1 #ifndef BLOCK_H 2 #define BLOCK_H 3 4 #include "block/aio.h" 5 #include "block/aio-wait.h" 6 #include "qemu/iov.h" 7 #include "qemu/coroutine.h" 8 #include "block/accounting.h" 9 #include "block/dirty-bitmap.h" 10 #include "block/blockjob.h" 11 #include "qemu/hbitmap.h" 12 13 /* 14 * generated_co_wrapper 15 * 16 * Function specifier, which does nothing but mark functions to be 17 * generated by scripts/block-coroutine-wrapper.py 18 * 19 * Read more in docs/devel/block-coroutine-wrapper.rst 20 */ 21 #define generated_co_wrapper 22 23 /* block.c */ 24 typedef struct BlockDriver BlockDriver; 25 typedef struct BdrvChild BdrvChild; 26 typedef struct BdrvChildClass BdrvChildClass; 27 28 typedef struct BlockDriverInfo { 29 /* in bytes, 0 if irrelevant */ 30 int cluster_size; 31 /* offset at which the VM state can be saved (0 if not possible) */ 32 int64_t vm_state_offset; 33 bool is_dirty; 34 /* 35 * True if this block driver only supports compressed writes 36 */ 37 bool needs_compressed_writes; 38 } BlockDriverInfo; 39 40 typedef struct BlockFragInfo { 41 uint64_t allocated_clusters; 42 uint64_t total_clusters; 43 uint64_t fragmented_clusters; 44 uint64_t compressed_clusters; 45 } BlockFragInfo; 46 47 typedef enum { 48 BDRV_REQ_COPY_ON_READ = 0x1, 49 BDRV_REQ_ZERO_WRITE = 0x2, 50 51 /* 52 * The BDRV_REQ_MAY_UNMAP flag is used in write_zeroes requests to indicate 53 * that the block driver should unmap (discard) blocks if it is guaranteed 54 * that the result will read back as zeroes. The flag is only passed to the 55 * driver if the block device is opened with BDRV_O_UNMAP. 56 */ 57 BDRV_REQ_MAY_UNMAP = 0x4, 58 59 BDRV_REQ_FUA = 0x10, 60 BDRV_REQ_WRITE_COMPRESSED = 0x20, 61 62 /* Signifies that this write request will not change the visible disk 63 * content. */ 64 BDRV_REQ_WRITE_UNCHANGED = 0x40, 65 66 /* Forces request serialisation. Use only with write requests. */ 67 BDRV_REQ_SERIALISING = 0x80, 68 69 /* Execute the request only if the operation can be offloaded or otherwise 70 * be executed efficiently, but return an error instead of using a slow 71 * fallback. */ 72 BDRV_REQ_NO_FALLBACK = 0x100, 73 74 /* 75 * BDRV_REQ_PREFETCH makes sense only in the context of copy-on-read 76 * (i.e., together with the BDRV_REQ_COPY_ON_READ flag or when a COR 77 * filter is involved), in which case it signals that the COR operation 78 * need not read the data into memory (qiov) but only ensure they are 79 * copied to the top layer (i.e., that COR operation is done). 80 */ 81 BDRV_REQ_PREFETCH = 0x200, 82 83 /* 84 * If we need to wait for other requests, just fail immediately. Used 85 * only together with BDRV_REQ_SERIALISING. 86 */ 87 BDRV_REQ_NO_WAIT = 0x400, 88 89 /* Mask of valid flags */ 90 BDRV_REQ_MASK = 0x7ff, 91 } BdrvRequestFlags; 92 93 typedef struct BlockSizes { 94 uint32_t phys; 95 uint32_t log; 96 } BlockSizes; 97 98 typedef struct HDGeometry { 99 uint32_t heads; 100 uint32_t sectors; 101 uint32_t cylinders; 102 } HDGeometry; 103 104 #define BDRV_O_RDWR 0x0002 105 #define BDRV_O_RESIZE 0x0004 /* request permission for resizing the node */ 106 #define BDRV_O_SNAPSHOT 0x0008 /* open the file read only and save writes in a snapshot */ 107 #define BDRV_O_TEMPORARY 0x0010 /* delete the file after use */ 108 #define BDRV_O_NOCACHE 0x0020 /* do not use the host page cache */ 109 #define BDRV_O_NATIVE_AIO 0x0080 /* use native AIO instead of the thread pool */ 110 #define BDRV_O_NO_BACKING 0x0100 /* don't open the backing file */ 111 #define BDRV_O_NO_FLUSH 0x0200 /* disable flushing on this disk */ 112 #define BDRV_O_COPY_ON_READ 0x0400 /* copy read backing sectors into image */ 113 #define BDRV_O_INACTIVE 0x0800 /* consistency hint for migration handoff */ 114 #define BDRV_O_CHECK 0x1000 /* open solely for consistency check */ 115 #define BDRV_O_ALLOW_RDWR 0x2000 /* allow reopen to change from r/o to r/w */ 116 #define BDRV_O_UNMAP 0x4000 /* execute guest UNMAP/TRIM operations */ 117 #define BDRV_O_PROTOCOL 0x8000 /* if no block driver is explicitly given: 118 select an appropriate protocol driver, 119 ignoring the format layer */ 120 #define BDRV_O_NO_IO 0x10000 /* don't initialize for I/O */ 121 #define