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/throttle.h" 37 38 #define BLOCK_FLAG_LAZY_REFCOUNTS 8 39 40 #define BLOCK_OPT_SIZE "size" 41 #define BLOCK_OPT_ENCRYPT "encryption" 42 #define BLOCK_OPT_ENCRYPT_FORMAT "encrypt.format" 43 #define BLOCK_OPT_COMPAT6 "compat6" 44 #define BLOCK_OPT_HWVERSION "hwversion" 45 #define BLOCK_OPT_BACKING_FILE "backing_file" 46 #define BLOCK_OPT_BACKING_FMT "backing_fmt" 47 #define BLOCK_OPT_CLUSTER_SIZE "cluster_size" 48 #define BLOCK_OPT_TABLE_SIZE "table_size" 49 #define BLOCK_OPT_PREALLOC "preallocation" 50 #define BLOCK_OPT_SUBFMT "subformat" 51 #define BLOCK_OPT_COMPAT_LEVEL "compat" 52 #define BLOCK_OPT_LAZY_REFCOUNTS "lazy_refcounts" 53 #define BLOCK_OPT_ADAPTER_TYPE "adapter_type" 54 #define BLOCK_OPT_REDUNDANCY "redundancy" 55 #define BLOCK_OPT_NOCOW "nocow" 56 #define BLOCK_OPT_OBJECT_SIZE "object_size" 57 #define BLOCK_OPT_REFCOUNT_BITS "refcount_bits" 58 #define BLOCK_OPT_DATA_FILE "data_file" 59 #define BLOCK_OPT_DATA_FILE_RAW "data_file_raw" 60 #define BLOCK_OPT_COMPRESSION_TYPE "compression_type" 61 62 #define BLOCK_PROBE_BUF_SIZE 512 63 64 enum BdrvTrackedRequestType { 65 BDRV_TRACKED_READ, 66 BDRV_TRACKED_WRITE, 67 BDRV_TRACKED_DISCARD, 68 BDRV_TRACKED_TRUNCATE, 69 }; 70 71 typedef struct BdrvTrackedRequest { 72 BlockDriverState *bs; 73 int64_t offset; 74 uint64_t bytes; 75 enum BdrvTrackedRequestType type; 76 77 bool serialising; 78 int64_t overlap_offset; 79 uint64_t overlap_bytes; 80 81 QLIST_ENTRY(BdrvTrackedRequest) list; 82 Coroutine *co; /* owner, used for deadlock detection */ 83 CoQueue wait_queue; /* coroutines blocked on this request */ 84 85 struct BdrvTrackedRequest *waiting_for; 86 } BdrvTrackedRequest; 87 88 struct BlockDriver { 89 const char *format_name; 90 int instance_size; 91 92 /* set to true if the BlockDriver is a block filter. Block filters pass 93 * certain callbacks that refer to data (see block.c) to their bs->file if 94 * the driver doesn't implement them. Drivers that do not wish to forward 95 * must implement them and return -ENOTSUP. 96 */ 97 bool is_filter; 98 /* 99 * Set to true if the BlockDriver is a format driver. Format nodes 100 * generally do not expect their children to be other format nodes 101 * (except for backing files), and so format probing is disabled 102 * on those children. 103 */ 104 bool is_format; 105 /* 106 * Return true if @to_replace can be replaced by a BDS with the 107 * same data as @bs without it affecting @bs's behavior (that is, 108 * without it being visible to @bs's parents). 109 */ 110 bool (*bdrv_recurse_can_replace)(BlockDriverState *bs, 111 BlockDriverState *to_replace); 112 113 int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename); 114 int (*bdrv_probe_device)(const char *filename); 115 116 /* Any driver implementing this callback is expected to be able to handle 117 * NULL file names in its .bdrv_open() implementation */ 118 void (*bdrv_parse_filename)(const char *filename, QDict *options, Error **errp); 119 /* Drivers not implementing bdrv_parse_filename nor bdrv_open should have 120 * this field set to true, except ones that are defined only by their 121 * child's bs. 122 * An example of the last type will be the quorum block driver. 123 */ 124 bool bdrv_needs_filename; 125 126 /* Set if a driver can support backing files */ 127 bool supports_backing; 128 129 /* For handling image reopen for split or non-split files */ 130 int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state, 131 BlockReopenQueue *queue, Error **errp); 132 void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state); 133 void (*bdrv_reopen_commit_post)(BDRVReopenState *reopen_state); 134 void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state); 135 void (*bdrv_join_options)(QDict *options, QDict *old_options); 136 137 int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags, 138 Error **errp); 139 140 /* Protocol drivers should implement this instead of bdrv_open */ 141 int (*bdrv_file_open)(BlockDriverState *bs, QDict *options, int flags, 142 Error **errp); 143 void (*bdrv_close)(BlockDriverState *bs); 144 int coroutine_fn (*bdrv_co_create)(BlockdevCreateOptions *opts, 145 Error **errp); 146 int coroutine_fn (*bdrv_co_create_opts)(BlockDriver *drv, 147 const char *filename, 148 QemuOpts *opts, 149 Error **errp); 150 int (*bdrv_make_empty)(BlockDriverState *bs); 151 152 /* 153 * Refreshes the bs->exact_filename field. If that is impossible, 154 * bs->exact_filename has to be left empty. 155 */ 156 void (*bdrv_refresh_filename)(BlockDriverState *bs); 157 158 /* 159 * Gathers the open options for all children into @target. 160 * A simple format driver (without backing file support) might 161 * implement this function like this: 162 * 163 * QINCREF(bs->file->bs->full_open_options); 164 * qdict_put(target, "file", bs->file->bs->full_open_options); 165 * 166 * If not specified, the generic implementation will simply put 167 * all children's options under their respective name. 168 * 169 * @backing_overridden is true when bs->backing seems not to be 170 * the child that would result from opening bs->backing_file. 171 * Therefore, if it is true, the backing child's options should be 172 * gathered; otherwise, there is no need since the backing child 173 * is the one implied by the image header. 174 * 175 * Note that ideally this function would not be needed. Every 176 * block driver which implements it is probably doing something 177 * shady regarding its runtime option structure. 178 */ 179 void (*bdrv_gather_child_options)(BlockDriverState *bs, QDict *target, 180 bool backing_overridden); 181 182 /* 183 * Returns an allocated string which is the directory name of this BDS: It 184 * will be used to make relative filenames absolute by prepending this 185 * function's return value to them. 186 */ 187 char *(*bdrv_dirname)(BlockDriverState *bs, Error **errp); 188 189 /* aio */ 190 BlockAIOCB *(*bdrv_aio_preadv)(BlockDriverState *bs, 191 uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags, 192 BlockCompletionFunc *cb, void *opaque); 193 BlockAIOCB *(*bdrv_aio_pwritev)(BlockDriverState *bs, 194 uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags, 195 BlockCompletionFunc *cb, void *opaque); 196 BlockAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs, 197 BlockCompletionFunc *cb, void *opaque); 198 BlockAIOCB *(*bdrv_aio_pdiscard)(BlockDriverState *bs, 199 int64_t offset, int bytes, 200 BlockCompletionFunc *cb, void *opaque); 201 202 int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs, 203 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov); 204 205 /** 206 * @offset: position in bytes to read at 207 * @bytes: number of bytes to read 208 * @qiov: the buffers to fill with read data 209 * @flags: currently unused, always 0 210 * 211 * @offset and @bytes will be a multiple of 'request_alignment', 212 * but the length of individual @qiov elements does not have to 213 * be a multiple. 