1 /* 2 * QEMU System Emulator block driver 3 * 4 * Copyright (c) 2003 Fabrice Bellard 5 * 6 * Permission is hereby granted, free of charge, to any person obtaining a copy 7 * of this software and associated documentation files (the "Software"), to deal 8 * in the Software without restriction, including without limitation the rights 9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell 10 * copies of the Software, and to permit persons to whom the Software is 11 * furnished to do so, subject to the following conditions: 12 * 13 * The above copyright notice and this permission notice shall be included in 14 * all copies or substantial portions of the Software. 15 * 16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN 22 * THE SOFTWARE. 23 */ 24 #ifndef BLOCK_INT_H 25 #define BLOCK_INT_H 26 27 #include "block/accounting.h" 28 #include "block/block.h" 29 #include "block/aio-wait.h" 30 #include "qemu/queue.h" 31 #include "qemu/coroutine.h" 32 #include "qemu/stats64.h" 33 #include "qemu/timer.h" 34 #include "qemu/hbitmap.h" 35 #include "block/snapshot.h" 36 #include "qemu/main-loop.h" 37 #include "qemu/throttle.h" 38 39 #define BLOCK_FLAG_LAZY_REFCOUNTS 8 40 41 #define BLOCK_OPT_SIZE "size" 42 #define BLOCK_OPT_ENCRYPT "encryption" 43 #define BLOCK_OPT_ENCRYPT_FORMAT "encrypt.format" 44 #define BLOCK_OPT_COMPAT6 "compat6" 45 #define BLOCK_OPT_HWVERSION "hwversion" 46 #define BLOCK_OPT_BACKING_FILE "backing_file" 47 #define BLOCK_OPT_BACKING_FMT "backing_fmt" 48 #define BLOCK_OPT_CLUSTER_SIZE "cluster_size" 49 #define BLOCK_OPT_TABLE_SIZE "table_size" 50 #define BLOCK_OPT_PREALLOC "preallocation" 51 #define BLOCK_OPT_SUBFMT "subformat" 52 #define BLOCK_OPT_COMPAT_LEVEL "compat" 53 #define BLOCK_OPT_LAZY_REFCOUNTS "lazy_refcounts" 54 #define BLOCK_OPT_ADAPTER_TYPE "adapter_type" 55 #define BLOCK_OPT_REDUNDANCY "redundancy" 56 #define BLOCK_OPT_NOCOW "nocow" 57 #define BLOCK_OPT_OBJECT_SIZE "object_size" 58 #define BLOCK_OPT_REFCOUNT_BITS "refcount_bits" 59 60 #define BLOCK_PROBE_BUF_SIZE 512 61 62 enum BdrvTrackedRequestType { 63 BDRV_TRACKED_READ, 64 BDRV_TRACKED_WRITE, 65 BDRV_TRACKED_DISCARD, 66 BDRV_TRACKED_TRUNCATE, 67 }; 68 69 typedef struct BdrvTrackedRequest { 70 BlockDriverState *bs; 71 int64_t offset; 72 uint64_t bytes; 73 enum BdrvTrackedRequestType type; 74 75 bool serialising; 76 int64_t overlap_offset; 77 uint64_t overlap_bytes; 78 79 QLIST_ENTRY(BdrvTrackedRequest) list; 80 Coroutine *co; /* owner, used for deadlock detection */ 81 CoQueue wait_queue; /* coroutines blocked on this request */ 82 83 struct BdrvTrackedRequest *waiting_for; 84 } BdrvTrackedRequest; 85 86 struct BlockDriver { 87 const char *format_name; 88 int instance_size; 89 90 /* set to true if the BlockDriver is a block filter. Block filters pass 91 * certain callbacks that refer to data (see block.c) to their bs->file if 92 * the driver doesn't implement them. Drivers that do not wish to forward 93 * must implement them and return -ENOTSUP. 94 */ 95 bool is_filter; 96 /* for snapshots block filter like Quorum can implement the 97 * following recursive callback. 98 * It's purpose is to recurse on the filter children while calling 99 * bdrv_recurse_is_first_non_filter on them. 100 * For a sample implementation look in the future Quorum block filter. 101 */ 102 bool (*bdrv_recurse_is_first_non_filter)(BlockDriverState *bs, 103 BlockDriverState *candidate); 104 105 int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename); 106 int (*bdrv_probe_device)(const char *filename); 107 108 /* Any driver implementing this callback is expected to be able to handle 109 * NULL file names in its .bdrv_open() implementation */ 110 void (*bdrv_parse_filename)(const char *filename, QDict *options, Error **errp); 111 /* Drivers not implementing bdrv_parse_filename nor bdrv_open should have 112 * this field set to true, except ones that are defined only by their 113 * child's bs. 114 * An example of the last type will be the quorum block driver. 115 */ 116 bool bdrv_needs_filename; 117 118 /* Set if a driver can support backing files */ 119 bool supports_backing; 120 121 /* For handling image reopen for split or non-split files */ 122 int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state, 123 BlockReopenQueue *queue, Error **errp); 124 void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state); 125 void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state); 126 void (*bdrv_join_options)(QDict *options, QDict *old_options); 127 128 int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags, 129 Error **errp); 130 131 /* Protocol drivers should implement this instead of bdrv_open */ 132 int (*bdrv_file_open)(BlockDriverState *bs, QDict *options, int flags, 133 Error **errp); 134 void (*bdrv_close)(BlockDriverState *bs); 135 int coroutine_fn (*bdrv_co_create)(BlockdevCreateOptions *opts, 136 Error **errp); 137 int coroutine_fn (*bdrv_co_create_opts)(const char *filename, 138 QemuOpts *opts, 139 Error **errp); 140 int (*bdrv_make_empty)(BlockDriverState *bs); 141 142 /* 143 * Refreshes the bs->exact_filename field. If that is impossible, 144 * bs->exact_filename has to be left empty. 145 */ 146 void (*bdrv_refresh_filename)(BlockDriverState *bs); 147 148 /* 149 * Gathers the open options for all children into @target. 150 * A simple format driver (without backing file support) might 151 * implement this function like this: 152 * 153 * QINCREF(bs->file->bs->full_open_options); 154 * qdict_put(target, "file", bs->file->bs->full_open_options); 155 * 156 * If not specified, the generic implementation will simply put 157 * all children's options under their respective name. 158 * 159 * @backing_overridden is true when bs->backing seems not to be 160 * the child that would result from opening bs->backing_file. 161 * Therefore, if it is true, the backing child's options should be 162 * gathered; otherwise, there is no need since the backing child 163 * is the one implied by the image header. 164 * 165 * Note that ideally this function would not be needed. Every 166 * block driver which implements it is probably doing something 167 * shady regarding its runtime option structure. 