include/block/block.h

   1 #ifndef BLOCK_H
   2 #define BLOCK_H
   3
   4 #include "block/aio.h"
   5 #include "qapi-types.h"
   6 #include "qemu/iov.h"
   7 #include "qemu/option.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 the driver can optimize writing zeroes by unmapping
  32      * sectors. This is equivalent to the BLKDISCARDZEROES ioctl in Linux
  33      * with the difference that in qemu a discard is allowed to silently
  34      * fail. Therefore we have to use bdrv_pwrite_zeroes with the
  35      * BDRV_REQ_MAY_UNMAP flag for an optimized zero write with unmapping.
  36      * After this call the driver has to guarantee that the contents read
  37      * back as zero. It is additionally required that the block device is
  38      * opened with BDRV_O_UNMAP flag for this to work.
  39      */
  40     bool can_write_zeroes_with_unmap;
  41     /*
  42      * True if this block driver only supports compressed writes
  43      */
  44     bool needs_compressed_writes;
  45 } BlockDriverInfo;
  46
  47 typedef struct BlockFragInfo {
  48     uint64_t allocated_clusters;
  49     uint64_t total_clusters;
  50     uint64_t fragmented_clusters;
  51     uint64_t compressed_clusters;
  52 } BlockFragInfo;
  53
  54 typedef enum {
  55     BDRV_REQ_COPY_ON_READ       = 0x1,
  56     BDRV_REQ_ZERO_WRITE         = 0x2,
  57     /* The BDRV_REQ_MAY_UNMAP flag is used to indicate that the block driver
  58      * is allowed to optimize a write zeroes request by unmapping (discarding)
  59      * blocks if it is guaranteed that the result will read back as
  60      * zeroes. The flag is only passed to the driver if the block device is
  61      * opened with BDRV_O_UNMAP.
  62      */
  63     BDRV_REQ_MAY_UNMAP          = 0x4,
  64     BDRV_REQ_NO_SERIALISING     = 0x8,
  65     BDRV_REQ_FUA                = 0x10,
  66     BDRV_REQ_WRITE_COMPRESSED   = 0x20,
  67
  68     /* Mask of valid flags */
  69     BDRV_REQ_MASK               = 0x3f,
  70 } BdrvRequestFlags;
  71
  72 typedef struct BlockSizes {
  73     uint32_t phys;
  74     uint32_t log;
  75 } BlockSizes;
  76
  77 typedef struct HDGeometry {
  78     uint32_t heads;
  79     uint32_t sectors;
  80     uint32_t cylinders;
  81 } HDGeometry;
  82
  83 #define BDRV_O_RDWR        0x0002
  84 #define BDRV_O_RESIZE      0x0004 /* request permission for resizing the node */
  85 #define BDRV_O_SNAPSHOT    0x0008 /* open the file read only and save writes in a snapshot */
  86 #define BDRV_O_TEMPORARY   0x0010 /* delete the file after use */
  87 #define BDRV_O_NOCACHE     0x0020 /* do not use the host page cache */
  88 #define BDRV_O_NATIVE_AIO  0x0080 /* use native AIO instead of the thread pool */
  89 #define BDRV_O_NO_BACKING  0x0100 /* don't open the backing file */
  90 #define BDRV_O_NO_FLUSH    0x0200 /* disable flushing on this disk */
  91 #define BDRV_O_COPY_ON_READ 0x0400 /* copy read backing sectors into image */
  92 #define BDRV_O_INACTIVE    0x0800  /* consistency hint for migration handoff */
  93 #define BDRV_O_CHECK       0x1000  /* open solely for consistency check */
  94 #define BDRV_O_ALLOW_RDWR  0x2000  /* allow reopen to change from r/o to r/w */
  95 #define BDRV_O_UNMAP       0x4000  /* execute guest UNMAP/TRIM operations */
  96 #define BDRV_O_PROTOCOL    0x8000  /* if no block driver is explicitly given:
  97                                       select an appropriate protocol driver,
  98                                       ignoring the format layer */
  99 #define BDRV_O_NO_IO       0x10000 /* don't initialize for I/O */
 100
 101 #define BDRV_O_CACHE_MASK  (BDRV_O_NOCACHE | BDRV_O_NO_FLUSH)
 102
 103
 104 /* Option names of options parsed by the block layer */
 105
 106 #define BDRV_OPT_CACHE_WB       "cache.writeback"
 107 #define BDRV_OPT_CACHE_DIRECT   "cache.direct"
 108 #define BDRV_OPT_CACHE_NO_FLUSH "cache.no-flush"
 109 #define BDRV_OPT_READ_ONLY      "read-only"
 110 #define BDRV_OPT_DISCARD        "discard"
 111 #define BDRV_OPT_FORCE_SHARE    "force-share"
 112
 113
 114 #define BDRV_SECTOR_BITS   9
 115 #define BDRV_SECTOR_SIZE   (1ULL << BDRV_SECTOR_BITS)
 116 #define BDRV_SECTOR_MASK   ~(BDRV_SECTOR_SIZE - 1)