214 * 215 * @bytes will always equal the total size of @qiov, and will be 216 * no larger than 'max_transfer'. 217 * 218 * The buffer in @qiov may point directly to guest memory. 219 */ 220 int coroutine_fn (*bdrv_co_preadv)(BlockDriverState *bs, 221 uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags); 222 int coroutine_fn (*bdrv_co_preadv_part)(BlockDriverState *bs, 223 uint64_t offset, uint64_t bytes, 224 QEMUIOVector *qiov, size_t qiov_offset, int flags); 225 int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs, 226 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, int flags); 227 /** 228 * @offset: position in bytes to write at 229 * @bytes: number of bytes to write 230 * @qiov: the buffers containing data to write 231 * @flags: zero or more bits allowed by 'supported_write_flags' 232 * 233 * @offset and @bytes will be a multiple of 'request_alignment', 234 * but the length of individual @qiov elements does not have to 235 * be a multiple. 236 * 237 * @bytes will always equal the total size of @qiov, and will be 238 * no larger than 'max_transfer'. 239 * 240 * The buffer in @qiov may point directly to guest memory. 241 */ 242 int coroutine_fn (*bdrv_co_pwritev)(BlockDriverState *bs, 243 uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags); 244 int coroutine_fn (*bdrv_co_pwritev_part)(BlockDriverState *bs, 245 uint64_t offset, uint64_t bytes, 246 QEMUIOVector *qiov, size_t qiov_offset, int flags); 247 248 /* 249 * Efficiently zero a region of the disk image. Typically an image format 250 * would use a compact metadata representation to implement this. This 251 * function pointer may be NULL or return -ENOSUP and .bdrv_co_writev() 252 * will be called instead. 253 */ 254 int coroutine_fn (*bdrv_co_pwrite_zeroes)(BlockDriverState *bs, 255 int64_t offset, int bytes, BdrvRequestFlags flags); 256 int coroutine_fn (*bdrv_co_pdiscard)(BlockDriverState *bs, 257 int64_t offset, int bytes); 258 259 /* Map [offset, offset + nbytes) range onto a child of @bs to copy from, 260 * and invoke bdrv_co_copy_range_from(child, ...), or invoke 261 * bdrv_co_copy_range_to() if @bs is the leaf child to copy data from. 262 * 263 * See the comment of bdrv_co_copy_range for the parameter and return value 264 * semantics. 265 */ 266 int coroutine_fn (*bdrv_co_copy_range_from)(BlockDriverState *bs, 267 BdrvChild *src, 268 uint64_t offset, 269 BdrvChild *dst, 270 uint64_t dst_offset, 271 uint64_t bytes, 272 BdrvRequestFlags read_flags, 273 BdrvRequestFlags write_flags); 274 275 /* Map [offset, offset + nbytes) range onto a child of bs to copy data to, 276 * and invoke bdrv_co_copy_range_to(child, src, ...), or perform the copy 277 * operation if @bs is the leaf and @src has the same BlockDriver. Return 278 * -ENOTSUP if @bs is the leaf but @src has a different BlockDriver. 279 * 280 * See the comment of bdrv_co_copy_range for the parameter and return value 281 * semantics. 282 */ 283 int coroutine_fn (*bdrv_co_copy_range_to)(BlockDriverState *bs, 284 BdrvChild *src, 285 uint64_t src_offset, 286 BdrvChild *dst, 287 uint64_t dst_offset, 288 uint64_t bytes, 289 BdrvRequestFlags read_flags, 290 BdrvRequestFlags write_flags); 291 292 /* 293 * Building block for bdrv_block_status[_above] and 294 * bdrv_is_allocated[_above]. The driver should answer only 295 * according to the current layer, and should only need to set 296 * BDRV_BLOCK_DATA, BDRV_BLOCK_ZERO, BDRV_BLOCK_OFFSET_VALID, 297 * and/or BDRV_BLOCK_RAW; if the current layer defers to a backing 298 * layer, the result should be 0 (and not BDRV_BLOCK_ZERO). See 299 * block.h for the overall meaning of the bits. As a hint, the 300 * flag want_zero is true if the caller cares more about precise 301 * mappings (favor accurate _OFFSET_VALID/_ZERO) or false for 302 * overall allocation (favor larger *pnum, perhaps by reporting 303 * _DATA instead of _ZERO). The block layer guarantees input 304 * clamped to bdrv_getlength() and aligned to request_alignment, 305 * as well as non-NULL pnum, map, and file; in turn, the driver 306 * must return an error or set pnum to an aligned non-zero value. 307 */ 308 int coroutine_fn (*bdrv_co_block_status)(BlockDriverState *bs, 309 bool want_zero, int64_t offset, int64_t bytes, int64_t *pnum, 310 int64_t *map, BlockDriverState **file); 311 312 /* 313 * Invalidate any cached meta-data. 314 */ 315 void coroutine_fn (*bdrv_co_invalidate_cache)(BlockDriverState *bs, 316 Error **errp); 317 int (*bdrv_inactivate)(BlockDriverState *bs); 318 319 /* 320 * Flushes all data for all layers by calling bdrv_co_flush for underlying 321 * layers, if needed. This function is needed for deterministic 322 * synchronization of the flush finishing callback. 323 */ 324 int coroutine_fn (*bdrv_co_flush)(BlockDriverState *bs); 325 326 /* Delete a created file. */ 327 int coroutine_fn (*bdrv_co_delete_file)(BlockDriverState *bs, 328 Error **errp); 329 330 /* 331 * Flushes all data that was already written to the OS all the way down to 332 * the disk (for example file-posix.c calls fsync()). 333 */ 334 int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs); 335 336 /* 337 * Flushes all internal caches to the OS. The data may still sit in a 338 * writeback cache of the host OS, but it will survive a crash of the qemu 339 * process. 340 */ 341 int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs); 342 343 /* 344 * Drivers setting this field must be able to work with just a plain 345 * filename with '<protocol_name>:' as a prefix, and no other options. 346 * Options may be extracted from the filename by implementing 347 * bdrv_parse_filename. 348 */ 349 const char *protocol_name; 350 351 /* 352 * Truncate @bs to @offset bytes using the given @prealloc mode 353 * when growing. Modes other than PREALLOC_MODE_OFF should be 354 * rejected when shrinking @bs. 355 * 356 * If @exact is true, @bs must be resized to exactly @offset. 357 * Otherwise, it is sufficient for @bs (if it is a host block 358 * device and thus there is no way to resize it) to be at least 359 * @offset bytes in length. 360 * 361 * If @exact is true and this function fails but would succeed 362 * with @exact = false, it should return -ENOTSUP. 