168 */ 169 void (*bdrv_gather_child_options)(BlockDriverState *bs, QDict *target, 170 bool backing_overridden); 171 172 /* 173 * Returns an allocated string which is the directory name of this BDS: It 174 * will be used to make relative filenames absolute by prepending this 175 * function's return value to them. 176 */ 177 char *(*bdrv_dirname)(BlockDriverState *bs, Error **errp); 178 179 /* aio */ 180 BlockAIOCB *(*bdrv_aio_preadv)(BlockDriverState *bs, 181 uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags, 182 BlockCompletionFunc *cb, void *opaque); 183 BlockAIOCB *(*bdrv_aio_pwritev)(BlockDriverState *bs, 184 uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags, 185 BlockCompletionFunc *cb, void *opaque); 186 BlockAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs, 187 BlockCompletionFunc *cb, void *opaque); 188 BlockAIOCB *(*bdrv_aio_pdiscard)(BlockDriverState *bs, 189 int64_t offset, int bytes, 190 BlockCompletionFunc *cb, void *opaque); 191 192 int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs, 193 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov); 194 195 /** 196 * @offset: position in bytes to read at 197 * @bytes: number of bytes to read 198 * @qiov: the buffers to fill with read data 199 * @flags: currently unused, always 0 200 * 201 * @offset and @bytes will be a multiple of 'request_alignment', 202 * but the length of individual @qiov elements does not have to 203 * be a multiple. 204 * 205 * @bytes will always equal the total size of @qiov, and will be 206 * no larger than 'max_transfer'. 207 * 208 * The buffer in @qiov may point directly to guest memory. 209 */ 210 int coroutine_fn (*bdrv_co_preadv)(BlockDriverState *bs, 211 uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags); 212 int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs, 213 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, int flags); 214 /** 215 * @offset: position in bytes to write at 216 * @bytes: number of bytes to write 217 * @qiov: the buffers containing data to write 218 * @flags: zero or more bits allowed by 'supported_write_flags' 219 * 220 * @offset and @bytes will be a multiple of 'request_alignment', 221 * but the length of individual @qiov elements does not have to 222 * be a multiple. 223 * 224 * @bytes will always equal the total size of @qiov, and will be 225 * no larger than 'max_transfer'. 226 * 227 * The buffer in @qiov may point directly to guest memory. 228 */ 229 int coroutine_fn (*bdrv_co_pwritev)(BlockDriverState *bs, 230 uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags); 231 232 /* 233 * Efficiently zero a region of the disk image. Typically an image format 234 * would use a compact metadata representation to implement this. This 235 * function pointer may be NULL or return -ENOSUP and .bdrv_co_writev() 236 * will be called instead. 237 */ 238 int coroutine_fn (*bdrv_co_pwrite_zeroes)(BlockDriverState *bs, 239 int64_t offset, int bytes, BdrvRequestFlags flags); 240 int coroutine_fn (*bdrv_co_pdiscard)(BlockDriverState *bs, 241 int64_t offset, int bytes); 242 243 /* Map [offset, offset + nbytes) range onto a child of @bs to copy from, 244 * and invoke bdrv_co_copy_range_from(child, ...), or invoke 245 * bdrv_co_copy_range_to() if @bs is the leaf child to copy data from. 246 * 247 * See the comment of bdrv_co_copy_range for the parameter and return value 248 * semantics. 249 */ 250 int coroutine_fn (*bdrv_co_copy_range_from)(BlockDriverState *bs, 251 BdrvChild *src, 252 uint64_t offset, 253 BdrvChild *dst, 254 uint64_t dst_offset, 255 uint64_t bytes, 256 BdrvRequestFlags read_flags, 257 BdrvRequestFlags write_flags); 258 259 /* Map [offset, offset + nbytes) range onto a child of bs to copy data to, 260 * and invoke bdrv_co_copy_range_to(child, src, ...), or perform the copy 261 * operation if @bs is the leaf and @src has the same BlockDriver. Return 262 * -ENOTSUP if @bs is the leaf but @src has a different BlockDriver. 263 * 264 * See the comment of bdrv_co_copy_range for the parameter and return value 265 * semantics. 266 */ 267 int coroutine_fn (*bdrv_co_copy_range_to)(BlockDriverState *bs, 268 BdrvChild *src, 269 uint64_t src_offset, 270 BdrvChild *dst, 271 uint64_t dst_offset, 272 uint64_t bytes, 273 BdrvRequestFlags read_flags, 274 BdrvRequestFlags write_flags); 275 276 /* 277 * Building block for bdrv_block_status[_above] and 278 * bdrv_is_allocated[_above]. The driver should answer only 279 * according to the current layer, and should only need to set 280 * BDRV_BLOCK_DATA, BDRV_BLOCK_ZERO, BDRV_BLOCK_OFFSET_VALID, 281 * and/or BDRV_BLOCK_RAW; if the current layer defers to a backing 282 * layer, the result should be 0 (and not BDRV_BLOCK_ZERO). See 283 * block.h for the overall meaning of the bits. As a hint, the 284 * flag want_zero is true if the caller cares more about precise 285 * mappings (favor accurate _OFFSET_VALID/_ZERO) or false for 286 * overall allocation (favor larger *pnum, perhaps by reporting 287 * _DATA instead of _ZERO). The block layer guarantees input 288 * clamped to bdrv_getlength() and aligned to request_alignment, 289 * as well as non-NULL pnum, map, and file; in turn, the driver 290 * must return an error or set pnum to an aligned non-zero value. 291 */ 292 int coroutine_fn (*bdrv_co_block_status)(BlockDriverState *bs, 293 bool want_zero, int64_t offset, int64_t bytes, int64_t *pnum, 294 int64_t *map, BlockDriverState **file); 295 296 /* 297 * Invalidate any cached meta-data. 298 */ 299 void coroutine_fn (*bdrv_co_invalidate_cache)(BlockDriverState *bs, 300 Error **errp); 301 int (*bdrv_inactivate)(BlockDriverState *bs); 302 303 /* 304 * Flushes all data for all layers by calling bdrv_co_flush for underlying 305 * layers, if needed. This function is needed for deterministic 306 * synchronization of the flush finishing callback. 307 */ 308 int coroutine_fn (*bdrv_co_flush)(BlockDriverState *bs); 309 310 /* 311 * Flushes all data that was already written to the OS all the way down to 312 * the disk (for example file-posix.c calls fsync()). 313 */ 314 int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs); 315 316 /* 317 * Flushes all internal caches to the OS. The data may still sit in a 318 * writeback cache of the host OS, but it will survive a crash of the qemu 319 * process. 