BDRV_O_AUTO_RDONLY 0x20000 /* degrade to read-only if opening read-write fails */ 122 #define BDRV_O_IO_URING 0x40000 /* use io_uring instead of the thread pool */ 123 124 #define BDRV_O_CACHE_MASK (BDRV_O_NOCACHE | BDRV_O_NO_FLUSH) 125 126 127 /* Option names of options parsed by the block layer */ 128 129 #define BDRV_OPT_CACHE_WB "cache.writeback" 130 #define BDRV_OPT_CACHE_DIRECT "cache.direct" 131 #define BDRV_OPT_CACHE_NO_FLUSH "cache.no-flush" 132 #define BDRV_OPT_READ_ONLY "read-only" 133 #define BDRV_OPT_AUTO_READ_ONLY "auto-read-only" 134 #define BDRV_OPT_DISCARD "discard" 135 #define BDRV_OPT_FORCE_SHARE "force-share" 136 137 138 #define BDRV_SECTOR_BITS 9 139 #define BDRV_SECTOR_SIZE (1ULL << BDRV_SECTOR_BITS) 140 141 #define BDRV_REQUEST_MAX_SECTORS MIN_CONST(SIZE_MAX >> BDRV_SECTOR_BITS, \ 142 INT_MAX >> BDRV_SECTOR_BITS) 143 #define BDRV_REQUEST_MAX_BYTES (BDRV_REQUEST_MAX_SECTORS << BDRV_SECTOR_BITS) 144 145 /* 146 * We want allow aligning requests and disk length up to any 32bit alignment 147 * and don't afraid of overflow. 148 * To achieve it, and in the same time use some pretty number as maximum disk 149 * size, let's define maximum "length" (a limit for any offset/bytes request and 150 * for disk size) to be the greatest power of 2 less than INT64_MAX. 151 */ 152 #define BDRV_MAX_ALIGNMENT (1L << 30) 153 #define BDRV_MAX_LENGTH (QEMU_ALIGN_DOWN(INT64_MAX, BDRV_MAX_ALIGNMENT)) 154 155 /* 156 * Allocation status flags for bdrv_block_status() and friends. 157 * 158 * Public flags: 159 * BDRV_BLOCK_DATA: allocation for data at offset is tied to this layer 160 * BDRV_BLOCK_ZERO: offset reads as zero 161 * BDRV_BLOCK_OFFSET_VALID: an associated offset exists for accessing raw data 162 * BDRV_BLOCK_ALLOCATED: the content of the block is determined by this 163 * layer rather than any backing, set by block layer 164 * BDRV_BLOCK_EOF: the returned pnum covers through end of file for this 165 * layer, set by block layer 166 * 167 * Internal flags: 168 * BDRV_BLOCK_RAW: for use by passthrough drivers, such as raw, to request 169 * that the block layer recompute the answer from the returned 170 * BDS; must be accompanied by just BDRV_BLOCK_OFFSET_VALID. 171 * BDRV_BLOCK_RECURSE: request that the block layer will recursively search for 172 * zeroes in file child of current block node inside 173 * returned region. Only valid together with both 174 * BDRV_BLOCK_DATA and BDRV_BLOCK_OFFSET_VALID. Should not 175 * appear with BDRV_BLOCK_ZERO. 176 * 177 * If BDRV_BLOCK_OFFSET_VALID is set, the map parameter represents the 178 * host offset within the returned BDS that is allocated for the 179 * corresponding raw guest data. However, whether that offset 180 * actually contains data also depends on BDRV_BLOCK_DATA, as follows: 181 * 182 * DATA ZERO OFFSET_VALID 183 * t t t sectors read as zero, returned file is zero at offset 184 * t f t sectors read as valid from file at offset 185 * f t t sectors preallocated, read as zero, returned file not 186 * necessarily zero at offset 187 * f f t sectors preallocated but read from backing_hd, 188 * returned file contains garbage at offset 189 * t t f sectors preallocated, read as zero, unknown offset 190 * t f f sectors read from unknown file or offset 191 * f t f not allocated or unknown offset, read as zero 192 * f f f not allocated or unknown offset, read from backing_hd 193 */ 194 #define BDRV_BLOCK_DATA 0x01 195 #define BDRV_BLOCK_ZERO 0x02 196 #define BDRV_BLOCK_OFFSET_VALID 0x04 197 #define BDRV_BLOCK_RAW 0x08 198 #define BDRV_BLOCK_ALLOCATED 0x10 199 #define BDRV_BLOCK_EOF 0x20 200 #define BDRV_BLOCK_RECURSE 0x40 201 202 typedef QTAILQ_HEAD(BlockReopenQueue, BlockReopenQueueEntry) BlockReopenQueue; 203 204 typedef struct BDRVReopenState { 205 BlockDriverState *bs; 206 int flags; 207 BlockdevDetectZeroesOptions detect_zeroes; 208 bool backing_missing; 209 bool replace_backing_bs; /* new_backing_bs is ignored if this is false */ 210 BlockDriverState *new_backing_bs; /* If NULL then detach the current bs */ 211 uint64_t perm, shared_perm; 212 QDict *options; 213 QDict *explicit_options; 214 void *opaque; 215 } BDRVReopenState; 216 217 /* 218 * Block operation types 