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
  47     /*
  48      * The BDRV_REQ_MAY_UNMAP flag is used in write_zeroes requests to indicate
  49      * that the block driver should unmap (discard) blocks if it is guaranteed
  50      * that the result will read back as zeroes. The flag is only passed to the
  51      * driver if the block device is opened with BDRV_O_UNMAP.
  52      */
  53     BDRV_REQ_MAY_UNMAP          = 0x4,
  54
  55     /*
  56      * The BDRV_REQ_NO_SERIALISING flag is only valid for reads and means that
  57      * we don't want wait_serialising_requests() during the read operation.
  58      *
  59      * This flag is used for backup copy-on-write operations, when we need to
  60      * read old data before write (write notifier triggered). It is okay since
  61      * we already waited for other serializing requests in the initiating write
  62      * (see bdrv_aligned_pwritev), and it is necessary if the initiating write
  63      * is already serializing (without the flag, the read would deadlock
  64      * waiting for the serialising write to complete).
  65      */
  66     BDRV_REQ_NO_SERIALISING     = 0x8,
  67     BDRV_REQ_FUA                = 0x10,
  68     BDRV_REQ_WRITE_COMPRESSED   = 0x20,
  69
  70     /* Signifies that this write request will not change the visible disk
  71      * content. */
  72     BDRV_REQ_WRITE_UNCHANGED    = 0x40,
  73
  74     /*
  75      * BDRV_REQ_SERIALISING forces request serialisation for writes.
  76      * It is used to ensure that writes to the backing file of a backup process
  77      * target cannot race with a read of the backup target that defers to the
  78      * backing file.
  79      *
  80      * Note, that BDRV_REQ_SERIALISING is _not_ opposite in meaning to
  81      * BDRV_REQ_NO_SERIALISING. A more descriptive name for the latter might be
  82      * _DO_NOT_WAIT_FOR_SERIALISING, except that is too long.
  83      */
  84     BDRV_REQ_SERIALISING        = 0x80,
  85
  86     /* Execute the request only if the operation can be offloaded or otherwise
  87      * be executed efficiently, but return an error instead of using a slow
  88      * fallback. */
  89     BDRV_REQ_NO_FALLBACK        = 0x100,
  90
  91     /* Mask of valid flags */
  92     BDRV_REQ_MASK               = 0x1ff,
  93 } BdrvRequestFlags;
  94
  95 typedef struct BlockSizes {
  96     uint32_t phys;
  97     uint32_t log;
  98 } BlockSizes;
  99
 100 typedef struct HDGeometry {
 101     uint32_t heads;
 102     uint32_t sectors;
 103     uint32_t cylinders;
 104 } HDGeometry;
 105
 106 #define BDRV_O_RDWR        0x0002
 107 #define BDRV_O_RESIZE      0x0004 /* request permission for resizing the node */
 108 #define BDRV_O_SNAPSHOT    0x0008 /* open the file read only and save writes in a snapshot */
 109 #define BDRV_O_TEMPORARY   0x0010 /* delete the file after use */
 110 #define BDRV_O_NOCACHE     0x0020 /* do not use the host page cache */
 111 #define BDRV_O_NATIVE_AIO  0x0080 /* use native AIO instead of the thread pool */
 112 #define BDRV_O_NO_BACKING  0x0100 /* don't open the backing file */
 113 #define BDRV_O_NO_FLUSH    0x0200 /* disable flushing on this disk */
 114 #define BDRV_O_COPY_ON_READ 0x0400 /* copy read backing sectors into image */
 115 #define BDRV_O_INACTIVE    0x0800  /* consistency hint for migration handoff */
