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 coroutine_fn (*bdrv_save_vmstate)(BlockDriverState *bs,
 228                                           QEMUIOVector *qiov,
 229                                           int64_t pos);
 230     int coroutine_fn (*bdrv_load_vmstate)(BlockDriverState *bs,
 231                                           QEMUIOVector *qiov,
 232                                           int64_t pos);
 233
 234     int (*bdrv_change_backing_file)(BlockDriverState *bs,
 235         const char *backing_file, const char *backing_fmt);
 236
 237     /* removable device specific */
 238     bool (*bdrv_is_inserted)(BlockDriverState *bs);
 239     int (*bdrv_media_changed)(BlockDriverState *bs);
 240     void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
 241     void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
 242
 243     /* to control generic scsi devices */
 244     BlockAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
 245         unsigned long int req, void *buf,
 246         BlockCompletionFunc *cb, void *opaque);
 247
 248     /* List of options for creating images, terminated by name == NULL */
 249     QemuOptsList *create_opts;
 250
 251     /*
 252      * Returns 0 for completed check, -errno for internal errors.
 253      * The check results are stored in result.
 254      */
 255     int (*bdrv_check)(BlockDriverState* bs, BdrvCheckResult *result,
 256         BdrvCheckMode fix);
 257
 258     int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts,
 259                               BlockDriverAmendStatusCB *status_cb,
 260                               void *cb_opaque);
 261
 262     void (*bdrv_debug_event)(BlockDriverState *bs, BlkdebugEvent event);
 263
 264     /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
 265     int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
 266         const char *tag);
 267     int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs,
 268         const char *tag);
 269     int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
 270     bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
 271
 272     void (*bdrv_refresh_limits)(BlockDriverState *bs, Error **errp);
 273
 274     /*
 275      * Returns 1 if newly created images are guaranteed to contain only
 276      * zeros, 0 otherwise.
 277      */
 278     int (*bdrv_has_zero_init)(BlockDriverState *bs);
 279
 280     /* Remove fd handlers, timers, and other event loop callbacks so the event
 281      * loop is no longer in use.  Called with no in-flight requests and in
 282      * depth-first traversal order with parents before child nodes.
 283      */
 284     void (*bdrv_detach_aio_context)(BlockDriverState *bs);
 285
 286     /* Add fd handlers, timers, and other event loop callbacks so I/O requests
 287      * can be processed again.  Called with no in-flight requests and in
 288      * depth-first traversal order with child nodes before parent nodes.
 289      */
 290     void (*bdrv_attach_aio_context)(BlockDriverState *bs,
 291                                     AioContext *new_context);
 292
 293     /* io queue for linux-aio */
 294     void (*bdrv_io_plug)(BlockDriverState *bs);
 295     void (*bdrv_io_unplug)(BlockDriverState *bs);
 296
 297     /**
 298      * Try to get @bs's logical and physical block size.
 299      * On success, store them in @bsz and return zero.
 300      * On failure, return negative errno.
 301      */
 302     int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz);
 303     /**
 304      * Try to get @bs's geometry (cyls, heads, sectors)
 305      * On success, store them in @geo and return 0.
 306      * On failure return -errno.
 307      * Only drivers that want to override guest geometry implement this
 308      * callback; see hd_geometry_guess().
 309      */
 310     int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo);
 311
 312     /**
 313      * Drain and stop any internal sources of requests in the driver, and
 314      * remain so until next I/O callback (e.g. bdrv_co_writev) is called.
