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_COMMON_H 25 #define BLOCK_INT_COMMON_H 26 27 #include "block/aio.h" 28 #include "block/block-common.h" 29 #include "block/block-global-state.h" 30 #include "block/snapshot.h" 31 #include "qemu/iov.h" 32 #include "qemu/rcu.h" 33 #include "qemu/stats64.h" 34 35 #define BLOCK_FLAG_LAZY_REFCOUNTS 8 36 37 #define BLOCK_OPT_SIZE "size" 38 #define BLOCK_OPT_ENCRYPT "encryption" 39 #define BLOCK_OPT_ENCRYPT_FORMAT "encrypt.format" 40 #define BLOCK_OPT_COMPAT6 "compat6" 41 #define BLOCK_OPT_HWVERSION "hwversion" 42 #define BLOCK_OPT_BACKING_FILE "backing_file" 43 #define BLOCK_OPT_BACKING_FMT "backing_fmt" 44 #define BLOCK_OPT_CLUSTER_SIZE "cluster_size" 45 #define BLOCK_OPT_TABLE_SIZE "table_size" 46 #define BLOCK_OPT_PREALLOC "preallocation" 47 #define BLOCK_OPT_SUBFMT "subformat" 48 #define BLOCK_OPT_COMPAT_LEVEL "compat" 49 #define BLOCK_OPT_LAZY_REFCOUNTS "lazy_refcounts" 50 #define BLOCK_OPT_ADAPTER_TYPE "adapter_type" 51 #define BLOCK_OPT_REDUNDANCY "redundancy" 52 #define BLOCK_OPT_NOCOW "nocow" 53 #define BLOCK_OPT_EXTENT_SIZE_HINT "extent_size_hint" 54 #define BLOCK_OPT_OBJECT_SIZE "object_size" 55 #define BLOCK_OPT_REFCOUNT_BITS "refcount_bits" 56 #define BLOCK_OPT_DATA_FILE "data_file" 57 #define BLOCK_OPT_DATA_FILE_RAW "data_file_raw" 58 #define BLOCK_OPT_COMPRESSION_TYPE "compression_type" 59 #define BLOCK_OPT_EXTL2 "extended_l2" 60 61 #define BLOCK_PROBE_BUF_SIZE 512 62 63 enum BdrvTrackedRequestType { 64 BDRV_TRACKED_READ, 65 BDRV_TRACKED_WRITE, 66 BDRV_TRACKED_DISCARD, 67 BDRV_TRACKED_TRUNCATE, 68 }; 69 70 /* 71 * That is not quite good that BdrvTrackedRequest structure is public, 72 * as block/io.c is very careful about incoming offset/bytes being 73 * correct. Be sure to assert bdrv_check_request() succeeded after any 74 * modification of BdrvTrackedRequest object out of block/io.c 75 */ 76 typedef struct BdrvTrackedRequest { 77 BlockDriverState *bs; 78 int64_t offset; 79 int64_t bytes; 80 enum BdrvTrackedRequestType type; 81 82 bool serialising; 83 int64_t overlap_offset; 84 int64_t overlap_bytes; 85 86 QLIST_ENTRY(BdrvTrackedRequest) list; 87 Coroutine *co; /* owner, used for deadlock detection */ 88 CoQueue wait_queue; /* coroutines blocked on this request */ 89 90 struct BdrvTrackedRequest *waiting_for; 91 } BdrvTrackedRequest; 92 93 94 struct BlockDriver { 95 /* 96 * These fields are initialized when this object is created, 97 * and are never changed afterwards. 98 */ 99 100 const char *format_name; 101 int instance_size; 102 103 /* 104 * Set to true if the BlockDriver is a block filter. Block filters pass 105 * certain callbacks that refer to data (see block.c) to their bs->file 106 * or bs->backing (whichever one exists) if the driver doesn't implement 107 * them. Drivers that do not wish to forward must implement them and return 108 * -ENOTSUP. 109 * Note that filters are not allowed to modify data. 110 * 111 * Filters generally cannot have more than a single filtered child, 112 * because the data they present must at all times be the same as 113 * that on their filtered child. That would be impossible to 114 * achieve for multiple filtered children. 115 * (And this filtered child must then be bs->file or bs->backing.) 116 */ 117 bool is_filter; 118 /* 119 * Only make sense for filter drivers, for others must be false. 120 * If true, filtered child is bs->backing. Otherwise it's bs->file. 121 * Two internal filters use bs->backing as filtered child and has this 122 * field set to true: mirror_top and commit_top. There also two such test 123 * filters in tests/unit/test-bdrv-graph-mod.c. 124 * 125 * Never create any more such filters! 126 * 127 * TODO: imagine how to deprecate this behavior and make all filters work 128 * similarly using bs->file as filtered child. 129 */ 130 bool filtered_child_is_backing; 131 132 /* 133 * Set to true if the BlockDriver is a format driver. Format nodes 134 * generally do not expect their children to be other format nodes 135 * (except for backing files), and so format probing is disabled 136 * on those children. 137 */ 138 bool is_format; 139 140 /* 141 * Set to true if the BlockDriver supports zoned children. 142 */ 143 bool supports_zoned_children; 144 145 /* 146 * Drivers not implementing bdrv_parse_filename nor bdrv_open should have 147 * this field set to true, except ones that are defined only by their 148 * child's bs. 149 * An example of the last type will be the quorum block driver. 150 */ 151 bool bdrv_needs_filename; 152 153 /* 154 * Set if a driver can support backing files. This also implies the 155 * following semantics: 156 * 157 * - Return status 0 of .bdrv_co_block_status means that corresponding 158 * blocks are not allocated in this layer of backing-chain 159 * - For such (unallocated) blocks, read will: 160 * - fill buffer with zeros if there is no backing file 161 * - read from the backing file otherwise, where the block layer 162 * takes care of reading zeros beyond EOF if backing file is short 163 */ 164 bool supports_backing; 165 166 /* 167 * Drivers setting this field must be able to work with just a plain 168 * filename with '<protocol_name>:' as a prefix, and no other options. 169 * Options may be extracted from the filename by implementing 170 * bdrv_parse_filename. 171 */ 172 const char *protocol_name; 173 174 /* List of options for creating images, terminated by name == NULL */ 175 QemuOptsList *create_opts; 176 177 /* List of options for image amend */ 178 QemuOptsList *amend_opts; 179 180 /* 181 * If this driver supports reopening images this contains a 182 * NULL-terminated list of the runtime options that can be 183 * modified. If an option in this list is unspecified during 184 * reopen then it _must_ be reset to its default value or return 185 * an error. 186 */ 187 const char *const *mutable_opts; 188 189 /* 190 * Pointer to a NULL-terminated array of names of strong options 191 * that can be specified for bdrv_open(). A strong option is one 192 * that changes the data of a BDS. 193 * If this pointer is NULL, the array is considered empty. 