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 GRAPH_UNLOCKED_PTR (*bdrv_reopen_prepare)( 239 BDRVReopenState *reopen_state, BlockReopenQueue *queue, Error **errp); 240 void GRAPH_UNLOCKED_PTR (*bdrv_reopen_commit)( 241 BDRVReopenState *reopen_state); 242 void GRAPH_UNLOCKED_PTR (*bdrv_reopen_commit_post)( 243 BDRVReopenState *reopen_state); 244 void GRAPH_UNLOCKED_PTR (*bdrv_reopen_abort)( 245 BDRVReopenState *reopen_state); 246 void (*bdrv_join_options)(QDict *options, QDict *old_options); 247 248 int GRAPH_UNLOCKED_PTR (*bdrv_open)( 249 BlockDriverState *bs, QDict *options, int flags, Error **errp); 250 251 /* Protocol drivers should implement this instead of bdrv_open */ 252 int GRAPH_UNLOCKED_PTR (*bdrv_file_open)( 253 BlockDriverState *bs, QDict *options, int flags, Error **errp); 254 void (*bdrv_close)(BlockDriverState *bs); 255 256 int coroutine_fn GRAPH_UNLOCKED_PTR (*bdrv_co_create)( 257 BlockdevCreateOptions *opts, Error **errp); 258 259 int coroutine_fn GRAPH_UNLOCKED_PTR (*bdrv_co_create_opts)( 260 BlockDriver *drv, const char *filename, QemuOpts *opts, Error **errp); 261 262 int GRAPH_RDLOCK_PTR (*bdrv_amend_options)( 263 BlockDriverState *bs, QemuOpts *opts, 264 BlockDriverAmendStatusCB *status_cb, void *cb_opaque, 265 bool force, Error **errp); 266 267 int GRAPH_RDLOCK_PTR (*bdrv_make_empty)(BlockDriverState *bs); 268 269 /* 270 * Refreshes the bs->exact_filename field. If that is impossible, 271 * bs->exact_filename has to be left empty. 272 */ 273 void GRAPH_RDLOCK_PTR (*bdrv_refresh_filename)(BlockDriverState *bs); 274 275 /* 276 * Gathers the open options for all children into @target. 277 * A simple format driver (without backing file support) might 278 * implement this function like this: 279 * 280 * QINCREF(bs->file->bs->full_open_options); 281 * qdict_put(target, "file", bs->file->bs->full_open_options); 282 * 283 * If not specified, the generic implementation will simply put 284 * all children's options under their respective name. 285 * 286 * @backing_overridden is true when bs->backing seems not to be 287 * the child that would result from opening bs->backing_file. 288 * Therefore, if it is true, the backing child's options should be 289 * gathered; otherwise, there is no need since the backing child 290 * is the one implied by the image header. 291 * 292 * Note that ideally this function would not be needed. Every 293 * block driver which implements it is probably doing something 294 * shady regarding its runtime option structure. 295 */ 296 void GRAPH_RDLOCK_PTR (*bdrv_gather_child_options)( 297 BlockDriverState *bs, QDict *target, bool backing_overridden); 298 299 /* 300 * Returns an allocated string which is the directory name of this BDS: It 301 * will be used to make relative filenames absolute by prepending this 302 * function's return value to them. 303 */ 304 char * GRAPH_RDLOCK_PTR (*bdrv_dirname)(BlockDriverState *bs, Error **errp); 305 306 /* 307 * This informs the driver that we are no longer interested in the result 308 * of in-flight requests, so don't waste the time if possible. 309 * 310 * One example usage is to avoid waiting for an nbd target node reconnect 311 * timeout during job-cancel with force=true. 312 */ 313 void (*bdrv_cancel_in_flight)(BlockDriverState *bs); 314 315 int GRAPH_RDLOCK_PTR (*bdrv_inactivate)(BlockDriverState *bs); 316 317 int GRAPH_RDLOCK_PTR (*bdrv_snapshot_create)( 318 BlockDriverState *bs, QEMUSnapshotInfo *sn_info); 319 320 int GRAPH_UNLOCKED_PTR (*bdrv_snapshot_goto)( 321 BlockDriverState *bs, const char *snapshot_id); 322 323 int GRAPH_RDLOCK_PTR (*bdrv_snapshot_delete)( 324 BlockDriverState *bs, const char *snapshot_id, const char *name, 325 Error **errp); 326 327 int (*bdrv_snapshot_list)(BlockDriverState *bs, 328 QEMUSnapshotInfo **psn_info); 329 int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs, 330 const char *snapshot_id, 331 const char *name, 332 Error **errp); 333 334 int (*bdrv_change_backing_file)(BlockDriverState *bs, 335 const char *backing_file, const char *backing_fmt); 336 337 /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */ 338 int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event, 339 const char *tag); 340 int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs, 341 const char *tag); 342 int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag); 343 bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag); 344 345 void GRAPH_RDLOCK_PTR (*bdrv_refresh_limits)( 346 BlockDriverState *bs, Error **errp); 347 348 /* 349 * Returns 1 if newly created images are guaranteed to contain only 350 * zeros, 0 otherwise. 351 */ 352 int (*bdrv_has_zero_init)(BlockDriverState *bs); 353 354 /* 355 * Remove fd handlers, timers, and other event loop callbacks so the event 356 * loop is no longer in use. Called with no in-flight requests and in 357 * depth-first traversal order with parents before child nodes. 