 315      */
 316     void (*bdrv_drain)(BlockDriverState *bs);
 317
 318     void (*bdrv_add_child)(BlockDriverState *parent, BlockDriverState *child,
 319                            Error **errp);
 320     void (*bdrv_del_child)(BlockDriverState *parent, BdrvChild *child,
 321                            Error **errp);
 322
 323     QLIST_ENTRY(BlockDriver) list;
 324 };
 325
 326 typedef struct BlockLimits {
 327     /* Alignment requirement, in bytes, for offset/length of I/O
 328      * requests. Must be a power of 2 less than INT_MAX; defaults to
 329      * 1 for drivers with modern byte interfaces, and to 512
 330      * otherwise. */
 331     uint32_t request_alignment;
 332
 333     /* maximum number of bytes that can be discarded at once (since it
 334      * is signed, it must be < 2G, if set), should be multiple of
 335      * pdiscard_alignment, but need not be power of 2. May be 0 if no
 336      * inherent 32-bit limit */
 337     int32_t max_pdiscard;
 338
 339     /* optimal alignment for discard requests in bytes, must be power
 340      * of 2, less than max_pdiscard if that is set, and multiple of
 341      * bl.request_alignment. May be 0 if bl.request_alignment is good
 342      * enough */
 343     uint32_t pdiscard_alignment;
 344
 345     /* maximum number of bytes that can zeroized at once (since it is
 346      * signed, it must be < 2G, if set), should be multiple of
 347      * pwrite_zeroes_alignment. May be 0 if no inherent 32-bit limit */
 348     int32_t max_pwrite_zeroes;
 349
 350     /* optimal alignment for write zeroes requests in bytes, must be
 351      * power of 2, less than max_pwrite_zeroes if that is set, and
 352      * multiple of bl.request_alignment. May be 0 if
 353      * bl.request_alignment is good enough */
 354     uint32_t pwrite_zeroes_alignment;
 355
 356     /* optimal transfer length in bytes (must be power of 2, and
 357      * multiple of bl.request_alignment), or 0 if no preferred size */
 358     uint32_t opt_transfer;
 359
 360     /* maximal transfer length in bytes (need not be power of 2, but
 361      * should be multiple of opt_transfer), or 0 for no 32-bit limit.
 362      * For now, anything larger than INT_MAX is clamped down. */
 363     uint32_t max_transfer;
 364
 365     /* memory alignment, in bytes so that no bounce buffer is needed */
 366     size_t min_mem_alignment;
 367
 368     /* memory alignment, in bytes, for bounce buffer */
 369     size_t opt_mem_alignment;
 370
 371     /* maximum number of iovec elements */
 372     int max_iov;
 373 } BlockLimits;
 374
 375 typedef struct BdrvOpBlocker BdrvOpBlocker;
 376
 377 typedef struct BdrvAioNotifier {
 378     void (*attached_aio_context)(AioContext *new_context, void *opaque);
 379     void (*detach_aio_context)(void *opaque);
 380
 381     void *opaque;
 382     bool deleted;
 383
 384     QLIST_ENTRY(BdrvAioNotifier) list;
 385 } BdrvAioNotifier;
 386
 387 struct BdrvChildRole {
 388     void (*inherit_options)(int *child_flags, QDict *child_options,
 389                             int parent_flags, QDict *parent_options);
 390
 391     void (*change_media)(BdrvChild *child, bool load);
 392     void (*resize)(BdrvChild *child);
 393
 394     /* Returns a name that is supposedly more useful for human users than the
 395      * node name for identifying the node in question (in particular, a BB
 396      * name), or NULL if the parent can't provide a better name. */
 397     const char* (*get_name)(BdrvChild *child);
 398
 399     /*
 400      * If this pair of functions is implemented, the parent doesn't issue new
 401      * requests after returning from .drained_begin() until .drained_end() is
 402      * called.
 403      *
 404      * Note that this can be nested. If drained_begin() was called twice, new
 405      * I/O is allowed only after drained_end() was called twice, too.
 406      */
 407     void (*drained_begin)(BdrvChild *child);
 408     void (*drained_end)(BdrvChild *child);
 409 };
 410
 411 extern const BdrvChildRole child_file;
 412 extern const BdrvChildRole child_format;
 413
 414 struct BdrvChild {
 415     BlockDriverState *bs;
 416     char *name;
 417     const BdrvChildRole *role;
 418     void *opaque;
 419     QLIST_ENTRY(BdrvChild) next;
 420     QLIST_ENTRY(BdrvChild) next_parent;
 421 };
 422
 423 /*
 424  * Note: the function bdrv_append() copies and swaps contents of
 425  * BlockDriverStates, so if you add new fields to this struct, please
 426  * inspect bdrv_append() to determine if the new fields need to be
 427  * copied as well.
