120                                BlockReopenQueue *queue, Error **errp);
 121     void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state);
 122     void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state);
 123     void (*bdrv_join_options)(QDict *options, QDict *old_options);
 124
 125     int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags,
 126                      Error **errp);
 127     int (*bdrv_file_open)(BlockDriverState *bs, QDict *options, int flags,
 128                           Error **errp);
 129     void (*bdrv_close)(BlockDriverState *bs);
 130     int (*bdrv_create)(const char *filename, QemuOpts *opts, Error **errp);
 131     int (*bdrv_set_key)(BlockDriverState *bs, const char *key);
 132     int (*bdrv_make_empty)(BlockDriverState *bs);
 133
 134     void (*bdrv_refresh_filename)(BlockDriverState *bs, QDict *options);
 135
 136     /* aio */
 137     BlockAIOCB *(*bdrv_aio_readv)(BlockDriverState *bs,
 138         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
 139         BlockCompletionFunc *cb, void *opaque);
 140     BlockAIOCB *(*bdrv_aio_writev)(BlockDriverState *bs,
 141         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
 142         BlockCompletionFunc *cb, void *opaque);
 143     BlockAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs,
 144         BlockCompletionFunc *cb, void *opaque);
 145     BlockAIOCB *(*bdrv_aio_discard)(BlockDriverState *bs,
 146         int64_t sector_num, int nb_sectors,
 147         BlockCompletionFunc *cb, void *opaque);
 148
 149     int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs,
 150         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
 151     int coroutine_fn (*bdrv_co_preadv)(BlockDriverState *bs,
 152         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags);
 153     int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs,
 154         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
 155     int coroutine_fn (*bdrv_co_writev_flags)(BlockDriverState *bs,
 156         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, int flags);
 157     int coroutine_fn (*bdrv_co_pwritev)(BlockDriverState *bs,
 158         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags);
 159
 160     /*
 161      * Efficiently zero a region of the disk image.  Typically an image format
 162      * would use a compact metadata representation to implement this.  This
 163      * function pointer may be NULL or return -ENOSUP and .bdrv_co_writev()
 164      * will be called instead.
 165      */
 166     int coroutine_fn (*bdrv_co_pwrite_zeroes)(BlockDriverState *bs,
 167         int64_t offset, int count, BdrvRequestFlags flags);
 168     int coroutine_fn (*bdrv_co_discard)(BlockDriverState *bs,
 169         int64_t sector_num, int nb_sectors);
 170     int64_t coroutine_fn (*bdrv_co_get_block_status)(BlockDriverState *bs,
 171         int64_t sector_num, int nb_sectors, int *pnum,
 172         BlockDriverState **file);
 173
 174     /*
 175      * Invalidate any cached meta-data.
 176      */
 177     void (*bdrv_invalidate_cache)(BlockDriverState *bs, Error **errp);
 178     int (*bdrv_inactivate)(BlockDriverState *bs);
 179
 180     /*
 181      * Flushes all data for all layers by calling bdrv_co_flush for underlying
 182      * layers, if needed. This function is needed for deterministic
 183      * synchronization of the flush finishing callback.
 184      */
 185     int coroutine_fn (*bdrv_co_flush)(BlockDriverState *bs);
 186
 187     /*
 188      * Flushes all data that was already written to the OS all the way down to
 189      * the disk (for example raw-posix calls fsync()).
 190      */
 191     int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs);
 192
 193     /*
 194      * Flushes all internal caches to the OS. The data may still sit in a
 195      * writeback cache of the host OS, but it will survive a crash of the qemu
 196      * process.