320 */ 321 int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs); 322 323 /* 324 * Drivers setting this field must be able to work with just a plain 325 * filename with '<protocol_name>:' as a prefix, and no other options. 326 * Options may be extracted from the filename by implementing 327 * bdrv_parse_filename. 328 */ 329 const char *protocol_name; 330 int coroutine_fn (*bdrv_co_truncate)(BlockDriverState *bs, int64_t offset, 331 PreallocMode prealloc, Error **errp); 332 333 int64_t (*bdrv_getlength)(BlockDriverState *bs); 334 bool has_variable_length; 335 int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs); 336 BlockMeasureInfo *(*bdrv_measure)(QemuOpts *opts, BlockDriverState *in_bs, 337 Error **errp); 338 339 int coroutine_fn (*bdrv_co_pwritev_compressed)(BlockDriverState *bs, 340 uint64_t offset, uint64_t bytes, QEMUIOVector *qiov); 341 342 int (*bdrv_snapshot_create)(BlockDriverState *bs, 343 QEMUSnapshotInfo *sn_info); 344 int (*bdrv_snapshot_goto)(BlockDriverState *bs, 345 const char *snapshot_id); 346 int (*bdrv_snapshot_delete)(BlockDriverState *bs, 347 const char *snapshot_id, 348 const char *name, 349 Error **errp); 350 int (*bdrv_snapshot_list)(BlockDriverState *bs, 351 QEMUSnapshotInfo **psn_info); 352 int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs, 353 const char *snapshot_id, 354 const char *name, 355 Error **errp); 356 int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi); 357 ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs, 358 Error **errp); 359 360 int coroutine_fn (*bdrv_save_vmstate)(BlockDriverState *bs, 361 QEMUIOVector *qiov, 362 int64_t pos); 363 int coroutine_fn (*bdrv_load_vmstate)(BlockDriverState *bs, 364 QEMUIOVector *qiov, 365 int64_t pos); 366 367 int (*bdrv_change_backing_file)(BlockDriverState *bs, 368 const char *backing_file, const char *backing_fmt); 369 370 /* removable device specific */ 371 bool (*bdrv_is_inserted)(BlockDriverState *bs); 372 void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag); 373 void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked); 374 375 /* to control generic scsi devices */ 376 BlockAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs, 377 unsigned long int req, void *buf, 378 BlockCompletionFunc *cb, void *opaque); 379 int coroutine_fn (*bdrv_co_ioctl)(BlockDriverState *bs, 380 unsigned long int req, void *buf); 381 382 /* List of options for creating images, terminated by name == NULL */ 383 QemuOptsList *create_opts; 384 385 /* 386 * Returns 0 for completed check, -errno for internal errors. 387 * The check results are stored in result. 388 */ 389 int coroutine_fn (*bdrv_co_check)(BlockDriverState *bs, 390 BdrvCheckResult *result, 391 BdrvCheckMode fix); 392 393 int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts, 394 BlockDriverAmendStatusCB *status_cb, 395 void *cb_opaque, 396 Error **errp); 397 398 void (*bdrv_debug_event)(BlockDriverState *bs, BlkdebugEvent event); 399 400 /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */ 401 int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event, 402 const char *tag); 403 int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs, 404 const char *tag); 405 int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag); 406 bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag); 407 408 void (*bdrv_refresh_limits)(BlockDriverState *bs, Error **errp); 409 410 /* 411 * Returns 1 if newly created images are guaranteed to contain only 412 * zeros, 0 otherwise. 413 */ 414 int (*bdrv_has_zero_init)(BlockDriverState *bs); 415 416 /* Remove fd handlers, timers, and other event loop callbacks so the event 417 * loop is no longer in use. Called with no in-flight requests and in 418 * depth-first traversal order with parents before child nodes. 419 */ 420 void (*bdrv_detach_aio_context)(BlockDriverState *bs); 421 422 /* Add fd handlers, timers, and other event loop callbacks so I/O requests 423 * can be processed again. Called with no in-flight requests and in 424 * depth-first traversal order with child nodes before parent nodes. 425 */ 426 void (*bdrv_attach_aio_context)(BlockDriverState *bs, 427 AioContext *new_context); 428 429 /* io queue for linux-aio */ 430 void (*bdrv_io_plug)(BlockDriverState *bs); 431 void (*bdrv_io_unplug)(BlockDriverState *bs); 432 433 /** 434 * Try to get @bs's logical and physical block size. 435 * On success, store them in @bsz and return zero. 436 * On failure, return negative errno. 437 */ 438 int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz); 439 /** 440 * Try to get @bs's geometry (cyls, heads, sectors) 441 * On success, store them in @geo and return 0. 442 * On failure return -errno. 443 * Only drivers that want to override guest geometry implement this 444 * callback; see hd_geometry_guess(). 445 */ 446 int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo); 447 448 /** 449 * bdrv_co_drain_begin is called if implemented in the beginning of a 450 * drain operation to drain and stop any internal sources of requests in 451 * the driver. 452 * bdrv_co_drain_end is called if implemented at the end of the drain. 453 * 454 * They should be used by the driver to e.g. manage scheduled I/O 455 * requests, or toggle an internal state. After the end of the drain new 456 * requests will continue normally. 457 */ 458 void coroutine_fn (*bdrv_co_drain_begin)(BlockDriverState *bs); 459 void coroutine_fn (*bdrv_co_drain_end)(BlockDriverState *bs); 460 461 void (*bdrv_add_child)(BlockDriverState *parent, BlockDriverState *child, 462 Error **errp); 463 void (*bdrv_del_child)(BlockDriverState *parent, BdrvChild *child, 464 Error **errp); 465 466 /** 467 * Informs the block driver that a permission change is intended. The 468 * driver checks whether the change is permissible and may take other 469 * preparations for the change (e.g. get file system locks). This operation 470 * is always followed either by a call to either .bdrv_set_perm or 471 * .bdrv_abort_perm_update. 