include/block/block.h

   1 #ifndef BLOCK_H
   2 #define BLOCK_H
   3
   4 #include "block/aio.h"
   5 #include "qapi/qapi-types-block-core.h"
   6 #include "qemu/iov.h"
   7 #include "qemu/coroutine.h"
   8 #include "block/accounting.h"
   9 #include "block/dirty-bitmap.h"
  10 #include "block/blockjob.h"
  11 #include "qemu/hbitmap.h"
  12
  13 /* block.c */
  14 typedef struct BlockDriver BlockDriver;
  15 typedef struct BdrvChild BdrvChild;
  16 typedef struct BdrvChildRole BdrvChildRole;
  17
  18 typedef struct BlockDriverInfo {
  19     /* in bytes, 0 if irrelevant */
  20     int cluster_size;
  21     /* offset at which the VM state can be saved (0 if not possible) */
  22     int64_t vm_state_offset;
  23     bool is_dirty;
  24     /*
  25      * True if unallocated blocks read back as zeroes. This is equivalent
  26      * to the LBPRZ flag in the SCSI logical block provisioning page.
  27      */
  28     bool unallocated_blocks_are_zero;
  29     /*
  30      * True if this block driver only supports compressed writes
  31      */
  32     bool needs_compressed_writes;
  33 } BlockDriverInfo;
  34
  35 typedef struct BlockFragInfo {
  36     uint64_t allocated_clusters;
  37     uint64_t total_clusters;
  38     uint64_t fragmented_clusters;
  39     uint64_t compressed_clusters;
  40 } BlockFragInfo;
  41
  42 typedef enum {
  43     BDRV_REQ_COPY_ON_READ       = 0x1,
  44     BDRV_REQ_ZERO_WRITE         = 0x2,
  45     /* The BDRV_REQ_MAY_UNMAP flag is used to indicate that the block driver
  46      * is allowed to optimize a write zeroes request by unmapping (discarding)
  47      * blocks if it is guaranteed that the result will read back as
  48      * zeroes. The flag is only passed to the driver if the block device is
  49      * opened with BDRV_O_UNMAP.
  50      */
  51     BDRV_REQ_MAY_UNMAP          = 0x4,
  52     BDRV_REQ_NO_SERIALISING     = 0x8,
  53     BDRV_REQ_FUA                = 0x10,
  54     BDRV_REQ_WRITE_COMPRESSED   = 0x20,
  55
  56     /* Mask of valid flags */
  57     BDRV_REQ_MASK               = 0x3f,
  58 } BdrvRequestFlags;
  59
  60 typedef struct BlockSizes {
  61     uint32_t phys;
  62     uint32_t log;
  63 } BlockSizes;
  64
  65 typedef struct HDGeometry {
  66     uint32_t heads;
  67     uint32_t sectors;
  68     uint32_t cylinders;
  69 } HDGeometry;
  70
  71 #define BDRV_O_RDWR        0x0002
  72 #define BDRV_O_RESIZE      0x0004 /* request permission for resizing the node */
  73 #define BDRV_O_SNAPSHOT    0x0008 /* open the file read only and save writes in a snapshot */
  74 #define BDRV_O_TEMPORARY   0x0010 /* delete the file after use */
  75 #define BDRV_O_NOCACHE     0x0020 /* do not use the host page cache */
  76 #define BDRV_O_NATIVE_AIO  0x0080 /* use native AIO instead of the thread pool */
  77 #define BDRV_O_NO_BACKING  0x0100 /* don't open the backing file */
  78 #define BDRV_O_NO_FLUSH    0x0200 /* disable flushing on this disk */
  79 #define BDRV_O_COPY_ON_READ 0x0400 /* copy read backing sectors into image */
  80 #define BDRV_O_INACTIVE    0x0800  /* consistency hint for migration handoff */
  81 #define BDRV_O_CHECK       0x1000  /* open solely for consistency check */
  82 #define BDRV_O_ALLOW_RDWR  0x2000  /* allow reopen to change from r/o to r/w */
  83 #define BDRV_O_UNMAP       0x4000  /* execute guest UNMAP/TRIM operations */
  84 #define BDRV_O_PROTOCOL    0x8000  /* if no block driver is explicitly given:
  85                                       select an appropriate protocol driver,
  86                                       ignoring the format layer */
  87 #define BDRV_O_NO_IO       0x10000 /* don't initialize for I/O */
  88
  89 #define BDRV_O_CACHE_MASK  (BDRV_O_NOCACHE | BDRV_O_NO_FLUSH)
  90
  91
  92 /* Option names of options parsed by the block layer */
  93
  94 #define BDRV_OPT_CACHE_WB       "cache.writeback"
  95 #define BDRV_OPT_CACHE_DIRECT   "cache.direct"
  96 #define BDRV_OPT_CACHE_NO_FLUSH "cache.no-flush"
  97 #define BDRV_OPT_READ_ONLY      "read-only"
  98 #define BDRV_OPT_DISCARD        "discard"
  99 #define BDRV_OPT_FORCE_SHARE    "force-share"
 100
 101
 102 #define BDRV_SECTOR_BITS   9