 428  */
 429 struct BlockDriverState {
 430     int64_t total_sectors; /* if we are reading a disk image, give its
 431                               size in sectors */
 432     int open_flags; /* flags used to open the file, re-used for re-open */
 433     bool read_only; /* if true, the media is read only */
 434     bool encrypted; /* if true, the media is encrypted */
 435     bool valid_key; /* if true, a valid encryption key has been set */
 436     bool sg;        /* if true, the device is a /dev/sg* */
 437     bool probed;    /* if true, format was probed rather than specified */
 438
 439     int copy_on_read; /* if nonzero, copy read backing sectors into image.
 440                          note this is a reference count */
 441
 442     BlockDriver *drv; /* NULL means no media */
 443     void *opaque;
 444
 445     AioContext *aio_context; /* event loop used for fd handlers, timers, etc */
 446     /* long-running tasks intended to always use the same AioContext as this
 447      * BDS may register themselves in this list to be notified of changes
 448      * regarding this BDS's context */
 449     QLIST_HEAD(, BdrvAioNotifier) aio_notifiers;
 450     bool walking_aio_notifiers; /* to make removal during iteration safe */
 451
 452     char filename[PATH_MAX];
 453     char backing_file[PATH_MAX]; /* if non zero, the image is a diff of
 454                                     this file image */
 455     char backing_format[16]; /* if non-zero and backing_file exists */
 456
 457     QDict *full_open_options;
 458     char exact_filename[PATH_MAX];
 459
 460     BdrvChild *backing;
 461     BdrvChild *file;
 462
 463     /* Callback before write request is processed */
 464     NotifierWithReturnList before_write_notifiers;
 465
 466     /* number of in-flight serialising requests */
 467     unsigned int serialising_in_flight;
 468
 469     /* Offset after the highest byte written to */
 470     uint64_t wr_highest_offset;
 471
 472     /* I/O Limits */
 473     BlockLimits bl;
 474
 475     /* Flags honored during pwrite (so far: BDRV_REQ_FUA) */
 476     unsigned int supported_write_flags;
 477     /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA,
 478      * BDRV_REQ_MAY_UNMAP) */
 479     unsigned int supported_zero_flags;
 480
 481     /* the following member gives a name to every node on the bs graph. */
 482     char node_name[32];
 483     /* element of the list of named nodes building the graph */
 484     QTAILQ_ENTRY(BlockDriverState) node_list;
 485     /* element of the list of all BlockDriverStates (all_bdrv_states) */
 486     QTAILQ_ENTRY(BlockDriverState) bs_list;
 487     /* element of the list of monitor-owned BDS */
 488     QTAILQ_ENTRY(BlockDriverState) monitor_list;
 489     QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps;
 490     int refcnt;
 491
 492     QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
 493
 494     /* operation blockers */
 495     QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX];
 496
 497     /* long-running background operation */
 498     BlockJob *job;
 499
 500     /* The node that this node inherited default options from (and a reopen on
 501      * which can affect this node by changing these defaults). This is always a
 502      * parent node of this node. */
 503     BlockDriverState *inherits_from;
 504     QLIST_HEAD(, BdrvChild) children;
 505     QLIST_HEAD(, BdrvChild) parents;
 506
 507     QDict *options;
 508     QDict *explicit_options;
 509     BlockdevDetectZeroesOptions detect_zeroes;
 510
 511     /* The error object in use for blocking operations on backing_hd */
 512     Error *backing_blocker;
 513
 514     /* threshold limit for writes, in bytes. "High water mark". */
 515     uint64_t write_threshold_offset;
 516     NotifierWithReturn write_threshold_notifier;
 517
 518     /* counters for nested bdrv_io_plug and bdrv_io_unplugged_begin */
 519     unsigned io_plugged;
 520     unsigned io_plug_disabled;
 521
 522     int quiesce_counter;
 523 };
 524
 525 struct BlockBackendRootState {
 526     int open_flags;
 527     bool read_only;
 528     BlockdevDetectZeroesOptions detect_zeroes;
 529 };
 530
 531 typedef enum BlockMirrorBackingMode {
 532     /* Reuse the existing backing chain from the source for the target.