219 */ 220 typedef enum BlockOpType { 221 BLOCK_OP_TYPE_BACKUP_SOURCE, 222 BLOCK_OP_TYPE_BACKUP_TARGET, 223 BLOCK_OP_TYPE_CHANGE, 224 BLOCK_OP_TYPE_COMMIT_SOURCE, 225 BLOCK_OP_TYPE_COMMIT_TARGET, 226 BLOCK_OP_TYPE_DATAPLANE, 227 BLOCK_OP_TYPE_DRIVE_DEL, 228 BLOCK_OP_TYPE_EJECT, 229 BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, 230 BLOCK_OP_TYPE_INTERNAL_SNAPSHOT, 231 BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE, 232 BLOCK_OP_TYPE_MIRROR_SOURCE, 233 BLOCK_OP_TYPE_MIRROR_TARGET, 234 BLOCK_OP_TYPE_RESIZE, 235 BLOCK_OP_TYPE_STREAM, 236 BLOCK_OP_TYPE_REPLACE, 237 BLOCK_OP_TYPE_MAX, 238 } BlockOpType; 239 240 /* Block node permission constants */ 241 enum { 242 /** 243 * A user that has the "permission" of consistent reads is guaranteed that 244 * their view of the contents of the block device is complete and 245 * self-consistent, representing the contents of a disk at a specific 246 * point. 247 * 248 * For most block devices (including their backing files) this is true, but 249 * the property cannot be maintained in a few situations like for 250 * intermediate nodes of a commit block job. 251 */ 252 BLK_PERM_CONSISTENT_READ = 0x01, 253 254 /** This permission is required to change the visible disk contents. */ 255 BLK_PERM_WRITE = 0x02, 256 257 /** 258 * This permission (which is weaker than BLK_PERM_WRITE) is both enough and 259 * required for writes to the block node when the caller promises that 260 * the visible disk content doesn't change. 261 * 262 * As the BLK_PERM_WRITE permission is strictly stronger, either is 263 * sufficient to perform an unchanging write. 264 */ 265 BLK_PERM_WRITE_UNCHANGED = 0x04, 266 267 /** This permission is required to change the size of a block node. */ 268 BLK_PERM_RESIZE = 0x08, 269 270 /** 271 * This permission is required to change the node that this BdrvChild 272 * points to. 273 */ 274 BLK_PERM_GRAPH_MOD = 0x10, 275 276 BLK_PERM_ALL = 0x1f, 277 278 DEFAULT_PERM_PASSTHROUGH = BLK_PERM_CONSISTENT_READ 279 | BLK_PERM_WRITE 280 | BLK_PERM_WRITE_UNCHANGED 281 | BLK_PERM_RESIZE, 282 283 DEFAULT_PERM_UNCHANGED = BLK_PERM_ALL & ~DEFAULT_PERM_PASSTHROUGH, 284 }; 285 286 /* 287 * Flags that parent nodes assign to child nodes to specify what kind of 288 * role(s) they take. 289 * 290 * At least one of DATA, METADATA, FILTERED, or COW must be set for 291 * every child. 292 */ 293 enum BdrvChildRoleBits { 294 /* 295 * This child stores data. 296 * Any node may have an arbitrary number of such children. 297 */ 298 BDRV_CHILD_DATA = (1 << 0), 299 300 /* 301 * This child stores metadata. 302 * Any node may have an arbitrary number of metadata-storing 303 * children. 304 */ 305 BDRV_CHILD_METADATA = (1 << 1), 306 307 /* 308 * A child that always presents exactly the same visible data as 309 * the parent, e.g. by virtue of the parent forwarding all reads 310 * and writes. 311 * This flag is mutually exclusive with DATA, METADATA, and COW. 312 * Any node may have at most one filtered child at a time. 313 */ 314 BDRV_CHILD_FILTERED = (1 << 2), 315 316 /* 317 * Child from which to read all data that isn't allocated in the 318 * parent (i.e., the backing child); such data is copied to the 319 * parent through COW (and optionally COR). 320 * This field is mutually exclusive with DATA, METADATA, and 321 * FILTERED. 322 * Any node may have at most one such backing child at a time. 323 */ 324 BDRV_CHILD_COW = (1 << 3), 325 326 /* 327 * The primary child. For most drivers, this is the child whose 328 * filename applies best to the parent node. 329 * Any node may have at most one primary child at a time. 330 */ 331 BDRV_CHILD_PRIMARY = (1 << 4), 332 333 /* Useful combination of flags */ 334 BDRV_CHILD_IMAGE = BDRV_CHILD_DATA 335 | BDRV_CHILD_METADATA 336 | BDRV_CHILD_PRIMARY, 337 }; 338 339 /* Mask of BdrvChildRoleBits values */ 340 typedef unsigned int BdrvChildRole; 341 342 char *bdrv_perm_names(uint64_t perm); 343 uint64_t bdrv_qapi_perm_to_blk_perm(BlockPermission qapi_perm); 344 345 /* disk I/O throttling */ 346 void bdrv_init(void); 347 void bdrv_init_with_whitelist(void); 348 bool bdrv_uses_whitelist(void); 349 int bdrv_is_whitelisted(BlockDriver *drv, bool read_only); 