194 * "filename" and "driver" are always considered strong. 195 */ 196 const char *const *strong_runtime_opts; 197 198 199 /* 200 * Global state (GS) API. These functions run under the BQL. 201 * 202 * See include/block/block-global-state.h for more information about 203 * the GS API. 204 */ 205 206 /* 207 * This function is invoked under BQL before .bdrv_co_amend() 208 * (which in contrast does not necessarily run under the BQL) 209 * to allow driver-specific initialization code that requires 210 * the BQL, like setting up specific permission flags. 211 */ 212 int GRAPH_RDLOCK_PTR (*bdrv_amend_pre_run)( 213 BlockDriverState *bs, Error **errp); 214 /* 215 * This function is invoked under BQL after .bdrv_co_amend() 216 * to allow cleaning up what was done in .bdrv_amend_pre_run(). 217 */ 218 void GRAPH_RDLOCK_PTR (*bdrv_amend_clean)(BlockDriverState *bs); 219 220 /* 221 * Return true if @to_replace can be replaced by a BDS with the 222 * same data as @bs without it affecting @bs's behavior (that is, 223 * without it being visible to @bs's parents). 224 */ 225 bool GRAPH_RDLOCK_PTR (*bdrv_recurse_can_replace)( 226 BlockDriverState *bs, BlockDriverState *to_replace); 227 228 int (*bdrv_probe_device)(const char *filename); 229 230 /* 231 * Any driver implementing this callback is expected to be able to handle 232 * NULL file names in its .bdrv_open() implementation. 233 */ 234 void (*bdrv_parse_filename)(const char *filename, QDict *options, 235 Error **errp); 236 237 /* For handling image reopen for split or non-split files. */ 238 int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state, 239 BlockReopenQueue *queue, Error **errp); 240 void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state); 241 void (*bdrv_reopen_commit_post)(BDRVReopenState *reopen_state); 242 void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state); 243 void (*bdrv_join_options)(QDict *options, QDict *old_options); 244 245 int GRAPH_UNLOCKED_PTR (*bdrv_open)( 246 BlockDriverState *bs, QDict *options, int flags, Error **errp); 247 248 /* Protocol drivers should implement this instead of bdrv_open */ 249 int GRAPH_UNLOCKED_PTR (*bdrv_file_open)( 250 BlockDriverState *bs, QDict *options, int flags, Error **errp); 251 void (*bdrv_close)(BlockDriverState *bs); 252 253 int coroutine_fn GRAPH_UNLOCKED_PTR (*bdrv_co_create)( 254 BlockdevCreateOptions *opts, Error **errp); 255 256 int coroutine_fn GRAPH_UNLOCKED_PTR (*bdrv_co_create_opts)( 257 BlockDriver *drv, const char *filename, QemuOpts *opts, Error **errp); 258 259 int (*bdrv_amend_options)(BlockDriverState *bs, 260 QemuOpts *opts, 261 BlockDriverAmendStatusCB *status_cb, 262 void *cb_opaque, 263 bool force, 264 Error **errp); 265 266 int (*bdrv_make_empty)(BlockDriverState *bs); 267 268 /* 269 * Refreshes the bs->exact_filename field. If that is impossible, 270 * bs->exact_filename has to be left empty. 271 */ 272 void (*bdrv_refresh_filename)(BlockDriverState *bs); 273 274 /* 275 * Gathers the open options for all children into @target. 276 * A simple format driver (without backing file support) might 277 * implement this function like this: 278 * 279 * QINCREF(bs->file->bs->full_open_options); 280 * qdict_put(target, "file", bs->file->bs->full_open_options); 281 * 282 * If not specified, the generic implementation will simply put 283 * all children's options under their respective name. 284 * 285 * @backing_overridden is true when bs->backing seems not to be 286 * the child that would result from opening bs->backing_file. 287 * Therefore, if it is true, the backing child's options should be 288 * gathered; otherwise, there is no need since the backing child 289 * is the one implied by the image header. 290 * 291 * Note that ideally this function would not be needed. Every 292 * block driver which implements it is probably doing something 293 * shady regarding its runtime option structure. 294 */ 295 void (*bdrv_gather_child_options)(BlockDriverState *bs, QDict *target, 296 bool backing_overridden); 297 298 /* 299 * Returns an allocated string which is the directory name of this BDS: It 300 * will be used to make relative filenames absolute by prepending this 301 * function's return value to them. 302 */ 303 char *(*bdrv_dirname)(BlockDriverState *bs, Error **errp); 304 305 /* 306 * This informs the driver that we are no longer interested in the result 307 * of in-flight requests, so don't waste the time if possible. 308 * 309 * One example usage is to avoid waiting for an nbd target node reconnect 310 * timeout during job-cancel with force=true. 311 */ 312 void (*bdrv_cancel_in_flight)(BlockDriverState *bs); 313 314 int GRAPH_RDLOCK_PTR (*bdrv_inactivate)(BlockDriverState *bs); 315 316 int GRAPH_RDLOCK_PTR (*bdrv_snapshot_create)( 317 BlockDriverState *bs, QEMUSnapshotInfo *sn_info); 318 319 int GRAPH_UNLOCKED_PTR (*bdrv_snapshot_goto)( 320 BlockDriverState *bs, const char *snapshot_id); 321 322 int GRAPH_RDLOCK_PTR (*bdrv_snapshot_delete)( 323 BlockDriverState *bs, const char *snapshot_id, const char *name, 324 Error **errp); 325 326 int (*bdrv_snapshot_list)(BlockDriverState *bs, 327 QEMUSnapshotInfo **psn_info); 328 int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs, 329 const char *snapshot_id, 330 const char *name, 331 Error **errp); 332 333 int (*bdrv_change_backing_file)(BlockDriverState *bs, 334 const char *backing_file, const char *backing_fmt); 335 336 /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */ 337 int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event, 338 const char *tag); 339 int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs, 340 const char *tag); 341 int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag); 342 bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag); 343 344 void GRAPH_RDLOCK_PTR (*bdrv_refresh_limits)( 345 BlockDriverState *bs, Error **errp); 346 347 /* 348 * Returns 1 if newly created images are guaranteed to contain only 349 * zeros, 0 otherwise. 350 */ 351 int (*bdrv_has_zero_init)(BlockDriverState *bs); 352 353 /* 354 * Remove fd handlers, timers, and other event loop callbacks so the event 355 * loop is no longer in use. Called with no in-flight requests and in 356 * depth-first traversal order with parents before child nodes. 