 103 #define BDRV_SECTOR_SIZE   (1ULL << BDRV_SECTOR_BITS)
 104 #define BDRV_SECTOR_MASK   ~(BDRV_SECTOR_SIZE - 1)
 105
 106 #define BDRV_REQUEST_MAX_SECTORS MIN(SIZE_MAX >> BDRV_SECTOR_BITS, \
 107                                      INT_MAX >> BDRV_SECTOR_BITS)
 108 #define BDRV_REQUEST_MAX_BYTES (BDRV_REQUEST_MAX_SECTORS << BDRV_SECTOR_BITS)
 109
 110 /*
 111  * Allocation status flags for bdrv_block_status() and friends.
 112  *
 113  * Public flags:
 114  * BDRV_BLOCK_DATA: allocation for data at offset is tied to this layer
 115  * BDRV_BLOCK_ZERO: offset reads as zero
 116  * BDRV_BLOCK_OFFSET_VALID: an associated offset exists for accessing raw data
 117  * BDRV_BLOCK_ALLOCATED: the content of the block is determined by this
 118  *                       layer (short for DATA || ZERO), set by block layer
 119  * BDRV_BLOCK_EOF: the returned pnum covers through end of file for this layer
 120  *
 121  * Internal flag:
 122  * BDRV_BLOCK_RAW: for use by passthrough drivers, such as raw, to request
 123  *                 that the block layer recompute the answer from the returned
 124  *                 BDS; must be accompanied by just BDRV_BLOCK_OFFSET_VALID.
 125  *
 126  * If BDRV_BLOCK_OFFSET_VALID is set, bits 9-62 (BDRV_BLOCK_OFFSET_MASK) of
 127  * the return value (old interface) or the entire map parameter (new
 128  * interface) represent the offset in the returned BDS that is allocated for
 129  * the corresponding raw data.  However, whether that offset actually
 130  * contains data also depends on BDRV_BLOCK_DATA, as follows:
 131  *
 132  * DATA ZERO OFFSET_VALID
 133  *  t    t        t       sectors read as zero, returned file is zero at offset
 134  *  t    f        t       sectors read as valid from file at offset
 135  *  f    t        t       sectors preallocated, read as zero, returned file not
 136  *                        necessarily zero at offset
 137  *  f    f        t       sectors preallocated but read from backing_hd,
 138  *                        returned file contains garbage at offset
 139  *  t    t        f       sectors preallocated, read as zero, unknown offset
 140  *  t    f        f       sectors read from unknown file or offset
 141  *  f    t        f       not allocated or unknown offset, read as zero
 142  *  f    f        f       not allocated or unknown offset, read from backing_hd
 143  */
 144 #define BDRV_BLOCK_DATA         0x01
 145 #define BDRV_BLOCK_ZERO         0x02
 146 #define BDRV_BLOCK_OFFSET_VALID 0x04
 147 #define BDRV_BLOCK_RAW          0x08
 148 #define BDRV_BLOCK_ALLOCATED    0x10
 149 #define BDRV_BLOCK_EOF          0x20
 150 #define BDRV_BLOCK_OFFSET_MASK  BDRV_SECTOR_MASK
 151
 152 typedef QSIMPLEQ_HEAD(BlockReopenQueue, BlockReopenQueueEntry) BlockReopenQueue;
 153
 154 typedef struct BDRVReopenState {
 155     BlockDriverState *bs;
 156     int flags;
 157     uint64_t perm, shared_perm;
 158     QDict *options;
 159     QDict *explicit_options;
 160     void *opaque;
 161 } BDRVReopenState;
 162
 163 /*
 164  * Block operation types
 165  */
 166 typedef enum BlockOpType {
 167     BLOCK_OP_TYPE_BACKUP_SOURCE,
 168     BLOCK_OP_TYPE_BACKUP_TARGET,
 169     BLOCK_OP_TYPE_CHANGE,
 170     BLOCK_OP_TYPE_COMMIT_SOURCE,
 171     BLOCK_OP_TYPE_COMMIT_TARGET,
 172     BLOCK_OP_TYPE_DATAPLANE,
 173     BLOCK_OP_TYPE_DRIVE_DEL,
 174     BLOCK_OP_TYPE_EJECT,
 175     BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
 176     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT,
 177     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE,
 178     BLOCK_OP_TYPE_MIRROR_SOURCE,
 179     BLOCK_OP_TYPE_MIRROR_TARGET,
 180     BLOCK_OP_TYPE_RESIZE,
 181     BLOCK_OP_TYPE_STREAM,
 182     BLOCK_OP_TYPE_REPLACE,
 183     BLOCK_OP_TYPE_MAX,
 184 } BlockOpType;
 185
 186 /* Block node permission constants */
 187 enum {
 188     /**
 189      * A user that has the "permission" of consistent reads is guaranteed that
 190      * their view of the contents of the block device is complete and
 191      * self-consistent, representing the contents of a disk at a specific
 192      * point.
 193      *
 194      * For most block devices (including their backing files) this is true, but