358 */ 359 void (*bdrv_detach_aio_context)(BlockDriverState *bs); 360 361 /* 362 * Add fd handlers, timers, and other event loop callbacks so I/O requests 363 * can be processed again. Called with no in-flight requests and in 364 * depth-first traversal order with child nodes before parent nodes. 365 */ 366 void (*bdrv_attach_aio_context)(BlockDriverState *bs, 367 AioContext *new_context); 368 369 /** 370 * bdrv_drain_begin is called if implemented in the beginning of a 371 * drain operation to drain and stop any internal sources of requests in 372 * the driver. 373 * bdrv_drain_end is called if implemented at the end of the drain. 374 * 375 * They should be used by the driver to e.g. manage scheduled I/O 376 * requests, or toggle an internal state. After the end of the drain new 377 * requests will continue normally. 378 * 379 * Implementations of both functions must not call aio_poll(). 380 */ 381 void (*bdrv_drain_begin)(BlockDriverState *bs); 382 void (*bdrv_drain_end)(BlockDriverState *bs); 383 384 /** 385 * Try to get @bs's logical and physical block size. 386 * On success, store them in @bsz and return zero. 387 * On failure, return negative errno. 388 */ 389 int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz); 390 /** 391 * Try to get @bs's geometry (cyls, heads, sectors) 392 * On success, store them in @geo and return 0. 393 * On failure return -errno. 394 * Only drivers that want to override guest geometry implement this 395 * callback; see hd_geometry_guess(). 396 */ 397 int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo); 398 399 void GRAPH_WRLOCK_PTR (*bdrv_add_child)( 400 BlockDriverState *parent, BlockDriverState *child, Error **errp); 401 402 void GRAPH_WRLOCK_PTR (*bdrv_del_child)( 403 BlockDriverState *parent, BdrvChild *child, Error **errp); 404 405 /** 406 * Informs the block driver that a permission change is intended. The 407 * driver checks whether the change is permissible and may take other 408 * preparations for the change (e.g. get file system locks). This operation 409 * is always followed either by a call to either .bdrv_set_perm or 410 * .bdrv_abort_perm_update. 411 * 412 * Checks whether the requested set of cumulative permissions in @perm 413 * can be granted for accessing @bs and whether no other users are using 414 * permissions other than those given in @shared (both arguments take 415 * BLK_PERM_* bitmasks). 416 * 417 * If both conditions are met, 0 is returned. Otherwise, -errno is returned 418 * and errp is set to an error describing the conflict. 419 */ 420 int GRAPH_RDLOCK_PTR (*bdrv_check_perm)(BlockDriverState *bs, uint64_t perm, 421 uint64_t shared, Error **errp); 422 423 /** 424 * Called to inform the driver that the set of cumulative set of used 425 * permissions for @bs has changed to @perm, and the set of shareable 426 * permission to @shared. The driver can use this to propagate changes to 427 * its children (i.e. request permissions only if a parent actually needs 428 * them). 429 * 430 * This function is only invoked after bdrv_check_perm(), so block drivers 431 * may rely on preparations made in their .bdrv_check_perm implementation. 432 */ 433 void GRAPH_RDLOCK_PTR (*bdrv_set_perm)( 434 BlockDriverState *bs, uint64_t perm, uint64_t shared); 435 436 /* 437 * Called to inform the driver that after a previous bdrv_check_perm() 438 * call, the permission update is not performed and any preparations made 439 * for it (e.g. taken file locks) need to be undone. 440 * 441 * This function can be called even for nodes that never saw a 442 * bdrv_check_perm() call. It is a no-op then. 443 */ 444 void GRAPH_RDLOCK_PTR (*bdrv_abort_perm_update)(BlockDriverState *bs); 445 446 /** 447 * Returns in @nperm and @nshared the permissions that the driver for @bs 448 * needs on its child @c, based on the cumulative permissions requested by 449 * the parents in @parent_perm and @parent_shared. 450 * 451 * If @c is NULL, return the permissions for attaching a new child for the 452 * given @child_class and @role. 453 * 454 * If @reopen_queue is non-NULL, don't return the currently needed 455 * permissions, but those that will be needed after applying the 456 * @reopen_queue. 457 */ 458 void GRAPH_RDLOCK_PTR (*bdrv_child_perm)( 459 BlockDriverState *bs, BdrvChild *c, BdrvChildRole role, 460 BlockReopenQueue *reopen_queue, 461 uint64_t parent_perm, uint64_t parent_shared, 462 uint64_t *nperm, uint64_t *nshared); 463 464 /** 465 * Register/unregister a buffer for I/O. For example, when the driver is 466 * interested to know the memory areas that will later be used in iovs, so 467 * that it can do IOMMU mapping with VFIO etc., in order to get better 468 * performance. In the case of VFIO drivers, this callback is used to do 469 * DMA mapping for hot buffers. 