 117
 118 #define BDRV_REQUEST_MAX_SECTORS MIN(SIZE_MAX >> BDRV_SECTOR_BITS, \
 119                                      INT_MAX >> BDRV_SECTOR_BITS)
 120 #define BDRV_REQUEST_MAX_BYTES (BDRV_REQUEST_MAX_SECTORS << BDRV_SECTOR_BITS)
 121
 122 /*
 123  * Allocation status flags for bdrv_block_status() and friends.
 124  *
 125  * Public flags:
 126  * BDRV_BLOCK_DATA: allocation for data at offset is tied to this layer
 127  * BDRV_BLOCK_ZERO: offset reads as zero
 128  * BDRV_BLOCK_OFFSET_VALID: an associated offset exists for accessing raw data
 129  * BDRV_BLOCK_ALLOCATED: the content of the block is determined by this
 130  *                       layer (short for DATA || ZERO), set by block layer
 131  * BDRV_BLOCK_EOF: the returned pnum covers through end of file for this layer
 132  *
 133  * Internal flag:
 134  * BDRV_BLOCK_RAW: for use by passthrough drivers, such as raw, to request
 135  *                 that the block layer recompute the answer from the returned
 136  *                 BDS; must be accompanied by just BDRV_BLOCK_OFFSET_VALID.
 137  *
 138  * If BDRV_BLOCK_OFFSET_VALID is set, bits 9-62 (BDRV_BLOCK_OFFSET_MASK) of
 139  * the return value (old interface) or the entire map parameter (new
 140  * interface) represent the offset in the returned BDS that is allocated for
 141  * the corresponding raw data.  However, whether that offset actually
 142  * contains data also depends on BDRV_BLOCK_DATA, as follows:
 143  *
 144  * DATA ZERO OFFSET_VALID
 145  *  t    t        t       sectors read as zero, returned file is zero at offset
 146  *  t    f        t       sectors read as valid from file at offset
 147  *  f    t        t       sectors preallocated, read as zero, returned file not
 148  *                        necessarily zero at offset
 149  *  f    f        t       sectors preallocated but read from backing_hd,
 150  *                        returned file contains garbage at offset
 151  *  t    t        f       sectors preallocated, read as zero, unknown offset
 152  *  t    f        f       sectors read from unknown file or offset
 153  *  f    t        f       not allocated or unknown offset, read as zero
 154  *  f    f        f       not allocated or unknown offset, read from backing_hd
 155  */
 156 #define BDRV_BLOCK_DATA         0x01
 157 #define BDRV_BLOCK_ZERO         0x02
 158 #define BDRV_BLOCK_OFFSET_VALID 0x04
 159 #define BDRV_BLOCK_RAW          0x08
 160 #define BDRV_BLOCK_ALLOCATED    0x10
 161 #define BDRV_BLOCK_EOF          0x20
 162 #define BDRV_BLOCK_OFFSET_MASK  BDRV_SECTOR_MASK
 163
 164 typedef QSIMPLEQ_HEAD(BlockReopenQueue, BlockReopenQueueEntry) BlockReopenQueue;
 165
 166 typedef struct BDRVReopenState {
 167     BlockDriverState *bs;
 168     int flags;
 169     uint64_t perm, shared_perm;
 170     QDict *options;
 171     QDict *explicit_options;
 172     void *opaque;
 173 } BDRVReopenState;
 174
 175 /*
 176  * Block operation types
 177  */
 178 typedef enum BlockOpType {
 179     BLOCK_OP_TYPE_BACKUP_SOURCE,
 180     BLOCK_OP_TYPE_BACKUP_TARGET,
 181     BLOCK_OP_TYPE_CHANGE,
 182     BLOCK_OP_TYPE_COMMIT_SOURCE,
 183     BLOCK_OP_TYPE_COMMIT_TARGET,
 184     BLOCK_OP_TYPE_DATAPLANE,
 185     BLOCK_OP_TYPE_DRIVE_DEL,
 186     BLOCK_OP_TYPE_EJECT,
 187     BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
 188     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT,
 189     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE,
 190     BLOCK_OP_TYPE_MIRROR_SOURCE,
 191     BLOCK_OP_TYPE_MIRROR_TARGET,
 192     BLOCK_OP_TYPE_RESIZE,
 193     BLOCK_OP_TYPE_STREAM,
 194     BLOCK_OP_TYPE_REPLACE,
 195     BLOCK_OP_TYPE_MAX,
 196 } BlockOpType;
 197
 198 /* Block node permission constants */
 199 enum {
 200     /**
 201      * A user that has the "permission" of consistent reads is guaranteed that
 202      * their view of the contents of the block device is complete and
 203      * self-consistent, representing the contents of a disk at a specific
 204      * point.
 205      *
 206      * For most block devices (including their backing files) this is true, but
 207      * the property cannot be maintained in a few situations like for
