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_EXTENT_SIZE_HINT "extent_size_hint" 57 #define BLOCK_OPT_OBJECT_SIZE "object_size" 58 #define BLOCK_OPT_REFCOUNT_BITS "refcount_bits" 59 #define BLOCK_OPT_DATA_FILE "data_file" 60 #define BLOCK_OPT_DATA_FILE_RAW "data_file_raw" 61 #define BLOCK_OPT_COMPRESSION_TYPE "compression_type" 62 #define BLOCK_OPT_EXTL2 "extended_l2" 63 64 #define BLOCK_PROBE_BUF_SIZE 512 65 66 enum BdrvTrackedRequestType { 67 BDRV_TRACKED_READ, 68 BDRV_TRACKED_WRITE, 69 BDRV_TRACKED_DISCARD, 70 BDRV_TRACKED_TRUNCATE, 71 }; 72 73 /* 74 * That is not quite good that BdrvTrackedRequest structure is public, 75 * as block/io.c is very careful about incoming offset/bytes being 76 * correct. Be sure to assert bdrv_check_request() succeeded after any 77 * modification of BdrvTrackedRequest object out of block/io.c 78 */ 79 typedef struct BdrvTrackedRequest { 80 BlockDriverState *bs; 81 int64_t offset; 82 int64_t bytes; 83 enum BdrvTrackedRequestType type; 84 85 bool serialising; 86 int64_t overlap_offset; 87 int64_t overlap_bytes; 88 89 QLIST_ENTRY(BdrvTrackedRequest) list; 90 Coroutine *co; /* owner, used for deadlock detection */ 91 CoQueue wait_queue; /* coroutines blocked on this request */ 92 93 struct BdrvTrackedRequest *waiting_for; 94 } BdrvTrackedRequest; 95 96 int bdrv_check_request(int64_t offset, int64_t bytes, Error **errp); 97 98 struct BlockDriver { 99 const char *format_name; 100 int instance_size; 101 102 /* set to true if the BlockDriver is a block filter. Block filters pass 103 * certain callbacks that refer to data (see block.c) to their bs->file 104 * or bs->backing (whichever one exists) if the driver doesn't implement 105 * them. Drivers that do not wish to forward must implement them and return 106 * -ENOTSUP. 107 * Note that filters are not allowed to modify data. 108 * 109 * Filters generally cannot have more than a single filtered child, 110 * because the data they present must at all times be the same as 111 * that on their filtered child. That would be impossible to 112 * achieve for multiple filtered children. 113 * (And this filtered child must then be bs->file or bs->backing.) 114 */ 115 bool is_filter; 116 /* 117 * Set to true if the BlockDriver is a format driver. Format nodes 118 * generally do not expect their children to be other format nodes 119 * (except for backing files), and so format probing is disabled 120 * on those children. 121 */ 122 bool is_format; 123 /* 124 * Return true if @to_replace can be replaced by a BDS with the 125 * same data as @bs without it affecting @bs's behavior (that is, 126 * without it being visible to @bs's parents). 127 */ 128 bool (*bdrv_recurse_can_replace)(BlockDriverState *bs, 129 BlockDriverState *to_replace); 130 131 int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename); 132 int (*bdrv_probe_device)(const char *filename); 133 134 /* Any driver implementing this callback is expected to be able to handle 135 * NULL file names in its .bdrv_open() implementation */ 136 void (*bdrv_parse_filename)(const char *filename, QDict *options, Error **errp); 137 /* Drivers not implementing bdrv_parse_filename nor bdrv_open should have 138 * this field set to true, except ones that are defined only by their 139 * child's bs. 140 * An example of the last type will be the quorum block driver. 141 */ 142 bool bdrv_needs_filename; 143 144 /* 145 * Set if a driver can support backing files. This also implies the 146 * following semantics: 147 * 148 * - Return status 0 of .bdrv_co_block_status means that corresponding 149 * blocks are not allocated in this layer of backing-chain 150 * - For such (unallocated) blocks, read will: 151 * - fill buffer with zeros if there is no backing file 152 * - read from the backing file otherwise, where the block layer 153 * takes care of reading zeros beyond EOF if backing file is short 154 */ 155 bool supports_backing; 156 157 /* For handling image reopen for split or non-split files */ 158 int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state, 159 BlockReopenQueue *queue, Error **errp); 160 void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state); 161 void (*bdrv_reopen_commit_post)(BDRVReopenState *reopen_state); 162 void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state); 163 void (*bdrv_join_options)(QDict *options, QDict *old_options); 164 165 int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags, 166 Error **errp); 167 168 /* Protocol drivers should implement this instead of bdrv_open */ 169 int (*bdrv_file_open)(BlockDriverState *bs, QDict *options, int flags, 170 Error **errp); 171 void (*bdrv_close)(BlockDriverState *bs); 172 173 174 int coroutine_fn (*bdrv_co_create)(BlockdevCreateOptions *opts, 175 Error **errp); 176 int coroutine_fn (*bdrv_co_create_opts)(BlockDriver *drv, 177 const char *filename, 178 QemuOpts *opts, 179 Error **errp); 180 181 int coroutine_fn (*bdrv_co_amend)(BlockDriverState *bs, 182 BlockdevAmendOptions *opts, 183 bool force, 184 Error **errp); 185 186 int (*bdrv_amend_options)(BlockDriverState *bs, 187 QemuOpts *opts, 188 BlockDriverAmendStatusCB *status_cb, 189 void *cb_opaque, 190 bool force, 191 Error **errp); 192 193 int (*bdrv_make_empty)(BlockDriverState *bs); 194 195 /* 196 * Refreshes the bs->exact_filename field. If that is impossible, 197 * bs->exact_filename has to be left empty. 198 */ 199 void (*bdrv_refresh_filename)(BlockDriverState *bs); 200 201 /* 202 * Gathers the open options for all children into @target. 203 * A simple format driver (without backing file support) might 204 * implement this function like this: 205 * 206 * QINCREF(bs->file->bs->full_open_options); 207 * qdict_put(target, "file", bs->file->bs->full_open_options); 208 * 209 * If not specified, the generic implementation will simply put 210 * all children's options under their respective name. 211 * 212 * @backing_overridden is true when bs->backing seems not to be 213 * the child that would result from opening bs->backing_file. 214 * Therefore, if it is true, the backing child's options should be 215 * gathered; otherwise, there is no need since the backing child 216 * is the one implied by the image header. 217 * 218 * Note that ideally this function would not be needed. Every 219 * block driver which implements it is probably doing something 220 * shady regarding its runtime option structure. 