 197      */
 198     int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs);
 199
 200     const char *protocol_name;
 201     int (*bdrv_truncate)(BlockDriverState *bs, int64_t offset);
 202
 203     int64_t (*bdrv_getlength)(BlockDriverState *bs);
 204     bool has_variable_length;
 205     int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
 206
 207     int (*bdrv_write_compressed)(BlockDriverState *bs, int64_t sector_num,
 208                                  const uint8_t *buf, int nb_sectors);
 209
 210     int (*bdrv_snapshot_create)(BlockDriverState *bs,
 211                                 QEMUSnapshotInfo *sn_info);
 212     int (*bdrv_snapshot_goto)(BlockDriverState *bs,
 213                               const char *snapshot_id);
 214     int (*bdrv_snapshot_delete)(BlockDriverState *bs,
 215                                 const char *snapshot_id,
 216                                 const char *name,
 217                                 Error **errp);
 218     int (*bdrv_snapshot_list)(BlockDriverState *bs,
 219                               QEMUSnapshotInfo **psn_info);
 220     int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
 221                                   const char *snapshot_id,
 222                                   const char *name,
 223                                   Error **errp);
 224     int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
 225     ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs);
 226
 227     int (*bdrv_save_vmstate)(BlockDriverState *bs, QEMUIOVector *qiov,
 228                              int64_t pos);
 229     int (*bdrv_load_vmstate)(BlockDriverState *bs, uint8_t *buf,
 230                              int64_t pos, int size);
 231
 232     int (*bdrv_change_backing_file)(BlockDriverState *bs,
 233         const char *backing_file, const char *backing_fmt);
 234
 235     /* removable device specific */
 236     bool (*bdrv_is_inserted)(BlockDriverState *bs);
 237     int (*bdrv_media_changed)(BlockDriverState *bs);
 238     void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
 239     void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
 240
 241     /* to control generic scsi devices */
 242     BlockAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
 243         unsigned long int req, void *buf,
 244         BlockCompletionFunc *cb, void *opaque);
 245
 246     /* List of options for creating images, terminated by name == NULL */
 247     QemuOptsList *create_opts;
 248
 249     /*
 250      * Returns 0 for completed check, -errno for internal errors.
 251      * The check results are stored in result.
 252      */
 253     int (*bdrv_check)(BlockDriverState* bs, BdrvCheckResult *result,
 254         BdrvCheckMode fix);
 255
 256     int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts,
 257                               BlockDriverAmendStatusCB *status_cb,
 258                               void *cb_opaque);
 259
 260     void (*bdrv_debug_event)(BlockDriverState *bs, BlkdebugEvent event);
 261
 262     /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
 263     int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
 264         const char *tag);
 265     int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs,
 266         const char *tag);
 267     int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
 268     bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
 269
 270     void (*bdrv_refresh_limits)(BlockDriverState *bs, Error **errp);
 271
 272     /*
 273      * Returns 1 if newly created images are guaranteed to contain only
 274      * zeros, 0 otherwise.
 275      */
 276     int (*bdrv_has_zero_init)(BlockDriverState *bs);
 277
 278     /* Remove fd handlers, timers, and other event loop callbacks so the event
 279      * loop is no longer in use.  Called with no in-flight requests and in
 280      * depth-first traversal order with parents before child nodes.
 281      */
 282     void (*bdrv_detach_aio_context)(BlockDriverState *bs);
 283
 284     /* Add fd handlers, timers, and other event loop callbacks so I/O requests
 285      * can be processed again.  Called with no in-flight requests and in
 286      * depth-first traversal order with child nodes before parent nodes.
 287      */
 288     void (*bdrv_attach_aio_context)(BlockDriverState *bs,
 289                                     AioContext *new_context);
 290
 291     /* io queue for linux-aio */
 292     void (*bdrv_io_plug)(BlockDriverState *bs);
 293     void (*bdrv_io_unplug)(BlockDriverState *bs);
 294
 295     /**
 296      * Try to get @bs's logical and physical block size.
 297      * On success, store them in @bsz and return zero.
 298      * On failure, return negative errno.
 299      */
 300     int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz);
 301     /**
 302      * Try to get @bs's geometry (cyls, heads, sectors)
 303      * On success, store them in @geo and return 0.
 304      * On failure return -errno.
 305      * Only drivers that want to override guest geometry implement this
 306      * callback; see hd_geometry_guess().
 307      */
 308     int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo);
 309
 310     /**
 311      * Drain and stop any internal sources of requests in the driver, and
 312      * remain so until next I/O callback (e.g. bdrv_co_writev) is called.