363 */ 364 int coroutine_fn (*bdrv_co_truncate)(BlockDriverState *bs, int64_t offset, 365 bool exact, PreallocMode prealloc, 366 BdrvRequestFlags flags, Error **errp); 367 368 int64_t (*bdrv_getlength)(BlockDriverState *bs); 369 bool has_variable_length; 370 int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs); 371 BlockMeasureInfo *(*bdrv_measure)(QemuOpts *opts, BlockDriverState *in_bs, 372 Error **errp); 373 374 int coroutine_fn (*bdrv_co_pwritev_compressed)(BlockDriverState *bs, 375 uint64_t offset, uint64_t bytes, QEMUIOVector *qiov); 376 int coroutine_fn (*bdrv_co_pwritev_compressed_part)(BlockDriverState *bs, 377 uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, 378 size_t qiov_offset); 379 380 int (*bdrv_snapshot_create)(BlockDriverState *bs, 381 QEMUSnapshotInfo *sn_info); 382 int (*bdrv_snapshot_goto)(BlockDriverState *bs, 383 const char *snapshot_id); 384 int (*bdrv_snapshot_delete)(BlockDriverState *bs, 385 const char *snapshot_id, 386 const char *name, 387 Error **errp); 388 int (*bdrv_snapshot_list)(BlockDriverState *bs, 389 QEMUSnapshotInfo **psn_info); 390 int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs, 391 const char *snapshot_id, 392 const char *name, 393 Error **errp); 394 int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi); 395 ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs, 396 Error **errp); 397 BlockStatsSpecific *(*bdrv_get_specific_stats)(BlockDriverState *bs); 398 399 int coroutine_fn (*bdrv_save_vmstate)(BlockDriverState *bs, 400 QEMUIOVector *qiov, 401 int64_t pos); 402 int coroutine_fn (*bdrv_load_vmstate)(BlockDriverState *bs, 403 QEMUIOVector *qiov, 404 int64_t pos); 405 406 int (*bdrv_change_backing_file)(BlockDriverState *bs, 407 const char *backing_file, const char *backing_fmt); 408 409 /* removable device specific */ 410 bool (*bdrv_is_inserted)(BlockDriverState *bs); 411 void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag); 412 void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked); 413 414 /* to control generic scsi devices */ 415 BlockAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs, 416 unsigned long int req, void *buf, 417 BlockCompletionFunc *cb, void *opaque); 418 int coroutine_fn (*bdrv_co_ioctl)(BlockDriverState *bs, 419 unsigned long int req, void *buf); 420 421 /* List of options for creating images, terminated by name == NULL */ 422 QemuOptsList *create_opts; 423 /* 424 * If this driver supports reopening images this contains a 425 * NULL-terminated list of the runtime options that can be 426 * modified. If an option in this list is unspecified during 427 * reopen then it _must_ be reset to its default value or return 428 * an error. 429 */ 430 const char *const *mutable_opts; 431 432 /* 433 * Returns 0 for completed check, -errno for internal errors. 434 * The check results are stored in result. 435 */ 436 int coroutine_fn (*bdrv_co_check)(BlockDriverState *bs, 437 BdrvCheckResult *result, 438 BdrvCheckMode fix); 439 440 int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts, 441 BlockDriverAmendStatusCB *status_cb, 442 void *cb_opaque, 443 Error **errp); 444 445 void (*bdrv_debug_event)(BlockDriverState *bs, BlkdebugEvent event); 446 447 /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */ 448 int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event, 449 const char *tag); 450 int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs, 451 const char *tag); 452 int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag); 453 bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag); 454 455 void (*bdrv_refresh_limits)(BlockDriverState *bs, Error **errp); 456 457 /* 458 * Returns 1 if newly created images are guaranteed to contain only 459 * zeros, 0 otherwise. 460 */ 461 int (*bdrv_has_zero_init)(BlockDriverState *bs); 462 463 /* Remove fd handlers, timers, and other event loop callbacks so the event 464 * loop is no longer in use. Called with no in-flight requests and in 465 * depth-first traversal order with parents before child nodes. 466 */ 467 void (*bdrv_detach_aio_context)(BlockDriverState *bs); 468 469 /* Add fd handlers, timers, and other event loop callbacks so I/O requests 470 * can be processed again. Called with no in-flight requests and in 471 * depth-first traversal order with child nodes before parent nodes. 472 */ 473 void (*bdrv_attach_aio_context)(BlockDriverState *bs, 474 AioContext *new_context); 475 476 /* io queue for linux-aio */ 477 void (*bdrv_io_plug)(BlockDriverState *bs); 478 void (*bdrv_io_unplug)(BlockDriverState *bs); 479 480 /** 481 * Try to get @bs's logical and physical block size. 482 * On success, store them in @bsz and return zero. 483 * On failure, return negative errno. 484 */ 485 int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz); 486 /** 487 * Try to get @bs's geometry (cyls, heads, sectors) 488 * On success, store them in @geo and return 0. 489 * On failure return -errno. 490 * Only drivers that want to override guest geometry implement this 491 * callback; see hd_geometry_guess(). 492 */ 493 int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo); 494 495 /** 496 * bdrv_co_drain_begin is called if implemented in the beginning of a 497 * drain operation to drain and stop any internal sources of requests in 498 * the driver. 499 * bdrv_co_drain_end is called if implemented at the end of the drain. 500 * 501 * They should be used by the driver to e.g. manage scheduled I/O 502 * requests, or toggle an internal state. After the end of the drain new 503 * requests will continue normally. 504 */ 505 void coroutine_fn (*bdrv_co_drain_begin)(BlockDriverState *bs); 506 void coroutine_fn (*bdrv_co_drain_end)(BlockDriverState *bs); 507 508 void (*bdrv_add_child)(BlockDriverState *parent, BlockDriverState *child, 509 Error **errp); 510 void (*bdrv_del_child)(BlockDriverState *parent, BdrvChild *child, 511 Error **errp); 512 513 /** 514 * Informs the block driver that a permission change is intended. The 515 * driver checks whether the change is permissible and may take other 516 * preparations for the change (e.g. get file system locks). This operation 517 * is always followed either by a call to either .bdrv_set_perm or 518 * .bdrv_abort_perm_update. 519 * 520 * Checks whether the requested set of cumulative permissions in @perm 521 * can be granted for accessing @bs and whether no other users are using 522 * permissions other than those given in @shared (both arguments take 523 * BLK_PERM_* bitmasks). 524 * 525 * If both conditions are met, 0 is returned. Otherwise, -errno is returned 526 * and errp is set to an error describing the conflict. 527 */ 528 int (*bdrv_check_perm)(BlockDriverState *bs, uint64_t perm, 529 uint64_t shared, Error **errp); 530 531 /** 532 * Called to inform the driver that the set of cumulative set of used 533 * permissions for @bs has changed to @perm, and the set of sharable 534 * permission to @shared. The driver can use this to propagate changes to 535 * its children (i.e. request permissions only if a parent actually needs 536 * them). 