357 */ 358 void (*bdrv_detach_aio_context)(BlockDriverState *bs); 359 360 /* 361 * Add fd handlers, timers, and other event loop callbacks so I/O requests 362 * can be processed again. Called with no in-flight requests and in 363 * depth-first traversal order with child nodes before parent nodes. 364 */ 365 void (*bdrv_attach_aio_context)(BlockDriverState *bs, 366 AioContext *new_context); 367 368 /** 369 * bdrv_drain_begin is called if implemented in the beginning of a 370 * drain operation to drain and stop any internal sources of requests in 371 * the driver. 372 * bdrv_drain_end is called if implemented at the end of the drain. 373 * 374 * They should be used by the driver to e.g. manage scheduled I/O 375 * requests, or toggle an internal state. After the end of the drain new 376 * requests will continue normally. 377 * 378 * Implementations of both functions must not call aio_poll(). 379 */ 380 void (*bdrv_drain_begin)(BlockDriverState *bs); 381 void (*bdrv_drain_end)(BlockDriverState *bs); 382 383 /** 384 * Try to get @bs's logical and physical block size. 385 * On success, store them in @bsz and return zero. 386 * On failure, return negative errno. 387 */ 388 int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz); 389 /** 390 * Try to get @bs's geometry (cyls, heads, sectors) 391 * On success, store them in @geo and return 0. 392 * On failure return -errno. 393 * Only drivers that want to override guest geometry implement this 394 * callback; see hd_geometry_guess(). 395 */ 396 int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo); 397 398 void GRAPH_WRLOCK_PTR (*bdrv_add_child)( 399 BlockDriverState *parent, BlockDriverState *child, Error **errp); 400 401 void GRAPH_WRLOCK_PTR (*bdrv_del_child)( 402 BlockDriverState *parent, BdrvChild *child, Error **errp); 403 404 /** 405 * Informs the block driver that a permission change is intended. The 406 * driver checks whether the change is permissible and may take other 407 * preparations for the change (e.g. get file system locks). This operation 408 * is always followed either by a call to either .bdrv_set_perm or 409 * .bdrv_abort_perm_update. 410 * 411 * Checks whether the requested set of cumulative permissions in @perm 412 * can be granted for accessing @bs and whether no other users are using 413 * permissions other than those given in @shared (both arguments take 414 * BLK_PERM_* bitmasks). 415 * 416 * If both conditions are met, 0 is returned. Otherwise, -errno is returned 417 * and errp is set to an error describing the conflict. 418 */ 419 int GRAPH_RDLOCK_PTR (*bdrv_check_perm)(BlockDriverState *bs, uint64_t perm, 420 uint64_t shared, Error **errp); 421 422 /** 423 * Called to inform the driver that the set of cumulative set of used 424 * permissions for @bs has changed to @perm, and the set of shareable 425 * permission to @shared. The driver can use this to propagate changes to 426 * its children (i.e. request permissions only if a parent actually needs 427 * them). 428 * 429 * This function is only invoked after bdrv_check_perm(), so block drivers 430 * may rely on preparations made in their .bdrv_check_perm implementation. 431 */ 432 void GRAPH_RDLOCK_PTR (*bdrv_set_perm)( 433 BlockDriverState *bs, uint64_t perm, uint64_t shared); 434 435 /* 436 * Called to inform the driver that after a previous bdrv_check_perm() 437 * call, the permission update is not performed and any preparations made 438 * for it (e.g. taken file locks) need to be undone. 439 * 440 * This function can be called even for nodes that never saw a 441 * bdrv_check_perm() call. It is a no-op then. 442 */ 443 void GRAPH_RDLOCK_PTR (*bdrv_abort_perm_update)(BlockDriverState *bs); 444 445 /** 446 * Returns in @nperm and @nshared the permissions that the driver for @bs 447 * needs on its child @c, based on the cumulative permissions requested by 448 * the parents in @parent_perm and @parent_shared. 449 * 450 * If @c is NULL, return the permissions for attaching a new child for the 451 * given @child_class and @role. 452 * 453 * If @reopen_queue is non-NULL, don't return the currently needed 454 * permissions, but those that will be needed after applying the 455 * @reopen_queue. 456 */ 457 void GRAPH_RDLOCK_PTR (*bdrv_child_perm)( 458 BlockDriverState *bs, BdrvChild *c, BdrvChildRole role, 459 BlockReopenQueue *reopen_queue, 460 uint64_t parent_perm, uint64_t parent_shared, 461 uint64_t *nperm, uint64_t *nshared); 462 463 /** 464 * Register/unregister a buffer for I/O. For example, when the driver is 465 * interested to know the memory areas that will later be used in iovs, so 466 * that it can do IOMMU mapping with VFIO etc., in order to get better 467 * performance. In the case of VFIO drivers, this callback is used to do 468 * DMA mapping for hot buffers. 469 * 470 * Returns: true on success, false on failure 471 */ 472 bool GRAPH_RDLOCK_PTR (*bdrv_register_buf)( 473 BlockDriverState *bs, void *host, size_t size, Error **errp); 474 void GRAPH_RDLOCK_PTR (*bdrv_unregister_buf)( 475 BlockDriverState *bs, void *host, size_t size); 476 477 /* 478 * This field is modified only under the BQL, and is part of 479 * the global state. 480 */ 481 QLIST_ENTRY(BlockDriver) list; 482 483 /* 484 * I/O API functions. These functions are thread-safe. 485 * 486 * See include/block/block-io.h for more information about 487 * the I/O API. 488 */ 489 490 int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename); 491 492 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_amend)( 493 BlockDriverState *bs, BlockdevAmendOptions *opts, bool force, 494 Error **errp); 495 496 /* aio */ 497 BlockAIOCB * GRAPH_RDLOCK_PTR (*bdrv_aio_preadv)(BlockDriverState *bs, 498 int64_t offset, int64_t bytes, QEMUIOVector *qiov, 499 BdrvRequestFlags flags, BlockCompletionFunc *cb, void *opaque); 500 501 BlockAIOCB * GRAPH_RDLOCK_PTR (*bdrv_aio_pwritev)(BlockDriverState *bs, 502 int64_t offset, int64_t bytes, QEMUIOVector *qiov, 503 BdrvRequestFlags flags, BlockCompletionFunc *cb, void *opaque); 504 505 BlockAIOCB * GRAPH_RDLOCK_PTR (*bdrv_aio_flush)( 506 BlockDriverState *bs, BlockCompletionFunc *cb, void *opaque); 507 508 BlockAIOCB * GRAPH_RDLOCK_PTR (*bdrv_aio_pdiscard)( 509 BlockDriverState *bs, int64_t offset, int bytes, 510 BlockCompletionFunc *cb, void *opaque); 511 512 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_readv)(BlockDriverState *bs, 513 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov); 514 515 /** 516 * @offset: position in bytes to read at 517 * @bytes: number of bytes to read 518 * @qiov: the buffers to fill with read data 519 * @flags: currently unused, always 0 520 * 521 * @offset and @bytes will be a multiple of 'request_alignment', 522 * but the length of individual @qiov elements does not have to 523 * be a multiple. 