472 * 473 * Checks whether the requested set of cumulative permissions in @perm 474 * can be granted for accessing @bs and whether no other users are using 475 * permissions other than those given in @shared (both arguments take 476 * BLK_PERM_* bitmasks). 477 * 478 * If both conditions are met, 0 is returned. Otherwise, -errno is returned 479 * and errp is set to an error describing the conflict. 480 */ 481 int (*bdrv_check_perm)(BlockDriverState *bs, uint64_t perm, 482 uint64_t shared, Error **errp); 483 484 /** 485 * Called to inform the driver that the set of cumulative set of used 486 * permissions for @bs has changed to @perm, and the set of sharable 487 * permission to @shared. The driver can use this to propagate changes to 488 * its children (i.e. request permissions only if a parent actually needs 489 * them). 490 * 491 * This function is only invoked after bdrv_check_perm(), so block drivers 492 * may rely on preparations made in their .bdrv_check_perm implementation. 493 */ 494 void (*bdrv_set_perm)(BlockDriverState *bs, uint64_t perm, uint64_t shared); 495 496 /* 497 * Called to inform the driver that after a previous bdrv_check_perm() 498 * call, the permission update is not performed and any preparations made 499 * for it (e.g. taken file locks) need to be undone. 500 * 501 * This function can be called even for nodes that never saw a 502 * bdrv_check_perm() call. It is a no-op then. 503 */ 504 void (*bdrv_abort_perm_update)(BlockDriverState *bs); 505 506 /** 507 * Returns in @nperm and @nshared the permissions that the driver for @bs 508 * needs on its child @c, based on the cumulative permissions requested by 509 * the parents in @parent_perm and @parent_shared. 510 * 511 * If @c is NULL, return the permissions for attaching a new child for the 512 * given @role. 513 * 514 * If @reopen_queue is non-NULL, don't return the currently needed 515 * permissions, but those that will be needed after applying the 516 * @reopen_queue. 517 */ 518 void (*bdrv_child_perm)(BlockDriverState *bs, BdrvChild *c, 519 const BdrvChildRole *role, 520 BlockReopenQueue *reopen_queue, 521 uint64_t parent_perm, uint64_t parent_shared, 522 uint64_t *nperm, uint64_t *nshared); 523 524 /** 525 * Bitmaps should be marked as 'IN_USE' in the image on reopening image 526 * as rw. This handler should realize it. It also should unset readonly 527 * field of BlockDirtyBitmap's in case of success. 528 */ 529 int (*bdrv_reopen_bitmaps_rw)(BlockDriverState *bs, Error **errp); 530 bool (*bdrv_can_store_new_dirty_bitmap)(BlockDriverState *bs, 531 const char *name, 532 uint32_t granularity, 533 Error **errp); 534 void (*bdrv_remove_persistent_dirty_bitmap)(BlockDriverState *bs, 535 const char *name, 536 Error **errp); 537 538 /** 539 * Register/unregister a buffer for I/O. For example, when the driver is 540 * interested to know the memory areas that will later be used in iovs, so 541 * that it can do IOMMU mapping with VFIO etc., in order to get better 542 * performance. In the case of VFIO drivers, this callback is used to do 543 * DMA mapping for hot buffers. 544 */ 545 void (*bdrv_register_buf)(BlockDriverState *bs, void *host, size_t size); 546 void (*bdrv_unregister_buf)(BlockDriverState *bs, void *host); 547 QLIST_ENTRY(BlockDriver) list; 548 549 /* Pointer to a NULL-terminated array of names of strong options 550 * that can be specified for bdrv_open(). A strong option is one 551 * that changes the data of a BDS. 552 * If this pointer is NULL, the array is considered empty. 553 * "filename" and "driver" are always considered strong. */ 554 const char *const *strong_runtime_opts; 555 }; 556 557 typedef struct BlockLimits { 558 /* Alignment requirement, in bytes, for offset/length of I/O 559 * requests. Must be a power of 2 less than INT_MAX; defaults to 560 * 1 for drivers with modern byte interfaces, and to 512 561 * otherwise. */ 562 uint32_t request_alignment; 563 564 /* Maximum number of bytes that can be discarded at once (since it 565 * is signed, it must be < 2G, if set). Must be multiple of 566 * pdiscard_alignment, but need not be power of 2. May be 0 if no 567 * inherent 32-bit limit */ 568 int32_t max_pdiscard; 569 570 /* Optimal alignment for discard requests in bytes. A power of 2 571 * is best but not mandatory. Must be a multiple of 572 * bl.request_alignment, and must be less than max_pdiscard if 573 * that is set. May be 0 if bl.request_alignment is good enough */ 574 uint32_t pdiscard_alignment; 575 576 /* Maximum number of bytes that can zeroized at once (since it is 577 * signed, it must be < 2G, if set). Must be multiple of 578 * pwrite_zeroes_alignment. May be 0 if no inherent 32-bit limit */ 579 int32_t max_pwrite_zeroes; 580 581 /* Optimal alignment for write zeroes requests in bytes. A power 582 * of 2 is best but not mandatory. Must be a multiple of 583 * bl.request_alignment, and must be less than max_pwrite_zeroes 584 * if that is set. May be 0 if bl.request_alignment is good 585 * enough */ 586 uint32_t pwrite_zeroes_alignment; 587 588 /* Optimal transfer length in bytes. A power of 2 is best but not 589 * mandatory. Must be a multiple of bl.request_alignment, or 0 if 590 * no preferred size */ 591 uint32_t opt_transfer; 592 593 /* Maximal transfer length in bytes. Need not be power of 2, but 594 * must be multiple of opt_transfer and bl.request_alignment, or 0 595 * for no 32-bit limit. For now, anything larger than INT_MAX is 596 * clamped down. */ 597 uint32_t max_transfer; 598 599 /* memory alignment, in bytes so that no bounce buffer is needed */ 600 size_t min_mem_alignment; 601 602 /* memory alignment, in bytes, for bounce buffer */ 603 size_t opt_mem_alignment; 604 605 /* maximum number of iovec elements */ 606 int max_iov; 607 } BlockLimits; 608 609 typedef struct BdrvOpBlocker BdrvOpBlocker; 610 611 typedef struct BdrvAioNotifier { 612 void (*attached_aio_context)(AioContext *new_context, void *opaque); 613 void (*detach_aio_context)(void *opaque); 614 615 void *opaque; 616 bool deleted; 617 618 QLIST_ENTRY(BdrvAioNotifier) list; 619 } BdrvAioNotifier; 620 621 struct BdrvChildRole { 622 /* If true, bdrv_replace_node() doesn't change the node this BdrvChild 623 * points to. */ 624 bool stay_at_node; 625 626 /* If true, the parent is a BlockDriverState and