221 */ 222 void (*bdrv_gather_child_options)(BlockDriverState *bs, QDict *target, 223 bool backing_overridden); 224 225 /* 226 * Returns an allocated string which is the directory name of this BDS: It 227 * will be used to make relative filenames absolute by prepending this 228 * function's return value to them. 229 */ 230 char *(*bdrv_dirname)(BlockDriverState *bs, Error **errp); 231 232 /* aio */ 233 BlockAIOCB *(*bdrv_aio_preadv)(BlockDriverState *bs, 234 uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags, 235 BlockCompletionFunc *cb, void *opaque); 236 BlockAIOCB *(*bdrv_aio_pwritev)(BlockDriverState *bs, 237 uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags, 238 BlockCompletionFunc *cb, void *opaque); 239 BlockAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs, 240 BlockCompletionFunc *cb, void *opaque); 241 BlockAIOCB *(*bdrv_aio_pdiscard)(BlockDriverState *bs, 242 int64_t offset, int bytes, 243 BlockCompletionFunc *cb, void *opaque); 244 245 int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs, 246 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov); 247 248 /** 249 * @offset: position in bytes to read at 250 * @bytes: number of bytes to read 251 * @qiov: the buffers to fill with read data 252 * @flags: currently unused, always 0 253 * 254 * @offset and @bytes will be a multiple of 'request_alignment', 255 * but the length of individual @qiov elements does not have to 256 * be a multiple. 257 * 258 * @bytes will always equal the total size of @qiov, and will be 259 * no larger than 'max_transfer'. 260 * 261 * The buffer in @qiov may point directly to guest memory. 262 */ 263 int coroutine_fn (*bdrv_co_preadv)(BlockDriverState *bs, 264 uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags); 265 int coroutine_fn (*bdrv_co_preadv_part)(BlockDriverState *bs, 266 uint64_t offset, uint64_t bytes, 267 QEMUIOVector *qiov, size_t qiov_offset, int flags); 268 int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs, 269 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, int flags); 270 /** 271 * @offset: position in bytes to write at 272 * @bytes: number of bytes to write 273 * @qiov: the buffers containing data to write 274 * @flags: zero or more bits allowed by 'supported_write_flags' 275 * 276 * @offset and @bytes will be a multiple of 'request_alignment', 277 * but the length of individual @qiov elements does not have to 278 * be a multiple. 279 * 280 * @bytes will always equal the total size of @qiov, and will be 281 * no larger than 'max_transfer'. 282 * 283 * The buffer in @qiov may point directly to guest memory. 284 */ 285 int coroutine_fn (*bdrv_co_pwritev)(BlockDriverState *bs, 286 uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags); 287 int coroutine_fn (*bdrv_co_pwritev_part)(BlockDriverState *bs, 288 uint64_t offset, uint64_t bytes, 289 QEMUIOVector *qiov, size_t qiov_offset, int flags); 290 291 /* 292 * Efficiently zero a region of the disk image. Typically an image format 293 * would use a compact metadata representation to implement this. This 294 * function pointer may be NULL or return -ENOSUP and .bdrv_co_writev() 295 * will be called instead. 296 */ 297 int coroutine_fn (*bdrv_co_pwrite_zeroes)(BlockDriverState *bs, 298 int64_t offset, int bytes, BdrvRequestFlags flags); 299 int coroutine_fn (*bdrv_co_pdiscard)(BlockDriverState *bs, 300 int64_t offset, int bytes); 301 302 /* Map [offset, offset + nbytes) range onto a child of @bs to copy from, 303 * and invoke bdrv_co_copy_range_from(child, ...), or invoke 304 * bdrv_co_copy_range_to() if @bs is the leaf child to copy data from. 305 * 306 * See the comment of bdrv_co_copy_range for the parameter and return value 307 * semantics. 308 */ 309 int coroutine_fn (*bdrv_co_copy_range_from)(BlockDriverState *bs, 310 BdrvChild *src, 311 uint64_t offset, 312 BdrvChild *dst, 313 uint64_t dst_offset, 314 uint64_t bytes, 315 BdrvRequestFlags read_flags, 316 BdrvRequestFlags write_flags); 317 318 /* Map [offset, offset + nbytes) range onto a child of bs to copy data to, 319 * and invoke bdrv_co_copy_range_to(child, src, ...), or perform the copy 320 * operation if @bs is the leaf and @src has the same BlockDriver. Return 321 * -ENOTSUP if @bs is the leaf but @src has a different BlockDriver. 322 * 323 * See the comment of bdrv_co_copy_range for the parameter and return value 324 * semantics. 325 */ 326 int coroutine_fn (*bdrv_co_copy_range_to)(BlockDriverState *bs, 327 BdrvChild *src, 328 uint64_t src_offset, 329 BdrvChild *dst, 330 uint64_t dst_offset, 331 uint64_t bytes, 332 BdrvRequestFlags read_flags, 333 BdrvRequestFlags write_flags); 334 335 /* 336 * Building block for bdrv_block_status[_above] and 337 * bdrv_is_allocated[_above]. The driver should answer only 338 * according to the current layer, and should only need to set 339 * BDRV_BLOCK_DATA, BDRV_BLOCK_ZERO, BDRV_BLOCK_OFFSET_VALID, 340 * and/or BDRV_BLOCK_RAW; if the current layer defers to a backing 341 * layer, the result should be 0 (and not BDRV_BLOCK_ZERO). See 342 * block.h for the overall meaning of the bits. As a hint, the 343 * flag want_zero is true if the caller cares more about precise 344 * mappings (favor accurate _OFFSET_VALID/_ZERO) or false for 345 * overall allocation (favor larger *pnum, perhaps by reporting 346 * _DATA instead of _ZERO). The block layer guarantees input 347 * clamped to bdrv_getlength() and aligned to request_alignment, 348 * as well as non-NULL pnum, map, and file; in turn, the driver 349 * must return an error or set pnum to an aligned non-zero value. 350 */ 351 int coroutine_fn (*bdrv_co_block_status)(BlockDriverState *bs, 352 bool want_zero, int64_t offset, int64_t bytes, int64_t *pnum, 353 int64_t *map, BlockDriverState **file); 354 355 /* 356 * This informs the driver that we are no longer interested in the result 357 * of in-flight requests, so don't waste the time if possible. 358 * 359 * One example usage is to avoid waiting for an nbd target node reconnect 360 * timeout during job-cancel with force=true. 361 */ 362 void (*bdrv_cancel_in_flight)(BlockDriverState *bs); 363 364 /* 365 * Invalidate any cached meta-data. 366 */ 367 void coroutine_fn (*bdrv_co_invalidate_cache)(BlockDriverState *bs, 368 Error **errp); 369 int (*bdrv_inactivate)(BlockDriverState *bs); 370 371 /* 372 * Flushes all data for all layers by calling bdrv_co_flush for underlying 373 * layers, if needed. This function is needed for deterministic 374 * synchronization of the flush finishing callback. 375 */ 376 int coroutine_fn (*bdrv_co_flush)(BlockDriverState *bs); 377 378 /* Delete a created file. */ 379 int coroutine_fn (*bdrv_co_delete_file)(BlockDriverState *bs, 380 Error **errp); 381 382 /* 383 * Flushes all data that was already written to the OS all the way down to 384 * the disk (for example file-posix.c calls fsync()). 385 */ 386 int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs); 387 388 /* 389 * Flushes all internal caches to the OS. The data may still sit in a 390 * writeback cache of the host OS, but it will survive a crash of the qemu 391 * process. 392 */ 393 int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs); 394 395 /* 396 * Drivers setting this field must be able to work with just a plain 397 * filename with '<protocol_name>:' as a prefix, and no other options. 398 * Options may be extracted from the filename by implementing 399 * bdrv_parse_filename. 400 */ 401 const char *protocol_name; 402 403 /* 404 * Truncate @bs to @offset bytes using the given @prealloc mode 405 * when growing. Modes other than PREALLOC_MODE_OFF should be 406 * rejected when shrinking @bs. 407 * 408 * If @exact is true, @bs must be resized to exactly @offset. 