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