 313      */
 314     void (*bdrv_drain)(BlockDriverState *bs);
 315
 316     void (*bdrv_add_child)(BlockDriverState *parent, BlockDriverState *child,
 317                            Error **errp);
 318     void (*bdrv_del_child)(BlockDriverState *parent, BdrvChild *child,
 319                            Error **errp);
 320
 321     QLIST_ENTRY(BlockDriver) list;
 322 };
 323
 324 typedef struct BlockLimits {
 325     /* maximum number of sectors that can be discarded at once */
 326     int max_discard;
 327
 328     /* optimal alignment for discard requests in sectors */
 329     int64_t discard_alignment;
 330
 331     /* maximum number of bytes that can zeroized at once (since it is
 332      * signed, it must be < 2G, if set) */
 333     int32_t max_pwrite_zeroes;
 334
 335     /* optimal alignment for write zeroes requests in bytes, must be
 336      * power of 2, and less than max_pwrite_zeroes if that is set */
 337     uint32_t pwrite_zeroes_alignment;
 338
 339     /* optimal transfer length in sectors */
 340     int opt_transfer_length;
 341
 342     /* maximal transfer length in sectors */
 343     int max_transfer_length;
 344
 345     /* memory alignment so that no bounce buffer is needed */
 346     size_t min_mem_alignment;
 347
 348     /* memory alignment for bounce buffer */
 349     size_t opt_mem_alignment;
 350
 351     /* maximum number of iovec elements */
 352     int max_iov;
 353 } BlockLimits;
 354
 355 typedef struct BdrvOpBlocker BdrvOpBlocker;
 356
 357 typedef struct BdrvAioNotifier {
 358     void (*attached_aio_context)(AioContext *new_context, void *opaque);
 359     void (*detach_aio_context)(void *opaque);
 360
 361     void *opaque;
 362
 363     QLIST_ENTRY(BdrvAioNotifier) list;
 364 } BdrvAioNotifier;
 365
 366 struct BdrvChildRole {
 367     void (*inherit_options)(int *child_flags, QDict *child_options,
 368                             int parent_flags, QDict *parent_options);
 369
 370     void (*change_media)(BdrvChild *child, bool load);
 371     void (*resize)(BdrvChild *child);
 372
 373     /* Returns a name that is supposedly more useful for human users than the
 374      * node name for identifying the node in question (in particular, a BB
 375      * name), or NULL if the parent can't provide a better name. */
 376     const char* (*get_name)(BdrvChild *child);
 377
 378     /*
 379      * If this pair of functions is implemented, the parent doesn't issue new
 380      * requests after returning from .drained_begin() until .drained_end() is
 381      * called.
 382      *
 383      * Note that this can be nested. If drained_begin() was called twice, new
 384      * I/O is allowed only after drained_end() was called twice, too.
 385      */
 386     void (*drained_begin)(BdrvChild *child);
 387     void (*drained_end)(BdrvChild *child);
 388 };
 389
 390 extern const BdrvChildRole child_file;
 391 extern const BdrvChildRole child_format;
 392
 393 struct BdrvChild {
 394     BlockDriverState *bs;
 395     char *name;
 396     const BdrvChildRole *role;
 397     void *opaque;
 398     QLIST_ENTRY(BdrvChild) next;
 399     QLIST_ENTRY(BdrvChild) next_parent;
 400 };
 401
 402 /*
 403  * Note: the function bdrv_append() copies and swaps contents of
 404  * BlockDriverStates, so if you add new fields to this struct, please
 405  * inspect bdrv_append() to determine if the new fields need to be
 406  * copied as well.