 116 #define BDRV_O_CHECK       0x1000  /* open solely for consistency check */
 117 #define BDRV_O_ALLOW_RDWR  0x2000  /* allow reopen to change from r/o to r/w */
 118 #define BDRV_O_UNMAP       0x4000  /* execute guest UNMAP/TRIM operations */
 119 #define BDRV_O_PROTOCOL    0x8000  /* if no block driver is explicitly given:
 120                                       select an appropriate protocol driver,
 121                                       ignoring the format layer */
 122 #define BDRV_O_NO_IO       0x10000 /* don't initialize for I/O */
 123 #define BDRV_O_AUTO_RDONLY 0x20000 /* degrade to read-only if opening read-write fails */
 124
 125 #define BDRV_O_CACHE_MASK  (BDRV_O_NOCACHE | BDRV_O_NO_FLUSH)
 126
 127
 128 /* Option names of options parsed by the block layer */
 129
 130 #define BDRV_OPT_CACHE_WB       "cache.writeback"
 131 #define BDRV_OPT_CACHE_DIRECT   "cache.direct"
 132 #define BDRV_OPT_CACHE_NO_FLUSH "cache.no-flush"
 133 #define BDRV_OPT_READ_ONLY      "read-only"
 134 #define BDRV_OPT_AUTO_READ_ONLY "auto-read-only"
 135 #define BDRV_OPT_DISCARD        "discard"
 136 #define BDRV_OPT_FORCE_SHARE    "force-share"
 137
 138
 139 #define BDRV_SECTOR_BITS   9
 140 #define BDRV_SECTOR_SIZE   (1ULL << BDRV_SECTOR_BITS)
 141 #define BDRV_SECTOR_MASK   ~(BDRV_SECTOR_SIZE - 1)
 142
 143 #define BDRV_REQUEST_MAX_SECTORS MIN(SIZE_MAX >> BDRV_SECTOR_BITS, \
 144                                      INT_MAX >> BDRV_SECTOR_BITS)
 145 #define BDRV_REQUEST_MAX_BYTES (BDRV_REQUEST_MAX_SECTORS << BDRV_SECTOR_BITS)
 146
 147 /*
 148  * Allocation status flags for bdrv_block_status() and friends.
 149  *
 150  * Public flags:
 151  * BDRV_BLOCK_DATA: allocation for data at offset is tied to this layer
 152  * BDRV_BLOCK_ZERO: offset reads as zero
 153  * BDRV_BLOCK_OFFSET_VALID: an associated offset exists for accessing raw data
 154  * BDRV_BLOCK_ALLOCATED: the content of the block is determined by this
 155  *                       layer rather than any backing, set by block layer
 156  * BDRV_BLOCK_EOF: the returned pnum covers through end of file for this
 157  *                 layer, set by block layer
 158  *
 159  * Internal flag:
 160  * BDRV_BLOCK_RAW: for use by passthrough drivers, such as raw, to request
 161  *                 that the block layer recompute the answer from the returned
 162  *                 BDS; must be accompanied by just BDRV_BLOCK_OFFSET_VALID.
 163  *
 164  * If BDRV_BLOCK_OFFSET_VALID is set, the map parameter represents the
 165  * host offset within the returned BDS that is allocated for the
 166  * corresponding raw guest data.  However, whether that offset
 167  * actually contains data also depends on BDRV_BLOCK_DATA, as follows:
 168  *
 169  * DATA ZERO OFFSET_VALID
 170  *  t    t        t       sectors read as zero, returned file is zero at offset
 171  *  t    f        t       sectors read as valid from file at offset
 172  *  f    t        t       sectors preallocated, read as zero, returned file not
 173  *                        necessarily zero at offset
 174  *  f    f        t       sectors preallocated but read from backing_hd,
 175  *                        returned file contains garbage at offset
 176  *  t    t        f       sectors preallocated, read as zero, unknown offset