409 * Otherwise, it is sufficient for @bs (if it is a host block 410 * device and thus there is no way to resize it) to be at least 411 * @offset bytes in length. 412 * 413 * If @exact is true and this function fails but would succeed 414 * with @exact = false, it should return -ENOTSUP. 415 */ 416 int coroutine_fn (*bdrv_co_truncate)(BlockDriverState *bs, int64_t offset, 417 bool exact, PreallocMode prealloc, 418 BdrvRequestFlags flags, Error **errp); 419 420 int64_t (*bdrv_getlength)(BlockDriverState *bs); 421 bool has_variable_length; 422 int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs); 423 BlockMeasureInfo *(*bdrv_measure)(QemuOpts *opts, BlockDriverState *in_bs, 424 Error **errp); 425 426 int coroutine_fn (*bdrv_co_pwritev_compressed)(BlockDriverState *bs, 427 uint64_t offset, uint64_t bytes, QEMUIOVector *qiov); 428 int coroutine_fn (*bdrv_co_pwritev_compressed_part)(BlockDriverState *bs, 429 uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, 430 size_t qiov_offset); 431 432 int (*bdrv_snapshot_create)(BlockDriverState *bs, 433 QEMUSnapshotInfo *sn_info); 434 int (*bdrv_snapshot_goto)(BlockDriverState *bs, 435 const char *snapshot_id); 436 int (*bdrv_snapshot_delete)(BlockDriverState *bs, 437 const char *snapshot_id, 438 const char *name, 439 Error **errp); 440 int (*bdrv_snapshot_list)(BlockDriverState *bs, 441 QEMUSnapshotInfo **psn_info); 442 int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs, 443 const char *snapshot_id, 444 const char *name, 445 Error **errp); 446 int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi); 447 ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs, 448 Error **errp); 449 BlockStatsSpecific *(*bdrv_get_specific_stats)(BlockDriverState *bs); 450 451 int coroutine_fn (*bdrv_save_vmstate)(BlockDriverState *bs, 452 QEMUIOVector *qiov, 453 int64_t pos); 454 int coroutine_fn (*bdrv_load_vmstate)(BlockDriverState *bs, 455 QEMUIOVector *qiov, 456 int64_t pos); 457 458 int (*bdrv_change_backing_file)(BlockDriverState *bs, 459 const char *backing_file, const char *backing_fmt); 460 461 /* removable device specific */ 462 bool (*bdrv_is_inserted)(BlockDriverState *bs); 463 void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag); 464 void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked); 465 466 /* to control generic scsi devices */ 467 BlockAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs, 468 unsigned long int req, void *buf, 469 BlockCompletionFunc *cb, void *opaque); 470 int coroutine_fn (*bdrv_co_ioctl)(BlockDriverState *bs, 471 unsigned long int req, void *buf); 472 473 /* List of options for creating images, terminated by name == NULL */ 474 QemuOptsList *create_opts; 475 476 /* List of options for image amend */ 477 QemuOptsList *amend_opts; 478 479 /* 480 * If this driver supports reopening images this contains a 481 * NULL-terminated list of the runtime options that can be 482 * modified. If an option in this list is unspecified during 483 * reopen then it _must_ be reset to its default value or return 484 * an error. 485 */ 486 const char *const *mutable_opts; 487 488 /* 489 * Returns 0 for completed check, -errno for internal errors. 490 * The check results are stored in result. 491 */ 492 int coroutine_fn (*bdrv_co_check)(BlockDriverState *bs, 493 BdrvCheckResult *result, 494 BdrvCheckMode fix); 495 496 void (*bdrv_debug_event)(BlockDriverState *bs, BlkdebugEvent event); 497 498 /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */ 499 int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event, 500 const char *tag); 501 int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs, 502 const char *tag); 503 int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag); 504 bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag); 505 506 void (*bdrv_refresh_limits)(BlockDriverState *bs, Error **errp); 507 508 /* 509 * Returns 1 if newly created images are guaranteed to contain only 510 * zeros, 0 otherwise. 511 */ 512 int (*bdrv_has_zero_init)(BlockDriverState *bs); 513 514 /* Remove fd handlers, timers, and other event loop callbacks so the event 515 * loop is no longer in use. Called with no in-flight requests and in 516 * depth-first traversal order with parents before child nodes. 517 */ 518 void (*bdrv_detach_aio_context)(BlockDriverState *bs); 519 520 /* Add fd handlers, timers, and other event loop callbacks so I/O requests 521 * can be processed again. Called with no in-flight requests and in 522 * depth-first traversal order with child nodes before parent nodes. 523 */ 524 void (*bdrv_attach_aio_context)(BlockDriverState *bs, 525 AioContext *new_context); 526 527 /* io queue for linux-aio */ 528 void (*bdrv_io_plug)(BlockDriverState *bs); 529 void (*bdrv_io_unplug)(BlockDriverState *bs); 530 531 /** 532 * Try to get @bs's logical and physical block size. 533 * On success, store them in @bsz and return zero. 534 * On failure, return negative errno. 535 */ 536 int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz); 537 /** 538 * Try to get @bs's geometry (cyls, heads, sectors) 539 * On success, store them in @geo and return 0. 540 * On failure return -errno. 541 * Only drivers that want to override guest geometry implement this 542 * callback; see hd_geometry_guess(). 543 */ 544 int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo); 545 546 /** 547 * bdrv_co_drain_begin is called if implemented in the beginning of a 548 * drain operation to drain and stop any internal sources of requests in 549 * the driver. 550 * bdrv_co_drain_end is called if implemented at the end of the drain. 551 * 552 * They should be used by the driver to e.g. manage scheduled I/O 553 * requests, or toggle an internal state. After the end of the drain new 554 * requests will continue normally. 555 */ 556 void coroutine_fn (*bdrv_co_drain_begin)(BlockDriverState *bs); 557 void coroutine_fn (*bdrv_co_drain_end)(BlockDriverState *bs); 558 559 void (*bdrv_add_child)(BlockDriverState *parent, BlockDriverState *child, 560 Error **errp); 561 void (*bdrv_del_child)(BlockDriverState *parent, BdrvChild *child, 562 Error **errp); 563 564 /** 565 * Informs the block driver that a permission change is intended. The 566 * driver checks whether the change is permissible and may take other 567 * preparations for the change (e.g. get file system locks). This operation 568 * is always followed either by a call to either .bdrv_set_perm or 569 * .bdrv_abort_perm_update. 570 * 571 * Checks whether the requested set of cumulative permissions in @perm 572 * can be granted for accessing @bs and whether no other users are using 573 * permissions other than those given in @shared (both arguments take 574 * BLK_PERM_* bitmasks). 575 * 576 * If both conditions are met, 0 is returned. Otherwise, -errno is returned 577 * and errp is set to an error describing the conflict. 