470 * 471 * Returns: true on success, false on failure 472 */ 473 bool GRAPH_RDLOCK_PTR (*bdrv_register_buf)( 474 BlockDriverState *bs, void *host, size_t size, Error **errp); 475 void GRAPH_RDLOCK_PTR (*bdrv_unregister_buf)( 476 BlockDriverState *bs, void *host, size_t size); 477 478 /* 479 * This field is modified only under the BQL, and is part of 480 * the global state. 481 */ 482 QLIST_ENTRY(BlockDriver) list; 483 484 /* 485 * I/O API functions. These functions are thread-safe. 486 * 487 * See include/block/block-io.h for more information about 488 * the I/O API. 489 */ 490 491 int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename); 492 493 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_amend)( 494 BlockDriverState *bs, BlockdevAmendOptions *opts, bool force, 495 Error **errp); 496 497 /* aio */ 498 BlockAIOCB * GRAPH_RDLOCK_PTR (*bdrv_aio_preadv)(BlockDriverState *bs, 499 int64_t offset, int64_t bytes, QEMUIOVector *qiov, 500 BdrvRequestFlags flags, BlockCompletionFunc *cb, void *opaque); 501 502 BlockAIOCB * GRAPH_RDLOCK_PTR (*bdrv_aio_pwritev)(BlockDriverState *bs, 503 int64_t offset, int64_t bytes, QEMUIOVector *qiov, 504 BdrvRequestFlags flags, BlockCompletionFunc *cb, void *opaque); 505 506 BlockAIOCB * GRAPH_RDLOCK_PTR (*bdrv_aio_flush)( 507 BlockDriverState *bs, BlockCompletionFunc *cb, void *opaque); 508 509 BlockAIOCB * GRAPH_RDLOCK_PTR (*bdrv_aio_pdiscard)( 510 BlockDriverState *bs, int64_t offset, int bytes, 511 BlockCompletionFunc *cb, void *opaque); 512 513 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_readv)(BlockDriverState *bs, 514 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov); 515 516 /** 517 * @offset: position in bytes to read at 518 * @bytes: number of bytes to read 519 * @qiov: the buffers to fill with read data 520 * @flags: currently unused, always 0 521 * 522 * @offset and @bytes will be a multiple of 'request_alignment', 523 * but the length of individual @qiov elements does not have to 524 * be a multiple. 525 * 526 * @bytes will always equal the total size of @qiov, and will be 527 * no larger than 'max_transfer'. 528 * 529 * The buffer in @qiov may point directly to guest memory. 530 */ 531 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_preadv)(BlockDriverState *bs, 532 int64_t offset, int64_t bytes, QEMUIOVector *qiov, 533 BdrvRequestFlags flags); 534 535 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_preadv_part)( 536 BlockDriverState *bs, int64_t offset, int64_t bytes, 537 QEMUIOVector *qiov, size_t qiov_offset, 538 BdrvRequestFlags flags); 539 540 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_writev)(BlockDriverState *bs, 541 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, 542 int flags); 543 /** 544 * @offset: position in bytes to write at 545 * @bytes: number of bytes to write 546 * @qiov: the buffers containing data to write 547 * @flags: zero or more bits allowed by 'supported_write_flags' 548 * 549 * @offset and @bytes will be a multiple of 'request_alignment', 550 * but the length of individual @qiov elements does not have to 551 * be a multiple. 552 * 553 * @bytes will always equal the total size of @qiov, and will be 554 * no larger than 'max_transfer'. 555 * 556 * The buffer in @qiov may point directly to guest memory. 557 */ 558 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_pwritev)( 559 BlockDriverState *bs, int64_t offset, int64_t bytes, QEMUIOVector *qiov, 560 BdrvRequestFlags flags); 561 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_pwritev_part)( 562 BlockDriverState *bs, int64_t offset, int64_t bytes, QEMUIOVector *qiov, 563 size_t qiov_offset, BdrvRequestFlags flags); 564 565 /* 566 * Efficiently zero a region of the disk image. Typically an image format 567 * would use a compact metadata representation to implement this. This 568 * function pointer may be NULL or return -ENOSUP and .bdrv_co_writev() 569 * will be called instead. 570 */ 571 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_pwrite_zeroes)( 572 BlockDriverState *bs, int64_t offset, int64_t bytes, 573 BdrvRequestFlags flags); 574 575 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_pdiscard)( 576 BlockDriverState *bs, int64_t offset, int64_t bytes); 577 578 /* 579 * Map [offset, offset + nbytes) range onto a child of @bs to copy from, 580 * and invoke bdrv_co_copy_range_from(child, ...), or invoke 581 * bdrv_co_copy_range_to() if @bs is the leaf child to copy data from. 582 * 583 * See the comment of bdrv_co_copy_range for the parameter and return value 584 * semantics. 585 */ 586 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_copy_range_from)( 587 BlockDriverState *bs, BdrvChild *src, int64_t offset, 588 BdrvChild *dst, int64_t dst_offset, int64_t bytes, 589 BdrvRequestFlags read_flags, BdrvRequestFlags write_flags); 590 591 /* 592 * Map [offset, offset + nbytes) range onto a child of bs to copy data to, 593 * and invoke bdrv_co_copy_range_to(child, src, ...), or perform the copy 594 * operation if @bs is the leaf and @src has the same BlockDriver. Return 595 * -ENOTSUP if @bs is the leaf but @src has a different BlockDriver. 