 407  */
 408 struct BlockDriverState {
 409     int64_t total_sectors; /* if we are reading a disk image, give its
 410                               size in sectors */
 411     int read_only; /* if true, the media is read only */
 412     int open_flags; /* flags used to open the file, re-used for re-open */
 413     int encrypted; /* if true, the media is encrypted */
 414     int valid_key; /* if true, a valid encryption key has been set */
 415     int sg;        /* if true, the device is a /dev/sg* */
 416     int copy_on_read; /* if true, copy read backing sectors into image
 417                          note this is a reference count */
 418     bool probed;
 419
 420     BlockDriver *drv; /* NULL means no media */
 421     void *opaque;
 422
 423     AioContext *aio_context; /* event loop used for fd handlers, timers, etc */
 424     /* long-running tasks intended to always use the same AioContext as this
 425      * BDS may register themselves in this list to be notified of changes
 426      * regarding this BDS's context */
 427     QLIST_HEAD(, BdrvAioNotifier) aio_notifiers;
 428
 429     char filename[PATH_MAX];
 430     char backing_file[PATH_MAX]; /* if non zero, the image is a diff of
 431                                     this file image */
 432     char backing_format[16]; /* if non-zero and backing_file exists */
 433
 434     QDict *full_open_options;
 435     char exact_filename[PATH_MAX];
 436
 437     BdrvChild *backing;
 438     BdrvChild *file;
 439
 440     /* Callback before write request is processed */
 441     NotifierWithReturnList before_write_notifiers;
 442
 443     /* number of in-flight serialising requests */
 444     unsigned int serialising_in_flight;
 445
 446     /* Offset after the highest byte written to */
 447     uint64_t wr_highest_offset;
 448
 449     /* I/O Limits */
 450     BlockLimits bl;
 451
 452     /* Whether produces zeros when read beyond eof */
 453     bool zero_beyond_eof;
 454
 455     /* Alignment requirement for offset/length of I/O requests */
 456     unsigned int request_alignment;
 457     /* Flags honored during pwrite (so far: BDRV_REQ_FUA) */
 458     unsigned int supported_write_flags;
 459     /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA,
 460      * BDRV_REQ_MAY_UNMAP) */
 461     unsigned int supported_zero_flags;
 462
 463     /* the following member gives a name to every node on the bs graph. */
 464     char node_name[32];
 465     /* element of the list of named nodes building the graph */
 466     QTAILQ_ENTRY(BlockDriverState) node_list;
 467     /* element of the list of all BlockDriverStates (all_bdrv_states) */
 468     QTAILQ_ENTRY(BlockDriverState) bs_list;
 469     /* element of the list of monitor-owned BDS */
 470     QTAILQ_ENTRY(BlockDriverState) monitor_list;
 471     QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps;
 472     int refcnt;
 473
 474     QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
 475
 476     /* operation blockers */
 477     QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX];
 478
 479     /* long-running background operation */
 480     BlockJob *job;
 481
 482     /* The node that this node inherited default options from (and a reopen on
 483      * which can affect this node by changing these defaults). This is always a
 484      * parent node of this node. */
 485     BlockDriverState *inherits_from;
 486     QLIST_HEAD(, BdrvChild) children;
 487     QLIST_HEAD(, BdrvChild) parents;
 488
 489     QDict *options;
 490     QDict *explicit_options;
 491     BlockdevDetectZeroesOptions detect_zeroes;
 492
 493     /* The error object in use for blocking operations on backing_hd */
 494     Error *backing_blocker;
 495
 496     /* threshold limit for writes, in bytes. "High water mark". */
 497     uint64_t write_threshold_offset;
 498     NotifierWithReturn write_threshold_notifier;
 499
 500     /* counters for nested bdrv_io_plug and bdrv_io_unplugged_begin */
 501     unsigned io_plugged;
 502     unsigned io_plug_disabled;
 503
 504     int quiesce_counter;
 505 };
 506
 507 struct BlockBackendRootState {
 508     int open_flags;
 509     bool read_only;
 510     BlockdevDetectZeroesOptions detect_zeroes;
 511 };
 512
 513 static inline BlockDriverState *backing_bs(BlockDriverState *bs)
 514 {
 515     return bs->backing ? bs->backing->bs : NULL;
 516 }
 517
 518
 519 /* Essential block drivers which must always be statically linked into qemu, and
 520  * which therefore can be accessed without using bdrv_find_format() */
 521 extern BlockDriver bdrv_file;
 522 extern BlockDriver bdrv_raw;
 523 extern BlockDriver bdrv_qcow2;
 524
 525 /**
 526  * bdrv_setup_io_funcs:
 527  *
 528  * Prepare a #BlockDriver for I/O request processing by populating
 529  * unimplemented coroutine and AIO interfaces with generic wrapper functions
 530  * that fall back to implemented interfaces.
 531  */
 532 void bdrv_setup_io_funcs(BlockDriver *bdrv);
 533
 534 int coroutine_fn bdrv_co_preadv(BlockDriverState *bs,
 535     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
 536     BdrvRequestFlags flags);
 537 int coroutine_fn bdrv_co_pwritev(BlockDriverState *bs,
 538     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
 539     BdrvRequestFlags flags);
 540
 541 int get_tmp_filename(char *filename, int size);
 542 BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size,
 543                             const char *filename);
 544
 545
 546 /**
 547  * bdrv_add_before_write_notifier:
 548  *
 549  * Register a callback that is invoked before write requests are processed but
 550  * after any throttling or waiting for overlapping requests.