bdrv_next_all_states() 627 * will return it. This information is used for drain_all, where every node 628 * will be drained separately, so the drain only needs to be propagated to 629 * non-BDS parents. */ 630 bool parent_is_bds; 631 632 void (*inherit_options)(int *child_flags, QDict *child_options, 633 int parent_flags, QDict *parent_options); 634 635 void (*change_media)(BdrvChild *child, bool load); 636 void (*resize)(BdrvChild *child); 637 638 /* Returns a name that is supposedly more useful for human users than the 639 * node name for identifying the node in question (in particular, a BB 640 * name), or NULL if the parent can't provide a better name. */ 641 const char *(*get_name)(BdrvChild *child); 642 643 /* Returns a malloced string that describes the parent of the child for a 644 * human reader. This could be a node-name, BlockBackend name, qdev ID or 645 * QOM path of the device owning the BlockBackend, job type and ID etc. The 646 * caller is responsible for freeing the memory. */ 647 char *(*get_parent_desc)(BdrvChild *child); 648 649 /* 650 * If this pair of functions is implemented, the parent doesn't issue new 651 * requests after returning from .drained_begin() until .drained_end() is 652 * called. 653 * 654 * These functions must not change the graph (and therefore also must not 655 * call aio_poll(), which could change the graph indirectly). 656 * 657 * Note that this can be nested. If drained_begin() was called twice, new 658 * I/O is allowed only after drained_end() was called twice, too. 659 */ 660 void (*drained_begin)(BdrvChild *child); 661 void (*drained_end)(BdrvChild *child); 662 663 /* 664 * Returns whether the parent has pending requests for the child. This 665 * callback is polled after .drained_begin() has been called until all 666 * activity on the child has stopped. 667 */ 668 bool (*drained_poll)(BdrvChild *child); 669 670 /* Notifies the parent that the child has been activated/inactivated (e.g. 671 * when migration is completing) and it can start/stop requesting 672 * permissions and doing I/O on it. */ 673 void (*activate)(BdrvChild *child, Error **errp); 674 int (*inactivate)(BdrvChild *child); 675 676 void (*attach)(BdrvChild *child); 677 void (*detach)(BdrvChild *child); 678 679 /* Notifies the parent that the filename of its child has changed (e.g. 680 * because the direct child was removed from the backing chain), so that it 681 * can update its reference. */ 682 int (*update_filename)(BdrvChild *child, BlockDriverState *new_base, 683 const char *filename, Error **errp); 684 }; 685 686 extern const BdrvChildRole child_file; 687 extern const BdrvChildRole child_format; 688 extern const BdrvChildRole child_backing; 689 690 struct BdrvChild { 691 BlockDriverState *bs; 692 char *name; 693 const BdrvChildRole *role; 694 void *opaque; 695 696 /** 697 * Granted permissions for operating on this BdrvChild (BLK_PERM_* bitmask) 698 */ 699 uint64_t perm; 700 701 /** 702 * Permissions that can still be granted to other users of @bs while this 703 * BdrvChild is still attached to it. (BLK_PERM_* bitmask) 704 */ 705 uint64_t shared_perm; 706 707 /* backup of permissions during permission update procedure */ 708 bool has_backup_perm; 709 uint64_t backup_perm; 710 uint64_t backup_shared_perm; 711 712 QLIST_ENTRY(BdrvChild) next; 713 QLIST_ENTRY(BdrvChild) next_parent; 714 }; 715 716 /* 717 * Note: the function bdrv_append() copies and swaps contents of 718 * BlockDriverStates, so if you add new fields to this struct, please 719 * inspect bdrv_append() to determine if the new fields need to be 720 * copied as well. 721 */ 722 struct BlockDriverState { 723 /* Protected by big QEMU lock or read-only after opening. No special 724 * locking needed during I/O... 725 */ 726 int open_flags; /* flags used to open the file, re-used for re-open */ 727 bool read_only; /* if true, the media is read only */ 728 bool encrypted; /* if true, the media is encrypted */ 729 bool sg; /* if true, the device is a /dev/sg* */ 730 bool probed; /* if true, format was probed rather than specified */ 731 bool force_share; /* if true, always allow all shared permissions */ 732 bool implicit; /* if true, this filter node was automatically inserted */ 733 734 BlockDriver *drv; /* NULL means no media */ 735 void *opaque; 736 737 AioContext *aio_context; /* event loop used for fd handlers, timers, etc */ 738 /* long-running tasks intended to always use the same AioContext as this 739 * BDS may register themselves in this list to be notified of changes 740 * regarding this BDS's context */ 741 QLIST_HEAD(, BdrvAioNotifier) aio_notifiers; 742 bool walking_aio_notifiers; /* to make removal during iteration safe */ 743 744 char filename[PATH_MAX]; 745 char backing_file[PATH_MAX]; /* if non zero, the image is a diff of 746 this file image */ 747 /* The backing filename indicated by the image header; if we ever 748 * open this file, then this is replaced by the resulting BDS's 749 * filename (i.e. after a bdrv_refresh_filename() run). */ 750 char auto_backing_file[PATH_MAX]; 751 char backing_format[16]; /* if non-zero and backing_file exists */ 752 753 QDict *full_open_options; 754 char exact_filename[PATH_MAX]; 755 756 BdrvChild *backing; 757 BdrvChild *file; 758 759 /* I/O Limits */ 760 BlockLimits bl; 761 762 /* Flags honored during pwrite (so far: BDRV_REQ_FUA, 763 * BDRV_REQ_WRITE_UNCHANGED). 