596 * 597 * See the comment of bdrv_co_copy_range for the parameter and return value 598 * semantics. 599 */ 600 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_copy_range_to)( 601 BlockDriverState *bs, BdrvChild *src, int64_t src_offset, 602 BdrvChild *dst, int64_t dst_offset, int64_t bytes, 603 BdrvRequestFlags read_flags, BdrvRequestFlags write_flags); 604 605 /* 606 * Building block for bdrv_block_status[_above] and 607 * bdrv_is_allocated[_above]. The driver should answer only 608 * according to the current layer, and should only need to set 609 * BDRV_BLOCK_DATA, BDRV_BLOCK_ZERO, BDRV_BLOCK_OFFSET_VALID, 610 * and/or BDRV_BLOCK_RAW; if the current layer defers to a backing 611 * layer, the result should be 0 (and not BDRV_BLOCK_ZERO). See 612 * block.h for the overall meaning of the bits. As a hint, the 613 * flag want_zero is true if the caller cares more about precise 614 * mappings (favor accurate _OFFSET_VALID/_ZERO) or false for 615 * overall allocation (favor larger *pnum, perhaps by reporting 616 * _DATA instead of _ZERO). The block layer guarantees input 617 * clamped to bdrv_getlength() and aligned to request_alignment, 618 * as well as non-NULL pnum, map, and file; in turn, the driver 619 * must return an error or set pnum to an aligned non-zero value. 620 * 621 * Note that @bytes is just a hint on how big of a region the 622 * caller wants to inspect. It is not a limit on *pnum. 623 * Implementations are free to return larger values of *pnum if 624 * doing so does not incur a performance penalty. 625 * 626 * block/io.c's bdrv_co_block_status() will utilize an unclamped 627 * *pnum value for the block-status cache on protocol nodes, prior 628 * to clamping *pnum for return to its caller. 629 */ 630 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_block_status)( 631 BlockDriverState *bs, 632 bool want_zero, int64_t offset, int64_t bytes, int64_t *pnum, 633 int64_t *map, BlockDriverState **file); 634 635 /* 636 * Snapshot-access API. 637 * 638 * Block-driver may provide snapshot-access API: special functions to access 639 * some internal "snapshot". The functions are similar with normal 640 * read/block_status/discard handler, but don't have any specific handling 641 * in generic block-layer: no serializing, no alignment, no tracked 642 * requests. So, block-driver that realizes these APIs is fully responsible 643 * for synchronization between snapshot-access API and normal IO requests. 644 * 645 * TODO: To be able to support qcow2's internal snapshots, this API will 646 * need to be extended to: 647 * - be able to select a specific snapshot 648 * - receive the snapshot's actual length (which may differ from bs's 649 * length) 650 */ 651 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_preadv_snapshot)( 652 BlockDriverState *bs, int64_t offset, int64_t bytes, 653 QEMUIOVector *qiov, size_t qiov_offset); 654 655 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_snapshot_block_status)( 656 BlockDriverState *bs, bool want_zero, int64_t offset, int64_t bytes, 657 int64_t *pnum, int64_t *map, BlockDriverState **file); 658 659 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_pdiscard_snapshot)( 660 BlockDriverState *bs, int64_t offset, int64_t bytes); 661 662 /* 663 * Invalidate any cached meta-data. 664 */ 665 void coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_invalidate_cache)( 666 BlockDriverState *bs, Error **errp); 667 668 /* 669 * Flushes all data for all layers by calling bdrv_co_flush for underlying 670 * layers, if needed. This function is needed for deterministic 671 * synchronization of the flush finishing callback. 672 */ 673 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_flush)(BlockDriverState *bs); 674 675 /* Delete a created file. */ 676 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_delete_file)( 677 BlockDriverState *bs, Error **errp); 678 679 /* 680 * Flushes all data that was already written to the OS all the way down to 681 * the disk (for example file-posix.c calls fsync()). 682 */ 683 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_flush_to_disk)( 684 BlockDriverState *bs); 685 686 /* 687 * Flushes all internal caches to the OS. The data may still sit in a 688 * writeback cache of the host OS, but it will survive a crash of the qemu 689 * process. 690 */ 691 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_flush_to_os)( 692 BlockDriverState *bs); 693 694 /* 695 * Truncate @bs to @offset bytes using the given @prealloc mode 696 * when growing. Modes other than PREALLOC_MODE_OFF should be 697 * rejected when shrinking @bs. 698 * 699 * If @exact is true, @bs must be resized to exactly @offset. 