 195      * the property cannot be maintained in a few situations like for
 196      * intermediate nodes of a commit block job.
 197      */
 198     BLK_PERM_CONSISTENT_READ    = 0x01,
 199
 200     /** This permission is required to change the visible disk contents. */
 201     BLK_PERM_WRITE              = 0x02,
 202
 203     /**
 204      * This permission (which is weaker than BLK_PERM_WRITE) is both enough and
 205      * required for writes to the block node when the caller promises that
 206      * the visible disk content doesn't change.
 207      */
 208     BLK_PERM_WRITE_UNCHANGED    = 0x04,
 209
 210     /** This permission is required to change the size of a block node. */
 211     BLK_PERM_RESIZE             = 0x08,
 212
 213     /**
 214      * This permission is required to change the node that this BdrvChild
 215      * points to.
 216      */
 217     BLK_PERM_GRAPH_MOD          = 0x10,
 218
 219     BLK_PERM_ALL                = 0x1f,
 220 };
 221
 222 char *bdrv_perm_names(uint64_t perm);
 223
 224 /* disk I/O throttling */
 225 void bdrv_init(void);
 226 void bdrv_init_with_whitelist(void);
 227 bool bdrv_uses_whitelist(void);
 228 BlockDriver *bdrv_find_protocol(const char *filename,
 229                                 bool allow_protocol_prefix,
 230                                 Error **errp);
 231 BlockDriver *bdrv_find_format(const char *format_name);
 232 int bdrv_create(BlockDriver *drv, const char* filename,
 233                 QemuOpts *opts, Error **errp);
 234 int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp);
 235 BlockDriverState *bdrv_new(void);
 236 void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top,
 237                  Error **errp);
 238 void bdrv_replace_node(BlockDriverState *from, BlockDriverState *to,
 239                        Error **errp);
 240
 241 int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough);
 242 int bdrv_parse_discard_flags(const char *mode, int *flags);
 243 BdrvChild *bdrv_open_child(const char *filename,
 244                            QDict *options, const char *bdref_key,
 245                            BlockDriverState* parent,
 246                            const BdrvChildRole *child_role,
 247                            bool allow_none, Error **errp);
 248 void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd,
 249                          Error **errp);
 250 int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options,
 251                            const char *bdref_key, Error **errp);
 252 BlockDriverState *bdrv_open(const char *filename, const char *reference,
 253                             QDict *options, int flags, Error **errp);
 254 BlockDriverState *bdrv_new_open_driver(BlockDriver *drv, const char *node_name,
 255                                        int flags, Error **errp);
 256 BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue,
 257                                     BlockDriverState *bs,
 258                                     QDict *options, int flags);
 259 int bdrv_reopen_multiple(AioContext *ctx, BlockReopenQueue *bs_queue, Error **errp);
 260 int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp);
 261 int bdrv_reopen_prepare(BDRVReopenState *reopen_state,
 262                         BlockReopenQueue *queue, Error **errp);
 263 void bdrv_reopen_commit(BDRVReopenState *reopen_state);
 264 void bdrv_reopen_abort(BDRVReopenState *reopen_state);
 265 int bdrv_read(BdrvChild *child, int64_t sector_num,
 266               uint8_t *buf, int nb_sectors);
 267 int bdrv_write(BdrvChild *child, int64_t sector_num,
 268                const uint8_t *buf, int nb_sectors);
 269 int bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset,
 270                        int bytes, BdrvRequestFlags flags);
 271 int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags);
 272 int bdrv_pread(BdrvChild *child, int64_t offset, void *buf, int bytes);
 273 int bdrv_preadv(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
 274 int bdrv_pwrite(BdrvChild *child, int64_t offset, const void *buf, int bytes);
 275 int bdrv_pwritev(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