764 * If a driver does not support BDRV_REQ_WRITE_UNCHANGED, those 765 * writes will be issued as normal writes without the flag set. 766 * This is important to note for drivers that do not explicitly 767 * request a WRITE permission for their children and instead take 768 * the same permissions as their parent did (this is commonly what 769 * block filters do). Such drivers have to be aware that the 770 * parent may have taken a WRITE_UNCHANGED permission only and is 771 * issuing such requests. Drivers either must make sure that 772 * these requests do not result in plain WRITE accesses (usually 773 * by supporting BDRV_REQ_WRITE_UNCHANGED, and then forwarding 774 * every incoming write request as-is, including potentially that 775 * flag), or they have to explicitly take the WRITE permission for 776 * their children. */ 777 unsigned int supported_write_flags; 778 /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA, 779 * BDRV_REQ_MAY_UNMAP, BDRV_REQ_WRITE_UNCHANGED) */ 780 unsigned int supported_zero_flags; 781 782 /* the following member gives a name to every node on the bs graph. */ 783 char node_name[32]; 784 /* element of the list of named nodes building the graph */ 785 QTAILQ_ENTRY(BlockDriverState) node_list; 786 /* element of the list of all BlockDriverStates (all_bdrv_states) */ 787 QTAILQ_ENTRY(BlockDriverState) bs_list; 788 /* element of the list of monitor-owned BDS */ 789 QTAILQ_ENTRY(BlockDriverState) monitor_list; 790 int refcnt; 791 792 /* operation blockers */ 793 QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX]; 794 795 /* long-running background operation */ 796 BlockJob *job; 797 798 /* The node that this node inherited default options from (and a reopen on 799 * which can affect this node by changing these defaults). This is always a 800 * parent node of this node. */ 801 BlockDriverState *inherits_from; 802 QLIST_HEAD(, BdrvChild) children; 803 QLIST_HEAD(, BdrvChild) parents; 804 805 QDict *options; 806 QDict *explicit_options; 807 BlockdevDetectZeroesOptions detect_zeroes; 808 809 /* The error object in use for blocking operations on backing_hd */ 810 Error *backing_blocker; 811 812 /* Protected by AioContext lock */ 813 814 /* If we are reading a disk image, give its size in sectors. 815 * Generally read-only; it is written to by load_snapshot and 816 * save_snaphost, but the block layer is quiescent during those. 817 */ 818 int64_t total_sectors; 819 820 /* Callback before write request is processed */ 821 NotifierWithReturnList before_write_notifiers; 822 823 /* threshold limit for writes, in bytes. "High water mark". */ 824 uint64_t write_threshold_offset; 825 NotifierWithReturn write_threshold_notifier; 826 827 /* Writing to the list requires the BQL _and_ the dirty_bitmap_mutex. 828 * Reading from the list can be done with either the BQL or the 829 * dirty_bitmap_mutex. Modifying a bitmap only requires 830 * dirty_bitmap_mutex. */ 831 QemuMutex dirty_bitmap_mutex; 832 QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps; 833 834 /* Offset after the highest byte written to */ 835 Stat64 wr_highest_offset; 836 837 /* If true, copy read backing sectors into image. Can be >1 if more 838 * than one client has requested copy-on-read. Accessed with atomic 839 * ops. 840 */ 841 int copy_on_read; 842 843 /* number of in-flight requests; overall and serialising. 844 * Accessed with atomic ops. 845 */ 846 unsigned int in_flight; 847 unsigned int serialising_in_flight; 848 849 /* counter for nested bdrv_io_plug. 850 * Accessed with atomic ops. 851 */ 852 unsigned io_plugged; 853 854 /* do we need to tell the quest if we have a volatile write cache? */ 855 int enable_write_cache; 856 857 /* Accessed with atomic ops. */ 858 int quiesce_counter; 859 int recursive_quiesce_counter; 860 861 unsigned int write_gen; /* Current data generation */ 862 863 /* Protected by reqs_lock. */ 864 CoMutex reqs_lock; 865 QLIST_HEAD(, BdrvTrackedRequest) tracked_requests; 866 CoQueue flush_queue; /* Serializing flush queue */ 867 bool active_flush_req; /* Flush request in flight? */ 868 869 /* Only read/written by whoever has set active_flush_req to true. */ 870 unsigned int flushed_gen; /* Flushed write generation */ 871 }; 872 873 struct BlockBackendRootState { 874 int open_flags; 875 bool read_only; 876 BlockdevDetectZeroesOptions detect_zeroes; 877 }; 878 879 typedef enum BlockMirrorBackingMode { 880 /* Reuse the existing backing chain from the source for the target. 881 * - sync=full: Set backing BDS to NULL. 882 * - sync=top: Use source's backing BDS. 883 * - sync=none: Use source as the backing BDS. */ 884 MIRROR_SOURCE_BACKING_CHAIN, 885 886 /* Open the target's backing chain completely anew */ 887 MIRROR_OPEN_BACKING_CHAIN, 888 889 /* Do not change the target's backing BDS after job completion */ 890 MIRROR_LEAVE_BACKING_CHAIN, 891 } BlockMirrorBackingMode; 892 893 static inline BlockDriverState *backing_bs(BlockDriverState *bs) 894 { 895 return bs->backing ? bs->backing->bs : NULL; 896 } 897 898 899 /* Essential block drivers which must always be statically linked into qemu, and 900 * which therefore can be accessed without using bdrv_find_format() */ 901 extern BlockDriver bdrv_file; 902 extern BlockDriver bdrv_raw; 903 extern BlockDriver bdrv_qcow2; 904 905 int coroutine_fn bdrv_co_preadv(BdrvChild *child, 906 int64_t offset, unsigned int bytes, QEMUIOVector *qiov, 907 BdrvRequestFlags flags); 908 int coroutine_fn bdrv_co_pwritev(BdrvChild *child, 909 int64_t offset, unsigned int bytes, QEMUIOVector *qiov, 910 BdrvRequestFlags flags); 911 912 extern unsigned int bdrv_drain_all_count; 913 void bdrv_apply_subtree_drain(BdrvChild *child, BlockDriverState *new_parent); 914 void bdrv_unapply_subtree_drain(BdrvChild *child, BlockDriverState *old_parent); 915 916 int get_tmp_filename(char *filename, int size); 917 BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size, 918 const char *filename); 919 920 void bdrv_parse_filename_strip_prefix(const char *filename, const char *prefix, 921 QDict *options); 922 923 924 /** 925 * bdrv_add_before_write_notifier: 926 * 927 * Register a callback that is invoked before write requests are processed but 928 * after any throttling or waiting for overlapping requests. 929 */ 930 void bdrv_add_before_write_notifier(BlockDriverState *bs, 931 NotifierWithReturn *notifier); 932 933 /** 934 * bdrv_detach_aio_context: 935 * 936 * May be called from .bdrv_detach_aio_context() to detach children from the 937 * current #AioContext. This is only needed by block drivers that manage their 938 * own children. Both ->file and ->backing are automatically handled and 939 * block drivers should not call this function on them explicitly. 940 */ 941 void bdrv_detach_aio_context(BlockDriverState *bs); 942 943 /** 944 * bdrv_attach_aio_context: 945 * 946 * May be called from .bdrv_attach_aio_context() to attach children to the new 947 * #AioContext. This is only needed by block drivers that manage their own 948 * children. Both ->file and ->backing are automatically handled and block 949 * drivers should not call this function on them explicitly. 