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