 276 int bdrv_pwrite_sync(BdrvChild *child, int64_t offset,
 277                      const void *buf, int count);
 278 int coroutine_fn bdrv_co_readv(BdrvChild *child, int64_t sector_num,
 279                                int nb_sectors, QEMUIOVector *qiov);
 280 int coroutine_fn bdrv_co_writev(BdrvChild *child, int64_t sector_num,
 281                                int nb_sectors, QEMUIOVector *qiov);
 282 /*
 283  * Efficiently zero a region of the disk image.  Note that this is a regular
 284  * I/O request like read or write and should have a reasonable size.  This
 285  * function is not suitable for zeroing the entire image in a single request
 286  * because it may allocate memory for the entire region.
 287  */
 288 int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset,
 289                                        int bytes, BdrvRequestFlags flags);
 290 BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,
 291     const char *backing_file);
 292 void bdrv_refresh_filename(BlockDriverState *bs);
 293 int bdrv_truncate(BdrvChild *child, int64_t offset, PreallocMode prealloc,
 294                   Error **errp);
 295 int64_t bdrv_nb_sectors(BlockDriverState *bs);
 296 int64_t bdrv_getlength(BlockDriverState *bs);
 297 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs);
 298 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
 299                                BlockDriverState *in_bs, Error **errp);
 300 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr);
 301 void bdrv_refresh_limits(BlockDriverState *bs, Error **errp);
 302 int bdrv_commit(BlockDriverState *bs);
 303 int bdrv_change_backing_file(BlockDriverState *bs,
 304     const char *backing_file, const char *backing_fmt);
 305 void bdrv_register(BlockDriver *bdrv);
 306 int bdrv_drop_intermediate(BlockDriverState *top, BlockDriverState *base,
 307                            const char *backing_file_str);
 308 BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
 309                                     BlockDriverState *bs);
 310 BlockDriverState *bdrv_find_base(BlockDriverState *bs);
 311
 312
 313 typedef struct BdrvCheckResult {
 314     int corruptions;
 315     int leaks;
 316     int check_errors;
 317     int corruptions_fixed;
 318     int leaks_fixed;
 319     int64_t image_end_offset;
 320     BlockFragInfo bfi;
 321 } BdrvCheckResult;
 322
 323 typedef enum {
 324     BDRV_FIX_LEAKS    = 1,
 325     BDRV_FIX_ERRORS   = 2,
 326 } BdrvCheckMode;
 327
 328 int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix);
 329
 330 /* The units of offset and total_work_size may be chosen arbitrarily by the
 331  * block driver; total_work_size may change during the course of the amendment
 332  * operation */
 333 typedef void BlockDriverAmendStatusCB(BlockDriverState *bs, int64_t offset,
 334                                       int64_t total_work_size, void *opaque);
 335 int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts,
 336                        BlockDriverAmendStatusCB *status_cb, void *cb_opaque);
 337
 338 /* external snapshots */
 339 bool bdrv_recurse_is_first_non_filter(BlockDriverState *bs,
 340                                       BlockDriverState *candidate);
 341 bool bdrv_is_first_non_filter(BlockDriverState *candidate);
 342
 343 /* check if a named node can be replaced when doing drive-mirror */
 344 BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs,
 345                                         const char *node_name, Error **errp);
 346
 347 /* async block I/O */
 348 void bdrv_aio_cancel(BlockAIOCB *acb);
 349 void bdrv_aio_cancel_async(BlockAIOCB *acb);
 350
 351 /* sg packet commands */
 352 int bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf);
 353
 354 /* Invalidate any cached metadata used by image formats */
 355 void bdrv_invalidate_cache(BlockDriverState *bs, Error **errp);
 356 void bdrv_invalidate_cache_all(Error **errp);
 357 int bdrv_inactivate_all(void);
 358
 359 /* Ensure contents are flushed to disk.  */
 360 int bdrv_flush(BlockDriverState *bs);
 361 int coroutine_fn bdrv_co_flush(BlockDriverState *bs);
 362 int bdrv_flush_all(void);