524 * 525 * @bytes will always equal the total size of @qiov, and will be 526 * no larger than 'max_transfer'. 527 * 528 * The buffer in @qiov may point directly to guest memory. 529 */ 530 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_preadv)(BlockDriverState *bs, 531 int64_t offset, int64_t bytes, QEMUIOVector *qiov, 532 BdrvRequestFlags flags); 533 534 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_preadv_part)( 535 BlockDriverState *bs, int64_t offset, int64_t bytes, 536 QEMUIOVector *qiov, size_t qiov_offset, 537 BdrvRequestFlags flags); 538 539 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_writev)(BlockDriverState *bs, 540 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, 541 int flags); 542 /** 543 * @offset: position in bytes to write at 544 * @bytes: number of bytes to write 545 * @qiov: the buffers containing data to write 546 * @flags: zero or more bits allowed by 'supported_write_flags' 547 * 548 * @offset and @bytes will be a multiple of 'request_alignment', 549 * but the length of individual @qiov elements does not have to 550 * be a multiple. 551 * 552 * @bytes will always equal the total size of @qiov, and will be 553 * no larger than 'max_transfer'. 554 * 555 * The buffer in @qiov may point directly to guest memory. 556 */ 557 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_pwritev)( 558 BlockDriverState *bs, int64_t offset, int64_t bytes, QEMUIOVector *qiov, 559 BdrvRequestFlags flags); 560 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_pwritev_part)( 561 BlockDriverState *bs, int64_t offset, int64_t bytes, QEMUIOVector *qiov, 562 size_t qiov_offset, BdrvRequestFlags flags); 563 564 /* 565 * Efficiently zero a region of the disk image. Typically an image format 566 * would use a compact metadata representation to implement this. This 567 * function pointer may be NULL or return -ENOSUP and .bdrv_co_writev() 568 * will be called instead. 569 */ 570 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_pwrite_zeroes)( 571 BlockDriverState *bs, int64_t offset, int64_t bytes, 572 BdrvRequestFlags flags); 573 574 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_pdiscard)( 575 BlockDriverState *bs, int64_t offset, int64_t bytes); 576 577 /* 578 * Map [offset, offset + nbytes) range onto a child of @bs to copy from, 579 * and invoke bdrv_co_copy_range_from(child, ...), or invoke 580 * bdrv_co_copy_range_to() if @bs is the leaf child to copy data from. 581 * 582 * See the comment of bdrv_co_copy_range for the parameter and return value 583 * semantics. 584 */ 585 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_copy_range_from)( 586 BlockDriverState *bs, BdrvChild *src, int64_t offset, 587 BdrvChild *dst, int64_t dst_offset, int64_t bytes, 588 BdrvRequestFlags read_flags, BdrvRequestFlags write_flags); 589 590 /* 591 * Map [offset, offset + nbytes) range onto a child of bs to copy data to, 592 * and invoke bdrv_co_copy_range_to(child, src, ...), or perform the copy 593 * operation if @bs is the leaf and @src has the same BlockDriver. Return 594 * -ENOTSUP if @bs is the leaf but @src has a different BlockDriver. 595 * 596 * See the comment of bdrv_co_copy_range for the parameter and return value 597 * semantics. 598 */ 599 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_copy_range_to)( 600 BlockDriverState *bs, BdrvChild *src, int64_t src_offset, 601 BdrvChild *dst, int64_t dst_offset, int64_t bytes, 602 BdrvRequestFlags read_flags, BdrvRequestFlags write_flags); 603 604 /* 605 * Building block for bdrv_block_status[_above] and 606 * bdrv_is_allocated[_above]. The driver should answer only 607 * according to the current layer, and should only need to set 608 * BDRV_BLOCK_DATA, BDRV_BLOCK_ZERO, BDRV_BLOCK_OFFSET_VALID, 609 * and/or BDRV_BLOCK_RAW; if the current layer defers to a backing 610 * layer, the result should be 0 (and not BDRV_BLOCK_ZERO). See 611 * block.h for the overall meaning of the bits. As a hint, the 612 * flag want_zero is true if the caller cares more about precise 613 * mappings (favor accurate _OFFSET_VALID/_ZERO) or false for 614 * overall allocation (favor larger *pnum, perhaps by reporting 615 * _DATA instead of _ZERO). The block layer guarantees input 616 * clamped to bdrv_getlength() and aligned to request_alignment, 617 * as well as non-NULL pnum, map, and file; in turn, the driver 618 * must return an error or set pnum to an aligned non-zero value. 619 * 620 * Note that @bytes is just a hint on how big of a region the 621 * caller wants to inspect. It is not a limit on *pnum. 622 * Implementations are free to return larger values of *pnum if 623 * doing so does not incur a performance penalty. 624 * 625 * block/io.c's bdrv_co_block_status() will utilize an unclamped 626 * *pnum value for the block-status cache on protocol nodes, prior 627 * to clamping *pnum for return to its caller. 628 */ 629 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_block_status)( 630 BlockDriverState *bs, 631 bool want_zero, int64_t offset, int64_t bytes, int64_t *pnum, 632 int64_t *map, BlockDriverState **file); 633 634 /* 635 * Snapshot-access API. 636 * 637 * Block-driver may provide snapshot-access API: special functions to access 638 * some internal "snapshot". The functions are similar with normal 639 * read/block_status/discard handler, but don't have any specific handling 640 * in generic block-layer: no serializing, no alignment, no tracked 641 * requests. So, block-driver that realizes these APIs is fully responsible 642 * for synchronization between snapshot-access API and normal IO requests. 