 551  */
 552 void bdrv_add_before_write_notifier(BlockDriverState *bs,
 553                                     NotifierWithReturn *notifier);
 554
 555 /**
 556  * bdrv_detach_aio_context:
 557  *
 558  * May be called from .bdrv_detach_aio_context() to detach children from the
 559  * current #AioContext.  This is only needed by block drivers that manage their
 560  * own children.  Both ->file and ->backing are automatically handled and
 561  * block drivers should not call this function on them explicitly.
 562  */
 563 void bdrv_detach_aio_context(BlockDriverState *bs);
 564
 565 /**
 566  * bdrv_attach_aio_context:
 567  *
 568  * May be called from .bdrv_attach_aio_context() to attach children to the new
 569  * #AioContext.  This is only needed by block drivers that manage their own
 570  * children.  Both ->file and ->backing are automatically handled and block
 571  * drivers should not call this function on them explicitly.
 572  */
 573 void bdrv_attach_aio_context(BlockDriverState *bs,
 574                              AioContext *new_context);
 575
 576 /**
 577  * bdrv_add_aio_context_notifier:
 578  *
 579  * If a long-running job intends to be always run in the same AioContext as a
 580  * certain BDS, it may use this function to be notified of changes regarding the
 581  * association of the BDS to an AioContext.
 582  *
 583  * attached_aio_context() is called after the target BDS has been attached to a
 584  * new AioContext; detach_aio_context() is called before the target BDS is being
 585  * detached from its old AioContext.
 586  */
 587 void bdrv_add_aio_context_notifier(BlockDriverState *bs,
 588         void (*attached_aio_context)(AioContext *new_context, void *opaque),
 589         void (*detach_aio_context)(void *opaque), void *opaque);
 590
 591 /**
 592  * bdrv_remove_aio_context_notifier:
 593  *
 594  * Unsubscribe of change notifications regarding the BDS's AioContext. The
 595  * parameters given here have to be the same as those given to
 596  * bdrv_add_aio_context_notifier().
 597  */
 598 void bdrv_remove_aio_context_notifier(BlockDriverState *bs,
 599                                       void (*aio_context_attached)(AioContext *,
 600                                                                    void *),
 601                                       void (*aio_context_detached)(void *),
 602                                       void *opaque);
 603
 604 #ifdef _WIN32
 605 int is_windows_drive(const char *filename);
 606 #endif
 607
 608 /**
 609  * stream_start:
 610  * @bs: Block device to operate on.
 611  * @base: Block device that will become the new base, or %NULL to
 612  * flatten the whole backing file chain onto @bs.
 613  * @base_id: The file name that will be written to @bs as the new
 614  * backing file if the job completes.  Ignored if @base is %NULL.
 615  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
 616  * @on_error: The action to take upon error.
 617  * @cb: Completion function for the job.
 618  * @opaque: Opaque pointer value passed to @cb.
 619  * @errp: Error object.
 620  *
 621  * Start a streaming operation on @bs.  Clusters that are unallocated
 622  * in @bs, but allocated in any image between @base and @bs (both
 623  * exclusive) will be written to @bs.  At the end of a successful
 624  * streaming job, the backing file of @bs will be changed to
 625  * @base_id in the written image and to @base in the live BlockDriverState.
 626  */
 627 void stream_start(BlockDriverState *bs, BlockDriverState *base,
 628                   const char *base_id, int64_t speed, BlockdevOnError on_error,
 629                   BlockCompletionFunc *cb,
 630                   void *opaque, Error **errp);
 631
 632 /**
 633  * commit_start:
 634  * @bs: Active block device.
 635  * @top: Top block device to be committed.
 636  * @base: Block device that will be written into, and become the new top.
 637  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
 638  * @on_error: The action to take upon error.
 639  * @cb: Completion function for the job.
 640  * @opaque: Opaque pointer value passed to @cb.
 641  * @backing_file_str: String to use as the backing file in @top's overlay
 642  * @errp: Error object.
 643  *
 644  */
 645 void commit_start(BlockDriverState *bs, BlockDriverState *base,
 646                  BlockDriverState *top, int64_t speed,
 647                  BlockdevOnError on_error, BlockCompletionFunc *cb,
 648                  void *opaque, const char *backing_file_str, Error **errp);
 649 /**
 650  * commit_active_start:
 651  * @bs: Active block device to be committed.
 652  * @base: Block device that will be written into, and become the new top.