700 * Otherwise, it is sufficient for @bs (if it is a host block 701 * device and thus there is no way to resize it) to be at least 702 * @offset bytes in length. 703 * 704 * If @exact is true and this function fails but would succeed 705 * with @exact = false, it should return -ENOTSUP. 706 */ 707 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_truncate)( 708 BlockDriverState *bs, int64_t offset, bool exact, 709 PreallocMode prealloc, BdrvRequestFlags flags, Error **errp); 710 711 int64_t coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_getlength)( 712 BlockDriverState *bs); 713 714 int64_t coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_get_allocated_file_size)( 715 BlockDriverState *bs); 716 717 BlockMeasureInfo *(*bdrv_measure)(QemuOpts *opts, BlockDriverState *in_bs, 718 Error **errp); 719 720 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_pwritev_compressed)( 721 BlockDriverState *bs, int64_t offset, int64_t bytes, 722 QEMUIOVector *qiov); 723 724 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_pwritev_compressed_part)( 725 BlockDriverState *bs, int64_t offset, int64_t bytes, 726 QEMUIOVector *qiov, size_t qiov_offset); 727 728 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_get_info)( 729 BlockDriverState *bs, BlockDriverInfo *bdi); 730 731 ImageInfoSpecific * GRAPH_RDLOCK_PTR (*bdrv_get_specific_info)( 732 BlockDriverState *bs, Error **errp); 733 BlockStatsSpecific *(*bdrv_get_specific_stats)(BlockDriverState *bs); 734 735 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_save_vmstate)( 736 BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos); 737 738 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_load_vmstate)( 739 BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos); 740 741 int coroutine_fn (*bdrv_co_zone_report)(BlockDriverState *bs, 742 int64_t offset, unsigned int *nr_zones, 743 BlockZoneDescriptor *zones); 744 int coroutine_fn (*bdrv_co_zone_mgmt)(BlockDriverState *bs, BlockZoneOp op, 745 int64_t offset, int64_t len); 746 int coroutine_fn (*bdrv_co_zone_append)(BlockDriverState *bs, 747 int64_t *offset, QEMUIOVector *qiov, 748 BdrvRequestFlags flags); 749 750 /* removable device specific */ 751 bool coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_is_inserted)( 752 BlockDriverState *bs); 753 void coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_eject)( 754 BlockDriverState *bs, bool eject_flag); 755 void coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_lock_medium)( 756 BlockDriverState *bs, bool locked); 757 758 /* to control generic scsi devices */ 759 BlockAIOCB *coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_aio_ioctl)( 760 BlockDriverState *bs, unsigned long int req, void *buf, 761 BlockCompletionFunc *cb, void *opaque); 762 763 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_ioctl)( 764 BlockDriverState *bs, unsigned long int req, void *buf); 765 766 /* 767 * Returns 0 for completed check, -errno for internal errors. 768 * The check results are stored in result. 769 */ 770 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_check)( 771 BlockDriverState *bs, BdrvCheckResult *result, BdrvCheckMode fix); 772 773 void coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_debug_event)( 774 BlockDriverState *bs, BlkdebugEvent event); 775 776 bool (*bdrv_supports_persistent_dirty_bitmap)(BlockDriverState *bs); 777 778 bool coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_can_store_new_dirty_bitmap)( 779 BlockDriverState *bs, const char *name, uint32_t granularity, 780 Error **errp); 781 782 int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_remove_persistent_dirty_bitmap)( 783 BlockDriverState *bs, const char *name, Error **errp); 784 }; 785 786 static inline bool TSA_NO_TSA block_driver_can_compress(BlockDriver *drv) 787 { 788 return drv->bdrv_co_pwritev_compressed || 789 drv->bdrv_co_pwritev_compressed_part; 790 } 791 792 typedef struct BlockLimits { 793 /* 794 * Alignment requirement, in bytes, for offset/length of I/O 795 * requests. Must be a power of 2 less than INT_MAX; defaults to 796 * 1 for drivers with modern byte interfaces, and to 512 797 * otherwise. 798 */ 799 uint32_t request_alignment; 800 801 /* 802 * Maximum number of bytes that can be discarded at once. Must be multiple 803 * of pdiscard_alignment, but need not be power of 2. May be 0 if no 804 * inherent 64-bit limit. 805 */ 806 int64_t max_pdiscard; 807 808 /* 809 * Optimal alignment for discard requests in bytes. A power of 2 810 * is best but not mandatory. Must be a multiple of 811 * bl.request_alignment, and must be less than max_pdiscard if 812 * that is set. May be 0 if bl.request_alignment is good enough 813 */ 814 uint32_t pdiscard_alignment; 815 816 /* 817 * Maximum number of bytes that can zeroized at once. Must be multiple of 818 * pwrite_zeroes_alignment. 