 208      * intermediate nodes of a commit block job.
 209      */
 210     BLK_PERM_CONSISTENT_READ    = 0x01,
 211
 212     /** This permission is required to change the visible disk contents. */
 213     BLK_PERM_WRITE              = 0x02,
 214
 215     /**
 216      * This permission (which is weaker than BLK_PERM_WRITE) is both enough and
 217      * required for writes to the block node when the caller promises that
 218      * the visible disk content doesn't change.
 219      */
 220     BLK_PERM_WRITE_UNCHANGED    = 0x04,
 221
 222     /** This permission is required to change the size of a block node. */
 223     BLK_PERM_RESIZE             = 0x08,
 224
 225     /**
 226      * This permission is required to change the node that this BdrvChild
 227      * points to.
 228      */
 229     BLK_PERM_GRAPH_MOD          = 0x10,
 230
 231     BLK_PERM_ALL                = 0x1f,
 232 };
 233
 234 char *bdrv_perm_names(uint64_t perm);
 235
 236 /* disk I/O throttling */
 237 void bdrv_init(void);
 238 void bdrv_init_with_whitelist(void);
 239 bool bdrv_uses_whitelist(void);
 240 BlockDriver *bdrv_find_protocol(const char *filename,
 241                                 bool allow_protocol_prefix,
 242                                 Error **errp);
 243 BlockDriver *bdrv_find_format(const char *format_name);
 244 int bdrv_create(BlockDriver *drv, const char* filename,
 245                 QemuOpts *opts, Error **errp);
 246 int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp);
 247 BlockDriverState *bdrv_new(void);
 248 void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top,
 249                  Error **errp);
 250 void bdrv_replace_node(BlockDriverState *from, BlockDriverState *to,
 251                        Error **errp);
 252
 253 int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough);
 254 int bdrv_parse_discard_flags(const char *mode, int *flags);
 255 BdrvChild *bdrv_open_child(const char *filename,
 256                            QDict *options, const char *bdref_key,
 257                            BlockDriverState* parent,
 258                            const BdrvChildRole *child_role,
 259                            bool allow_none, Error **errp);
 260 void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd,
 261                          Error **errp);
 262 int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options,
 263                            const char *bdref_key, Error **errp);
 264 BlockDriverState *bdrv_open(const char *filename, const char *reference,
 265                             QDict *options, int flags, Error **errp);
 266 BlockDriverState *bdrv_new_open_driver(BlockDriver *drv, const char *node_name,
 267                                        int flags, Error **errp);
 268 BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue,
 269                                     BlockDriverState *bs,
 270                                     QDict *options, int flags);
 271 int bdrv_reopen_multiple(AioContext *ctx, BlockReopenQueue *bs_queue, Error **errp);
 272 int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp);
 273 int bdrv_reopen_prepare(BDRVReopenState *reopen_state,
 274                         BlockReopenQueue *queue, Error **errp);
 275 void bdrv_reopen_commit(BDRVReopenState *reopen_state);
 276 void bdrv_reopen_abort(BDRVReopenState *reopen_state);
 277 int bdrv_read(BdrvChild *child, int64_t sector_num,
 278               uint8_t *buf, int nb_sectors);
 279 int bdrv_write(BdrvChild *child, int64_t sector_num,
 280                const uint8_t *buf, int nb_sectors);
 281 int bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset,
 282                        int bytes, BdrvRequestFlags flags);
 283 int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags);
 284 int bdrv_pread(BdrvChild *child, int64_t offset, void *buf, int bytes);
 285 int bdrv_preadv(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
 286 int bdrv_pwrite(BdrvChild *child, int64_t offset, const void *buf, int bytes);
 287 int bdrv_pwritev(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