350 BlockDriver *bdrv_find_protocol(const char *filename, 351 bool allow_protocol_prefix, 352 Error **errp); 353 BlockDriver *bdrv_find_format(const char *format_name); 354 int bdrv_create(BlockDriver *drv, const char* filename, 355 QemuOpts *opts, Error **errp); 356 int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp); 357 358 BlockDriverState *bdrv_new(void); 359 int bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top, 360 Error **errp); 361 int bdrv_replace_node(BlockDriverState *from, BlockDriverState *to, 362 Error **errp); 363 BlockDriverState *bdrv_insert_node(BlockDriverState *bs, QDict *node_options, 364 int flags, Error **errp); 365 366 int bdrv_parse_aio(const char *mode, int *flags); 367 int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough); 368 int bdrv_parse_discard_flags(const char *mode, int *flags); 369 BdrvChild *bdrv_open_child(const char *filename, 370 QDict *options, const char *bdref_key, 371 BlockDriverState* parent, 372 const BdrvChildClass *child_class, 373 BdrvChildRole child_role, 374 bool allow_none, Error **errp); 375 BlockDriverState *bdrv_open_blockdev_ref(BlockdevRef *ref, Error **errp); 376 int bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd, 377 Error **errp); 378 int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options, 379 const char *bdref_key, Error **errp); 380 BlockDriverState *bdrv_open(const char *filename, const char *reference, 381 QDict *options, int flags, Error **errp); 382 BlockDriverState *bdrv_new_open_driver(BlockDriver *drv, const char *node_name, 383 int flags, Error **errp); 384 BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue, 385 BlockDriverState *bs, QDict *options, 386 bool keep_old_opts); 387 int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp); 388 int bdrv_reopen_set_read_only(BlockDriverState *bs, bool read_only, 389 Error **errp); 390 int bdrv_reopen_prepare(BDRVReopenState *reopen_state, 391 BlockReopenQueue *queue, Error **errp); 392 void bdrv_reopen_commit(BDRVReopenState *reopen_state); 393 void bdrv_reopen_abort(BDRVReopenState *reopen_state); 394 int bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset, 395 int64_t bytes, BdrvRequestFlags flags); 396 int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags); 397 int bdrv_pread(BdrvChild *child, int64_t offset, void *buf, int64_t bytes); 398 int bdrv_pwrite(BdrvChild *child, int64_t offset, const void *buf, 399 int64_t bytes); 400 int bdrv_pwrite_sync(BdrvChild *child, int64_t offset, 401 const void *buf, int64_t bytes); 402 /* 403 * Efficiently zero a region of the disk image. Note that this is a regular 404 * I/O request like read or write and should have a reasonable size. This 405 * function is not suitable for zeroing the entire image in a single request 406 * because it may allocate memory for the entire region. 407 */ 408 int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset, 409 int64_t bytes, BdrvRequestFlags flags); 410 BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs, 411 const char *backing_file); 412 void bdrv_refresh_filename(BlockDriverState *bs); 413 414 int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset, bool exact, 415 PreallocMode prealloc, BdrvRequestFlags flags, 416 Error **errp); 417 int generated_co_wrapper 418 bdrv_truncate(BdrvChild *child, int64_t offset, bool exact, 419 PreallocMode prealloc, BdrvRequestFlags flags, Error **errp); 420 421 int64_t bdrv_nb_sectors(BlockDriverState *bs); 422 int64_t bdrv_getlength(BlockDriverState *bs); 423 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs); 424 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts, 425 BlockDriverState *in_bs, Error **errp); 426 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr); 427 void bdrv_refresh_limits(BlockDriverState *bs, Error **errp); 428 int bdrv_commit(BlockDriverState *bs); 429 int bdrv_make_empty(BdrvChild *c, Error **errp); 430 int bdrv_change_backing_file(BlockDriverState *bs, const char *backing_file, 431 const char *backing_fmt, bool warn); 432 void bdrv_register(BlockDriver *bdrv); 433 int bdrv_drop_intermediate(BlockDriverState *top, BlockDriverState *base, 434 const char *backing_file_str); 