 363 void bdrv_close_all(void);
 364 void bdrv_drain(BlockDriverState *bs);
 365 void coroutine_fn bdrv_co_drain(BlockDriverState *bs);
 366 void bdrv_drain_all_begin(void);
 367 void bdrv_drain_all_end(void);
 368 void bdrv_drain_all(void);
 369
 370 #define BDRV_POLL_WHILE(bs, cond) ({                       \
 371     bool waited_ = false;                                  \
 372     bool busy_ = true;                                     \
 373     BlockDriverState *bs_ = (bs);                          \
 374     AioContext *ctx_ = bdrv_get_aio_context(bs_);          \
 375     if (aio_context_in_iothread(ctx_)) {                   \
 376         while ((cond) || busy_) {                          \
 377             busy_ = aio_poll(ctx_, (cond));                \
 378             waited_ |= !!(cond) | busy_;                   \
 379         }                                                  \
 380     } else {                                               \
 381         assert(qemu_get_current_aio_context() ==           \
 382                qemu_get_aio_context());                    \
 383         /* Ask bdrv_dec_in_flight to wake up the main      \
 384          * QEMU AioContext.  Extra I/O threads never take  \
 385          * other I/O threads' AioContexts (see for example \
 386          * block_job_defer_to_main_loop for how to do it). \
 387          */                                                \
 388         assert(!bs_->wakeup);                              \
 389         /* Set bs->wakeup before evaluating cond.  */      \
 390         atomic_mb_set(&bs_->wakeup, true);                 \
 391         while (busy_) {                                    \
 392             if ((cond)) {                                  \
 393                 waited_ = busy_ = true;                    \
 394                 aio_context_release(ctx_);                 \
 395                 aio_poll(qemu_get_aio_context(), true);    \
 396                 aio_context_acquire(ctx_);                 \
 397             } else {                                       \
 398                 busy_ = aio_poll(ctx_, false);             \
 399                 waited_ |= busy_;                          \
 400             }                                              \
 401         }                                                  \
 402         atomic_set(&bs_->wakeup, false);                   \
 403     }                                                      \
 404     waited_; })
 405
 406 int bdrv_pdiscard(BlockDriverState *bs, int64_t offset, int bytes);
 407 int bdrv_co_pdiscard(BlockDriverState *bs, int64_t offset, int bytes);
 408 int bdrv_has_zero_init_1(BlockDriverState *bs);
 409 int bdrv_has_zero_init(BlockDriverState *bs);
 410 bool bdrv_unallocated_blocks_are_zero(BlockDriverState *bs);
 411 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs);
 412 int bdrv_block_status(BlockDriverState *bs, int64_t offset,
 413                       int64_t bytes, int64_t *pnum, int64_t *map,
 414                       BlockDriverState **file);
 415 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base,
 416                             int64_t offset, int64_t bytes, int64_t *pnum,
 417                             int64_t *map, BlockDriverState **file);
 418 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
 419                       int64_t *pnum);
 420 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
 421                             int64_t offset, int64_t bytes, int64_t *pnum);
 422
 423 bool bdrv_is_read_only(BlockDriverState *bs);
 424 int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only,
 425                            bool ignore_allow_rdw, Error **errp);
 426 int bdrv_set_read_only(BlockDriverState *bs, bool read_only, Error **errp);
 427 bool bdrv_is_sg(BlockDriverState *bs);
 428 bool bdrv_is_inserted(BlockDriverState *bs);
 429 void bdrv_lock_medium(BlockDriverState *bs, bool locked);
 430 void bdrv_eject(BlockDriverState *bs, bool eject_flag);
 431 const char *bdrv_get_format_name(BlockDriverState *bs);
 432 BlockDriverState *bdrv_find_node(const char *node_name);
 433 BlockDeviceInfoList *bdrv_named_nodes_list(Error **errp);