578 */ 579 int (*bdrv_check_perm)(BlockDriverState *bs, uint64_t perm, 580 uint64_t shared, Error **errp); 581 582 /** 583 * Called to inform the driver that the set of cumulative set of used 584 * permissions for @bs has changed to @perm, and the set of sharable 585 * permission to @shared. The driver can use this to propagate changes to 586 * its children (i.e. request permissions only if a parent actually needs 587 * them). 588 * 589 * This function is only invoked after bdrv_check_perm(), so block drivers 590 * may rely on preparations made in their .bdrv_check_perm implementation. 591 */ 592 void (*bdrv_set_perm)(BlockDriverState *bs, uint64_t perm, uint64_t shared); 593 594 /* 595 * Called to inform the driver that after a previous bdrv_check_perm() 596 * call, the permission update is not performed and any preparations made 597 * for it (e.g. taken file locks) need to be undone. 598 * 599 * This function can be called even for nodes that never saw a 600 * bdrv_check_perm() call. It is a no-op then. 601 */ 602 void (*bdrv_abort_perm_update)(BlockDriverState *bs); 603 604 /** 605 * Returns in @nperm and @nshared the permissions that the driver for @bs 606 * needs on its child @c, based on the cumulative permissions requested by 607 * the parents in @parent_perm and @parent_shared. 608 * 609 * If @c is NULL, return the permissions for attaching a new child for the 610 * given @child_class and @role. 611 * 612 * If @reopen_queue is non-NULL, don't return the currently needed 613 * permissions, but those that will be needed after applying the 614 * @reopen_queue. 615 */ 616 void (*bdrv_child_perm)(BlockDriverState *bs, BdrvChild *c, 617 BdrvChildRole role, 618 BlockReopenQueue *reopen_queue, 619 uint64_t parent_perm, uint64_t parent_shared, 620 uint64_t *nperm, uint64_t *nshared); 621 622 bool (*bdrv_supports_persistent_dirty_bitmap)(BlockDriverState *bs); 623 bool (*bdrv_co_can_store_new_dirty_bitmap)(BlockDriverState *bs, 624 const char *name, 625 uint32_t granularity, 626 Error **errp); 627 int (*bdrv_co_remove_persistent_dirty_bitmap)(BlockDriverState *bs, 628 const char *name, 629 Error **errp); 630 631 /** 632 * Register/unregister a buffer for I/O. For example, when the driver is 633 * interested to know the memory areas that will later be used in iovs, so 634 * that it can do IOMMU mapping with VFIO etc., in order to get better 635 * performance. In the case of VFIO drivers, this callback is used to do 636 * DMA mapping for hot buffers. 637 */ 638 void (*bdrv_register_buf)(BlockDriverState *bs, void *host, size_t size); 639 void (*bdrv_unregister_buf)(BlockDriverState *bs, void *host); 640 QLIST_ENTRY(BlockDriver) list; 641 642 /* Pointer to a NULL-terminated array of names of strong options 643 * that can be specified for bdrv_open(). A strong option is one 644 * that changes the data of a BDS. 645 * If this pointer is NULL, the array is considered empty. 646 * "filename" and "driver" are always considered strong. */ 647 const char *const *strong_runtime_opts; 648 }; 649 650 static inline bool block_driver_can_compress(BlockDriver *drv) 651 { 652 return drv->bdrv_co_pwritev_compressed || 653 drv->bdrv_co_pwritev_compressed_part; 654 } 655 656 typedef struct BlockLimits { 657 /* Alignment requirement, in bytes, for offset/length of I/O 658 * requests. Must be a power of 2 less than INT_MAX; defaults to 659 * 1 for drivers with modern byte interfaces, and to 512 660 * otherwise. */ 661 uint32_t request_alignment; 662 663 /* Maximum number of bytes that can be discarded at once (since it 664 * is signed, it must be < 2G, if set). Must be multiple of 665 * pdiscard_alignment, but need not be power of 2. May be 0 if no 666 * inherent 32-bit limit */ 667 int32_t max_pdiscard; 668 669 /* Optimal alignment for discard requests in bytes. A power of 2 670 * is best but not mandatory. Must be a multiple of 671 * bl.request_alignment, and must be less than max_pdiscard if 672 * that is set. May be 0 if bl.request_alignment is good enough */ 673 uint32_t pdiscard_alignment; 674 675 /* Maximum number of bytes that can zeroized at once (since it is 676 * signed, it must be < 2G, if set). Must be multiple of 677 * pwrite_zeroes_alignment. May be 0 if no inherent 32-bit limit */ 678 int32_t max_pwrite_zeroes; 679 680 /* Optimal alignment for write zeroes requests in bytes. A power 681 * of 2 is best but not mandatory. Must be a multiple of 682 * bl.request_alignment, and must be less than max_pwrite_zeroes 683 * if that is set. May be 0 if bl.request_alignment is good 684 * enough */ 685 uint32_t pwrite_zeroes_alignment; 686 687 /* Optimal transfer length in bytes. A power of 2 is best but not 688 * mandatory. Must be a multiple of bl.request_alignment, or 0 if 689 * no preferred size */ 690 uint32_t opt_transfer; 691 692 /* Maximal transfer length in bytes. Need not be power of 2, but 693 * must be multiple of opt_transfer and bl.request_alignment, or 0 694 * for no 32-bit limit. For now, anything larger than INT_MAX is 695 * clamped down. */ 696 uint32_t max_transfer; 697 698 /* Maximal hardware transfer length in bytes. Applies whenever 699 * transfers to the device bypass the kernel I/O scheduler, for 700 * example with SG_IO. If larger than max_transfer or if zero, 701 * blk_get_max_hw_transfer will fall back to max_transfer. 702 */ 703 uint64_t max_hw_transfer; 704 705 /* memory alignment, in bytes so that no bounce buffer is needed */ 706 size_t min_mem_alignment; 707 708 /* memory alignment, in bytes, for bounce buffer */ 709 size_t opt_mem_alignment; 710 711 /* maximum number of iovec elements */ 712 int max_iov; 713 } BlockLimits; 714 715 typedef struct BdrvOpBlocker BdrvOpBlocker; 716 717 typedef struct BdrvAioNotifier { 718 void (*attached_aio_context)(AioContext *new_context, void *opaque); 719 void (*detach_aio_context)(void *opaque); 720 721 void *opaque; 722 bool deleted; 723 724 QLIST_ENTRY(BdrvAioNotifier) list; 725 } BdrvAioNotifier; 726 727 struct BdrvChildClass { 728 /* If true, bdrv_replace_node() doesn't change the node this BdrvChild 729 * points to. */ 730 bool stay_at_node; 731 732 /* If true, the parent is a BlockDriverState and bdrv_next_all_states() 733 * will return it. This information is used for drain_all, where every node 734 * will be drained separately, so the drain only needs to be propagated to 735 * non-BDS parents. */ 736 bool parent_is_bds; 737 738 void (*inherit_options)(BdrvChildRole role, bool parent_is_format, 739 int *child_flags, QDict *child_options, 740 int parent_flags, QDict *parent_options); 741 742 void (*change_media)(BdrvChild *child, bool load); 743 void (*resize)(BdrvChild *child); 744 745 /* Returns a name that is supposedly more useful for human users than the 746 * node name for identifying the node in question (in particular, a BB 747 * name), or NULL if the parent can't provide a better name. */ 748 const char *(*get_name)(BdrvChild *child); 749 750 /* Returns a malloced string that describes the parent of the child for a 751 * human reader. This could be a node-name, BlockBackend name, qdev ID or 752 * QOM path of the device owning the BlockBackend, job type and ID etc. The 753 * caller is responsible for freeing the memory. */ 754 char *(*get_parent_desc)(BdrvChild *child); 755 756 /* 757 * If this pair of functions is implemented, the parent doesn't issue new 758 * requests after returning from .drained_begin() until .drained_end() is 759 * called. 