 533      * - sync=full: Set backing BDS to NULL.
 534      * - sync=top:  Use source's backing BDS.
 535      * - sync=none: Use source as the backing BDS. */
 536     MIRROR_SOURCE_BACKING_CHAIN,
 537
 538     /* Open the target's backing chain completely anew */
 539     MIRROR_OPEN_BACKING_CHAIN,
 540
 541     /* Do not change the target's backing BDS after job completion */
 542     MIRROR_LEAVE_BACKING_CHAIN,
 543 } BlockMirrorBackingMode;
 544
 545 static inline BlockDriverState *backing_bs(BlockDriverState *bs)
 546 {
 547     return bs->backing ? bs->backing->bs : NULL;
 548 }
 549
 550
 551 /* Essential block drivers which must always be statically linked into qemu, and
 552  * which therefore can be accessed without using bdrv_find_format() */
 553 extern BlockDriver bdrv_file;
 554 extern BlockDriver bdrv_raw;
 555 extern BlockDriver bdrv_qcow2;
 556
 557 /**
 558  * bdrv_setup_io_funcs:
 559  *
 560  * Prepare a #BlockDriver for I/O request processing by populating
 561  * unimplemented coroutine and AIO interfaces with generic wrapper functions
 562  * that fall back to implemented interfaces.
 563  */
 564 void bdrv_setup_io_funcs(BlockDriver *bdrv);
 565
 566 int coroutine_fn bdrv_co_preadv(BdrvChild *child,
 567     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
 568     BdrvRequestFlags flags);
 569 int coroutine_fn bdrv_co_pwritev(BdrvChild *child,
 570     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
 571     BdrvRequestFlags flags);
 572
 573 int get_tmp_filename(char *filename, int size);
 574 BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size,
 575                             const char *filename);
 576
 577
 578 /**
 579  * bdrv_add_before_write_notifier:
 580  *
 581  * Register a callback that is invoked before write requests are processed but
 582  * after any throttling or waiting for overlapping requests.
 583  */
 584 void bdrv_add_before_write_notifier(BlockDriverState *bs,
 585                                     NotifierWithReturn *notifier);
 586
 587 /**
 588  * bdrv_detach_aio_context:
 589  *
 590  * May be called from .bdrv_detach_aio_context() to detach children from the
 591  * current #AioContext.  This is only needed by block drivers that manage their
 592  * own children.  Both ->file and ->backing are automatically handled and
 593  * block drivers should not call this function on them explicitly.
 594  */
 595 void bdrv_detach_aio_context(BlockDriverState *bs);
 596
 597 /**
 598  * bdrv_attach_aio_context:
 599  *
 600  * May be called from .bdrv_attach_aio_context() to attach children to the new
 601  * #AioContext.  This is only needed by block drivers that manage their own
 602  * children.  Both ->file and ->backing are automatically handled and block
 603  * drivers should not call this function on them explicitly.
 604  */
 605 void bdrv_attach_aio_context(BlockDriverState *bs,
 606                              AioContext *new_context);
 607
 608 /**
 609  * bdrv_add_aio_context_notifier:
 610  *
 611  * If a long-running job intends to be always run in the same AioContext as a
 612  * certain BDS, it may use this function to be notified of changes regarding the
 613  * association of the BDS to an AioContext.
 614  *
 615  * attached_aio_context() is called after the target BDS has been attached to a
 616  * new AioContext; detach_aio_context() is called before the target BDS is being
 617  * detached from its old AioContext.