 177  *  t    f        f       sectors read from unknown file or offset
 178  *  f    t        f       not allocated or unknown offset, read as zero
 179  *  f    f        f       not allocated or unknown offset, read from backing_hd
 180  */
 181 #define BDRV_BLOCK_DATA         0x01
 182 #define BDRV_BLOCK_ZERO         0x02
 183 #define BDRV_BLOCK_OFFSET_VALID 0x04
 184 #define BDRV_BLOCK_RAW          0x08
 185 #define BDRV_BLOCK_ALLOCATED    0x10
 186 #define BDRV_BLOCK_EOF          0x20
 187 #define BDRV_BLOCK_OFFSET_MASK  BDRV_SECTOR_MASK
 188
 189 typedef QSIMPLEQ_HEAD(BlockReopenQueue, BlockReopenQueueEntry) BlockReopenQueue;
 190
 191 typedef struct BDRVReopenState {
 192     BlockDriverState *bs;
 193     int flags;
 194     BlockdevDetectZeroesOptions detect_zeroes;
 195     bool backing_missing;
 196     bool replace_backing_bs;  /* new_backing_bs is ignored if this is false */
 197     BlockDriverState *new_backing_bs; /* If NULL then detach the current bs */
 198     uint64_t perm, shared_perm;
 199     QDict *options;
 200     QDict *explicit_options;
 201     void *opaque;
 202 } BDRVReopenState;
 203
 204 /*
 205  * Block operation types
 206  */
 207 typedef enum BlockOpType {
 208     BLOCK_OP_TYPE_BACKUP_SOURCE,
 209     BLOCK_OP_TYPE_BACKUP_TARGET,
 210     BLOCK_OP_TYPE_CHANGE,
 211     BLOCK_OP_TYPE_COMMIT_SOURCE,
 212     BLOCK_OP_TYPE_COMMIT_TARGET,
 213     BLOCK_OP_TYPE_DATAPLANE,
 214     BLOCK_OP_TYPE_DRIVE_DEL,
 215     BLOCK_OP_TYPE_EJECT,
 216     BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
 217     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT,
 218     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE,
 219     BLOCK_OP_TYPE_MIRROR_SOURCE,
 220     BLOCK_OP_TYPE_MIRROR_TARGET,
 221     BLOCK_OP_TYPE_RESIZE,
 222     BLOCK_OP_TYPE_STREAM,
 223     BLOCK_OP_TYPE_REPLACE,
 224     BLOCK_OP_TYPE_MAX,
 225 } BlockOpType;
 226
 227 /* Block node permission constants */
 228 enum {
 229     /**
 230      * A user that has the "permission" of consistent reads is guaranteed that
 231      * their view of the contents of the block device is complete and
 232      * self-consistent, representing the contents of a disk at a specific
 233      * point.
 234      *
 235      * For most block devices (including their backing files) this is true, but
 236      * the property cannot be maintained in a few situations like for
 237      * intermediate nodes of a commit block job.
 238      */
 239     BLK_PERM_CONSISTENT_READ    = 0x01,
 240
 241     /** This permission is required to change the visible disk contents. */
 242     BLK_PERM_WRITE              = 0x02,
 243
 244     /**
 245      * This permission (which is weaker than BLK_PERM_WRITE) is both enough and
 246      * required for writes to the block node when the caller promises that
 247      * the visible disk content doesn't change.
 248      *
 249      * As the BLK_PERM_WRITE permission is strictly stronger, either is
 250      * sufficient to perform an unchanging write.
 251      */
 252     BLK_PERM_WRITE_UNCHANGED    = 0x04,
 253
 254     /** This permission is required to change the size of a block node. */
 255     BLK_PERM_RESIZE             = 0x08,
 256
 257     /**
 258      * This permission is required to change the node that this BdrvChild
 259      * points to.