 288 int bdrv_pwrite_sync(BdrvChild *child, int64_t offset,
 289                      const void *buf, int count);
 290 int coroutine_fn bdrv_co_readv(BdrvChild *child, int64_t sector_num,
 291                                int nb_sectors, QEMUIOVector *qiov);
 292 int coroutine_fn bdrv_co_writev(BdrvChild *child, int64_t sector_num,
 293                                int nb_sectors, QEMUIOVector *qiov);
 294 /*
 295  * Efficiently zero a region of the disk image.  Note that this is a regular
 296  * I/O request like read or write and should have a reasonable size.  This
 297  * function is not suitable for zeroing the entire image in a single request
 298  * because it may allocate memory for the entire region.
 299  */
 300 int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset,
 301                                        int bytes, BdrvRequestFlags flags);
 302 BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,
 303     const char *backing_file);
 304 void bdrv_refresh_filename(BlockDriverState *bs);
 305 int bdrv_truncate(BdrvChild *child, int64_t offset, PreallocMode prealloc,
 306                   Error **errp);
 307 int64_t bdrv_nb_sectors(BlockDriverState *bs);
 308 int64_t bdrv_getlength(BlockDriverState *bs);
 309 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs);
 310 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
 311                                BlockDriverState *in_bs, Error **errp);
 312 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr);
 313 void bdrv_refresh_limits(BlockDriverState *bs, Error **errp);
 314 int bdrv_commit(BlockDriverState *bs);
 315 int bdrv_change_backing_file(BlockDriverState *bs,
 316     const char *backing_file, const char *backing_fmt);
 317 void bdrv_register(BlockDriver *bdrv);
 318 int bdrv_drop_intermediate(BlockDriverState *top, BlockDriverState *base,
 319                            const char *backing_file_str);
 320 BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
 321                                     BlockDriverState *bs);
 322 BlockDriverState *bdrv_find_base(BlockDriverState *bs);
 323
 324
 325 typedef struct BdrvCheckResult {
 326     int corruptions;
 327     int leaks;
 328     int check_errors;
 329     int corruptions_fixed;
 330     int leaks_fixed;
 331     int64_t image_end_offset;
 332     BlockFragInfo bfi;
 333 } BdrvCheckResult;
 334
 335 typedef enum {
 336     BDRV_FIX_LEAKS    = 1,
 337     BDRV_FIX_ERRORS   = 2,
 338 } BdrvCheckMode;
 339
 340 int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix);
 341
 342 /* The units of offset and total_work_size may be chosen arbitrarily by the
 343  * block driver; total_work_size may change during the course of the amendment
 344  * operation */
 345 typedef void BlockDriverAmendStatusCB(BlockDriverState *bs, int64_t offset,
 346                                       int64_t total_work_size, void *opaque);
 347 int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts,
 348                        BlockDriverAmendStatusCB *status_cb, void *cb_opaque);
 349
 350 /* external snapshots */
 351 bool bdrv_recurse_is_first_non_filter(BlockDriverState *bs,
 352                                       BlockDriverState *candidate);
 353 bool bdrv_is_first_non_filter(BlockDriverState *candidate);
 354
 355 /* check if a named node can be replaced when doing drive-mirror */
 356 BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs,
 357                                         const char *node_name, Error **errp);
 358
 359 /* async block I/O */
 360 void bdrv_aio_cancel(BlockAIOCB *acb);
 361 void bdrv_aio_cancel_async(BlockAIOCB *acb);
 362
 363 /* sg packet commands */
 364 int bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf);
 365
 366 /* Invalidate any cached metadata used by image formats */
 367 void bdrv_invalidate_cache(BlockDriverState *bs, Error **errp);
 368 void bdrv_invalidate_cache_all(Error **errp);
 369 int bdrv_inactivate_all(void);
 370
 371 /* Ensure contents are flushed to disk.  */
 372 int bdrv_flush(BlockDriverState *bs);
 373 int coroutine_fn bdrv_co_flush(BlockDriverState *bs);
 374 int bdrv_flush_all(void);
 375 void bdrv_close_all(void);