 618  */
 619 void bdrv_add_aio_context_notifier(BlockDriverState *bs,
 620         void (*attached_aio_context)(AioContext *new_context, void *opaque),
 621         void (*detach_aio_context)(void *opaque), void *opaque);
 622
 623 /**
 624  * bdrv_remove_aio_context_notifier:
 625  *
 626  * Unsubscribe of change notifications regarding the BDS's AioContext. The
 627  * parameters given here have to be the same as those given to
 628  * bdrv_add_aio_context_notifier().
 629  */
 630 void bdrv_remove_aio_context_notifier(BlockDriverState *bs,
 631                                       void (*aio_context_attached)(AioContext *,
 632                                                                    void *),
 633                                       void (*aio_context_detached)(void *),
 634                                       void *opaque);
 635
 636 #ifdef _WIN32
 637 int is_windows_drive(const char *filename);
 638 #endif
 639
 640 /**
 641  * stream_start:
 642  * @job_id: The id of the newly-created job, or %NULL to use the
 643  * device name of @bs.
 644  * @bs: Block device to operate on.
 645  * @base: Block device that will become the new base, or %NULL to
 646  * flatten the whole backing file chain onto @bs.
 647  * @backing_file_str: The file name that will be written to @bs as the
 648  * the new backing file if the job completes. Ignored if @base is %NULL.
 649  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
 650  * @on_error: The action to take upon error.
 651  * @cb: Completion function for the job.
 652  * @opaque: Opaque pointer value passed to @cb.
 653  * @errp: Error object.
 654  *
 655  * Start a streaming operation on @bs.  Clusters that are unallocated
 656  * in @bs, but allocated in any image between @base and @bs (both
 657  * exclusive) will be written to @bs.  At the end of a successful
 658  * streaming job, the backing file of @bs will be changed to
 659  * @backing_file_str in the written image and to @base in the live
 660  * BlockDriverState.
 661  */
 662 void stream_start(const char *job_id, BlockDriverState *bs,
 663                   BlockDriverState *base, const char *backing_file_str,
 664                   int64_t speed, BlockdevOnError on_error,
 665                   BlockCompletionFunc *cb, void *opaque, Error **errp);
 666
 667 /**
 668  * commit_start:
 669  * @job_id: The id of the newly-created job, or %NULL to use the
 670  * device name of @bs.
 671  * @bs: Active block device.
 672  * @top: Top block device to be committed.
 673  * @base: Block device that will be written into, and become the new top.
 674  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
 675  * @on_error: The action to take upon error.
 676  * @cb: Completion function for the job.
 677  * @opaque: Opaque pointer value passed to @cb.
 678  * @backing_file_str: String to use as the backing file in @top's overlay
 679  * @errp: Error object.
 680  *
 681  */
 682 void commit_start(const char *job_id, BlockDriverState *bs,
 683                   BlockDriverState *base, BlockDriverState *top, int64_t speed,
 684                   BlockdevOnError on_error, BlockCompletionFunc *cb,
 685                   void *opaque, const char *backing_file_str, Error **errp);
 686 /**
 687  * commit_active_start:
 688  * @job_id: The id of the newly-created job, or %NULL to use the
 689  * device name of @bs.
 690  * @bs: Active block device to be committed.
 691  * @base: Block device that will be written into, and become the new top.
 692  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
 693  * @on_error: The action to take upon error.
 694  * @cb: Completion function for the job.
 695  * @opaque: Opaque pointer value passed to @cb.
 696  * @errp: Error object.
 697  *
 698  */
 699 void commit_active_start(const char *job_id, BlockDriverState *bs,
 700                          BlockDriverState *base, int64_t speed,
 701                          BlockdevOnError on_error,
 702                          BlockCompletionFunc *cb,
 703                          void *opaque, Error **errp);
 704 /*
 705  * mirror_start:
 706  * @job_id: The id of the newly-created job, or %NULL to use the
 707  * device name of @bs.
 708  * @bs: Block device to operate on.
 709  * @target: Block device to write to.