0 means no limit. 819 */ 820 int64_t max_pwrite_zeroes; 821 822 /* 823 * Optimal alignment for write zeroes requests in bytes. A power 824 * of 2 is best but not mandatory. Must be a multiple of 825 * bl.request_alignment, and must be less than max_pwrite_zeroes 826 * if that is set. May be 0 if bl.request_alignment is good 827 * enough 828 */ 829 uint32_t pwrite_zeroes_alignment; 830 831 /* 832 * Optimal transfer length in bytes. A power of 2 is best but not 833 * mandatory. Must be a multiple of bl.request_alignment, or 0 if 834 * no preferred size 835 */ 836 uint32_t opt_transfer; 837 838 /* 839 * Maximal transfer length in bytes. Need not be power of 2, but 840 * must be multiple of opt_transfer and bl.request_alignment, or 0 841 * for no 32-bit limit. For now, anything larger than INT_MAX is 842 * clamped down. 843 */ 844 uint32_t max_transfer; 845 846 /* 847 * Maximal hardware transfer length in bytes. Applies whenever 848 * transfers to the device bypass the kernel I/O scheduler, for 849 * example with SG_IO. If larger than max_transfer or if zero, 850 * blk_get_max_hw_transfer will fall back to max_transfer. 851 */ 852 uint64_t max_hw_transfer; 853 854 /* 855 * Maximal number of scatter/gather elements allowed by the hardware. 856 * Applies whenever transfers to the device bypass the kernel I/O 857 * scheduler, for example with SG_IO. If larger than max_iov 858 * or if zero, blk_get_max_hw_iov will fall back to max_iov. 859 */ 860 int max_hw_iov; 861 862 863 /* memory alignment, in bytes so that no bounce buffer is needed */ 864 size_t min_mem_alignment; 865 866 /* memory alignment, in bytes, for bounce buffer */ 867 size_t opt_mem_alignment; 868 869 /* maximum number of iovec elements */ 870 int max_iov; 871 872 /* 873 * true if the length of the underlying file can change, and QEMU 874 * is expected to adjust automatically. Mostly for CD-ROM drives, 875 * whose length is zero when the tray is empty (they don't need 876 * an explicit monitor command to load the disk inside the guest). 877 */ 878 bool has_variable_length; 879 880 /* device zone model */ 881 BlockZoneModel zoned; 882 883 /* zone size expressed in bytes */ 884 uint32_t zone_size; 885 886 /* total number of zones */ 887 uint32_t nr_zones; 888 889 /* maximum sectors of a zone append write operation */ 890 uint32_t max_append_sectors; 891 892 /* maximum number of open zones */ 893 uint32_t max_open_zones; 894 895 /* maximum number of active zones */ 896 uint32_t max_active_zones; 897 898 uint32_t write_granularity; 899 } BlockLimits; 900 901 typedef struct BdrvOpBlocker BdrvOpBlocker; 902 903 typedef struct BdrvAioNotifier { 904 void (*attached_aio_context)(AioContext *new_context, void *opaque); 905 void (*detach_aio_context)(void *opaque); 906 907 void *opaque; 908 bool deleted; 909 910 QLIST_ENTRY(BdrvAioNotifier) list; 911 } BdrvAioNotifier; 912 913 struct BdrvChildClass { 914 /* 915 * If true, bdrv_replace_node() doesn't change the node this BdrvChild 916 * points to. 917 */ 918 bool stay_at_node; 919 920 /* 921 * If true, the parent is a BlockDriverState and bdrv_next_all_states() 922 * will return it. This information is used for drain_all, where every node 923 * will be drained separately, so the drain only needs to be propagated to 924 * non-BDS parents. 925 */ 926 bool parent_is_bds; 927 928 /* 929 * Global state (GS) API. These functions run under the BQL. 930 * 931 * See include/block/block-global-state.h for more information about 932 * the GS API. 933 */ 934 void (*inherit_options)(BdrvChildRole role, bool parent_is_format, 935 int *child_flags, QDict *child_options, 936 int parent_flags, QDict *parent_options); 937 void (*change_media)(BdrvChild *child, bool load); 938 939 /* 940 * Returns a malloced string that describes the parent of the child for a 941 * human reader. This could be a node-name, BlockBackend name, qdev ID or 942 * QOM path of the device owning the BlockBackend, job type and ID etc. The 943 * caller is responsible for freeing the memory. 944 */ 945 char *(*get_parent_desc)(BdrvChild *child); 946 947 /* 948 * Notifies the parent that the child has been activated/inactivated (e.g. 949 * when migration is completing) and it can start/stop requesting 950 * permissions and doing I/O on it. 951 */ 952 void GRAPH_RDLOCK_PTR (*activate)(BdrvChild *child, Error **errp); 953 int GRAPH_RDLOCK_PTR (*inactivate)(BdrvChild *child); 954 955 void GRAPH_WRLOCK_PTR (*attach)(BdrvChild *child); 956 void GRAPH_WRLOCK_PTR (*detach)(BdrvChild *child); 957 958 /* 959 * If this pair of functions is implemented, the parent doesn't issue new 960 * requests after returning from .drained_begin() until .drained_end() is 961 * called. 962 * 963 * These functions must not change the graph (and therefore also must not 964 * call aio_poll(), which could change the graph indirectly). 