 260      */
 261     BLK_PERM_GRAPH_MOD          = 0x10,
 262
 263     BLK_PERM_ALL                = 0x1f,
 264
 265     DEFAULT_PERM_PASSTHROUGH    = BLK_PERM_CONSISTENT_READ
 266                                  | BLK_PERM_WRITE
 267                                  | BLK_PERM_WRITE_UNCHANGED
 268                                  | BLK_PERM_RESIZE,
 269
 270     DEFAULT_PERM_UNCHANGED      = BLK_PERM_ALL & ~DEFAULT_PERM_PASSTHROUGH,
 271 };
 272
 273 char *bdrv_perm_names(uint64_t perm);
 274
 275 /* disk I/O throttling */
 276 void bdrv_init(void);
 277 void bdrv_init_with_whitelist(void);
 278 bool bdrv_uses_whitelist(void);
 279 int bdrv_is_whitelisted(BlockDriver *drv, bool read_only);
 280 BlockDriver *bdrv_find_protocol(const char *filename,
 281                                 bool allow_protocol_prefix,
 282                                 Error **errp);
 283 BlockDriver *bdrv_find_format(const char *format_name);
 284 int bdrv_create(BlockDriver *drv, const char* filename,
 285                 QemuOpts *opts, Error **errp);
 286 int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp);
 287 BlockDriverState *bdrv_new(void);
 288 void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top,
 289                  Error **errp);
 290 void bdrv_replace_node(BlockDriverState *from, BlockDriverState *to,
 291                        Error **errp);
 292
 293 int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough);
 294 int bdrv_parse_discard_flags(const char *mode, int *flags);
 295 BdrvChild *bdrv_open_child(const char *filename,
 296                            QDict *options, const char *bdref_key,
 297                            BlockDriverState* parent,
 298                            const BdrvChildRole *child_role,
 299                            bool allow_none, Error **errp);
 300 BlockDriverState *bdrv_open_blockdev_ref(BlockdevRef *ref, Error **errp);
 301 void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd,
 302                          Error **errp);
 303 int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options,
 304                            const char *bdref_key, Error **errp);
 305 BlockDriverState *bdrv_open(const char *filename, const char *reference,
 306                             QDict *options, int flags, Error **errp);
 307 BlockDriverState *bdrv_new_open_driver(BlockDriver *drv, const char *node_name,
 308                                        int flags, Error **errp);
 309 BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue,
 310                                     BlockDriverState *bs, QDict *options,
 311                                     bool keep_old_opts);
 312 int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp);
 313 int bdrv_reopen_set_read_only(BlockDriverState *bs, bool read_only,
 314                               Error **errp);
 315 int bdrv_reopen_prepare(BDRVReopenState *reopen_state,
 316                         BlockReopenQueue *queue, Error **errp);
 317 void bdrv_reopen_commit(BDRVReopenState *reopen_state);
 318 void bdrv_reopen_abort(BDRVReopenState *reopen_state);
 319 int bdrv_read(BdrvChild *child, int64_t sector_num,
 320               uint8_t *buf, int nb_sectors);
 321 int bdrv_write(BdrvChild *child, int64_t sector_num,
 322                const uint8_t *buf, int nb_sectors);
 323 int bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset,
 324                        int bytes, BdrvRequestFlags flags);
 325 int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags);
 326 int bdrv_pread(BdrvChild *child, int64_t offset, void *buf, int bytes);
 327 int bdrv_preadv(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
 328 int bdrv_pwrite(BdrvChild *child, int64_t offset, const void *buf, int bytes);
 329 int bdrv_pwritev(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
 330 int bdrv_pwrite_sync(BdrvChild *child, int64_t offset,
 331                      const void *buf, int count);
 332 /*
 333  * Efficiently zero a region of the disk image.  Note that this is a regular
 334  * I/O request like read or write and should have a reasonable size.  This
 335  * function is not suitable for zeroing the entire image in a single request
 336  * because it may allocate memory for the entire region.