 434 BlockDriverState *bdrv_lookup_bs(const char *device,
 435                                  const char *node_name,
 436                                  Error **errp);
 437 bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base);
 438 BlockDriverState *bdrv_next_node(BlockDriverState *bs);
 439
 440 typedef struct BdrvNextIterator {
 441     enum {
 442         BDRV_NEXT_BACKEND_ROOTS,
 443         BDRV_NEXT_MONITOR_OWNED,
 444     } phase;
 445     BlockBackend *blk;
 446     BlockDriverState *bs;
 447 } BdrvNextIterator;
 448
 449 BlockDriverState *bdrv_first(BdrvNextIterator *it);
 450 BlockDriverState *bdrv_next(BdrvNextIterator *it);
 451 void bdrv_next_cleanup(BdrvNextIterator *it);
 452
 453 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs);
 454 bool bdrv_is_encrypted(BlockDriverState *bs);
 455 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
 456                          void *opaque);
 457 const char *bdrv_get_node_name(const BlockDriverState *bs);
 458 const char *bdrv_get_device_name(const BlockDriverState *bs);
 459 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
 460 int bdrv_get_flags(BlockDriverState *bs);
 461 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
 462 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs);
 463 void bdrv_round_to_clusters(BlockDriverState *bs,
 464                             int64_t offset, int64_t bytes,
 465                             int64_t *cluster_offset,
 466                             int64_t *cluster_bytes);
 467
 468 const char *bdrv_get_encrypted_filename(BlockDriverState *bs);
 469 void bdrv_get_backing_filename(BlockDriverState *bs,
 470                                char *filename, int filename_size);
 471 void bdrv_get_full_backing_filename(BlockDriverState *bs,
 472                                     char *dest, size_t sz, Error **errp);
 473 void bdrv_get_full_backing_filename_from_filename(const char *backed,
 474                                                   const char *backing,
 475                                                   char *dest, size_t sz,
 476                                                   Error **errp);
 477
 478 int path_has_protocol(const char *path);
 479 int path_is_absolute(const char *path);
 480 void path_combine(char *dest, int dest_size,
 481                   const char *base_path,
 482                   const char *filename);
 483
 484 int bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
 485 int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
 486 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
 487                       int64_t pos, int size);
 488
 489 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
 490                       int64_t pos, int size);
 491
 492 void bdrv_img_create(const char *filename, const char *fmt,
 493                      const char *base_filename, const char *base_fmt,
 494                      char *options, uint64_t img_size, int flags,
 495                      bool quiet, Error **errp);
 496
 497 /* Returns the alignment in bytes that is required so that no bounce buffer
 498  * is required throughout the stack */
 499 size_t bdrv_min_mem_align(BlockDriverState *bs);
 500 /* Returns optimal alignment in bytes for bounce buffer */
 501 size_t bdrv_opt_mem_align(BlockDriverState *bs);
 502 void *qemu_blockalign(BlockDriverState *bs, size_t size);
 503 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
 504 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
 505 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
 506 bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov);
 507
 508 void bdrv_enable_copy_on_read(BlockDriverState *bs);
 509 void bdrv_disable_copy_on_read(BlockDriverState *bs);
 510
 511 void bdrv_ref(BlockDriverState *bs);
 512 void bdrv_unref(BlockDriverState *bs);
 513 void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child);
 514 BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
 515                              BlockDriverState *child_bs,
 516                              const char *child_name,
 517                              const BdrvChildRole *child_role,