 376 void bdrv_drain(BlockDriverState *bs);
 377 void coroutine_fn bdrv_co_drain(BlockDriverState *bs);
 378 void bdrv_drain_all_begin(void);
 379 void bdrv_drain_all_end(void);
 380 void bdrv_drain_all(void);
 381
 382 #define BDRV_POLL_WHILE(bs, cond) ({                       \
 383     bool waited_ = false;                                  \
 384     bool busy_ = true;                                     \
 385     BlockDriverState *bs_ = (bs);                          \
 386     AioContext *ctx_ = bdrv_get_aio_context(bs_);          \
 387     if (aio_context_in_iothread(ctx_)) {                   \
 388         while ((cond) || busy_) {                          \
 389             busy_ = aio_poll(ctx_, (cond));                \
 390             waited_ |= !!(cond) | busy_;                   \
 391         }                                                  \
 392     } else {                                               \
 393         assert(qemu_get_current_aio_context() ==           \
 394                qemu_get_aio_context());                    \
 395         /* Ask bdrv_dec_in_flight to wake up the main      \
 396          * QEMU AioContext.  Extra I/O threads never take  \
 397          * other I/O threads' AioContexts (see for example \
 398          * block_job_defer_to_main_loop for how to do it). \
 399          */                                                \
 400         assert(!bs_->wakeup);                              \
 401         /* Set bs->wakeup before evaluating cond.  */      \
 402         atomic_mb_set(&bs_->wakeup, true);                 \
 403         while (busy_) {                                    \
 404             if ((cond)) {                                  \
 405                 waited_ = busy_ = true;                    \
 406                 aio_context_release(ctx_);                 \
 407                 aio_poll(qemu_get_aio_context(), true);    \
 408                 aio_context_acquire(ctx_);                 \
 409             } else {                                       \
 410                 busy_ = aio_poll(ctx_, false);             \
 411                 waited_ |= busy_;                          \
 412             }                                              \
 413         }                                                  \
 414         atomic_set(&bs_->wakeup, false);                   \
 415     }                                                      \
 416     waited_; })
 417
 418 int bdrv_pdiscard(BlockDriverState *bs, int64_t offset, int bytes);
 419 int bdrv_co_pdiscard(BlockDriverState *bs, int64_t offset, int bytes);
 420 int bdrv_has_zero_init_1(BlockDriverState *bs);
 421 int bdrv_has_zero_init(BlockDriverState *bs);
 422 bool bdrv_unallocated_blocks_are_zero(BlockDriverState *bs);
 423 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs);
 424 int bdrv_block_status(BlockDriverState *bs, int64_t offset,
 425                       int64_t bytes, int64_t *pnum, int64_t *map,
 426                       BlockDriverState **file);
 427 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base,
 428                             int64_t offset, int64_t bytes, int64_t *pnum,
 429                             int64_t *map, BlockDriverState **file);
 430 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
 431                       int64_t *pnum);
 432 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
 433                             int64_t offset, int64_t bytes, int64_t *pnum);
 434
 435 bool bdrv_is_read_only(BlockDriverState *bs);
 436 int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only,
 437                            bool ignore_allow_rdw, Error **errp);
 438 int bdrv_set_read_only(BlockDriverState *bs, bool read_only, Error **errp);
 439 bool bdrv_is_sg(BlockDriverState *bs);
 440 bool bdrv_is_inserted(BlockDriverState *bs);
 441 void bdrv_lock_medium(BlockDriverState *bs, bool locked);
 442 void bdrv_eject(BlockDriverState *bs, bool eject_flag);
 443 const char *bdrv_get_format_name(BlockDriverState *bs);
 444 BlockDriverState *bdrv_find_node(const char *node_name);
 445 BlockDeviceInfoList *bdrv_named_nodes_list(Error **errp);