 337  */
 338 int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset,
 339                                        int bytes, BdrvRequestFlags flags);
 340 BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,
 341     const char *backing_file);
 342 void bdrv_refresh_filename(BlockDriverState *bs);
 343
 344 int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset,
 345                                   PreallocMode prealloc, Error **errp);
 346 int bdrv_truncate(BdrvChild *child, int64_t offset, PreallocMode prealloc,
 347                   Error **errp);
 348
 349 int64_t bdrv_nb_sectors(BlockDriverState *bs);
 350 int64_t bdrv_getlength(BlockDriverState *bs);
 351 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs);
 352 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
 353                                BlockDriverState *in_bs, Error **errp);
 354 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr);
 355 void bdrv_refresh_limits(BlockDriverState *bs, Error **errp);
 356 int bdrv_commit(BlockDriverState *bs);
 357 int bdrv_change_backing_file(BlockDriverState *bs,
 358     const char *backing_file, const char *backing_fmt);
 359 void bdrv_register(BlockDriver *bdrv);
 360 int bdrv_drop_intermediate(BlockDriverState *top, BlockDriverState *base,
 361                            const char *backing_file_str);
 362 BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
 363                                     BlockDriverState *bs);
 364 BlockDriverState *bdrv_find_base(BlockDriverState *bs);
 365 bool bdrv_is_backing_chain_frozen(BlockDriverState *bs, BlockDriverState *base,
 366                                   Error **errp);
 367 int bdrv_freeze_backing_chain(BlockDriverState *bs, BlockDriverState *base,
 368                               Error **errp);
 369 void bdrv_unfreeze_backing_chain(BlockDriverState *bs, BlockDriverState *base);
 370
 371
 372 typedef struct BdrvCheckResult {
 373     int corruptions;
 374     int leaks;
 375     int check_errors;
 376     int corruptions_fixed;
 377     int leaks_fixed;
 378     int64_t image_end_offset;
 379     BlockFragInfo bfi;
 380 } BdrvCheckResult;
 381
 382 typedef enum {
 383     BDRV_FIX_LEAKS    = 1,
 384     BDRV_FIX_ERRORS   = 2,
 385 } BdrvCheckMode;
 386
 387 int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix);
 388
 389 /* The units of offset and total_work_size may be chosen arbitrarily by the
 390  * block driver; total_work_size may change during the course of the amendment
 391  * operation */
 392 typedef void BlockDriverAmendStatusCB(BlockDriverState *bs, int64_t offset,
 393                                       int64_t total_work_size, void *opaque);
 394 int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts,
 395                        BlockDriverAmendStatusCB *status_cb, void *cb_opaque,
 396                        Error **errp);
 397
 398 /* external snapshots */
 399 bool bdrv_recurse_is_first_non_filter(BlockDriverState *bs,
 400                                       BlockDriverState *candidate);
 401 bool bdrv_is_first_non_filter(BlockDriverState *candidate);
 402
 403 /* check if a named node can be replaced when doing drive-mirror */
 404 BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs,
 405                                         const char *node_name, Error **errp);
 406
 407 /* async block I/O */
 408 void bdrv_aio_cancel(BlockAIOCB *acb);
 409 void bdrv_aio_cancel_async(BlockAIOCB *acb);
 410
 411 /* sg packet commands */
 412 int bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf);
 413
 414 /* Invalidate any cached metadata used by image formats */
 415 void bdrv_invalidate_cache(BlockDriverState *bs, Error **errp);
 416 void bdrv_invalidate_cache_all(Error **errp);
 417 int bdrv_inactivate_all(void);
 418
 419 /* Ensure contents are flushed to disk.  */
 420 int bdrv_flush(BlockDriverState *bs);
 421 int coroutine_fn bdrv_co_flush(BlockDriverState *bs);
 422 int bdrv_flush_all(void);
 423 void bdrv_close_all(void);
 424 void bdrv_drain(BlockDriverState *bs);
 425 void coroutine_fn bdrv_co_drain(BlockDriverState *bs);
 426 void bdrv_drain_all_begin(void);
 427 void bdrv_drain_all_end(void);
 428 void bdrv_drain_all(void);
 429
 430 #define BDRV_POLL_WHILE(bs, cond) ({                       \
 431     BlockDriverState *bs_ = (bs);                          \
 432     AIO_WAIT_WHILE(bdrv_get_aio_context(bs_),              \
 433                    cond); })
 434
 435 int bdrv_pdiscard(BdrvChild *child, int64_t offset, int bytes);