 518                              Error **errp);
 519
 520 bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp);
 521 void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason);
 522 void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason);
 523 void bdrv_op_block_all(BlockDriverState *bs, Error *reason);
 524 void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason);
 525 bool bdrv_op_blocker_is_empty(BlockDriverState *bs);
 526
 527 #define BLKDBG_EVENT(child, evt) \
 528     do { \
 529         if (child) { \
 530             bdrv_debug_event(child->bs, evt); \
 531         } \
 532     } while (0)
 533
 534 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
 535
 536 int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event,
 537                            const char *tag);
 538 int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag);
 539 int bdrv_debug_resume(BlockDriverState *bs, const char *tag);
 540 bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag);
 541
 542 /**
 543  * bdrv_get_aio_context:
 544  *
 545  * Returns: the currently bound #AioContext
 546  */
 547 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
 548
 549 /**
 550  * Transfer control to @co in the aio context of @bs
 551  */
 552 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co);
 553
 554 /**
 555  * bdrv_set_aio_context:
 556  *
 557  * Changes the #AioContext used for fd handlers, timers, and BHs by this
 558  * BlockDriverState and all its children.
 559  *
 560  * This function must be called with iothread lock held.
 561  */
 562 void bdrv_set_aio_context(BlockDriverState *bs, AioContext *new_context);
 563 int bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz);
 564 int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo);
 565
 566 void bdrv_io_plug(BlockDriverState *bs);
 567 void bdrv_io_unplug(BlockDriverState *bs);
 568
 569 /**
 570  * bdrv_parent_drained_begin:
 571  *
 572  * Begin a quiesced section of all users of @bs. This is part of
 573  * bdrv_drained_begin.
 574  */
 575 void bdrv_parent_drained_begin(BlockDriverState *bs, BdrvChild *ignore);
 576
 577 /**
 578  * bdrv_parent_drained_end:
 579  *
 580  * End a quiesced section of all users of @bs. This is part of
 581  * bdrv_drained_end.
 582  */
 583 void bdrv_parent_drained_end(BlockDriverState *bs, BdrvChild *ignore);
 584
 585 /**
 586  * bdrv_drained_begin:
 587  *
 588  * Begin a quiesced section for exclusive access to the BDS, by disabling
 589  * external request sources including NBD server and device model. Note that
 590  * this doesn't block timers or coroutines from submitting more requests, which
 591  * means block_job_pause is still necessary.
 592  *
 593  * This function can be recursive.
 594  */
 595 void bdrv_drained_begin(BlockDriverState *bs);
 596
 597 /**
 598  * Like bdrv_drained_begin, but recursively begins a quiesced section for
 599  * exclusive access to all child nodes as well.
 600  */
 601 void bdrv_subtree_drained_begin(BlockDriverState *bs);
 602
 603 /**
 604  * bdrv_drained_end:
 605  *
 606  * End a quiescent section started by bdrv_drained_begin().
 607  */
 608 void bdrv_drained_end(BlockDriverState *bs);
 609
 610 /**
 611  * End a quiescent section started by bdrv_subtree_drained_begin().
 612  */
 613 void bdrv_subtree_drained_end(BlockDriverState *bs);
 614
 615 void bdrv_add_child(BlockDriverState *parent, BlockDriverState *child,
 616                     Error **errp);
 617 void bdrv_del_child(BlockDriverState *parent, BdrvChild *child, Error **errp);
 618
 619 bool bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
 620                                      uint32_t granularity, Error **errp);
 621 /**
 622  *
 623  * bdrv_register_buf/bdrv_unregister_buf:
 624  *
 625  * Register/unregister a buffer for I/O. For example, VFIO drivers are
 626  * interested to know the memory areas that would later be used for I/O, so
 627  * that they can prepare IOMMU mapping etc., to get better performance.
 628  */
 629 void bdrv_register_buf(BlockDriverState *bs, void *host, size_t size);
 630 void bdrv_unregister_buf(BlockDriverState *bs, void *host);
 631 #endif