643 * 644 * TODO: To be able to support qcow2's internal snapshots, this API will 645 * need to be extended to: 646 * - be able to select a specific snapshot 647 * - receive the snapshot's actual length (which may differ from bs's 648 * length) 649 */ 650 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_preadv_snapshot)( 651 BlockDriverState *bs, int64_t offset, int64_t bytes, 652 QEMUIOVector *qiov, size_t qiov_offset); 653 654 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_snapshot_block_status)( 655 BlockDriverState *bs, bool want_zero, int64_t offset, int64_t bytes, 656 int64_t *pnum, int64_t *map, BlockDriverState **file); 657 658 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_pdiscard_snapshot)( 659 BlockDriverState *bs, int64_t offset, int64_t bytes); 660 661 /* 662 * Invalidate any cached meta-data. 663 */ 664 void coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_invalidate_cache)( 665 BlockDriverState *bs, Error **errp); 666 667 /* 668 * Flushes all data for all layers by calling bdrv_co_flush for underlying 669 * layers, if needed. This function is needed for deterministic 670 * synchronization of the flush finishing callback. 671 */ 672 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_flush)(BlockDriverState *bs); 673 674 /* Delete a created file. */ 675 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_delete_file)( 676 BlockDriverState *bs, Error **errp); 677 678 /* 679 * Flushes all data that was already written to the OS all the way down to 680 * the disk (for example file-posix.c calls fsync()). 681 */ 682 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_flush_to_disk)( 683 BlockDriverState *bs); 684 685 /* 686 * Flushes all internal caches to the OS. The data may still sit in a 687 * writeback cache of the host OS, but it will survive a crash of the qemu 688 * process. 689 */ 690 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_flush_to_os)( 691 BlockDriverState *bs); 692 693 /* 694 * Truncate @bs to @offset bytes using the given @prealloc mode 695 * when growing. Modes other than PREALLOC_MODE_OFF should be 696 * rejected when shrinking @bs. 697 * 698 * If @exact is true, @bs must be resized to exactly @offset. 699 * Otherwise, it is sufficient for @bs (if it is a host block 700 * device and thus there is no way to resize it) to be at least 701 * @offset bytes in length. 702 * 703 * If @exact is true and this function fails but would succeed 704 * with @exact = false, it should return -ENOTSUP. 705 */ 706 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_truncate)( 707 BlockDriverState *bs, int64_t offset, bool exact, 708 PreallocMode prealloc, BdrvRequestFlags flags, Error **errp); 709 710 int64_t coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_getlength)( 711 BlockDriverState *bs); 712 713 int64_t coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_get_allocated_file_size)( 714 BlockDriverState *bs); 715 716 BlockMeasureInfo *(*bdrv_measure)(QemuOpts *opts, BlockDriverState *in_bs, 717 Error **errp); 718 719 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_pwritev_compressed)( 720 BlockDriverState *bs, int64_t offset, int64_t bytes, 721 QEMUIOVector *qiov); 722 723 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_pwritev_compressed_part)( 724 BlockDriverState *bs, int64_t offset, int64_t bytes, 725 QEMUIOVector *qiov, size_t qiov_offset); 726 727 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_get_info)( 728 BlockDriverState *bs, BlockDriverInfo *bdi); 729 730 ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs, 731 Error **errp); 732 BlockStatsSpecific *(*bdrv_get_specific_stats)(BlockDriverState *bs); 733 734 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_save_vmstate)( 735 BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos); 736 737 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_load_vmstate)( 738 BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos); 739 740 int coroutine_fn (*bdrv_co_zone_report)(BlockDriverState *bs, 741 int64_t offset, unsigned int *nr_zones, 742 BlockZoneDescriptor *zones); 743 int coroutine_fn (*bdrv_co_zone_mgmt)(BlockDriverState *bs, BlockZoneOp op, 744 int64_t offset, int64_t len); 745 int coroutine_fn (*bdrv_co_zone_append)(BlockDriverState *bs, 746 int64_t *offset, QEMUIOVector *qiov, 747 BdrvRequestFlags flags); 748 749 /* removable device specific */ 750 bool coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_is_inserted)( 751 BlockDriverState *bs); 752 void coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_eject)( 753 BlockDriverState *bs, bool eject_flag); 754 void coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_lock_medium)( 755 BlockDriverState *bs, bool locked); 756 757 /* to control generic scsi devices */ 758 BlockAIOCB *coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_aio_ioctl)( 759 BlockDriverState *bs, unsigned long int req, void *buf, 760 BlockCompletionFunc *cb, void *opaque); 761 762 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_ioctl)( 763 BlockDriverState *bs, unsigned long int req, void *buf); 764 765 /* 766 * Returns 0 for completed check, -errno for internal errors. 767 * The check results are stored in result. 768 */ 769 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_check)( 770 BlockDriverState *bs, BdrvCheckResult *result, BdrvCheckMode fix); 771 772 void coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_debug_event)( 773 BlockDriverState *bs, BlkdebugEvent event); 774 775 bool (*bdrv_supports_persistent_dirty_bitmap)(BlockDriverState *bs); 776 777 bool coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_can_store_new_dirty_bitmap)( 778 BlockDriverState *bs, const char *name, uint32_t granularity, 779 Error **errp); 780 781 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_remove_persistent_dirty_bitmap)( 782 BlockDriverState *bs, const char *name, Error **errp); 783 }; 784 785 static inline bool TSA_NO_TSA block_driver_can_compress(BlockDriver *drv) 786 { 787 return drv->bdrv_co_pwritev_compressed || 788 drv->bdrv_co_pwritev_compressed_part; 789 } 790 791 typedef struct BlockLimits { 792 /* 793 * Alignment requirement, in bytes, for offset/length of I/O 794 * requests. Must be a power of 2 less than INT_MAX; defaults to 795 * 1 for drivers with modern byte interfaces, and to 512 796 * otherwise. 