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