537 * 538 * This function is only invoked after bdrv_check_perm(), so block drivers 539 * may rely on preparations made in their .bdrv_check_perm implementation. 540 */ 541 void (*bdrv_set_perm)(BlockDriverState *bs, uint64_t perm, uint64_t shared); 542 543 /* 544 * Called to inform the driver that after a previous bdrv_check_perm() 545 * call, the permission update is not performed and any preparations made 546 * for it (e.g. taken file locks) need to be undone. 547 * 548 * This function can be called even for nodes that never saw a 549 * bdrv_check_perm() call. It is a no-op then. 550 */ 551 void (*bdrv_abort_perm_update)(BlockDriverState *bs); 552 553 /** 554 * Returns in @nperm and @nshared the permissions that the driver for @bs 555 * needs on its child @c, based on the cumulative permissions requested by 556 * the parents in @parent_perm and @parent_shared. 557 * 558 * If @c is NULL, return the permissions for attaching a new child for the 559 * given @child_class and @role. 560 * 561 * If @reopen_queue is non-NULL, don't return the currently needed 562 * permissions, but those that will be needed after applying the 563 * @reopen_queue. 564 */ 565 void (*bdrv_child_perm)(BlockDriverState *bs, BdrvChild *c, 566 BdrvChildRole role, 567 BlockReopenQueue *reopen_queue, 568 uint64_t parent_perm, uint64_t parent_shared, 569 uint64_t *nperm, uint64_t *nshared); 570 571 bool (*bdrv_supports_persistent_dirty_bitmap)(BlockDriverState *bs); 572 bool (*bdrv_co_can_store_new_dirty_bitmap)(BlockDriverState *bs, 573 const char *name, 574 uint32_t granularity, 575 Error **errp); 576 int (*bdrv_co_remove_persistent_dirty_bitmap)(BlockDriverState *bs, 577 const char *name, 578 Error **errp); 579 580 /** 581 * Register/unregister a buffer for I/O. For example, when the driver is 582 * interested to know the memory areas that will later be used in iovs, so 583 * that it can do IOMMU mapping with VFIO etc., in order to get better 584 * performance. In the case of VFIO drivers, this callback is used to do 585 * DMA mapping for hot buffers. 586 */ 587 void (*bdrv_register_buf)(BlockDriverState *bs, void *host, size_t size); 588 void (*bdrv_unregister_buf)(BlockDriverState *bs, void *host); 589 QLIST_ENTRY(BlockDriver) list; 590 591 /* Pointer to a NULL-terminated array of names of strong options 592 * that can be specified for bdrv_open(). A strong option is one 593 * that changes the data of a BDS. 594 * If this pointer is NULL, the array is considered empty. 595 * "filename" and "driver" are always considered strong. */ 596 const char *const *strong_runtime_opts; 597 }; 598 599 static inline bool block_driver_can_compress(BlockDriver *drv) 600 { 601 return drv->bdrv_co_pwritev_compressed || 602 drv->bdrv_co_pwritev_compressed_part; 603 } 604 605 typedef struct BlockLimits { 606 /* Alignment requirement, in bytes, for offset/length of I/O 607 * requests. Must be a power of 2 less than INT_MAX; defaults to 608 * 1 for drivers with modern byte interfaces, and to 512 609 * otherwise. */ 610 uint32_t request_alignment; 611 612 /* Maximum number of bytes that can be discarded at once (since it 613 * is signed, it must be < 2G, if set). Must be multiple of 614 * pdiscard_alignment, but need not be power of 2. May be 0 if no 615 * inherent 32-bit limit */ 616 int32_t max_pdiscard; 617 618 /* Optimal alignment for discard requests in bytes. A power of 2 619 * is best but not mandatory. Must be a multiple of 620 * bl.request_alignment, and must be less than max_pdiscard if 621 * that is set. May be 0 if bl.request_alignment is good enough */ 622 uint32_t pdiscard_alignment; 623 624 /* Maximum number of bytes that can zeroized at once (since it is 625 * signed, it must be < 2G, if set). Must be multiple of 626 * pwrite_zeroes_alignment. May be 0 if no inherent 32-bit limit */ 627 int32_t max_pwrite_zeroes; 628 629 /* Optimal alignment for write zeroes requests in bytes. A power 630 * of 2 is best but not mandatory. Must be a multiple of 631 * bl.request_alignment, and must be less than max_pwrite_zeroes 632 * if that is set. May be 0 if bl.request_alignment is good 633 * enough */ 634 uint32_t pwrite_zeroes_alignment; 635 636 /* Optimal transfer length in bytes. A power of 2 is best but not 637 * mandatory. Must be a multiple of bl.request_alignment, or 0 if 638 * no preferred size */ 639 uint32_t opt_transfer; 640 641 /* Maximal transfer length in bytes. Need not be power of 2, but 642 * must be multiple of opt_transfer and bl.request_alignment, or 0 643 * for no 32-bit limit. For now, anything larger than INT_MAX is 644 * clamped down. */ 645 uint32_t max_transfer; 646 647 /* memory alignment, in bytes so that no bounce buffer is needed */ 648 size_t min_mem_alignment; 649 650 /* memory alignment, in bytes, for bounce buffer */ 651 size_t opt_mem_alignment; 652 653 /* maximum number of iovec elements */ 654 int max_iov; 655 } BlockLimits; 656 657 typedef struct BdrvOpBlocker BdrvOpBlocker; 658 659 typedef struct BdrvAioNotifier { 660 void (*attached_aio_context)(AioContext *new_context, void *opaque); 661 void (*detach_aio_context)(void *opaque); 662 663 void *opaque; 664 bool deleted; 665 666 QLIST_ENTRY(BdrvAioNotifier) list; 667 } BdrvAioNotifier; 668 669 struct BdrvChildClass { 670 /* If true, bdrv_replace_node() doesn't change the node this BdrvChild 671 * points to. */ 672 bool stay_at_node; 673 674 /* If true, the parent is a BlockDriverState and bdrv_next_all_states() 675 * will return it. This information is used for drain_all, where every node 676 * will be drained separately, so the drain only needs to be propagated to 677 * non-BDS parents. */ 678 bool parent_is_bds; 679 680 void (*inherit_options)(BdrvChildRole role, bool parent_is_format, 681 int *child_flags, QDict *child_options, 682 int parent_flags, QDict *parent_options); 683 684 void (*change_media)(BdrvChild *child, bool load); 685 void (*resize)(BdrvChild *child); 686 687 /* Returns a name that is supposedly more useful for human users than the 688 * node name for identifying the node in question (in particular, a BB 689 * name), or NULL if the parent can't provide a better name. */ 690 const char *(*get_name)(BdrvChild *child); 691 692 /* Returns a malloced string that describes the parent of the child for a 693 * human reader. This could be a node-name, BlockBackend name, qdev ID or 694 * QOM path of the device owning the BlockBackend, job type and ID etc. The 695 * caller is responsible for freeing the memory. */ 696 char *(*get_parent_desc)(BdrvChild *child); 697 698 /* 699 * If this pair of functions is implemented, the parent doesn't issue new 700 * requests after returning from .drained_begin() until .drained_end() is 701 * called. 