797 */ 798 uint32_t request_alignment; 799 800 /* 801 * Maximum number of bytes that can be discarded at once. Must be multiple 802 * of pdiscard_alignment, but need not be power of 2. May be 0 if no 803 * inherent 64-bit limit. 804 */ 805 int64_t max_pdiscard; 806 807 /* 808 * Optimal alignment for discard requests in bytes. A power of 2 809 * is best but not mandatory. Must be a multiple of 810 * bl.request_alignment, and must be less than max_pdiscard if 811 * that is set. May be 0 if bl.request_alignment is good enough 812 */ 813 uint32_t pdiscard_alignment; 814 815 /* 816 * Maximum number of bytes that can zeroized at once. Must be multiple of 817 * pwrite_zeroes_alignment. 0 means no limit. 818 */ 819 int64_t max_pwrite_zeroes; 820 821 /* 822 * Optimal alignment for write zeroes requests in bytes. A power 823 * of 2 is best but not mandatory. Must be a multiple of 824 * bl.request_alignment, and must be less than max_pwrite_zeroes 825 * if that is set. May be 0 if bl.request_alignment is good 826 * enough 827 */ 828 uint32_t pwrite_zeroes_alignment; 829 830 /* 831 * Optimal transfer length in bytes. A power of 2 is best but not 832 * mandatory. Must be a multiple of bl.request_alignment, or 0 if 833 * no preferred size 834 */ 835 uint32_t opt_transfer; 836 837 /* 838 * Maximal transfer length in bytes. Need not be power of 2, but 839 * must be multiple of opt_transfer and bl.request_alignment, or 0 840 * for no 32-bit limit. For now, anything larger than INT_MAX is 841 * clamped down. 842 */ 843 uint32_t max_transfer; 844 845 /* 846 * Maximal hardware transfer length in bytes. Applies whenever 847 * transfers to the device bypass the kernel I/O scheduler, for 848 * example with SG_IO. If larger than max_transfer or if zero, 849 * blk_get_max_hw_transfer will fall back to max_transfer. 850 */ 851 uint64_t max_hw_transfer; 852 853 /* 854 * Maximal number of scatter/gather elements allowed by the hardware. 855 * Applies whenever transfers to the device bypass the kernel I/O 856 * scheduler, for example with SG_IO. If larger than max_iov 857 * or if zero, blk_get_max_hw_iov will fall back to max_iov. 858 */ 859 int max_hw_iov; 860 861 862 /* memory alignment, in bytes so that no bounce buffer is needed */ 863 size_t min_mem_alignment; 864 865 /* memory alignment, in bytes, for bounce buffer */ 866 size_t opt_mem_alignment; 867 868 /* maximum number of iovec elements */ 869 int max_iov; 870 871 /* 872 * true if the length of the underlying file can change, and QEMU 873 * is expected to adjust automatically. Mostly for CD-ROM drives, 874 * whose length is zero when the tray is empty (they don't need 875 * an explicit monitor command to load the disk inside the guest). 876 */ 877 bool has_variable_length; 878 879 /* device zone model */ 880 BlockZoneModel zoned; 881 882 /* zone size expressed in bytes */ 883 uint32_t zone_size; 884 885 /* total number of zones */ 886 uint32_t nr_zones; 887 888 /* maximum sectors of a zone append write operation */ 889 uint32_t max_append_sectors; 890 891 /* maximum number of open zones */ 892 uint32_t max_open_zones; 893 894 /* maximum number of active zones */ 895 uint32_t max_active_zones; 896 897 uint32_t write_granularity; 898 } BlockLimits; 899 900 typedef struct BdrvOpBlocker BdrvOpBlocker; 901 902 typedef struct BdrvAioNotifier { 903 void (*attached_aio_context)(AioContext *new_context, void *opaque); 904 void (*detach_aio_context)(void *opaque); 905 906 void *opaque; 907 bool deleted; 908 909 QLIST_ENTRY(BdrvAioNotifier) list; 910 } BdrvAioNotifier; 911 912 struct BdrvChildClass { 913 /* 914 * If true, bdrv_replace_node() doesn't change the node this BdrvChild 915 * points to. 916 */ 917 bool stay_at_node; 918 919 /* 920 * If true, the parent is a BlockDriverState and bdrv_next_all_states() 921 * will return it. This information is used for drain_all, where every node 922 * will be drained separately, so the drain only needs to be propagated to 923 * non-BDS parents. 924 */ 925 bool parent_is_bds; 926 927 /* 928 * Global state (GS) API. These functions run under the BQL. 929 * 930 * See include/block/block-global-state.h for more information about 931 * the GS API. 932 */ 933 void (*inherit_options)(BdrvChildRole role, bool parent_is_format, 934 int *child_flags, QDict *child_options, 935 int parent_flags, QDict *parent_options); 936 void (*change_media)(BdrvChild *child, bool load); 937 938 /* 939 * Returns a malloced string that describes the parent of the child for a 940 * human reader. This could be a node-name, BlockBackend name, qdev ID or 941 * QOM path of the device owning the BlockBackend, job type and ID etc. The 942 * caller is responsible for freeing the memory. 943 */ 944 char *(*get_parent_desc)(BdrvChild *child); 945 946 /* 947 * Notifies the parent that the child has been activated/inactivated (e.g. 948 * when migration is completing) and it can start/stop requesting 949 * permissions and doing I/O on it. 950 */ 951 void GRAPH_RDLOCK_PTR (*activate)(BdrvChild *child, Error **errp); 952 int GRAPH_RDLOCK_PTR (*inactivate)(BdrvChild *child); 953 954 void GRAPH_WRLOCK_PTR (*attach)(BdrvChild *child); 955 void GRAPH_WRLOCK_PTR (*detach)(BdrvChild *child); 956 957 /* 958 * If this pair of functions is implemented, the parent doesn't issue new 959 * requests after returning from .drained_begin() until .drained_end() is 960 * called. 961 * 962 * These functions must not change the graph (and therefore also must not 963 * call aio_poll(), which could change the graph indirectly). 