 446 BlockDriverState *bdrv_lookup_bs(const char *device,
 447                                  const char *node_name,
 448                                  Error **errp);
 449 bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base);
 450 BlockDriverState *bdrv_next_node(BlockDriverState *bs);
 451
 452 typedef struct BdrvNextIterator {
 453     enum {
 454         BDRV_NEXT_BACKEND_ROOTS,
 455         BDRV_NEXT_MONITOR_OWNED,
 456     } phase;
 457     BlockBackend *blk;
 458     BlockDriverState *bs;
 459 } BdrvNextIterator;
 460
 461 BlockDriverState *bdrv_first(BdrvNextIterator *it);
 462 BlockDriverState *bdrv_next(BdrvNextIterator *it);
 463 void bdrv_next_cleanup(BdrvNextIterator *it);
 464
 465 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs);
 466 bool bdrv_is_encrypted(BlockDriverState *bs);
 467 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
 468                          void *opaque);
 469 const char *bdrv_get_node_name(const BlockDriverState *bs);
 470 const char *bdrv_get_device_name(const BlockDriverState *bs);
 471 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
 472 int bdrv_get_flags(BlockDriverState *bs);
 473 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
 474 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs);
 475 void bdrv_round_to_clusters(BlockDriverState *bs,
 476                             int64_t offset, int64_t bytes,
 477                             int64_t *cluster_offset,
 478                             int64_t *cluster_bytes);
 479
 480 const char *bdrv_get_encrypted_filename(BlockDriverState *bs);
 481 void bdrv_get_backing_filename(BlockDriverState *bs,
 482                                char *filename, int filename_size);
 483 void bdrv_get_full_backing_filename(BlockDriverState *bs,
 484                                     char *dest, size_t sz, Error **errp);
 485 void bdrv_get_full_backing_filename_from_filename(const char *backed,
 486                                                   const char *backing,
 487                                                   char *dest, size_t sz,
 488                                                   Error **errp);
 489
 490 int path_has_protocol(const char *path);
 491 int path_is_absolute(const char *path);
 492 void path_combine(char *dest, int dest_size,
 493                   const char *base_path,
 494                   const char *filename);
 495
 496 int bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
 497 int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
 498 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
 499                       int64_t pos, int size);
 500
 501 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
 502                       int64_t pos, int size);
 503
 504 void bdrv_img_create(const char *filename, const char *fmt,
 505                      const char *base_filename, const char *base_fmt,
 506                      char *options, uint64_t img_size, int flags,
 507                      bool quiet, Error **errp);
 508
 509 /* Returns the alignment in bytes that is required so that no bounce buffer
 510  * is required throughout the stack */
 511 size_t bdrv_min_mem_align(BlockDriverState *bs);
 512 /* Returns optimal alignment in bytes for bounce buffer */
 513 size_t bdrv_opt_mem_align(BlockDriverState *bs);
 514 void *qemu_blockalign(BlockDriverState *bs, size_t size);
 515 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
 516 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
 517 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
 518 bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov);
 519
 520 void bdrv_enable_copy_on_read(BlockDriverState *bs);
 521 void bdrv_disable_copy_on_read(BlockDriverState *bs);
 522
 523 void bdrv_ref(BlockDriverState *bs);
 524 void bdrv_unref(BlockDriverState *bs);
 525 void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child);
 526 BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
 527                              BlockDriverState *child_bs,
 528                              const char *child_name,
 529                              const BdrvChildRole *child_role,