702 * 703 * These functions must not change the graph (and therefore also must not 704 * call aio_poll(), which could change the graph indirectly). 705 * 706 * If drained_end() schedules background operations, it must atomically 707 * increment *drained_end_counter for each such operation and atomically 708 * decrement it once the operation has settled. 709 * 710 * Note that this can be nested. If drained_begin() was called twice, new 711 * I/O is allowed only after drained_end() was called twice, too. 712 */ 713 void (*drained_begin)(BdrvChild *child); 714 void (*drained_end)(BdrvChild *child, int *drained_end_counter); 715 716 /* 717 * Returns whether the parent has pending requests for the child. This 718 * callback is polled after .drained_begin() has been called until all 719 * activity on the child has stopped. 720 */ 721 bool (*drained_poll)(BdrvChild *child); 722 723 /* Notifies the parent that the child has been activated/inactivated (e.g. 724 * when migration is completing) and it can start/stop requesting 725 * permissions and doing I/O on it. */ 726 void (*activate)(BdrvChild *child, Error **errp); 727 int (*inactivate)(BdrvChild *child); 728 729 void (*attach)(BdrvChild *child); 730 void (*detach)(BdrvChild *child); 731 732 /* Notifies the parent that the filename of its child has changed (e.g. 733 * because the direct child was removed from the backing chain), so that it 734 * can update its reference. */ 735 int (*update_filename)(BdrvChild *child, BlockDriverState *new_base, 736 const char *filename, Error **errp); 737 738 bool (*can_set_aio_ctx)(BdrvChild *child, AioContext *ctx, 739 GSList **ignore, Error **errp); 740 void (*set_aio_ctx)(BdrvChild *child, AioContext *ctx, GSList **ignore); 741 }; 742 743 extern const BdrvChildClass child_of_bds; 744 745 struct BdrvChild { 746 BlockDriverState *bs; 747 char *name; 748 const BdrvChildClass *klass; 749 BdrvChildRole role; 750 void *opaque; 751 752 /** 753 * Granted permissions for operating on this BdrvChild (BLK_PERM_* bitmask) 754 */ 755 uint64_t perm; 756 757 /** 758 * Permissions that can still be granted to other users of @bs while this 759 * BdrvChild is still attached to it. (BLK_PERM_* bitmask) 760 */ 761 uint64_t shared_perm; 762 763 /* backup of permissions during permission update procedure */ 764 bool has_backup_perm; 765 uint64_t backup_perm; 766 uint64_t backup_shared_perm; 767 768 /* 769 * This link is frozen: the child can neither be replaced nor 770 * detached from the parent. 771 */ 772 bool frozen; 773 774 /* 775 * How many times the parent of this child has been drained 776 * (through klass->drained_*). 777 * Usually, this is equal to bs->quiesce_counter (potentially 778 * reduced by bdrv_drain_all_count). It may differ while the 779 * child is entering or leaving a drained section. 780 */ 781 int parent_quiesce_counter; 782 783 QLIST_ENTRY(BdrvChild) next; 784 QLIST_ENTRY(BdrvChild) next_parent; 785 }; 786 787 /* 788 * Note: the function bdrv_append() copies and swaps contents of 789 * BlockDriverStates, so if you add new fields to this struct, please 790 * inspect bdrv_append() to determine if the new fields need to be 791 * copied as well. 792 */ 793 struct BlockDriverState { 794 /* Protected by big QEMU lock or read-only after opening. No special 795 * locking needed during I/O... 796 */ 797 int open_flags; /* flags used to open the file, re-used for re-open */ 798 bool read_only; /* if true, the media is read only */ 799 bool encrypted; /* if true, the media is encrypted */ 800 bool sg; /* if true, the device is a /dev/sg* */ 801 bool probed; /* if true, format was probed rather than specified */ 802 bool force_share; /* if true, always allow all shared permissions */ 803 bool implicit; /* if true, this filter node was automatically inserted */ 804 805 BlockDriver *drv; /* NULL means no media */ 806 void *opaque; 807 808 AioContext *aio_context; /* event loop used for fd handlers, timers, etc */ 809 /* long-running tasks intended to always use the same AioContext as this 810 * BDS may register themselves in this list to be notified of changes 811 * regarding this BDS's context */ 812 QLIST_HEAD(, BdrvAioNotifier) aio_notifiers; 813 bool walking_aio_notifiers; /* to make removal during iteration safe */ 814 815 char filename[PATH_MAX]; 816 char backing_file[PATH_MAX]; /* if non zero, the image is a diff of 817 this file image */ 818 /* The backing filename indicated by the image header; if we ever 819 * open this file, then this is replaced by the resulting BDS's 820 * filename (i.e. after a bdrv_refresh_filename() run). */ 821 char auto_backing_file[PATH_MAX]; 822 char backing_format[16]; /* if non-zero and backing_file exists */ 823 824 QDict *full_open_options; 825 char exact_filename[PATH_MAX]; 826 827 BdrvChild *backing; 828 BdrvChild *file; 829 830 /* I/O Limits */ 831 BlockLimits bl; 832 833 /* Flags honored during pwrite (so far: BDRV_REQ_FUA, 834 * BDRV_REQ_WRITE_UNCHANGED). 835 * If a driver does not support BDRV_REQ_WRITE_UNCHANGED, those 836 * writes will be issued as normal writes without the flag set. 837 * This is important to note for drivers that do not explicitly 838 * request a WRITE permission for their children and instead take 839 * the same permissions as their parent did (this is commonly what 840 * block filters do). Such drivers have to be aware that the 841 * parent may have taken a WRITE_UNCHANGED permission only and is 842 * issuing such requests. Drivers either must make sure that 843 * these requests do not result in plain WRITE accesses (usually 844 * by supporting BDRV_REQ_WRITE_UNCHANGED, and then forwarding 845 * every incoming write request as-is, including potentially that 846 * flag), or they have to explicitly take the WRITE permission for 847 * their children. */ 848 unsigned int supported_write_flags; 849 /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA, 850 * BDRV_REQ_MAY_UNMAP, BDRV_REQ_WRITE_UNCHANGED) */ 851 unsigned int supported_zero_flags; 852 /* 853 * Flags honoured during truncate (so far: BDRV_REQ_ZERO_WRITE). 