 436 int bdrv_co_pdiscard(BdrvChild *child, int64_t offset, int bytes);
 437 int bdrv_has_zero_init_1(BlockDriverState *bs);
 438 int bdrv_has_zero_init(BlockDriverState *bs);
 439 bool bdrv_unallocated_blocks_are_zero(BlockDriverState *bs);
 440 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs);
 441 int bdrv_block_status(BlockDriverState *bs, int64_t offset,
 442                       int64_t bytes, int64_t *pnum, int64_t *map,
 443                       BlockDriverState **file);
 444 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base,
 445                             int64_t offset, int64_t bytes, int64_t *pnum,
 446                             int64_t *map, BlockDriverState **file);
 447 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
 448                       int64_t *pnum);
 449 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
 450                             int64_t offset, int64_t bytes, int64_t *pnum);
 451
 452 bool bdrv_is_read_only(BlockDriverState *bs);
 453 int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only,
 454                            bool ignore_allow_rdw, Error **errp);
 455 int bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg,
 456                               Error **errp);
 457 bool bdrv_is_writable(BlockDriverState *bs);
 458 bool bdrv_is_sg(BlockDriverState *bs);
 459 bool bdrv_is_inserted(BlockDriverState *bs);
 460 void bdrv_lock_medium(BlockDriverState *bs, bool locked);
 461 void bdrv_eject(BlockDriverState *bs, bool eject_flag);
 462 const char *bdrv_get_format_name(BlockDriverState *bs);
 463 BlockDriverState *bdrv_find_node(const char *node_name);
 464 BlockDeviceInfoList *bdrv_named_nodes_list(Error **errp);
 465 XDbgBlockGraph *bdrv_get_xdbg_block_graph(Error **errp);
 466 BlockDriverState *bdrv_lookup_bs(const char *device,
 467                                  const char *node_name,
 468                                  Error **errp);
 469 bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base);
 470 BlockDriverState *bdrv_next_node(BlockDriverState *bs);
 471 BlockDriverState *bdrv_next_all_states(BlockDriverState *bs);
 472
 473 typedef struct BdrvNextIterator {
 474     enum {
 475         BDRV_NEXT_BACKEND_ROOTS,
 476         BDRV_NEXT_MONITOR_OWNED,
 477     } phase;
 478     BlockBackend *blk;
 479     BlockDriverState *bs;
 480 } BdrvNextIterator;
 481
 482 BlockDriverState *bdrv_first(BdrvNextIterator *it);
 483 BlockDriverState *bdrv_next(BdrvNextIterator *it);
 484 void bdrv_next_cleanup(BdrvNextIterator *it);
 485
 486 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs);
 487 bool bdrv_is_encrypted(BlockDriverState *bs);
 488 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
 489                          void *opaque, bool read_only);
 490 const char *bdrv_get_node_name(const BlockDriverState *bs);
 491 const char *bdrv_get_device_name(const BlockDriverState *bs);
 492 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
 493 int bdrv_get_flags(BlockDriverState *bs);
 494 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
 495 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs,
 496                                           Error **errp);
 497 void bdrv_round_to_clusters(BlockDriverState *bs,
 498                             int64_t offset, int64_t bytes,
 499                             int64_t *cluster_offset,
 500                             int64_t *cluster_bytes);
 501
 502 void bdrv_get_backing_filename(BlockDriverState *bs,
 503                                char *filename, int filename_size);
 504 char *bdrv_get_full_backing_filename(BlockDriverState *bs, Error **errp);
 505 char *bdrv_get_full_backing_filename_from_filename(const char *backed,
 506                                                    const char *backing,
 507                                                    Error **errp);
 508 char *bdrv_dirname(BlockDriverState *bs, Error **errp);
 509
 510 int path_has_protocol(const char *path);
 511 int path_is_absolute(const char *path);
 512 char *path_combine(const char *base_path, const char *filename);
 513
 514 int bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
 515 int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
 516 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
 517                       int64_t pos, int size);
 518
 519 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
 520                       int64_t pos, int size);
 521