 710  * @replaces: Block graph node name to replace once the mirror is done. Can
 711  *            only be used when full mirroring is selected.
 712  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
 713  * @granularity: The chosen granularity for the dirty bitmap.
 714  * @buf_size: The amount of data that can be in flight at one time.
 715  * @mode: Whether to collapse all images in the chain to the target.
 716  * @backing_mode: How to establish the target's backing chain after completion.
 717  * @on_source_error: The action to take upon error reading from the source.
 718  * @on_target_error: The action to take upon error writing to the target.
 719  * @unmap: Whether to unmap target where source sectors only contain zeroes.
 720  * @cb: Completion function for the job.
 721  * @opaque: Opaque pointer value passed to @cb.
 722  * @errp: Error object.
 723  *
 724  * Start a mirroring operation on @bs.  Clusters that are allocated
 725  * in @bs will be written to @bs until the job is cancelled or
 726  * manually completed.  At the end of a successful mirroring job,
 727  * @bs will be switched to read from @target.
 728  */
 729 void mirror_start(const char *job_id, BlockDriverState *bs,
 730                   BlockDriverState *target, const char *replaces,
 731                   int64_t speed, uint32_t granularity, int64_t buf_size,
 732                   MirrorSyncMode mode, BlockMirrorBackingMode backing_mode,
 733                   BlockdevOnError on_source_error,
 734                   BlockdevOnError on_target_error,
 735                   bool unmap,
 736                   BlockCompletionFunc *cb,
 737                   void *opaque, Error **errp);
 738
 739 /*
 740  * backup_start:
 741  * @job_id: The id of the newly-created job, or %NULL to use the
 742  * device name of @bs.
 743  * @bs: Block device to operate on.
 744  * @target: Block device to write to.
 745  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
 746  * @sync_mode: What parts of the disk image should be copied to the destination.
 747  * @sync_bitmap: The dirty bitmap if sync_mode is MIRROR_SYNC_MODE_INCREMENTAL.
 748  * @on_source_error: The action to take upon error reading from the source.
 749  * @on_target_error: The action to take upon error writing to the target.
 750  * @cb: Completion function for the job.
 751  * @opaque: Opaque pointer value passed to @cb.
 752  * @txn: Transaction that this job is part of (may be NULL).
 753  *
 754  * Start a backup operation on @bs.  Clusters in @bs are written to @target
 755  * until the job is cancelled or manually completed.
 756  */
 757 void backup_start(const char *job_id, BlockDriverState *bs,
 758                   BlockDriverState *target, int64_t speed,
 759                   MirrorSyncMode sync_mode, BdrvDirtyBitmap *sync_bitmap,
 760                   BlockdevOnError on_source_error,
 761                   BlockdevOnError on_target_error,
 762                   BlockCompletionFunc *cb, void *opaque,
 763                   BlockJobTxn *txn, Error **errp);
 764
 765 void hmp_drive_add_node(Monitor *mon, const char *optstr);
 766
 767 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
 768                                   const char *child_name,
 769                                   const BdrvChildRole *child_role,
 770                                   void *opaque);
 771 void bdrv_root_unref_child(BdrvChild *child);
 772
 773 const char *bdrv_get_parent_name(const BlockDriverState *bs);
 774 void blk_dev_change_media_cb(BlockBackend *blk, bool load);
 775 bool blk_dev_has_removable_media(BlockBackend *blk);
 776 bool blk_dev_has_tray(BlockBackend *blk);
 777 void blk_dev_eject_request(BlockBackend *blk, bool force);
 778 bool blk_dev_is_tray_open(BlockBackend *blk);
 779 bool blk_dev_is_medium_locked(BlockBackend *blk);
 780
 781 void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector, int nr_sectors);
 782 bool bdrv_requests_pending(BlockDriverState *bs);
 783
 784 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap **out);
 785 void bdrv_undo_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap *in);
 786
 787 void blockdev_close_all_bdrv_states(void);
 788
 789 #endif /* BLOCK_INT_H */