965 * 966 * Note that this can be nested. If drained_begin() was called twice, new 967 * I/O is allowed only after drained_end() was called twice, too. 968 */ 969 void GRAPH_RDLOCK_PTR (*drained_begin)(BdrvChild *child); 970 void GRAPH_RDLOCK_PTR (*drained_end)(BdrvChild *child); 971 972 /* 973 * Returns whether the parent has pending requests for the child. This 974 * callback is polled after .drained_begin() has been called until all 975 * activity on the child has stopped. 976 */ 977 bool GRAPH_RDLOCK_PTR (*drained_poll)(BdrvChild *child); 978 979 /* 980 * Notifies the parent that the filename of its child has changed (e.g. 981 * because the direct child was removed from the backing chain), so that it 982 * can update its reference. 983 */ 984 int (*update_filename)(BdrvChild *child, BlockDriverState *new_base, 985 const char *filename, Error **errp); 986 987 bool (*change_aio_ctx)(BdrvChild *child, AioContext *ctx, 988 GHashTable *visited, Transaction *tran, 989 Error **errp); 990 991 /* 992 * I/O API functions. These functions are thread-safe. 993 * 994 * See include/block/block-io.h for more information about 995 * the I/O API. 996 */ 997 998 void (*resize)(BdrvChild *child); 999 1000 /* 1001 * Returns a name that is supposedly more useful for human users than the 1002 * node name for identifying the node in question (in particular, a BB 1003 * name), or NULL if the parent can't provide a better name. 1004 */ 1005 const char *(*get_name)(BdrvChild *child); 1006 1007 AioContext *(*get_parent_aio_context)(BdrvChild *child); 1008 }; 1009 1010 extern const BdrvChildClass child_of_bds; 1011 1012 struct BdrvChild { 1013 BlockDriverState *bs; 1014 char *name; 1015 const BdrvChildClass *klass; 1016 BdrvChildRole role; 1017 void *opaque; 1018 1019 /** 1020 * Granted permissions for operating on this BdrvChild (BLK_PERM_* bitmask) 1021 */ 1022 uint64_t perm; 1023 1024 /** 1025 * Permissions that can still be granted to other users of @bs while this 1026 * BdrvChild is still attached to it. (BLK_PERM_* bitmask) 1027 */ 1028 uint64_t shared_perm; 1029 1030 /* 1031 * This link is frozen: the child can neither be replaced nor 1032 * detached from the parent. 1033 */ 1034 bool frozen; 1035 1036 /* 1037 * True if the parent of this child has been drained by this BdrvChild 1038 * (through klass->drained_*). 1039 * 1040 * It is generally true if bs->quiesce_counter > 0. It may differ while the 1041 * child is entering or leaving a drained section. 1042 */ 1043 bool quiesced_parent; 1044 1045 QLIST_ENTRY(BdrvChild GRAPH_RDLOCK_PTR) next; 1046 QLIST_ENTRY(BdrvChild GRAPH_RDLOCK_PTR) next_parent; 1047 }; 1048 1049 /* 1050 * Allows bdrv_co_block_status() to cache one data region for a 1051 * protocol node. 1052 * 1053 * @valid: Whether the cache is valid (should be accessed with atomic 1054 * functions so this can be reset by RCU readers) 1055 * @data_start: Offset where we know (or strongly assume) is data 1056 * @data_end: Offset where the data region ends (which is not necessarily 1057 * the start of a zeroed region) 1058 */ 1059 typedef struct BdrvBlockStatusCache { 1060 struct rcu_head rcu; 1061 1062 bool valid; 1063 int64_t data_start; 1064 int64_t data_end; 1065 } BdrvBlockStatusCache; 1066 1067 struct BlockDriverState { 1068 /* 1069 * Protected by big QEMU lock or read-only after opening. No special 1070 * locking needed during I/O... 1071 */ 1072 int open_flags; /* flags used to open the file, re-used for re-open */ 1073 bool encrypted; /* if true, the media is encrypted */ 1074 bool sg; /* if true, the device is a /dev/sg* */ 1075 bool probed; /* if true, format was probed rather than specified */ 1076 bool force_share; /* if true, always allow all shared permissions */ 1077 bool implicit; /* if true, this filter node was automatically inserted */ 1078 1079 BlockDriver *drv; /* NULL means no media */ 1080 void *opaque; 1081 1082 AioContext *aio_context; /* event loop used for fd handlers, timers, etc */ 1083 /* 1084 * long-running tasks intended to always use the same AioContext as this 1085 * BDS may register themselves in this list to be notified of changes 1086 * regarding this BDS's context 1087 */ 1088 QLIST_HEAD(, BdrvAioNotifier) aio_notifiers; 1089 bool walking_aio_notifiers; /* to make removal during iteration safe */ 1090 1091 char filename[PATH_MAX]; 1092 /* 1093 * If not empty, this image is a diff in relation to backing_file. 1094 * Note that this is the name given in the image header and 1095 * therefore may or may not be equal to .backing->bs->filename. 1096 * If this field contains a relative path, it is to be resolved 1097 * relatively to the overlay's location. 