 522 void bdrv_img_create(const char *filename, const char *fmt,
 523                      const char *base_filename, const char *base_fmt,
 524                      char *options, uint64_t img_size, int flags,
 525                      bool quiet, Error **errp);
 526
 527 /* Returns the alignment in bytes that is required so that no bounce buffer
 528  * is required throughout the stack */
 529 size_t bdrv_min_mem_align(BlockDriverState *bs);
 530 /* Returns optimal alignment in bytes for bounce buffer */
 531 size_t bdrv_opt_mem_align(BlockDriverState *bs);
 532 void *qemu_blockalign(BlockDriverState *bs, size_t size);
 533 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
 534 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
 535 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
 536 bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov);
 537
 538 void bdrv_enable_copy_on_read(BlockDriverState *bs);
 539 void bdrv_disable_copy_on_read(BlockDriverState *bs);
 540
 541 void bdrv_ref(BlockDriverState *bs);
 542 void bdrv_unref(BlockDriverState *bs);
 543 void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child);
 544 BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
 545                              BlockDriverState *child_bs,
 546                              const char *child_name,
 547                              const BdrvChildRole *child_role,
 548                              Error **errp);
 549
 550 bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp);
 551 void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason);
 552 void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason);
 553 void bdrv_op_block_all(BlockDriverState *bs, Error *reason);
 554 void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason);
 555 bool bdrv_op_blocker_is_empty(BlockDriverState *bs);
 556
 557 #define BLKDBG_EVENT(child, evt) \
 558     do { \
 559         if (child) { \
 560             bdrv_debug_event(child->bs, evt); \
 561         } \
 562     } while (0)
 563
 564 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
 565
 566 int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event,
 567                            const char *tag);
 568 int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag);
 569 int bdrv_debug_resume(BlockDriverState *bs, const char *tag);
 570 bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag);
 571
 572 /**
 573  * bdrv_get_aio_context:
 574  *
 575  * Returns: the currently bound #AioContext
 576  */
 577 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
 578
 579 /**
 580  * Transfer control to @co in the aio context of @bs
 581  */
 582 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co);
 583
 584 /**
 585  * bdrv_set_aio_context:
 586  *
 587  * Changes the #AioContext used for fd handlers, timers, and BHs by this
 588  * BlockDriverState and all its children.
 589  *
 590  * This function must be called with iothread lock held.
 591  */
 592 void bdrv_set_aio_context(BlockDriverState *bs, AioContext *new_context);
 593 int bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz);
 594 int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo);
 595
 596 void bdrv_io_plug(BlockDriverState *bs);
 597 void bdrv_io_unplug(BlockDriverState *bs);
 598
 599 /**
 600  * bdrv_parent_drained_begin:
 601  *
 602  * Begin a quiesced section of all users of @bs. This is part of
 603  * bdrv_drained_begin.
 604  */
 605 void bdrv_parent_drained_begin(BlockDriverState *bs, BdrvChild *ignore,
 606                                bool ignore_bds_parents);
 607
 608 /**
 609  * bdrv_parent_drained_begin_single:
 610  *
 611  * Begin a quiesced section for the parent of @c. If @poll is true, wait for
 612  * any pending activity to cease.
 613  */
 614 void bdrv_parent_drained_begin_single(BdrvChild *c, bool poll);
 615
 616 /**
 617  * bdrv_parent_drained_end:
 618  *
 619  * End a quiesced section of all users of @bs. This is part of
 620  * bdrv_drained_end.
 621  */
 622 void bdrv_parent_drained_end(BlockDriverState *bs, BdrvChild *ignore,
 623                              bool ignore_bds_parents);
 624
 625 /**
 626  * bdrv_drain_poll:
 627  *
 628  * Poll for pending requests in @bs, its parents (except for @ignore_parent),
 629  * and if @recursive is true its children as well (used for subtree drain).
 630  *
 631  * If @ignore_bds_parents is true, parents that are BlockDriverStates must
 632  * ignore the drain request because they will be drained separately (used for
 633  * drain_all).