964 * 965 * Note that this can be nested. If drained_begin() was called twice, new 966 * I/O is allowed only after drained_end() was called twice, too. 967 */ 968 void GRAPH_RDLOCK_PTR (*drained_begin)(BdrvChild *child); 969 void GRAPH_RDLOCK_PTR (*drained_end)(BdrvChild *child); 970 971 /* 972 * Returns whether the parent has pending requests for the child. This 973 * callback is polled after .drained_begin() has been called until all 974 * activity on the child has stopped. 975 */ 976 bool GRAPH_RDLOCK_PTR (*drained_poll)(BdrvChild *child); 977 978 /* 979 * Notifies the parent that the filename of its child has changed (e.g. 980 * because the direct child was removed from the backing chain), so that it 981 * can update its reference. 982 */ 983 int (*update_filename)(BdrvChild *child, BlockDriverState *new_base, 984 const char *filename, Error **errp); 985 986 bool (*change_aio_ctx)(BdrvChild *child, AioContext *ctx, 987 GHashTable *visited, Transaction *tran, 988 Error **errp); 989 990 /* 991 * I/O API functions. These functions are thread-safe. 992 * 993 * See include/block/block-io.h for more information about 994 * the I/O API. 995 */ 996 997 void (*resize)(BdrvChild *child); 998 999 /* 1000 * Returns a name that is supposedly more useful for human users than the 1001 * node name for identifying the node in question (in particular, a BB 1002 * name), or NULL if the parent can't provide a better name. 1003 */ 1004 const char *(*get_name)(BdrvChild *child); 1005 1006 AioContext *(*get_parent_aio_context)(BdrvChild *child); 1007 }; 1008 1009 extern const BdrvChildClass child_of_bds; 1010 1011 struct BdrvChild { 1012 BlockDriverState *bs; 1013 char *name; 1014 const BdrvChildClass *klass; 1015 BdrvChildRole role; 1016 void *opaque; 1017 1018 /** 1019 * Granted permissions for operating on this BdrvChild (BLK_PERM_* bitmask) 1020 */ 1021 uint64_t perm; 1022 1023 /** 1024 * Permissions that can still be granted to other users of @bs while this 1025 * BdrvChild is still attached to it. (BLK_PERM_* bitmask) 1026 */ 1027 uint64_t shared_perm; 1028 1029 /* 1030 * This link is frozen: the child can neither be replaced nor 1031 * detached from the parent. 1032 */ 1033 bool frozen; 1034 1035 /* 1036 * True if the parent of this child has been drained by this BdrvChild 1037 * (through klass->drained_*). 1038 * 1039 * It is generally true if bs->quiesce_counter > 0. It may differ while the 1040 * child is entering or leaving a drained section. 1041 */ 1042 bool quiesced_parent; 1043 1044 QLIST_ENTRY(BdrvChild) next; 1045 QLIST_ENTRY(BdrvChild) next_parent; 1046 }; 1047 1048 /* 1049 * Allows bdrv_co_block_status() to cache one data region for a 1050 * protocol node. 1051 * 1052 * @valid: Whether the cache is valid (should be accessed with atomic 1053 * functions so this can be reset by RCU readers) 1054 * @data_start: Offset where we know (or strongly assume) is data 1055 * @data_end: Offset where the data region ends (which is not necessarily 1056 * the start of a zeroed region) 1057 */ 1058 typedef struct BdrvBlockStatusCache { 1059 struct rcu_head rcu; 1060 1061 bool valid; 1062 int64_t data_start; 1063 int64_t data_end; 1064 } BdrvBlockStatusCache; 1065 1066 struct BlockDriverState { 1067 /* 1068 * Protected by big QEMU lock or read-only after opening. No special 1069 * locking needed during I/O... 1070 */ 1071 int open_flags; /* flags used to open the file, re-used for re-open */ 1072 bool encrypted; /* if true, the media is encrypted */ 1073 bool sg; /* if true, the device is a /dev/sg* */ 1074 bool probed; /* if true, format was probed rather than specified */ 1075 bool force_share; /* if true, always allow all shared permissions */ 1076 bool implicit; /* if true, this filter node was automatically inserted */ 1077 1078 BlockDriver *drv; /* NULL means no media */ 1079 void *opaque; 1080 1081 AioContext *aio_context; /* event loop used for fd handlers, timers, etc */ 1082 /* 1083 * long-running tasks intended to always use the same AioContext as this 1084 * BDS may register themselves in this list to be notified of changes 1085 * regarding this BDS's context 1086 */ 1087 QLIST_HEAD(, BdrvAioNotifier) aio_notifiers; 1088 bool walking_aio_notifiers; /* to make removal during iteration safe */ 1089 1090 char filename[PATH_MAX]; 1091 /* 1092 * If not empty, this image is a diff in relation to backing_file. 1093 * Note that this is the name given in the image header and 1094 * therefore may or may not be equal to .backing->bs->filename. 1095 * If this field contains a relative path, it is to be resolved 1096 * relatively to the overlay's location. 1097 */ 1098 char backing_file[PATH_MAX]; 1099 /* 1100 * The backing filename indicated by the image header. Contrary 1101 * to backing_file, if we ever open this file, auto_backing_file 1102 * is replaced by the resulting BDS's filename (i.e. after a 1103 * bdrv_refresh_filename() run). 1104 */ 1105 char auto_backing_file[PATH_MAX]; 1106 char backing_format[16]; /* if non-zero and backing_file exists */ 1107 1108 QDict *full_open_options; 1109 char exact_filename[PATH_MAX]; 1110 1111 /* I/O Limits */ 1112 BlockLimits bl; 1113 1114 /* 1115 * Flags honored during pread 1116 */ 1117 BdrvRequestFlags supported_read_flags; 1118 /* 1119 * Flags honored during pwrite (so far: BDRV_REQ_FUA, 1120 * BDRV_REQ_WRITE_UNCHANGED). 1121 * If a driver does not support BDRV_REQ_WRITE_UNCHANGED, those 1122 * writes will be issued as normal writes without the flag set. 1123 * This is important to note for drivers that do not explicitly 1124 * request a WRITE permission for their children and instead take 1125 * the same permissions as their parent did (this is commonly what 1126 * block filters do). Such drivers have to be aware that the 1127 * parent may have taken a WRITE_UNCHANGED permission only and is 1128 * issuing such requests. Drivers either must make sure that 1129 * these requests do not result in plain WRITE accesses (usually 1130 * by supporting BDRV_REQ_WRITE_UNCHANGED, and then forwarding 1131 * every incoming write request as-is, including potentially that 1132 * flag), or they have to explicitly take the WRITE permission for 1133 * their children. 