 653  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
 654  * @on_error: The action to take upon error.
 655  * @cb: Completion function for the job.
 656  * @opaque: Opaque pointer value passed to @cb.
 657  * @errp: Error object.
 658  *
 659  */
 660 void commit_active_start(BlockDriverState *bs, BlockDriverState *base,
 661                          int64_t speed,
 662                          BlockdevOnError on_error,
 663                          BlockCompletionFunc *cb,
 664                          void *opaque, Error **errp);
 665 /*
 666  * mirror_start:
 667  * @bs: Block device to operate on.
 668  * @target: Block device to write to.
 669  * @replaces: Block graph node name to replace once the mirror is done. Can
 670  *            only be used when full mirroring is selected.
 671  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
 672  * @granularity: The chosen granularity for the dirty bitmap.
 673  * @buf_size: The amount of data that can be in flight at one time.
 674  * @mode: Whether to collapse all images in the chain to the target.
 675  * @on_source_error: The action to take upon error reading from the source.
 676  * @on_target_error: The action to take upon error writing to the target.
 677  * @unmap: Whether to unmap target where source sectors only contain zeroes.
 678  * @cb: Completion function for the job.
 679  * @opaque: Opaque pointer value passed to @cb.
 680  * @errp: Error object.
 681  *
 682  * Start a mirroring operation on @bs.  Clusters that are allocated
 683  * in @bs will be written to @bs until the job is cancelled or
 684  * manually completed.  At the end of a successful mirroring job,
 685  * @bs will be switched to read from @target.
 686  */
 687 void mirror_start(BlockDriverState *bs, BlockDriverState *target,
 688                   const char *replaces,
 689                   int64_t speed, uint32_t granularity, int64_t buf_size,
 690                   MirrorSyncMode mode, BlockdevOnError on_source_error,
 691                   BlockdevOnError on_target_error,
 692                   bool unmap,
 693                   BlockCompletionFunc *cb,
 694                   void *opaque, Error **errp);
 695
 696 /*
 697  * backup_start:
 698  * @bs: Block device to operate on.
 699  * @target: Block device to write to.
 700  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
 701  * @sync_mode: What parts of the disk image should be copied to the destination.
 702  * @sync_bitmap: The dirty bitmap if sync_mode is MIRROR_SYNC_MODE_INCREMENTAL.
 703  * @on_source_error: The action to take upon error reading from the source.
 704  * @on_target_error: The action to take upon error writing to the target.
 705  * @cb: Completion function for the job.
 706  * @opaque: Opaque pointer value passed to @cb.
 707  * @txn: Transaction that this job is part of (may be NULL).
 708  *
 709  * Start a backup operation on @bs.  Clusters in @bs are written to @target
 710  * until the job is cancelled or manually completed.
 711  */
 712 void backup_start(BlockDriverState *bs, BlockDriverState *target,
 713                   int64_t speed, MirrorSyncMode sync_mode,
 714                   BdrvDirtyBitmap *sync_bitmap,
 715                   BlockdevOnError on_source_error,
 716                   BlockdevOnError on_target_error,
 717                   BlockCompletionFunc *cb, void *opaque,
 718                   BlockJobTxn *txn, Error **errp);
 719
 720 void hmp_drive_add_node(Monitor *mon, const char *optstr);
 721
 722 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
 723                                   const char *child_name,
 724                                   const BdrvChildRole *child_role,
 725                                   void *opaque);
 726 void bdrv_root_unref_child(BdrvChild *child);
 727
 728 const char *bdrv_get_parent_name(const BlockDriverState *bs);
 729 void blk_dev_change_media_cb(BlockBackend *blk, bool load);
 730 bool blk_dev_has_removable_media(BlockBackend *blk);
 731 bool blk_dev_has_tray(BlockBackend *blk);
 732 void blk_dev_eject_request(BlockBackend *blk, bool force);
 733 bool blk_dev_is_tray_open(BlockBackend *blk);
 734 bool blk_dev_is_medium_locked(BlockBackend *blk);
 735
 736 void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector, int nr_sectors);
 737 bool bdrv_requests_pending(BlockDriverState *bs);
 738
 739 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap **out);
 740 void bdrv_undo_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap *in);
 741
 742 void blockdev_close_all_bdrv_states(void);
 743
 744 #endif /* BLOCK_INT_H */