760 * 761 * These functions must not change the graph (and therefore also must not 762 * call aio_poll(), which could change the graph indirectly). 763 * 764 * If drained_end() schedules background operations, it must atomically 765 * increment *drained_end_counter for each such operation and atomically 766 * decrement it once the operation has settled. 767 * 768 * Note that this can be nested. If drained_begin() was called twice, new 769 * I/O is allowed only after drained_end() was called twice, too. 770 */ 771 void (*drained_begin)(BdrvChild *child); 772 void (*drained_end)(BdrvChild *child, int *drained_end_counter); 773 774 /* 775 * Returns whether the parent has pending requests for the child. This 776 * callback is polled after .drained_begin() has been called until all 777 * activity on the child has stopped. 778 */ 779 bool (*drained_poll)(BdrvChild *child); 780 781 /* Notifies the parent that the child has been activated/inactivated (e.g. 782 * when migration is completing) and it can start/stop requesting 783 * permissions and doing I/O on it. */ 784 void (*activate)(BdrvChild *child, Error **errp); 785 int (*inactivate)(BdrvChild *child); 786 787 void (*attach)(BdrvChild *child); 788 void (*detach)(BdrvChild *child); 789 790 /* Notifies the parent that the filename of its child has changed (e.g. 791 * because the direct child was removed from the backing chain), so that it 792 * can update its reference. */ 793 int (*update_filename)(BdrvChild *child, BlockDriverState *new_base, 794 const char *filename, Error **errp); 795 796 bool (*can_set_aio_ctx)(BdrvChild *child, AioContext *ctx, 797 GSList **ignore, Error **errp); 798 void (*set_aio_ctx)(BdrvChild *child, AioContext *ctx, GSList **ignore); 799 800 AioContext *(*get_parent_aio_context)(BdrvChild *child); 801 }; 802 803 extern const BdrvChildClass child_of_bds; 804 805 struct BdrvChild { 806 BlockDriverState *bs; 807 char *name; 808 const BdrvChildClass *klass; 809 BdrvChildRole role; 810 void *opaque; 811 812 /** 813 * Granted permissions for operating on this BdrvChild (BLK_PERM_* bitmask) 814 */ 815 uint64_t perm; 816 817 /** 818 * Permissions that can still be granted to other users of @bs while this 819 * BdrvChild is still attached to it. (BLK_PERM_* bitmask) 820 */ 821 uint64_t shared_perm; 822 823 /* 824 * This link is frozen: the child can neither be replaced nor 825 * detached from the parent. 826 */ 827 bool frozen; 828 829 /* 830 * How many times the parent of this child has been drained 831 * (through klass->drained_*). 832 * Usually, this is equal to bs->quiesce_counter (potentially 833 * reduced by bdrv_drain_all_count). It may differ while the 834 * child is entering or leaving a drained section. 835 */ 836 int parent_quiesce_counter; 837 838 QLIST_ENTRY(BdrvChild) next; 839 QLIST_ENTRY(BdrvChild) next_parent; 840 }; 841 842 /* 843 * Note: the function bdrv_append() copies and swaps contents of 844 * BlockDriverStates, so if you add new fields to this struct, please 845 * inspect bdrv_append() to determine if the new fields need to be 846 * copied as well. 847 */ 848 struct BlockDriverState { 849 /* Protected by big QEMU lock or read-only after opening. No special 850 * locking needed during I/O... 851 */ 852 int open_flags; /* flags used to open the file, re-used for re-open */ 853 bool encrypted; /* if true, the media is encrypted */ 854 bool sg; /* if true, the device is a /dev/sg* */ 855 bool probed; /* if true, format was probed rather than specified */ 856 bool force_share; /* if true, always allow all shared permissions */ 857 bool implicit; /* if true, this filter node was automatically inserted */ 858 859 BlockDriver *drv; /* NULL means no media */ 860 void *opaque; 861 862 AioContext *aio_context; /* event loop used for fd handlers, timers, etc */ 863 /* long-running tasks intended to always use the same AioContext as this 864 * BDS may register themselves in this list to be notified of changes 865 * regarding this BDS's context */ 866 QLIST_HEAD(, BdrvAioNotifier) aio_notifiers; 867 bool walking_aio_notifiers; /* to make removal during iteration safe */ 868 869 char filename[PATH_MAX]; 870 /* 871 * If not empty, this image is a diff in relation to backing_file. 872 * Note that this is the name given in the image header and 873 * therefore may or may not be equal to .backing->bs->filename. 874 * If this field contains a relative path, it is to be resolved 875 * relatively to the overlay's location. 876 */ 877 char backing_file[PATH_MAX]; 878 /* 879 * The backing filename indicated by the image header. Contrary 880 * to backing_file, if we ever open this file, auto_backing_file 881 * is replaced by the resulting BDS's filename (i.e. after a 882 * bdrv_refresh_filename() run). 883 */ 884 char auto_backing_file[PATH_MAX]; 885 char backing_format[16]; /* if non-zero and backing_file exists */ 886 887 QDict *full_open_options; 888 char exact_filename[PATH_MAX]; 889 890 BdrvChild *backing; 891 BdrvChild *file; 892 893 /* I/O Limits */ 894 BlockLimits bl; 895 896 /* 897 * Flags honored during pread 898 */ 899 unsigned int supported_read_flags; 900 /* Flags honored during pwrite (so far: BDRV_REQ_FUA, 901 * BDRV_REQ_WRITE_UNCHANGED). 902 * If a driver does not support BDRV_REQ_WRITE_UNCHANGED, those 903 * writes will be issued as normal writes without the flag set. 904 * This is important to note for drivers that do not explicitly 905 * request a WRITE permission for their children and instead take 906 * the same permissions as their parent did (this is commonly what 907 * block filters do). Such drivers have to be aware that the 908 * parent may have taken a WRITE_UNCHANGED permission only and is 909 * issuing such requests. Drivers either must make sure that 910 * these requests do not result in plain WRITE accesses (usually 911 * by supporting BDRV_REQ_WRITE_UNCHANGED, and then forwarding 912 * every incoming write request as-is, including potentially that 913 * flag), or they have to explicitly take the WRITE permission for 914 * their children. */ 915 unsigned int supported_write_flags; 916 /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA, 917 * BDRV_REQ_MAY_UNMAP, BDRV_REQ_WRITE_UNCHANGED) */ 918 unsigned int supported_zero_flags; 919 /* 920 * Flags honoured during truncate (so far: BDRV_REQ_ZERO_WRITE). 