950 */ 951 void bdrv_attach_aio_context(BlockDriverState *bs, 952 AioContext *new_context); 953 954 /** 955 * bdrv_add_aio_context_notifier: 956 * 957 * If a long-running job intends to be always run in the same AioContext as a 958 * certain BDS, it may use this function to be notified of changes regarding the 959 * association of the BDS to an AioContext. 960 * 961 * attached_aio_context() is called after the target BDS has been attached to a 962 * new AioContext; detach_aio_context() is called before the target BDS is being 963 * detached from its old AioContext. 964 */ 965 void bdrv_add_aio_context_notifier(BlockDriverState *bs, 966 void (*attached_aio_context)(AioContext *new_context, void *opaque), 967 void (*detach_aio_context)(void *opaque), void *opaque); 968 969 /** 970 * bdrv_remove_aio_context_notifier: 971 * 972 * Unsubscribe of change notifications regarding the BDS's AioContext. The 973 * parameters given here have to be the same as those given to 974 * bdrv_add_aio_context_notifier(). 975 */ 976 void bdrv_remove_aio_context_notifier(BlockDriverState *bs, 977 void (*aio_context_attached)(AioContext *, 978 void *), 979 void (*aio_context_detached)(void *), 980 void *opaque); 981 982 /** 983 * bdrv_wakeup: 984 * @bs: The BlockDriverState for which an I/O operation has been completed. 985 * 986 * Wake up the main thread if it is waiting on BDRV_POLL_WHILE. During 987 * synchronous I/O on a BlockDriverState that is attached to another 988 * I/O thread, the main thread lets the I/O thread's event loop run, 989 * waiting for the I/O operation to complete. A bdrv_wakeup will wake 990 * up the main thread if necessary. 991 * 992 * Manual calls to bdrv_wakeup are rarely necessary, because 993 * bdrv_dec_in_flight already calls it. 994 */ 995 void bdrv_wakeup(BlockDriverState *bs); 996 997 #ifdef _WIN32 998 int is_windows_drive(const char *filename); 999 #endif 1000 1001 /** 1002 * stream_start: 1003 * @job_id: The id of the newly-created job, or %NULL to use the 1004 * device name of @bs. 1005 * @bs: Block device to operate on. 1006 * @base: Block device that will become the new base, or %NULL to 1007 * flatten the whole backing file chain onto @bs. 1008 * @backing_file_str: The file name that will be written to @bs as the 1009 * the new backing file if the job completes. Ignored if @base is %NULL. 1010 * @creation_flags: Flags that control the behavior of the Job lifetime. 1011 * See @BlockJobCreateFlags 1012 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 1013 * @on_error: The action to take upon error. 1014 * @errp: Error object. 1015 * 1016 * Start a streaming operation on @bs. Clusters that are unallocated 1017 * in @bs, but allocated in any image between @base and @bs (both 1018 * exclusive) will be written to @bs. At the end of a successful 1019 * streaming job, the backing file of @bs will be changed to 1020 * @backing_file_str in the written image and to @base in the live 1021 * BlockDriverState. 1022 */ 1023 void stream_start(const char *job_id, BlockDriverState *bs, 1024 BlockDriverState *base, const char *backing_file_str, 1025 int creation_flags, int64_t speed, 1026 BlockdevOnError on_error, Error **errp); 1027 1028 /** 1029 * commit_start: 1030 * @job_id: The id of the newly-created job, or %NULL to use the 1031 * device name of @bs. 1032 * @bs: Active block device. 1033 * @top: Top block device to be committed. 1034 * @base: Block device that will be written into, and become the new top. 1035 * @creation_flags: Flags that control the behavior of the Job lifetime. 1036 * See @BlockJobCreateFlags 1037 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 1038 * @on_error: The action to take upon error. 1039 * @backing_file_str: String to use as the backing file in @top's overlay 1040 * @filter_node_name: The node name that should be assigned to the filter 1041 * driver that the commit job inserts into the graph above @top. NULL means 1042 * that a node name should be autogenerated. 1043 * @errp: Error object. 1044 * 1045 */ 1046 void commit_start(const char *job_id, BlockDriverState *bs, 1047 BlockDriverState *base, BlockDriverState *top, 1048 int creation_flags, int64_t speed, 1049 BlockdevOnError on_error, const char *backing_file_str, 1050 const char *filter_node_name, Error **errp); 1051 /** 1052 * commit_active_start: 1053 * @job_id: The id of the newly-created job, or %NULL to use the 1054 * device name of @bs. 1055 * @bs: Active block device to be committed. 1056 * @base: Block device that will be written into, and become the new top. 1057 * @creation_flags: Flags that control the behavior of the Job lifetime. 1058 * See @BlockJobCreateFlags 1059 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 1060 * @on_error: The action to take upon error. 1061 * @filter_node_name: The node name that should be assigned to the filter 1062 * driver that the commit job inserts into the graph above @bs. NULL means that 1063 * a node name should be autogenerated. 1064 * @cb: Completion function for the job. 1065 * @opaque: Opaque pointer value passed to @cb. 1066 * @auto_complete: Auto complete the job. 1067 * @errp: Error object. 1068 * 1069 */ 1070 void commit_active_start(const char *job_id, BlockDriverState *bs, 1071 BlockDriverState *base, int creation_flags, 1072 int64_t speed, BlockdevOnError on_error, 1073 const char *filter_node_name, 1074 BlockCompletionFunc *cb, void *opaque, 1075 bool auto_complete, Error **errp); 1076 /* 1077 * mirror_start: 1078 * @job_id: The id of the newly-created job, or %NULL to use the 1079 * device name of @bs. 1080 * @bs: Block device to operate on. 1081 * @target: Block device to write to. 1082 * @replaces: Block graph node name to replace once the mirror is done. Can 1083 * only be used when full mirroring is selected. 1084 * @creation_flags: Flags that control the behavior of the Job lifetime. 1085 * See @BlockJobCreateFlags 1086 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 1087 * @granularity: The chosen granularity for the dirty bitmap. 