854 * 855 * If BDRV_REQ_ZERO_WRITE is given, the truncate operation must make sure 856 * that any added space reads as all zeros. If this can't be guaranteed, 857 * the operation must fail. 858 */ 859 unsigned int supported_truncate_flags; 860 861 /* the following member gives a name to every node on the bs graph. */ 862 char node_name[32]; 863 /* element of the list of named nodes building the graph */ 864 QTAILQ_ENTRY(BlockDriverState) node_list; 865 /* element of the list of all BlockDriverStates (all_bdrv_states) */ 866 QTAILQ_ENTRY(BlockDriverState) bs_list; 867 /* element of the list of monitor-owned BDS */ 868 QTAILQ_ENTRY(BlockDriverState) monitor_list; 869 int refcnt; 870 871 /* operation blockers */ 872 QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX]; 873 874 /* The node that this node inherited default options from (and a reopen on 875 * which can affect this node by changing these defaults). This is always a 876 * parent node of this node. */ 877 BlockDriverState *inherits_from; 878 QLIST_HEAD(, BdrvChild) children; 879 QLIST_HEAD(, BdrvChild) parents; 880 881 QDict *options; 882 QDict *explicit_options; 883 BlockdevDetectZeroesOptions detect_zeroes; 884 885 /* The error object in use for blocking operations on backing_hd */ 886 Error *backing_blocker; 887 888 /* Protected by AioContext lock */ 889 890 /* If we are reading a disk image, give its size in sectors. 891 * Generally read-only; it is written to by load_snapshot and 892 * save_snaphost, but the block layer is quiescent during those. 893 */ 894 int64_t total_sectors; 895 896 /* Callback before write request is processed */ 897 NotifierWithReturnList before_write_notifiers; 898 899 /* threshold limit for writes, in bytes. "High water mark". */ 900 uint64_t write_threshold_offset; 901 NotifierWithReturn write_threshold_notifier; 902 903 /* Writing to the list requires the BQL _and_ the dirty_bitmap_mutex. 904 * Reading from the list can be done with either the BQL or the 905 * dirty_bitmap_mutex. Modifying a bitmap only requires 906 * dirty_bitmap_mutex. */ 907 QemuMutex dirty_bitmap_mutex; 908 QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps; 909 910 /* Offset after the highest byte written to */ 911 Stat64 wr_highest_offset; 912 913 /* If true, copy read backing sectors into image. Can be >1 if more 914 * than one client has requested copy-on-read. Accessed with atomic 915 * ops. 916 */ 917 int copy_on_read; 918 919 /* number of in-flight requests; overall and serialising. 920 * Accessed with atomic ops. 921 */ 922 unsigned int in_flight; 923 unsigned int serialising_in_flight; 924 925 /* counter for nested bdrv_io_plug. 926 * Accessed with atomic ops. 927 */ 928 unsigned io_plugged; 929 930 /* do we need to tell the quest if we have a volatile write cache? */ 931 int enable_write_cache; 932 933 /* Accessed with atomic ops. */ 934 int quiesce_counter; 935 int recursive_quiesce_counter; 936 937 unsigned int write_gen; /* Current data generation */ 938 939 /* Protected by reqs_lock. */ 940 CoMutex reqs_lock; 941 QLIST_HEAD(, BdrvTrackedRequest) tracked_requests; 942 CoQueue flush_queue; /* Serializing flush queue */ 943 bool active_flush_req; /* Flush request in flight? */ 944 945 /* Only read/written by whoever has set active_flush_req to true. */ 946 unsigned int flushed_gen; /* Flushed write generation */ 947 948 /* BdrvChild links to this node may never be frozen */ 949 bool never_freeze; 950 }; 951 952 struct BlockBackendRootState { 953 int open_flags; 954 bool read_only; 955 BlockdevDetectZeroesOptions detect_zeroes; 956 }; 957 958 typedef enum BlockMirrorBackingMode { 959 /* Reuse the existing backing chain from the source for the target. 960 * - sync=full: Set backing BDS to NULL. 961 * - sync=top: Use source's backing BDS. 962 * - sync=none: Use source as the backing BDS. */ 963 MIRROR_SOURCE_BACKING_CHAIN, 964 965 /* Open the target's backing chain completely anew */ 966 MIRROR_OPEN_BACKING_CHAIN, 967 968 /* Do not change the target's backing BDS after job completion */ 969 MIRROR_LEAVE_BACKING_CHAIN, 970 } BlockMirrorBackingMode; 971 972 static inline BlockDriverState *backing_bs(BlockDriverState *bs) 973 { 974 return bs->backing ? bs->backing->bs : NULL; 975 } 976 977 978 /* Essential block drivers which must always be statically linked into qemu, and 979 * which therefore can be accessed without using bdrv_find_format() */ 980 extern BlockDriver bdrv_file; 981 extern BlockDriver bdrv_raw; 982 extern BlockDriver bdrv_qcow2; 983 984 int coroutine_fn bdrv_co_preadv(BdrvChild *child, 985 int64_t offset, unsigned int bytes, QEMUIOVector *qiov, 986 BdrvRequestFlags flags); 987 int coroutine_fn bdrv_co_preadv_part(BdrvChild *child, 988 int64_t offset, unsigned int bytes, 989 QEMUIOVector *qiov, size_t qiov_offset, BdrvRequestFlags flags); 990 int coroutine_fn bdrv_co_pwritev(BdrvChild *child, 991 int64_t offset, unsigned int bytes, QEMUIOVector *qiov, 992 BdrvRequestFlags flags); 993 int coroutine_fn bdrv_co_pwritev_part(BdrvChild *child, 994 int64_t offset, unsigned int bytes, 995 QEMUIOVector *qiov, size_t qiov_offset, BdrvRequestFlags flags); 996 997 static inline int coroutine_fn bdrv_co_pread(BdrvChild *child, 998 int64_t offset, unsigned int bytes, void *buf, BdrvRequestFlags flags) 999 { 1000 QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes); 1001 1002 return bdrv_co_preadv(child, offset, bytes, &qiov, flags); 1003 } 1004 1005 static inline int coroutine_fn bdrv_co_pwrite(BdrvChild *child, 1006 int64_t offset, unsigned int bytes, void *buf, BdrvRequestFlags flags) 1007 { 1008 QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes); 1009 1010 return bdrv_co_pwritev(child, offset, bytes, &qiov, flags); 1011 } 1012 1013 extern unsigned int bdrv_drain_all_count; 1014 void bdrv_apply_subtree_drain(BdrvChild *child, BlockDriverState *new_parent); 1015 void bdrv_unapply_subtree_drain(BdrvChild *child, BlockDriverState *old_parent); 1016 1017 bool coroutine_fn bdrv_mark_request_serialising(BdrvTrackedRequest *req, uint64_t align); 1018 BdrvTrackedRequest *coroutine_fn bdrv_co_get_self_request(BlockDriverState *bs); 1019 1020 int get_tmp_filename(char *filename, int size); 1021 BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size, 1022 const char *filename); 1023 1024 void bdrv_parse_filename_strip_prefix(const char *filename, const char *prefix, 1025 QDict *options); 1026 1027 1028 /** 1029 * bdrv_add_before_write_notifier: 1030 * 1031 * Register a callback that is invoked before write requests are processed but 1032 * after any throttling or waiting for overlapping requests. 1033 */ 1034 void bdrv_add_before_write_notifier(BlockDriverState *bs, 1035 NotifierWithReturn *notifier); 1036 1037 /** 1038 * bdrv_add_aio_context_notifier: 1039 * 1040 * If a long-running job intends to be always run in the same AioContext as a 1041 * certain BDS, it may use this function to be notified of changes regarding the 1042 * association of the BDS to an AioContext. 1043 * 1044 * attached_aio_context() is called after the target BDS has been attached to a 1045 * new AioContext; detach_aio_context() is called before the target BDS is being 1046 * detached from its old AioContext. 