921 * 922 * If BDRV_REQ_ZERO_WRITE is given, the truncate operation must make sure 923 * that any added space reads as all zeros. If this can't be guaranteed, 924 * the operation must fail. 925 */ 926 unsigned int supported_truncate_flags; 927 928 /* the following member gives a name to every node on the bs graph. */ 929 char node_name[32]; 930 /* element of the list of named nodes building the graph */ 931 QTAILQ_ENTRY(BlockDriverState) node_list; 932 /* element of the list of all BlockDriverStates (all_bdrv_states) */ 933 QTAILQ_ENTRY(BlockDriverState) bs_list; 934 /* element of the list of monitor-owned BDS */ 935 QTAILQ_ENTRY(BlockDriverState) monitor_list; 936 int refcnt; 937 938 /* operation blockers */ 939 QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX]; 940 941 /* The node that this node inherited default options from (and a reopen on 942 * which can affect this node by changing these defaults). This is always a 943 * parent node of this node. */ 944 BlockDriverState *inherits_from; 945 QLIST_HEAD(, BdrvChild) children; 946 QLIST_HEAD(, BdrvChild) parents; 947 948 QDict *options; 949 QDict *explicit_options; 950 BlockdevDetectZeroesOptions detect_zeroes; 951 952 /* The error object in use for blocking operations on backing_hd */ 953 Error *backing_blocker; 954 955 /* Protected by AioContext lock */ 956 957 /* If we are reading a disk image, give its size in sectors. 958 * Generally read-only; it is written to by load_snapshot and 959 * save_snaphost, but the block layer is quiescent during those. 960 */ 961 int64_t total_sectors; 962 963 /* threshold limit for writes, in bytes. "High water mark". */ 964 uint64_t write_threshold_offset; 965 966 /* Writing to the list requires the BQL _and_ the dirty_bitmap_mutex. 967 * Reading from the list can be done with either the BQL or the 968 * dirty_bitmap_mutex. Modifying a bitmap only requires 969 * dirty_bitmap_mutex. */ 970 QemuMutex dirty_bitmap_mutex; 971 QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps; 972 973 /* Offset after the highest byte written to */ 974 Stat64 wr_highest_offset; 975 976 /* If true, copy read backing sectors into image. Can be >1 if more 977 * than one client has requested copy-on-read. Accessed with atomic 978 * ops. 979 */ 980 int copy_on_read; 981 982 /* number of in-flight requests; overall and serialising. 983 * Accessed with atomic ops. 984 */ 985 unsigned int in_flight; 986 unsigned int serialising_in_flight; 987 988 /* counter for nested bdrv_io_plug. 989 * Accessed with atomic ops. 990 */ 991 unsigned io_plugged; 992 993 /* do we need to tell the quest if we have a volatile write cache? */ 994 int enable_write_cache; 995 996 /* Accessed with atomic ops. */ 997 int quiesce_counter; 998 int recursive_quiesce_counter; 999 1000 unsigned int write_gen; /* Current data generation */ 1001 1002 /* Protected by reqs_lock. */ 1003 CoMutex reqs_lock; 1004 QLIST_HEAD(, BdrvTrackedRequest) tracked_requests; 1005 CoQueue flush_queue; /* Serializing flush queue */ 1006 bool active_flush_req; /* Flush request in flight? */ 1007 1008 /* Only read/written by whoever has set active_flush_req to true. */ 1009 unsigned int flushed_gen; /* Flushed write generation */ 1010 1011 /* BdrvChild links to this node may never be frozen */ 1012 bool never_freeze; 1013 }; 1014 1015 struct BlockBackendRootState { 1016 int open_flags; 1017 BlockdevDetectZeroesOptions detect_zeroes; 1018 }; 1019 1020 typedef enum BlockMirrorBackingMode { 1021 /* Reuse the existing backing chain from the source for the target. 1022 * - sync=full: Set backing BDS to NULL. 1023 * - sync=top: Use source's backing BDS. 1024 * - sync=none: Use source as the backing BDS. */ 1025 MIRROR_SOURCE_BACKING_CHAIN, 1026 1027 /* Open the target's backing chain completely anew */ 1028 MIRROR_OPEN_BACKING_CHAIN, 1029 1030 /* Do not change the target's backing BDS after job completion */ 1031 MIRROR_LEAVE_BACKING_CHAIN, 1032 } BlockMirrorBackingMode; 1033 1034 1035 /* Essential block drivers which must always be statically linked into qemu, and 1036 * which therefore can be accessed without using bdrv_find_format() */ 1037 extern BlockDriver bdrv_file; 1038 extern BlockDriver bdrv_raw; 1039 extern BlockDriver bdrv_qcow2; 1040 1041 int coroutine_fn bdrv_co_preadv(BdrvChild *child, 1042 int64_t offset, int64_t bytes, QEMUIOVector *qiov, 1043 BdrvRequestFlags flags); 1044 int coroutine_fn bdrv_co_preadv_part(BdrvChild *child, 1045 int64_t offset, int64_t bytes, 1046 QEMUIOVector *qiov, size_t qiov_offset, BdrvRequestFlags flags); 1047 int coroutine_fn bdrv_co_pwritev(BdrvChild *child, 1048 int64_t offset, int64_t bytes, QEMUIOVector *qiov, 1049 BdrvRequestFlags flags); 1050 int coroutine_fn bdrv_co_pwritev_part(BdrvChild *child, 1051 int64_t offset, int64_t bytes, 1052 QEMUIOVector *qiov, size_t qiov_offset, BdrvRequestFlags flags); 1053 1054 static inline int coroutine_fn bdrv_co_pread(BdrvChild *child, 1055 int64_t offset, unsigned int bytes, void *buf, BdrvRequestFlags flags) 1056 { 1057 QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes); 1058 1059 return bdrv_co_preadv(child, offset, bytes, &qiov, flags); 1060 } 1061 1062 static inline int coroutine_fn bdrv_co_pwrite(BdrvChild *child, 1063 int64_t offset, unsigned int bytes, void *buf, BdrvRequestFlags flags) 1064 { 1065 QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes); 1066 1067 return bdrv_co_pwritev(child, offset, bytes, &qiov, flags); 1068 } 1069 1070 extern unsigned int bdrv_drain_all_count; 1071 void bdrv_apply_subtree_drain(BdrvChild *child, BlockDriverState *new_parent); 1072 void bdrv_unapply_subtree_drain(BdrvChild *child, BlockDriverState *old_parent); 1073 1074 bool coroutine_fn bdrv_make_request_serialising(BdrvTrackedRequest *req, 1075 uint64_t align); 1076 BdrvTrackedRequest *coroutine_fn bdrv_co_get_self_request(BlockDriverState *bs); 1077 1078 int get_tmp_filename(char *filename, int size); 1079 BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size, 1080 const char *filename); 1081 1082 void bdrv_parse_filename_strip_prefix(const char *filename, const char *prefix, 1083 QDict *options); 1084 1085 bool bdrv_backing_overridden(BlockDriverState *bs); 1086 1087 1088 /** 1089 * bdrv_add_aio_context_notifier: 1090 * 1091 * If a long-running job intends to be always run in the same AioContext as a 1092 * certain BDS, it may use this function to be notified of changes regarding the 1093 * association of the BDS to an AioContext. 1094 * 1095 * attached_aio_context() is called after the target BDS has been attached to a 1096 * new AioContext; detach_aio_context() is called before the target BDS is being 1097 * detached from its old AioContext. 1098 */ 1099 void bdrv_add_aio_context_notifier(BlockDriverState *bs, 1100 void (*attached_aio_context)(AioContext *new_context, void *opaque), 1101 void (*detach_aio_context)(void *opaque), void *opaque); 1102 1103 /** 1104 * bdrv_remove_aio_context_notifier: 1105 * 1106 * Unsubscribe of change notifications regarding the BDS's AioContext. The 1107 * parameters given here have to be the same as those given to 1108 * bdrv_add_aio_context_notifier(). 1109 */ 1110 void bdrv_remove_aio_context_notifier(BlockDriverState *bs, 1111 void (*aio_context_attached)(AioContext *, 1112 void *), 1113 void (*aio_context_detached)(void *), 1114 void *opaque); 1115 1116 /** 1117 * bdrv_wakeup: 1118 * @bs: The BlockDriverState for which an I/O operation has been completed. 