 530                              Error **errp);
 531
 532 bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp);
 533 void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason);
 534 void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason);
 535 void bdrv_op_block_all(BlockDriverState *bs, Error *reason);
 536 void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason);
 537 bool bdrv_op_blocker_is_empty(BlockDriverState *bs);
 538
 539 #define BLKDBG_EVENT(child, evt) \
 540     do { \
 541         if (child) { \
 542             bdrv_debug_event(child->bs, evt); \
 543         } \
 544     } while (0)
 545
 546 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
 547
 548 int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event,
 549                            const char *tag);
 550 int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag);
 551 int bdrv_debug_resume(BlockDriverState *bs, const char *tag);
 552 bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag);
 553
 554 /**
 555  * bdrv_get_aio_context:
 556  *
 557  * Returns: the currently bound #AioContext
 558  */
 559 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
 560
 561 /**
 562  * Transfer control to @co in the aio context of @bs
 563  */
 564 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co);
 565
 566 /**
 567  * bdrv_set_aio_context:
 568  *
 569  * Changes the #AioContext used for fd handlers, timers, and BHs by this
 570  * BlockDriverState and all its children.
 571  *
 572  * This function must be called with iothread lock held.
 573  */
 574 void bdrv_set_aio_context(BlockDriverState *bs, AioContext *new_context);
 575 int bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz);
 576 int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo);
 577
 578 void bdrv_io_plug(BlockDriverState *bs);
 579 void bdrv_io_unplug(BlockDriverState *bs);
 580
 581 /**
 582  * bdrv_parent_drained_begin:
 583  *
 584  * Begin a quiesced section of all users of @bs. This is part of
 585  * bdrv_drained_begin.
 586  */
 587 void bdrv_parent_drained_begin(BlockDriverState *bs, BdrvChild *ignore);
 588
 589 /**
 590  * bdrv_parent_drained_end:
 591  *
 592  * End a quiesced section of all users of @bs. This is part of
 593  * bdrv_drained_end.
 594  */
 595 void bdrv_parent_drained_end(BlockDriverState *bs, BdrvChild *ignore);
 596
 597 /**
 598  * bdrv_drained_begin:
 599  *
 600  * Begin a quiesced section for exclusive access to the BDS, by disabling
 601  * external request sources including NBD server and device model. Note that
 602  * this doesn't block timers or coroutines from submitting more requests, which
 603  * means block_job_pause is still necessary.
 604  *
 605  * This function can be recursive.
 606  */
 607 void bdrv_drained_begin(BlockDriverState *bs);
 608
 609 /**
 610  * Like bdrv_drained_begin, but recursively begins a quiesced section for
 611  * exclusive access to all child nodes as well.
 612  */
 613 void bdrv_subtree_drained_begin(BlockDriverState *bs);
 614
 615 /**
 616  * bdrv_drained_end:
 617  *
 618  * End a quiescent section started by bdrv_drained_begin().
 619  */
 620 void bdrv_drained_end(BlockDriverState *bs);
 621
 622 /**
 623  * End a quiescent section started by bdrv_subtree_drained_begin().
 624  */
 625 void bdrv_subtree_drained_end(BlockDriverState *bs);
 626
 627 void bdrv_add_child(BlockDriverState *parent, BlockDriverState *child,
 628                     Error **errp);
 629 void bdrv_del_child(BlockDriverState *parent, BdrvChild *child, Error **errp);
 630
 631 bool bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
 632                                      uint32_t granularity, Error **errp);
 633 /**
 634  *
 635  * bdrv_register_buf/bdrv_unregister_buf:
 636  *
 637  * Register/unregister a buffer for I/O. For example, VFIO drivers are
 638  * interested to know the memory areas that would later be used for I/O, so
 639  * that they can prepare IOMMU mapping etc., to get better performance.
 640  */
 641 void bdrv_register_buf(BlockDriverState *bs, void *host, size_t size);
 642 void bdrv_unregister_buf(BlockDriverState *bs, void *host);
 643 #endif