1088 * @buf_size: The amount of data that can be in flight at one time. 1089 * @mode: Whether to collapse all images in the chain to the target. 1090 * @backing_mode: How to establish the target's backing chain after completion. 1091 * @on_source_error: The action to take upon error reading from the source. 1092 * @on_target_error: The action to take upon error writing to the target. 1093 * @unmap: Whether to unmap target where source sectors only contain zeroes. 1094 * @filter_node_name: The node name that should be assigned to the filter 1095 * driver that the mirror job inserts into the graph above @bs. NULL means that 1096 * a node name should be autogenerated. 1097 * @copy_mode: When to trigger writes to the target. 1098 * @errp: Error object. 1099 * 1100 * Start a mirroring operation on @bs. Clusters that are allocated 1101 * in @bs will be written to @target until the job is cancelled or 1102 * manually completed. At the end of a successful mirroring job, 1103 * @bs will be switched to read from @target. 1104 */ 1105 void mirror_start(const char *job_id, BlockDriverState *bs, 1106 BlockDriverState *target, const char *replaces, 1107 int creation_flags, int64_t speed, 1108 uint32_t granularity, int64_t buf_size, 1109 MirrorSyncMode mode, BlockMirrorBackingMode backing_mode, 1110 BlockdevOnError on_source_error, 1111 BlockdevOnError on_target_error, 1112 bool unmap, const char *filter_node_name, 1113 MirrorCopyMode copy_mode, Error **errp); 1114 1115 /* 1116 * backup_job_create: 1117 * @job_id: The id of the newly-created job, or %NULL to use the 1118 * device name of @bs. 1119 * @bs: Block device to operate on. 1120 * @target: Block device to write to. 1121 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 1122 * @sync_mode: What parts of the disk image should be copied to the destination. 1123 * @sync_bitmap: The dirty bitmap if sync_mode is MIRROR_SYNC_MODE_INCREMENTAL. 1124 * @on_source_error: The action to take upon error reading from the source. 1125 * @on_target_error: The action to take upon error writing to the target. 1126 * @creation_flags: Flags that control the behavior of the Job lifetime. 1127 * See @BlockJobCreateFlags 1128 * @cb: Completion function for the job. 1129 * @opaque: Opaque pointer value passed to @cb. 1130 * @txn: Transaction that this job is part of (may be NULL). 1131 * 1132 * Create a backup operation on @bs. Clusters in @bs are written to @target 1133 * until the job is cancelled or manually completed. 1134 */ 1135 BlockJob *backup_job_create(const char *job_id, BlockDriverState *bs, 1136 BlockDriverState *target, int64_t speed, 1137 MirrorSyncMode sync_mode, 1138 BdrvDirtyBitmap *sync_bitmap, 1139 bool compress, 1140 BlockdevOnError on_source_error, 1141 BlockdevOnError on_target_error, 1142 int creation_flags, 1143 BlockCompletionFunc *cb, void *opaque, 1144 JobTxn *txn, Error **errp); 1145 1146 void hmp_drive_add_node(Monitor *mon, const char *optstr); 1147 1148 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs, 1149 const char *child_name, 1150 const BdrvChildRole *child_role, 1151 uint64_t perm, uint64_t shared_perm, 1152 void *opaque, Error **errp); 1153 void bdrv_root_unref_child(BdrvChild *child); 1154 1155 int bdrv_child_try_set_perm(BdrvChild *c, uint64_t perm, uint64_t shared, 1156 Error **errp); 1157 1158 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by 1159 * block filters: Forward CONSISTENT_READ, WRITE, WRITE_UNCHANGED and RESIZE to 1160 * all children */ 1161 void bdrv_filter_default_perms(BlockDriverState *bs, BdrvChild *c, 1162 const BdrvChildRole *role, 1163 BlockReopenQueue *reopen_queue, 1164 uint64_t perm, uint64_t shared, 1165 uint64_t *nperm, uint64_t *nshared); 1166 1167 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by 1168 * (non-raw) image formats: Like above for bs->backing, but for bs->file it 1169 * requires WRITE | RESIZE for read-write images, always requires 1170 * CONSISTENT_READ and doesn't share WRITE. */ 1171 void bdrv_format_default_perms(BlockDriverState *bs, BdrvChild *c, 1172 const BdrvChildRole *role, 1173 BlockReopenQueue *reopen_queue, 1174 uint64_t perm, uint64_t shared, 1175 uint64_t *nperm, uint64_t *nshared); 1176 1177 /* 1178 * Default implementation for drivers to pass bdrv_co_block_status() to 1179 * their file. 1180 */ 1181 int coroutine_fn bdrv_co_block_status_from_file(BlockDriverState *bs, 1182 bool want_zero, 1183 int64_t offset, 1184 int64_t bytes, 1185 int64_t *pnum, 1186 int64_t *map, 1187 BlockDriverState **file); 1188 /* 1189 * Default implementation for drivers to pass bdrv_co_block_status() to 1190 * their backing file. 1191 */ 1192 int coroutine_fn bdrv_co_block_status_from_backing(BlockDriverState *bs, 1193 bool want_zero, 1194 int64_t offset, 1195 int64_t bytes, 1196 int64_t *pnum, 1197 int64_t *map, 1198 BlockDriverState **file); 1199 const char *bdrv_get_parent_name(const BlockDriverState *bs); 1200 void blk_dev_change_media_cb(BlockBackend *blk, bool load, Error **errp); 1201 bool blk_dev_has_removable_media(BlockBackend *blk); 1202 bool blk_dev_has_tray(BlockBackend *blk); 1203 void blk_dev_eject_request(BlockBackend *blk, bool force); 1204 bool blk_dev_is_tray_open(BlockBackend *blk); 1205 bool blk_dev_is_medium_locked(BlockBackend *blk); 1206 1207 void bdrv_set_dirty(BlockDriverState *bs, int64_t offset, int64_t bytes); 1208 1209 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap **out); 1210 void bdrv_restore_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap *backup); 1211 1212 void bdrv_inc_in_flight(BlockDriverState *bs); 1213 void bdrv_dec_in_flight(BlockDriverState *bs); 1214 1215 void blockdev_close_all_bdrv_states(void); 1216 1217 int coroutine_fn bdrv_co_copy_range_from(BdrvChild *src, uint64_t src_offset, 1218 BdrvChild *dst, uint64_t dst_offset, 1219 uint64_t bytes, 1220 BdrvRequestFlags read_flags, 1221 BdrvRequestFlags write_flags); 1222 int coroutine_fn bdrv_co_copy_range_to(BdrvChild *src, uint64_t src_offset, 1223 BdrvChild *dst, uint64_t dst_offset, 1224 uint64_t bytes, 1225 BdrvRequestFlags read_flags, 1226 BdrvRequestFlags write_flags); 1227 1228 int refresh_total_sectors(BlockDriverState *bs, int64_t hint); 1229 1230 #endif /* BLOCK_INT_H */ 1231