1119 * 1120 * Wake up the main thread if it is waiting on BDRV_POLL_WHILE. During 1121 * synchronous I/O on a BlockDriverState that is attached to another 1122 * I/O thread, the main thread lets the I/O thread's event loop run, 1123 * waiting for the I/O operation to complete. A bdrv_wakeup will wake 1124 * up the main thread if necessary. 1125 * 1126 * Manual calls to bdrv_wakeup are rarely necessary, because 1127 * bdrv_dec_in_flight already calls it. 1128 */ 1129 void bdrv_wakeup(BlockDriverState *bs); 1130 1131 #ifdef _WIN32 1132 int is_windows_drive(const char *filename); 1133 #endif 1134 1135 /** 1136 * stream_start: 1137 * @job_id: The id of the newly-created job, or %NULL to use the 1138 * device name of @bs. 1139 * @bs: Block device to operate on. 1140 * @base: Block device that will become the new base, or %NULL to 1141 * flatten the whole backing file chain onto @bs. 1142 * @backing_file_str: The file name that will be written to @bs as the 1143 * the new backing file if the job completes. Ignored if @base is %NULL. 1144 * @creation_flags: Flags that control the behavior of the Job lifetime. 1145 * See @BlockJobCreateFlags 1146 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 1147 * @on_error: The action to take upon error. 1148 * @filter_node_name: The node name that should be assigned to the filter 1149 * driver that the stream job inserts into the graph above 1150 * @bs. NULL means that a node name should be autogenerated. 1151 * @errp: Error object. 1152 * 1153 * Start a streaming operation on @bs. Clusters that are unallocated 1154 * in @bs, but allocated in any image between @base and @bs (both 1155 * exclusive) will be written to @bs. At the end of a successful 1156 * streaming job, the backing file of @bs will be changed to 1157 * @backing_file_str in the written image and to @base in the live 1158 * BlockDriverState. 1159 */ 1160 void stream_start(const char *job_id, BlockDriverState *bs, 1161 BlockDriverState *base, const char *backing_file_str, 1162 BlockDriverState *bottom, 1163 int creation_flags, int64_t speed, 1164 BlockdevOnError on_error, 1165 const char *filter_node_name, 1166 Error **errp); 1167 1168 /** 1169 * commit_start: 1170 * @job_id: The id of the newly-created job, or %NULL to use the 1171 * device name of @bs. 1172 * @bs: Active block device. 1173 * @top: Top block device to be committed. 1174 * @base: Block device that will be written into, and become the new top. 1175 * @creation_flags: Flags that control the behavior of the Job lifetime. 1176 * See @BlockJobCreateFlags 1177 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 1178 * @on_error: The action to take upon error. 1179 * @backing_file_str: String to use as the backing file in @top's overlay 1180 * @filter_node_name: The node name that should be assigned to the filter 1181 * driver that the commit job inserts into the graph above @top. NULL means 1182 * that a node name should be autogenerated. 1183 * @errp: Error object. 1184 * 1185 */ 1186 void commit_start(const char *job_id, BlockDriverState *bs, 1187 BlockDriverState *base, BlockDriverState *top, 1188 int creation_flags, int64_t speed, 1189 BlockdevOnError on_error, const char *backing_file_str, 1190 const char *filter_node_name, Error **errp); 1191 /** 1192 * commit_active_start: 1193 * @job_id: The id of the newly-created job, or %NULL to use the 1194 * device name of @bs. 1195 * @bs: Active block device to be committed. 1196 * @base: Block device that will be written into, and become the new top. 1197 * @creation_flags: Flags that control the behavior of the Job lifetime. 1198 * See @BlockJobCreateFlags 1199 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 1200 * @on_error: The action to take upon error. 1201 * @filter_node_name: The node name that should be assigned to the filter 1202 * driver that the commit job inserts into the graph above @bs. NULL means that 1203 * a node name should be autogenerated. 1204 * @cb: Completion function for the job. 1205 * @opaque: Opaque pointer value passed to @cb. 1206 * @auto_complete: Auto complete the job. 1207 * @errp: Error object. 1208 * 1209 */ 1210 BlockJob *commit_active_start(const char *job_id, BlockDriverState *bs, 1211 BlockDriverState *base, int creation_flags, 1212 int64_t speed, BlockdevOnError on_error, 1213 const char *filter_node_name, 1214 BlockCompletionFunc *cb, void *opaque, 1215 bool auto_complete, Error **errp); 1216 /* 1217 * mirror_start: 1218 * @job_id: The id of the newly-created job, or %NULL to use the 1219 * device name of @bs. 1220 * @bs: Block device to operate on. 1221 * @target: Block device to write to. 1222 * @replaces: Block graph node name to replace once the mirror is done. Can 1223 * only be used when full mirroring is selected. 1224 * @creation_flags: Flags that control the behavior of the Job lifetime. 1225 * See @BlockJobCreateFlags 1226 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 1227 * @granularity: The chosen granularity for the dirty bitmap. 1228 * @buf_size: The amount of data that can be in flight at one time. 1229 * @mode: Whether to collapse all images in the chain to the target. 1230 * @backing_mode: How to establish the target's backing chain after completion. 1231 * @zero_target: Whether the target should be explicitly zero-initialized 1232 * @on_source_error: The action to take upon error reading from the source. 1233 * @on_target_error: The action to take upon error writing to the target. 1234 * @unmap: Whether to unmap target where source sectors only contain zeroes. 1235 * @filter_node_name: The node name that should be assigned to the filter 1236 * driver that the mirror job inserts into the graph above @bs. NULL means that 1237 * a node name should be autogenerated. 1238 * @copy_mode: When to trigger writes to the target. 1239 * @errp: Error object. 1240 * 1241 * Start a mirroring operation on @bs. Clusters that are allocated 1242 * in @bs will be written to @target until the job is cancelled or 1243 * manually completed. At the end of a successful mirroring job, 1244 * @bs will be switched to read from @target. 1245 */ 1246 void mirror_start(const char *job_id, BlockDriverState *bs, 1247 BlockDriverState *target, const char *replaces, 1248 int creation_flags, int64_t speed, 1249 uint32_t granularity, int64_t buf_size, 1250 MirrorSyncMode mode, BlockMirrorBackingMode backing_mode, 1251 bool zero_target, 1252 BlockdevOnError on_source_error, 1253 BlockdevOnError on_target_error, 1254 bool unmap, const char *filter_node_name, 1255 MirrorCopyMode copy_mode, Error **errp); 1256 1257 /* 1258 * backup_job_create: 1259 * @job_id: The id of the newly-created job, or %NULL to use the 1260 * device name of @bs. 1261 * @bs: Block device to operate on. 1262 * @target: Block device to write to. 1263 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 1264 * @sync_mode: What parts of the disk image should be copied to the destination. 