435 BlockDriverState *bdrv_find_overlay(BlockDriverState *active, 436 BlockDriverState *bs); 437 BlockDriverState *bdrv_find_base(BlockDriverState *bs); 438 bool bdrv_is_backing_chain_frozen(BlockDriverState *bs, BlockDriverState *base, 439 Error **errp); 440 int bdrv_freeze_backing_chain(BlockDriverState *bs, BlockDriverState *base, 441 Error **errp); 442 void bdrv_unfreeze_backing_chain(BlockDriverState *bs, BlockDriverState *base); 443 int coroutine_fn bdrv_co_delete_file(BlockDriverState *bs, Error **errp); 444 void coroutine_fn bdrv_co_delete_file_noerr(BlockDriverState *bs); 445 446 447 typedef struct BdrvCheckResult { 448 int corruptions; 449 int leaks; 450 int check_errors; 451 int corruptions_fixed; 452 int leaks_fixed; 453 int64_t image_end_offset; 454 BlockFragInfo bfi; 455 } BdrvCheckResult; 456 457 typedef enum { 458 BDRV_FIX_LEAKS = 1, 459 BDRV_FIX_ERRORS = 2, 460 } BdrvCheckMode; 461 462 int generated_co_wrapper bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, 463 BdrvCheckMode fix); 464 465 /* The units of offset and total_work_size may be chosen arbitrarily by the 466 * block driver; total_work_size may change during the course of the amendment 467 * operation */ 468 typedef void BlockDriverAmendStatusCB(BlockDriverState *bs, int64_t offset, 469 int64_t total_work_size, void *opaque); 470 int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts, 471 BlockDriverAmendStatusCB *status_cb, void *cb_opaque, 472 bool force, 473 Error **errp); 474 475 /* check if a named node can be replaced when doing drive-mirror */ 476 BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs, 477 const char *node_name, Error **errp); 478 479 /* async block I/O */ 480 void bdrv_aio_cancel(BlockAIOCB *acb); 481 void bdrv_aio_cancel_async(BlockAIOCB *acb); 482 483 /* sg packet commands */ 484 int bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf); 485 486 /* Invalidate any cached metadata used by image formats */ 487 int generated_co_wrapper bdrv_invalidate_cache(BlockDriverState *bs, 488 Error **errp); 489 void bdrv_invalidate_cache_all(Error **errp); 490 int bdrv_inactivate_all(void); 491 492 /* Ensure contents are flushed to disk. */ 493 int generated_co_wrapper bdrv_flush(BlockDriverState *bs); 494 int coroutine_fn bdrv_co_flush(BlockDriverState *bs); 495 int bdrv_flush_all(void); 496 void bdrv_close_all(void); 497 void bdrv_drain(BlockDriverState *bs); 498 void coroutine_fn bdrv_co_drain(BlockDriverState *bs); 499 void bdrv_drain_all_begin(void); 500 void bdrv_drain_all_end(void); 501 void bdrv_drain_all(void); 502 503 #define BDRV_POLL_WHILE(bs, cond) ({ \ 504 BlockDriverState *bs_ = (bs); \ 505 AIO_WAIT_WHILE(bdrv_get_aio_context(bs_), \ 506 cond); }) 507 508 int generated_co_wrapper bdrv_pdiscard(BdrvChild *child, int64_t offset, 509 int64_t bytes); 510 int bdrv_co_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes); 511 int bdrv_has_zero_init_1(BlockDriverState *bs); 512 int bdrv_has_zero_init(BlockDriverState *bs); 513 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs); 514 int bdrv_block_status(BlockDriverState *bs, int64_t offset, 515 int64_t bytes, int64_t *pnum, int64_t *map, 516 BlockDriverState **file); 517 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base, 518 int64_t offset, int64_t bytes, int64_t *pnum, 519 int64_t *map, BlockDriverState **file); 520 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes, 521 int64_t *pnum); 522 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base, 523 bool include_base, int64_t offset, int64_t bytes, 524 int64_t *pnum); 525 int coroutine_fn bdrv_co_is_zero_fast(BlockDriverState *bs, int64_t offset, 526 int64_t bytes); 527 528 bool bdrv_is_read_only(BlockDriverState *bs); 529 int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only, 530 bool ignore_allow_rdw, Error **errp); 531 int bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg, 532 Error **errp); 533 bool bdrv_is_writable(BlockDriverState *bs); 534 bool bdrv_is_sg(BlockDriverState *bs); 535 bool bdrv_is_inserted(BlockDriverState *bs); 536 void bdrv_lock_medium(BlockDriverState *bs, bool locked); 537 void bdrv_eject(BlockDriverState *bs, bool eject_flag); 