1047 */ 1048 void bdrv_add_aio_context_notifier(BlockDriverState *bs, 1049 void (*attached_aio_context)(AioContext *new_context, void *opaque), 1050 void (*detach_aio_context)(void *opaque), void *opaque); 1051 1052 /** 1053 * bdrv_remove_aio_context_notifier: 1054 * 1055 * Unsubscribe of change notifications regarding the BDS's AioContext. The 1056 * parameters given here have to be the same as those given to 1057 * bdrv_add_aio_context_notifier(). 1058 */ 1059 void bdrv_remove_aio_context_notifier(BlockDriverState *bs, 1060 void (*aio_context_attached)(AioContext *, 1061 void *), 1062 void (*aio_context_detached)(void *), 1063 void *opaque); 1064 1065 /** 1066 * bdrv_wakeup: 1067 * @bs: The BlockDriverState for which an I/O operation has been completed. 1068 * 1069 * Wake up the main thread if it is waiting on BDRV_POLL_WHILE. During 1070 * synchronous I/O on a BlockDriverState that is attached to another 1071 * I/O thread, the main thread lets the I/O thread's event loop run, 1072 * waiting for the I/O operation to complete. A bdrv_wakeup will wake 1073 * up the main thread if necessary. 1074 * 1075 * Manual calls to bdrv_wakeup are rarely necessary, because 1076 * bdrv_dec_in_flight already calls it. 1077 */ 1078 void bdrv_wakeup(BlockDriverState *bs); 1079 1080 #ifdef _WIN32 1081 int is_windows_drive(const char *filename); 1082 #endif 1083 1084 /** 1085 * stream_start: 1086 * @job_id: The id of the newly-created job, or %NULL to use the 1087 * device name of @bs. 1088 * @bs: Block device to operate on. 1089 * @base: Block device that will become the new base, or %NULL to 1090 * flatten the whole backing file chain onto @bs. 1091 * @backing_file_str: The file name that will be written to @bs as the 1092 * the new backing file if the job completes. Ignored if @base is %NULL. 1093 * @creation_flags: Flags that control the behavior of the Job lifetime. 1094 * See @BlockJobCreateFlags 1095 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 1096 * @on_error: The action to take upon error. 1097 * @errp: Error object. 1098 * 1099 * Start a streaming operation on @bs. Clusters that are unallocated 1100 * in @bs, but allocated in any image between @base and @bs (both 1101 * exclusive) will be written to @bs. At the end of a successful 1102 * streaming job, the backing file of @bs will be changed to 1103 * @backing_file_str in the written image and to @base in the live 1104 * BlockDriverState. 1105 */ 1106 void stream_start(const char *job_id, BlockDriverState *bs, 1107 BlockDriverState *base, const char *backing_file_str, 1108 int creation_flags, int64_t speed, 1109 BlockdevOnError on_error, Error **errp); 1110 1111 /** 1112 * commit_start: 1113 * @job_id: The id of the newly-created job, or %NULL to use the 1114 * device name of @bs. 1115 * @bs: Active block device. 1116 * @top: Top block device to be committed. 1117 * @base: Block device that will be written into, and become the new top. 1118 * @creation_flags: Flags that control the behavior of the Job lifetime. 1119 * See @BlockJobCreateFlags 1120 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 1121 * @on_error: The action to take upon error. 1122 * @backing_file_str: String to use as the backing file in @top's overlay 1123 * @filter_node_name: The node name that should be assigned to the filter 1124 * driver that the commit job inserts into the graph above @top. NULL means 1125 * that a node name should be autogenerated. 1126 * @errp: Error object. 1127 * 1128 */ 1129 void commit_start(const char *job_id, BlockDriverState *bs, 1130 BlockDriverState *base, BlockDriverState *top, 1131 int creation_flags, int64_t speed, 1132 BlockdevOnError on_error, const char *backing_file_str, 1133 const char *filter_node_name, Error **errp); 1134 /** 1135 * commit_active_start: 1136 * @job_id: The id of the newly-created job, or %NULL to use the 1137 * device name of @bs. 1138 * @bs: Active block device to be committed. 1139 * @base: Block device that will be written into, and become the new top. 1140 * @creation_flags: Flags that control the behavior of the Job lifetime. 1141 * See @BlockJobCreateFlags 1142 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 1143 * @on_error: The action to take upon error. 1144 * @filter_node_name: The node name that should be assigned to the filter 1145 * driver that the commit job inserts into the graph above @bs. NULL means that 1146 * a node name should be autogenerated. 1147 * @cb: Completion function for the job. 1148 * @opaque: Opaque pointer value passed to @cb. 1149 * @auto_complete: Auto complete the job. 1150 * @errp: Error object. 1151 * 1152 */ 1153 BlockJob *commit_active_start(const char *job_id, BlockDriverState *bs, 1154 BlockDriverState *base, int creation_flags, 1155 int64_t speed, BlockdevOnError on_error, 1156 const char *filter_node_name, 1157 BlockCompletionFunc *cb, void *opaque, 1158 bool auto_complete, Error **errp); 1159 /* 1160 * mirror_start: 1161 * @job_id: The id of the newly-created job, or %NULL to use the 1162 * device name of @bs. 1163 * @bs: Block device to operate on. 1164 * @target: Block device to write to. 1165 * @replaces: Block graph node name to replace once the mirror is done. Can 1166 * only be used when full mirroring is selected. 1167 * @creation_flags: Flags that control the behavior of the Job lifetime. 1168 * See @BlockJobCreateFlags 1169 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 1170 * @granularity: The chosen granularity for the dirty bitmap. 1171 * @buf_size: The amount of data that can be in flight at one time. 1172 * @mode: Whether to collapse all images in the chain to the target. 1173 * @backing_mode: How to establish the target's backing chain after completion. 1174 * @zero_target: Whether the target should be explicitly zero-initialized 1175 * @on_source_error: The action to take upon error reading from the source. 1176 * @on_target_error: The action to take upon error writing to the target. 1177 * @unmap: Whether to unmap target where source sectors only contain zeroes. 1178 * @filter_node_name: The node name that should be assigned to the filter 1179 * driver that the mirror job inserts into the graph above @bs. NULL means that 1180 * a node name should be autogenerated. 1181 * @copy_mode: When to trigger writes to the target. 1182 * @errp: Error object. 1183 * 1184 * Start a mirroring operation on @bs. Clusters that are allocated 1185 * in @bs will be written to @target until the job is cancelled or 1186 * manually completed. At the end of a successful mirroring job, 1187 * @bs will be switched to read from @target. 