1134 */ 1135 BdrvRequestFlags supported_write_flags; 1136 /* 1137 * Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA, 1138 * BDRV_REQ_MAY_UNMAP, BDRV_REQ_WRITE_UNCHANGED) 1139 */ 1140 BdrvRequestFlags supported_zero_flags; 1141 /* 1142 * Flags honoured during truncate (so far: BDRV_REQ_ZERO_WRITE). 1143 * 1144 * If BDRV_REQ_ZERO_WRITE is given, the truncate operation must make sure 1145 * that any added space reads as all zeros. If this can't be guaranteed, 1146 * the operation must fail. 1147 */ 1148 BdrvRequestFlags supported_truncate_flags; 1149 1150 /* the following member gives a name to every node on the bs graph. */ 1151 char node_name[32]; 1152 /* element of the list of named nodes building the graph */ 1153 QTAILQ_ENTRY(BlockDriverState) node_list; 1154 /* element of the list of all BlockDriverStates (all_bdrv_states) */ 1155 QTAILQ_ENTRY(BlockDriverState) bs_list; 1156 /* element of the list of monitor-owned BDS */ 1157 QTAILQ_ENTRY(BlockDriverState) monitor_list; 1158 int refcnt; 1159 1160 /* operation blockers. Protected by BQL. */ 1161 QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX]; 1162 1163 /* 1164 * The node that this node inherited default options from (and a reopen on 1165 * which can affect this node by changing these defaults). This is always a 1166 * parent node of this node. 1167 */ 1168 BlockDriverState *inherits_from; 1169 1170 /* 1171 * @backing and @file are some of @children or NULL. All these three fields 1172 * (@file, @backing and @children) are modified only in 1173 * bdrv_child_cb_attach() and bdrv_child_cb_detach(). 1174 * 1175 * See also comment in include/block/block.h, to learn how backing and file 1176 * are connected with BdrvChildRole. 1177 */ 1178 QLIST_HEAD(, BdrvChild) children; 1179 BdrvChild *backing; 1180 BdrvChild *file; 1181 1182 QLIST_HEAD(, BdrvChild) parents; 1183 1184 QDict *options; 1185 QDict *explicit_options; 1186 BlockdevDetectZeroesOptions detect_zeroes; 1187 1188 /* The error object in use for blocking operations on backing_hd */ 1189 Error *backing_blocker; 1190 1191 /* Protected by AioContext lock */ 1192 1193 /* 1194 * If we are reading a disk image, give its size in sectors. 1195 * Generally read-only; it is written to by load_snapshot and 1196 * save_snaphost, but the block layer is quiescent during those. 1197 */ 1198 int64_t total_sectors; 1199 1200 /* threshold limit for writes, in bytes. "High water mark". */ 1201 uint64_t write_threshold_offset; 1202 1203 /* 1204 * Writing to the list requires the BQL _and_ the dirty_bitmap_mutex. 1205 * Reading from the list can be done with either the BQL or the 1206 * dirty_bitmap_mutex. Modifying a bitmap only requires 1207 * dirty_bitmap_mutex. 1208 */ 1209 QemuMutex dirty_bitmap_mutex; 1210 QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps; 1211 1212 /* Offset after the highest byte written to */ 1213 Stat64 wr_highest_offset; 1214 1215 /* 1216 * If true, copy read backing sectors into image. Can be >1 if more 1217 * than one client has requested copy-on-read. Accessed with atomic 1218 * ops. 1219 */ 1220 int copy_on_read; 1221 1222 /* 1223 * number of in-flight requests; overall and serialising. 1224 * Accessed with atomic ops. 1225 */ 1226 unsigned int in_flight; 1227 unsigned int serialising_in_flight; 1228 1229 /* do we need to tell the quest if we have a volatile write cache? */ 1230 int enable_write_cache; 1231 1232 /* Accessed with atomic ops. */ 1233 int quiesce_counter; 1234 1235 unsigned int write_gen; /* Current data generation */ 1236 1237 /* Protected by reqs_lock. */ 1238 QemuMutex reqs_lock; 1239 QLIST_HEAD(, BdrvTrackedRequest) tracked_requests; 1240 CoQueue flush_queue; /* Serializing flush queue */ 1241 bool active_flush_req; /* Flush request in flight? */ 1242 1243 /* Only read/written by whoever has set active_flush_req to true. */ 1244 unsigned int flushed_gen; /* Flushed write generation */ 1245 1246 /* BdrvChild links to this node may never be frozen */ 1247 bool never_freeze; 1248 1249 /* Lock for block-status cache RCU writers */ 1250 CoMutex bsc_modify_lock; 1251 /* Always non-NULL, but must only be dereferenced under an RCU read guard */ 1252 BdrvBlockStatusCache *block_status_cache; 1253 1254 /* array of write pointers' location of each zone in the zoned device. */ 1255 BlockZoneWps *wps; 1256 }; 1257 1258 struct BlockBackendRootState { 1259 int open_flags; 1260 BlockdevDetectZeroesOptions detect_zeroes; 1261 }; 1262 1263 typedef enum BlockMirrorBackingMode { 1264 /* 1265 * Reuse the existing backing chain from the source for the target. 1266 * - sync=full: Set backing BDS to NULL. 1267 * - sync=top: Use source's backing BDS. 1268 * - sync=none: Use source as the backing BDS. 1269 */ 1270 MIRROR_SOURCE_BACKING_CHAIN, 1271 1272 /* Open the target's backing chain completely anew */ 1273 MIRROR_OPEN_BACKING_CHAIN, 1274 1275 /* Do not change the target's backing BDS after job completion */ 1276 MIRROR_LEAVE_BACKING_CHAIN, 1277 } BlockMirrorBackingMode; 1278 1279 1280 /* 1281 * Essential block drivers which must always be statically linked into qemu, and 1282 * which therefore can be accessed without using bdrv_find_format() 1283 */ 1284 extern BlockDriver bdrv_file; 1285 extern BlockDriver bdrv_raw; 1286 extern BlockDriver bdrv_qcow2; 1287 1288 extern unsigned int bdrv_drain_all_count; 1289 extern QemuOptsList bdrv_create_opts_simple; 1290 1291 /* 1292 * Common functions that are neither I/O nor Global State. 1293 * 1294 * See include/block/block-common.h for more information about 1295 * the Common API. 1296 */ 1297 1298 static inline BlockDriverState *child_bs(BdrvChild *child) 1299 { 1300 return child ? child->bs : NULL; 1301 } 1302 1303 int bdrv_check_request(int64_t offset, int64_t bytes, Error **errp); 1304 char *create_tmp_file(Error **errp); 1305 void bdrv_parse_filename_strip_prefix(const char *filename, const char *prefix, 1306 QDict *options); 1307 1308 1309 int bdrv_check_qiov_request(int64_t offset, int64_t bytes, 1310 QEMUIOVector *qiov, size_t qiov_offset, 1311 Error **errp); 1312 1313 #ifdef _WIN32 1314 int is_windows_drive(const char *filename); 1315 #endif 1316 1317 #endif /* BLOCK_INT_COMMON_H */ 1318