538 const char *bdrv_get_format_name(BlockDriverState *bs); 539 BlockDriverState *bdrv_find_node(const char *node_name); 540 BlockDeviceInfoList *bdrv_named_nodes_list(bool flat, Error **errp); 541 XDbgBlockGraph *bdrv_get_xdbg_block_graph(Error **errp); 542 BlockDriverState *bdrv_lookup_bs(const char *device, 543 const char *node_name, 544 Error **errp); 545 bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base); 546 BlockDriverState *bdrv_next_node(BlockDriverState *bs); 547 BlockDriverState *bdrv_next_all_states(BlockDriverState *bs); 548 549 typedef struct BdrvNextIterator { 550 enum { 551 BDRV_NEXT_BACKEND_ROOTS, 552 BDRV_NEXT_MONITOR_OWNED, 553 } phase; 554 BlockBackend *blk; 555 BlockDriverState *bs; 556 } BdrvNextIterator; 557 558 BlockDriverState *bdrv_first(BdrvNextIterator *it); 559 BlockDriverState *bdrv_next(BdrvNextIterator *it); 560 void bdrv_next_cleanup(BdrvNextIterator *it); 561 562 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs); 563 bool bdrv_supports_compressed_writes(BlockDriverState *bs); 564 void bdrv_iterate_format(void (*it)(void *opaque, const char *name), 565 void *opaque, bool read_only); 566 const char *bdrv_get_node_name(const BlockDriverState *bs); 567 const char *bdrv_get_device_name(const BlockDriverState *bs); 568 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs); 569 int bdrv_get_flags(BlockDriverState *bs); 570 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi); 571 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs, 572 Error **errp); 573 BlockStatsSpecific *bdrv_get_specific_stats(BlockDriverState *bs); 574 void bdrv_round_to_clusters(BlockDriverState *bs, 575 int64_t offset, int64_t bytes, 576 int64_t *cluster_offset, 577 int64_t *cluster_bytes); 578 579 void bdrv_get_backing_filename(BlockDriverState *bs, 580 char *filename, int filename_size); 581 char *bdrv_get_full_backing_filename(BlockDriverState *bs, Error **errp); 582 char *bdrv_get_full_backing_filename_from_filename(const char *backed, 583 const char *backing, 584 Error **errp); 585 char *bdrv_dirname(BlockDriverState *bs, Error **errp); 586 587 int path_has_protocol(const char *path); 588 int path_is_absolute(const char *path); 589 char *path_combine(const char *base_path, const char *filename); 590 591 int generated_co_wrapper 592 bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos); 593 int generated_co_wrapper 594 bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos); 595 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf, 596 int64_t pos, int size); 597 598 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf, 599 int64_t pos, int size); 600 601 void bdrv_img_create(const char *filename, const char *fmt, 602 const char *base_filename, const char *base_fmt, 603 char *options, uint64_t img_size, int flags, 604 bool quiet, Error **errp); 605 606 /* Returns the alignment in bytes that is required so that no bounce buffer 607 * is required throughout the stack */ 608 size_t bdrv_min_mem_align(BlockDriverState *bs); 609 /* Returns optimal alignment in bytes for bounce buffer */ 610 size_t bdrv_opt_mem_align(BlockDriverState *bs); 611 void *qemu_blockalign(BlockDriverState *bs, size_t size); 612 void *qemu_blockalign0(BlockDriverState *bs, size_t size); 613 void *qemu_try_blockalign(BlockDriverState *bs, size_t size); 614 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size); 615 bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov); 616 617 void bdrv_enable_copy_on_read(BlockDriverState *bs); 618 void bdrv_disable_copy_on_read(BlockDriverState *bs); 619 620 void bdrv_ref(BlockDriverState *bs); 621 void bdrv_unref(BlockDriverState *bs); 622 void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child); 623 BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs, 624 BlockDriverState *child_bs, 625 const char *child_name, 626 const BdrvChildClass *child_class, 627 BdrvChildRole child_role, 628 Error **errp); 629 630 bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp); 631 void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason); 632 void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason); 633 void bdrv_op_block_all(BlockDriverState *bs, Error *reason); 634 void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason); 635 bool bdrv_op_blocker_is_empty(BlockDriverState *bs); 636 637 #define BLKDBG_EVENT(child, evt) \ 638 do { \ 639 if (child) { \ 640 bdrv_debug_event(child->bs, evt); \ 641 } \ 642 } while (0) 643 644 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event); 645 646 int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event, 647 const char *tag); 648 int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag); 649 int bdrv_debug_resume(BlockDriverState *bs, const char *tag); 650 bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag); 651 652 /** 653 * bdrv_get_aio_context: 654 * 655 * Returns: the currently bound #AioContext 656 */ 657 AioContext *bdrv_get_aio_context(BlockDriverState *bs); 658 659 /** 660 * Move the current coroutine to the AioContext of @bs and return the old 661 * AioContext of the coroutine. Increase bs->in_flight so that draining @bs 662 * will wait for the operation to proceed until the corresponding 663 * bdrv_co_leave(). 664 * 665 * Consequently, you can't call drain inside a bdrv_co_enter/leave() section as 666 * this will deadlock. 667 */ 668 AioContext *coroutine_fn bdrv_co_enter(BlockDriverState *bs); 669 670 /** 671 * Ends a section started by bdrv_co_enter(). Move the current coroutine back 672 * to old_ctx and decrease bs->in_flight again. 673 */ 674 void coroutine_fn bdrv_co_leave(BlockDriverState *bs, AioContext *old_ctx); 675 676 /** 677 * Locks the AioContext of @bs if it's not the current AioContext. This avoids 678 * double locking which could lead to deadlocks: This is a coroutine_fn, so we 679 * know we already own the lock of the current AioContext. 680 * 681 * May only be called in the main thread. 682 */ 683 void coroutine_fn bdrv_co_lock(BlockDriverState *bs); 684 685 /** 686 * Unlocks the AioContext of @bs if it's not the current AioContext. 687 */ 688 void coroutine_fn bdrv_co_unlock(BlockDriverState *bs); 689 690 /** 691 * Transfer control to @co in the aio context of @bs 692 */ 693 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co); 694 695 void bdrv_set_aio_context_ignore(BlockDriverState *bs, 696 AioContext *new_context, GSList **ignore); 697 int bdrv_try_set_aio_context(BlockDriverState *bs, AioContext *ctx, 698 Error **errp); 699 int bdrv_child_try_set_aio_context(BlockDriverState *bs, AioContext *ctx, 700 BdrvChild *ignore_child, Error **errp); 701 bool bdrv_child_can_set_aio_context(BdrvChild *c, AioContext *ctx, 702 GSList **ignore, Error **errp); 703 bool bdrv_can_set_aio_context(BlockDriverState *bs, AioContext *ctx, 704 GSList **ignore, Error **errp); 705 int bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz); 706 int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo); 707 708 void bdrv_io_plug(BlockDriverState *bs); 709 void bdrv_io_unplug(BlockDriverState *bs); 710 711 /** 712 * bdrv_parent_drained_begin_single: 713 * 714 * Begin a quiesced section for the parent of @c. If @poll is true, wait for 715 * any pending activity to cease. 716 */ 717 void bdrv_parent_drained_begin_single(BdrvChild *c, bool poll); 718 719 /** 720 * bdrv_parent_drained_end_single: 721 * 722 * End a quiesced section for the parent of @c. 723 * 724 * This polls @bs's AioContext until all scheduled sub-drained_ends 725 * have settled, which may result in graph changes. 726 */ 727 void bdrv_parent_drained_end_single(BdrvChild *c); 728 729 /** 730 * bdrv_drain_poll: 731 * 732 * Poll for pending requests in @bs, its parents (except for @ignore_parent), 733 * and if @recursive is true its children as well (used for subtree drain). 734 * 735 * If @ignore_bds_parents is true, parents that are BlockDriverStates must 736 * ignore the drain request because they will be drained separately (used for 737 * drain_all). 738 * 739 * This is part of bdrv_drained_begin. 740 */ 741 bool bdrv_drain_poll(BlockDriverState *bs, bool recursive, 742 BdrvChild *ignore_parent, bool ignore_bds_parents); 743 744 /** 745 * bdrv_drained_begin: 746 * 747 * Begin a quiesced section for exclusive access to the BDS, by disabling 748 * external request sources including NBD server and device model. Note that 749 * this doesn't block timers or coroutines from submitting more requests, which 750 * means block_job_pause is still necessary. 