1188 */ 1189 void mirror_start(const char *job_id, BlockDriverState *bs, 1190 BlockDriverState *target, const char *replaces, 1191 int creation_flags, int64_t speed, 1192 uint32_t granularity, int64_t buf_size, 1193 MirrorSyncMode mode, BlockMirrorBackingMode backing_mode, 1194 bool zero_target, 1195 BlockdevOnError on_source_error, 1196 BlockdevOnError on_target_error, 1197 bool unmap, const char *filter_node_name, 1198 MirrorCopyMode copy_mode, Error **errp); 1199 1200 /* 1201 * backup_job_create: 1202 * @job_id: The id of the newly-created job, or %NULL to use the 1203 * device name of @bs. 1204 * @bs: Block device to operate on. 1205 * @target: Block device to write to. 1206 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 1207 * @sync_mode: What parts of the disk image should be copied to the destination. 1208 * @sync_bitmap: The dirty bitmap if sync_mode is 'bitmap' or 'incremental' 1209 * @bitmap_mode: The bitmap synchronization policy to use. 1210 * @on_source_error: The action to take upon error reading from the source. 1211 * @on_target_error: The action to take upon error writing to the target. 1212 * @creation_flags: Flags that control the behavior of the Job lifetime. 1213 * See @BlockJobCreateFlags 1214 * @cb: Completion function for the job. 1215 * @opaque: Opaque pointer value passed to @cb. 1216 * @txn: Transaction that this job is part of (may be NULL). 1217 * 1218 * Create a backup operation on @bs. Clusters in @bs are written to @target 1219 * until the job is cancelled or manually completed. 1220 */ 1221 BlockJob *backup_job_create(const char *job_id, BlockDriverState *bs, 1222 BlockDriverState *target, int64_t speed, 1223 MirrorSyncMode sync_mode, 1224 BdrvDirtyBitmap *sync_bitmap, 1225 BitmapSyncMode bitmap_mode, 1226 bool compress, 1227 const char *filter_node_name, 1228 BlockdevOnError on_source_error, 1229 BlockdevOnError on_target_error, 1230 int creation_flags, 1231 BlockCompletionFunc *cb, void *opaque, 1232 JobTxn *txn, Error **errp); 1233 1234 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs, 1235 const char *child_name, 1236 const BdrvChildClass *child_class, 1237 BdrvChildRole child_role, 1238 AioContext *ctx, 1239 uint64_t perm, uint64_t shared_perm, 1240 void *opaque, Error **errp); 1241 void bdrv_root_unref_child(BdrvChild *child); 1242 1243 void bdrv_get_cumulative_perm(BlockDriverState *bs, uint64_t *perm, 1244 uint64_t *shared_perm); 1245 1246 /** 1247 * Sets a BdrvChild's permissions. Avoid if the parent is a BDS; use 1248 * bdrv_child_refresh_perms() instead and make the parent's 1249 * .bdrv_child_perm() implementation return the correct values. 1250 */ 1251 int bdrv_child_try_set_perm(BdrvChild *c, uint64_t perm, uint64_t shared, 1252 Error **errp); 1253 1254 /** 1255 * Calls bs->drv->bdrv_child_perm() and updates the child's permission 1256 * masks with the result. 1257 * Drivers should invoke this function whenever an event occurs that 1258 * makes their .bdrv_child_perm() implementation return different 1259 * values than before, but which will not result in the block layer 1260 * automatically refreshing the permissions. 1261 */ 1262 int bdrv_child_refresh_perms(BlockDriverState *bs, BdrvChild *c, Error **errp); 1263 1264 bool bdrv_recurse_can_replace(BlockDriverState *bs, 1265 BlockDriverState *to_replace); 1266 1267 /* 1268 * Default implementation for BlockDriver.bdrv_child_perm() that can 1269 * be used by block filters and image formats, as long as they use the 1270 * child_of_bds child class and set an appropriate BdrvChildRole. 1271 */ 1272 void bdrv_default_perms(BlockDriverState *bs, BdrvChild *c, 1273 BdrvChildRole role, BlockReopenQueue *reopen_queue, 1274 uint64_t perm, uint64_t shared, 1275 uint64_t *nperm, uint64_t *nshared); 1276 1277 /* 1278 * Default implementation for drivers to pass bdrv_co_block_status() to 1279 * their file. 1280 */ 1281 int coroutine_fn bdrv_co_block_status_from_file(BlockDriverState *bs, 1282 bool want_zero, 1283 int64_t offset, 1284 int64_t bytes, 1285 int64_t *pnum, 1286 int64_t *map, 1287 BlockDriverState **file); 1288 /* 1289 * Default implementation for drivers to pass bdrv_co_block_status() to 1290 * their backing file. 1291 */ 1292 int coroutine_fn bdrv_co_block_status_from_backing(BlockDriverState *bs, 1293 bool want_zero, 1294 int64_t offset, 1295 int64_t bytes, 1296 int64_t *pnum, 1297 int64_t *map, 1298 BlockDriverState **file); 1299 const char *bdrv_get_parent_name(const BlockDriverState *bs); 1300 void blk_dev_change_media_cb(BlockBackend *blk, bool load, Error **errp); 1301 bool blk_dev_has_removable_media(BlockBackend *blk); 1302 bool blk_dev_has_tray(BlockBackend *blk); 1303 void blk_dev_eject_request(BlockBackend *blk, bool force); 1304 bool blk_dev_is_tray_open(BlockBackend *blk); 1305 bool blk_dev_is_medium_locked(BlockBackend *blk); 1306 1307 void bdrv_set_dirty(BlockDriverState *bs, int64_t offset, int64_t bytes); 1308 1309 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap **out); 1310 void bdrv_restore_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap *backup); 1311 bool bdrv_dirty_bitmap_merge_internal(BdrvDirtyBitmap *dest, 1312 const BdrvDirtyBitmap *src, 1313 HBitmap **backup, bool lock); 1314 1315 void bdrv_inc_in_flight(BlockDriverState *bs); 1316 void bdrv_dec_in_flight(BlockDriverState *bs); 1317 1318 void blockdev_close_all_bdrv_states(void); 1319 1320 int coroutine_fn bdrv_co_copy_range_from(BdrvChild *src, uint64_t src_offset, 1321 BdrvChild *dst, uint64_t dst_offset, 1322 uint64_t bytes, 1323 BdrvRequestFlags read_flags, 1324 BdrvRequestFlags write_flags); 1325 int coroutine_fn bdrv_co_copy_range_to(BdrvChild *src, uint64_t src_offset, 1326 BdrvChild *dst, uint64_t dst_offset, 1327 uint64_t bytes, 1328 BdrvRequestFlags read_flags, 1329 BdrvRequestFlags write_flags); 1330 1331 int refresh_total_sectors(BlockDriverState *bs, int64_t hint); 1332 1333 void bdrv_set_monitor_owned(BlockDriverState *bs); 1334 BlockDriverState *bds_tree_init(QDict *bs_opts, Error **errp); 1335 1336 /** 1337 * Simple implementation of bdrv_co_create_opts for protocol drivers 1338 * which only support creation via opening a file 1339 * (usually existing raw storage device) 1340 */ 1341 int coroutine_fn bdrv_co_create_opts_simple(BlockDriver *drv, 1342 const char *filename, 1343 QemuOpts *opts, 1344 Error **errp); 1345 extern QemuOptsList bdrv_create_opts_simple; 1346 1347 BdrvDirtyBitmap *block_dirty_bitmap_lookup(const char *node, 1348 const char *name, 1349 BlockDriverState **pbs, 1350 Error **errp); 1351 BdrvDirtyBitmap *block_dirty_bitmap_merge(const char *node, const char *target, 1352 BlockDirtyBitmapMergeSourceList *bms, 1353 HBitmap **backup, Error **errp); 1354 BdrvDirtyBitmap *block_dirty_bitmap_remove(const char *node, const char *name, 1355 bool release, 1356 BlockDriverState **bitmap_bs, 1357 Error **errp); 1358 1359 #endif /* BLOCK_INT_H */ 1360