 634  *
 635  * This is part of bdrv_drained_begin.
 636  */
 637 bool bdrv_drain_poll(BlockDriverState *bs, bool recursive,
 638                      BdrvChild *ignore_parent, bool ignore_bds_parents);
 639
 640 /**
 641  * bdrv_drained_begin:
 642  *
 643  * Begin a quiesced section for exclusive access to the BDS, by disabling
 644  * external request sources including NBD server and device model. Note that
 645  * this doesn't block timers or coroutines from submitting more requests, which
 646  * means block_job_pause is still necessary.
 647  *
 648  * This function can be recursive.
 649  */
 650 void bdrv_drained_begin(BlockDriverState *bs);
 651
 652 /**
 653  * bdrv_do_drained_begin_quiesce:
 654  *
 655  * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already
 656  * running requests to complete.
 657  */
 658 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs,
 659                                    BdrvChild *parent, bool ignore_bds_parents);
 660
 661 /**
 662  * Like bdrv_drained_begin, but recursively begins a quiesced section for
 663  * exclusive access to all child nodes as well.
 664  */
 665 void bdrv_subtree_drained_begin(BlockDriverState *bs);
 666
 667 /**
 668  * bdrv_drained_end:
 669  *
 670  * End a quiescent section started by bdrv_drained_begin().
 671  */
 672 void bdrv_drained_end(BlockDriverState *bs);
 673
 674 /**
 675  * End a quiescent section started by bdrv_subtree_drained_begin().
 676  */
 677 void bdrv_subtree_drained_end(BlockDriverState *bs);
 678
 679 void bdrv_add_child(BlockDriverState *parent, BlockDriverState *child,
 680                     Error **errp);
 681 void bdrv_del_child(BlockDriverState *parent, BdrvChild *child, Error **errp);
 682
 683 bool bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
 684                                      uint32_t granularity, Error **errp);
 685 /**
 686  *
 687  * bdrv_register_buf/bdrv_unregister_buf:
 688  *
 689  * Register/unregister a buffer for I/O. For example, VFIO drivers are
 690  * interested to know the memory areas that would later be used for I/O, so
 691  * that they can prepare IOMMU mapping etc., to get better performance.
 692  */
 693 void bdrv_register_buf(BlockDriverState *bs, void *host, size_t size);
 694 void bdrv_unregister_buf(BlockDriverState *bs, void *host);
 695
 696 /**
 697  *
 698  * bdrv_co_copy_range:
 699  *
 700  * Do offloaded copy between two children. If the operation is not implemented
 701  * by the driver, or if the backend storage doesn't support it, a negative
 702  * error code will be returned.
 703  *
 704  * Note: block layer doesn't emulate or fallback to a bounce buffer approach
 705  * because usually the caller shouldn't attempt offloaded copy any more (e.g.
 706  * calling copy_file_range(2)) after the first error, thus it should fall back
 707  * to a read+write path in the caller level.
 708  *
 709  * @src: Source child to copy data from
 710  * @src_offset: offset in @src image to read data
 711  * @dst: Destination child to copy data to
 712  * @dst_offset: offset in @dst image to write data
 713  * @bytes: number of bytes to copy
 714  * @flags: request flags. Supported flags:
 715  *         BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero
 716  *                               write on @dst as if bdrv_co_pwrite_zeroes is
 717  *                               called. Used to simplify caller code, or
 718  *                               during BlockDriver.bdrv_co_copy_range_from()
 719  *                               recursion.
 720  *         BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping
 721  *                                   requests currently in flight.
 722  *
 723  * Returns: 0 if succeeded; negative error code if failed.
 724  **/
 725 int coroutine_fn bdrv_co_copy_range(BdrvChild *src, uint64_t src_offset,
 726                                     BdrvChild *dst, uint64_t dst_offset,
 727                                     uint64_t bytes, BdrvRequestFlags read_flags,
 728                                     BdrvRequestFlags write_flags);
 729 #endif