1265 * @sync_bitmap: The dirty bitmap if sync_mode is 'bitmap' or 'incremental' 1266 * @bitmap_mode: The bitmap synchronization policy to use. 1267 * @perf: Performance options. All actual fields assumed to be present, 1268 * all ".has_*" fields are ignored. 1269 * @on_source_error: The action to take upon error reading from the source. 1270 * @on_target_error: The action to take upon error writing to the target. 1271 * @creation_flags: Flags that control the behavior of the Job lifetime. 1272 * See @BlockJobCreateFlags 1273 * @cb: Completion function for the job. 1274 * @opaque: Opaque pointer value passed to @cb. 1275 * @txn: Transaction that this job is part of (may be NULL). 1276 * 1277 * Create a backup operation on @bs. Clusters in @bs are written to @target 1278 * until the job is cancelled or manually completed. 1279 */ 1280 BlockJob *backup_job_create(const char *job_id, BlockDriverState *bs, 1281 BlockDriverState *target, int64_t speed, 1282 MirrorSyncMode sync_mode, 1283 BdrvDirtyBitmap *sync_bitmap, 1284 BitmapSyncMode bitmap_mode, 1285 bool compress, 1286 const char *filter_node_name, 1287 BackupPerf *perf, 1288 BlockdevOnError on_source_error, 1289 BlockdevOnError on_target_error, 1290 int creation_flags, 1291 BlockCompletionFunc *cb, void *opaque, 1292 JobTxn *txn, Error **errp); 1293 1294 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs, 1295 const char *child_name, 1296 const BdrvChildClass *child_class, 1297 BdrvChildRole child_role, 1298 uint64_t perm, uint64_t shared_perm, 1299 void *opaque, Error **errp); 1300 void bdrv_root_unref_child(BdrvChild *child); 1301 1302 void bdrv_get_cumulative_perm(BlockDriverState *bs, uint64_t *perm, 1303 uint64_t *shared_perm); 1304 1305 /** 1306 * Sets a BdrvChild's permissions. Avoid if the parent is a BDS; use 1307 * bdrv_child_refresh_perms() instead and make the parent's 1308 * .bdrv_child_perm() implementation return the correct values. 1309 */ 1310 int bdrv_child_try_set_perm(BdrvChild *c, uint64_t perm, uint64_t shared, 1311 Error **errp); 1312 1313 /** 1314 * Calls bs->drv->bdrv_child_perm() and updates the child's permission 1315 * masks with the result. 1316 * Drivers should invoke this function whenever an event occurs that 1317 * makes their .bdrv_child_perm() implementation return different 1318 * values than before, but which will not result in the block layer 1319 * automatically refreshing the permissions. 1320 */ 1321 int bdrv_child_refresh_perms(BlockDriverState *bs, BdrvChild *c, Error **errp); 1322 1323 bool bdrv_recurse_can_replace(BlockDriverState *bs, 1324 BlockDriverState *to_replace); 1325 1326 /* 1327 * Default implementation for BlockDriver.bdrv_child_perm() that can 1328 * be used by block filters and image formats, as long as they use the 1329 * child_of_bds child class and set an appropriate BdrvChildRole. 1330 */ 1331 void bdrv_default_perms(BlockDriverState *bs, BdrvChild *c, 1332 BdrvChildRole role, BlockReopenQueue *reopen_queue, 1333 uint64_t perm, uint64_t shared, 1334 uint64_t *nperm, uint64_t *nshared); 1335 1336 const char *bdrv_get_parent_name(const BlockDriverState *bs); 1337 void blk_dev_change_media_cb(BlockBackend *blk, bool load, Error **errp); 1338 bool blk_dev_has_removable_media(BlockBackend *blk); 1339 bool blk_dev_has_tray(BlockBackend *blk); 1340 void blk_dev_eject_request(BlockBackend *blk, bool force); 1341 bool blk_dev_is_tray_open(BlockBackend *blk); 1342 bool blk_dev_is_medium_locked(BlockBackend *blk); 1343 1344 void bdrv_set_dirty(BlockDriverState *bs, int64_t offset, int64_t bytes); 1345 1346 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap **out); 1347 void bdrv_restore_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap *backup); 1348 bool bdrv_dirty_bitmap_merge_internal(BdrvDirtyBitmap *dest, 1349 const BdrvDirtyBitmap *src, 1350 HBitmap **backup, bool lock); 1351 1352 void bdrv_inc_in_flight(BlockDriverState *bs); 1353 void bdrv_dec_in_flight(BlockDriverState *bs); 1354 1355 void blockdev_close_all_bdrv_states(void); 1356 1357 int coroutine_fn bdrv_co_copy_range_from(BdrvChild *src, int64_t src_offset, 1358 BdrvChild *dst, int64_t dst_offset, 1359 int64_t bytes, 1360 BdrvRequestFlags read_flags, 1361 BdrvRequestFlags write_flags); 1362 int coroutine_fn bdrv_co_copy_range_to(BdrvChild *src, int64_t src_offset, 1363 BdrvChild *dst, int64_t dst_offset, 1364 int64_t bytes, 1365 BdrvRequestFlags read_flags, 1366 BdrvRequestFlags write_flags); 1367 1368 int refresh_total_sectors(BlockDriverState *bs, int64_t hint); 1369 1370 void bdrv_set_monitor_owned(BlockDriverState *bs); 1371 BlockDriverState *bds_tree_init(QDict *bs_opts, Error **errp); 1372 1373 /** 1374 * Simple implementation of bdrv_co_create_opts for protocol drivers 1375 * which only support creation via opening a file 1376 * (usually existing raw storage device) 1377 */ 1378 int coroutine_fn bdrv_co_create_opts_simple(BlockDriver *drv, 1379 const char *filename, 1380 QemuOpts *opts, 1381 Error **errp); 1382 extern QemuOptsList bdrv_create_opts_simple; 1383 1384 BdrvDirtyBitmap *block_dirty_bitmap_lookup(const char *node, 1385 const char *name, 1386 BlockDriverState **pbs, 1387 Error **errp); 1388 BdrvDirtyBitmap *block_dirty_bitmap_merge(const char *node, const char *target, 1389 BlockDirtyBitmapMergeSourceList *bms, 1390 HBitmap **backup, Error **errp); 1391 BdrvDirtyBitmap *block_dirty_bitmap_remove(const char *node, const char *name, 1392 bool release, 1393 BlockDriverState **bitmap_bs, 1394 Error **errp); 1395 1396 BdrvChild *bdrv_cow_child(BlockDriverState *bs); 1397 BdrvChild *bdrv_filter_child(BlockDriverState *bs); 1398 BdrvChild *bdrv_filter_or_cow_child(BlockDriverState *bs); 1399 BdrvChild *bdrv_primary_child(BlockDriverState *bs); 1400 BlockDriverState *bdrv_skip_implicit_filters(BlockDriverState *bs); 1401 BlockDriverState *bdrv_skip_filters(BlockDriverState *bs); 1402 BlockDriverState *bdrv_backing_chain_next(BlockDriverState *bs); 1403 1404 static inline BlockDriverState *child_bs(BdrvChild *child) 1405 { 1406 return child ? child->bs : NULL; 1407 } 1408 1409 static inline BlockDriverState *bdrv_cow_bs(BlockDriverState *bs) 1410 { 1411 return child_bs(bdrv_cow_child(bs)); 1412 } 1413 1414 static inline BlockDriverState *bdrv_filter_bs(BlockDriverState *bs) 1415 { 1416 return child_bs(bdrv_filter_child(bs)); 1417 } 1418 1419 static inline BlockDriverState *bdrv_filter_or_cow_bs(BlockDriverState *bs) 1420 { 1421 return child_bs(bdrv_filter_or_cow_child(bs)); 1422 } 1423 1424 static inline BlockDriverState *bdrv_primary_bs(BlockDriverState *bs) 1425 { 1426 return child_bs(bdrv_primary_child(bs)); 1427 } 1428 1429 /** 1430 * End all quiescent sections started by bdrv_drain_all_begin(). This is 1431 * needed when deleting a BDS before bdrv_drain_all_end() is called. 1432 * 1433 * NOTE: this is an internal helper for bdrv_close() *only*. No one else 1434 * should call it. 1435 */ 1436 void bdrv_drain_all_end_quiesce(BlockDriverState *bs); 1437 1438 #endif /* BLOCK_INT_H */ 1439