751 * 752 * This function can be recursive. 753 */ 754 void bdrv_drained_begin(BlockDriverState *bs); 755 756 /** 757 * bdrv_do_drained_begin_quiesce: 758 * 759 * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already 760 * running requests to complete. 761 */ 762 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs, 763 BdrvChild *parent, bool ignore_bds_parents); 764 765 /** 766 * Like bdrv_drained_begin, but recursively begins a quiesced section for 767 * exclusive access to all child nodes as well. 768 */ 769 void bdrv_subtree_drained_begin(BlockDriverState *bs); 770 771 /** 772 * bdrv_drained_end: 773 * 774 * End a quiescent section started by bdrv_drained_begin(). 775 * 776 * This polls @bs's AioContext until all scheduled sub-drained_ends 777 * have settled. On one hand, that may result in graph changes. On 778 * the other, this requires that the caller either runs in the main 779 * loop; or that all involved nodes (@bs and all of its parents) are 780 * in the caller's AioContext. 781 */ 782 void bdrv_drained_end(BlockDriverState *bs); 783 784 /** 785 * bdrv_drained_end_no_poll: 786 * 787 * Same as bdrv_drained_end(), but do not poll for the subgraph to 788 * actually become unquiesced. Therefore, no graph changes will occur 789 * with this function. 790 * 791 * *drained_end_counter is incremented for every background operation 792 * that is scheduled, and will be decremented for every operation once 793 * it settles. The caller must poll until it reaches 0. The counter 794 * should be accessed using atomic operations only. 795 */ 796 void bdrv_drained_end_no_poll(BlockDriverState *bs, int *drained_end_counter); 797 798 /** 799 * End a quiescent section started by bdrv_subtree_drained_begin(). 800 */ 801 void bdrv_subtree_drained_end(BlockDriverState *bs); 802 803 void bdrv_add_child(BlockDriverState *parent, BlockDriverState *child, 804 Error **errp); 805 void bdrv_del_child(BlockDriverState *parent, BdrvChild *child, Error **errp); 806 807 bool bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name, 808 uint32_t granularity, Error **errp); 809 /** 810 * 811 * bdrv_register_buf/bdrv_unregister_buf: 812 * 813 * Register/unregister a buffer for I/O. For example, VFIO drivers are 814 * interested to know the memory areas that would later be used for I/O, so 815 * that they can prepare IOMMU mapping etc., to get better performance. 816 */ 817 void bdrv_register_buf(BlockDriverState *bs, void *host, size_t size); 818 void bdrv_unregister_buf(BlockDriverState *bs, void *host); 819 820 /** 821 * 822 * bdrv_co_copy_range: 823 * 824 * Do offloaded copy between two children. If the operation is not implemented 825 * by the driver, or if the backend storage doesn't support it, a negative 826 * error code will be returned. 827 * 828 * Note: block layer doesn't emulate or fallback to a bounce buffer approach 829 * because usually the caller shouldn't attempt offloaded copy any more (e.g. 830 * calling copy_file_range(2)) after the first error, thus it should fall back 831 * to a read+write path in the caller level. 832 * 833 * @src: Source child to copy data from 834 * @src_offset: offset in @src image to read data 835 * @dst: Destination child to copy data to 836 * @dst_offset: offset in @dst image to write data 837 * @bytes: number of bytes to copy 838 * @flags: request flags. Supported flags: 839 * BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero 840 * write on @dst as if bdrv_co_pwrite_zeroes is 841 * called. Used to simplify caller code, or 842 * during BlockDriver.bdrv_co_copy_range_from() 843 * recursion. 844 * BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping 845 * requests currently in flight. 846 * 847 * Returns: 0 if succeeded; negative error code if failed. 848 **/ 849 int coroutine_fn bdrv_co_copy_range(BdrvChild *src, int64_t src_offset, 850 BdrvChild *dst, int64_t dst_offset, 851 int64_t bytes, BdrvRequestFlags read_flags, 852 BdrvRequestFlags write_flags); 853 854 void bdrv_cancel_in_flight(BlockDriverState *bs); 855 856 #endif 857