1098 */ 1099 char backing_file[PATH_MAX]; 1100 /* 1101 * The backing filename indicated by the image header. Contrary 1102 * to backing_file, if we ever open this file, auto_backing_file 1103 * is replaced by the resulting BDS's filename (i.e. after a 1104 * bdrv_refresh_filename() run). 1105 */ 1106 char auto_backing_file[PATH_MAX]; 1107 char backing_format[16]; /* if non-zero and backing_file exists */ 1108 1109 QDict *full_open_options; 1110 char exact_filename[PATH_MAX]; 1111 1112 /* I/O Limits */ 1113 BlockLimits bl; 1114 1115 /* 1116 * Flags honored during pread 1117 */ 1118 BdrvRequestFlags supported_read_flags; 1119 /* 1120 * Flags honored during pwrite (so far: BDRV_REQ_FUA, 1121 * BDRV_REQ_WRITE_UNCHANGED). 1122 * If a driver does not support BDRV_REQ_WRITE_UNCHANGED, those 1123 * writes will be issued as normal writes without the flag set. 1124 * This is important to note for drivers that do not explicitly 1125 * request a WRITE permission for their children and instead take 1126 * the same permissions as their parent did (this is commonly what 1127 * block filters do). Such drivers have to be aware that the 1128 * parent may have taken a WRITE_UNCHANGED permission only and is 1129 * issuing such requests. Drivers either must make sure that 1130 * these requests do not result in plain WRITE accesses (usually 1131 * by supporting BDRV_REQ_WRITE_UNCHANGED, and then forwarding 1132 * every incoming write request as-is, including potentially that 1133 * flag), or they have to explicitly take the WRITE permission for 1134 * their children. 1135 */ 1136 BdrvRequestFlags supported_write_flags; 1137 /* 1138 * Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA, 1139 * BDRV_REQ_MAY_UNMAP, BDRV_REQ_WRITE_UNCHANGED) 1140 */ 1141 BdrvRequestFlags supported_zero_flags; 1142 /* 1143 * Flags honoured during truncate (so far: BDRV_REQ_ZERO_WRITE). 1144 * 1145 * If BDRV_REQ_ZERO_WRITE is given, the truncate operation must make sure 1146 * that any added space reads as all zeros. If this can't be guaranteed, 1147 * the operation must fail. 1148 */ 1149 BdrvRequestFlags supported_truncate_flags; 1150 1151 /* the following member gives a name to every node on the bs graph. */ 1152 char node_name[32]; 1153 /* element of the list of named nodes building the graph */ 1154 QTAILQ_ENTRY(BlockDriverState) node_list; 1155 /* element of the list of all BlockDriverStates (all_bdrv_states) */ 1156 QTAILQ_ENTRY(BlockDriverState) bs_list; 1157 /* element of the list of monitor-owned BDS */ 1158 QTAILQ_ENTRY(BlockDriverState) monitor_list; 1159 int refcnt; 1160 1161 /* operation blockers. Protected by BQL. */ 1162 QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX]; 1163 1164 /* 1165 * The node that this node inherited default options from (and a reopen on 1166 * which can affect this node by changing these defaults). This is always a 1167 * parent node of this node. 1168 */ 1169 BlockDriverState *inherits_from; 1170 1171 /* 1172 * @backing and @file are some of @children or NULL. All these three fields 1173 * (@file, @backing and @children) are modified only in 1174 * bdrv_child_cb_attach() and bdrv_child_cb_detach(). 1175 * 1176 * See also comment in include/block/block.h, to learn how backing and file 1177 * are connected with BdrvChildRole. 1178 */ 1179 QLIST_HEAD(, BdrvChild GRAPH_RDLOCK_PTR) children; 1180 BdrvChild *backing; 1181 BdrvChild *file; 1182 1183 QLIST_HEAD(, BdrvChild GRAPH_RDLOCK_PTR) parents; 1184 1185 QDict *options; 1186 QDict *explicit_options; 1187 BlockdevDetectZeroesOptions detect_zeroes; 1188 1189 /* The error object in use for blocking operations on backing_hd */ 1190 Error *backing_blocker; 1191 1192 /* Protected by AioContext lock */ 1193 1194 /* 1195 * If we are reading a disk image, give its size in sectors. 1196 * Generally read-only; it is written to by load_snapshot and 1197 * save_snaphost, but the block layer is quiescent during those. 1198 */ 1199 int64_t total_sectors; 1200 1201 /* threshold limit for writes, in bytes. "High water mark". */ 1202 uint64_t write_threshold_offset; 1203 1204 /* 1205 * Writing to the list requires the BQL _and_ the dirty_bitmap_mutex. 1206 * Reading from the list can be done with either the BQL or the 1207 * dirty_bitmap_mutex. Modifying a bitmap only requires 1208 * dirty_bitmap_mutex. 1209 */ 1210 QemuMutex dirty_bitmap_mutex; 1211 QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps; 1212 1213 /* Offset after the highest byte written to */ 1214 Stat64 wr_highest_offset; 1215 1216 /* 1217 * If true, copy read backing sectors into image. Can be >1 if more 1218 * than one client has requested copy-on-read. Accessed with atomic 1219 * ops. 1220 */ 1221 int copy_on_read; 1222 1223 /* 1224 * number of in-flight requests; overall and serialising. 1225 * Accessed with atomic ops. 1226 */ 1227 unsigned int in_flight; 1228 unsigned int serialising_in_flight; 1229 1230 /* do we need to tell the quest if we have a volatile write cache? */ 1231 int enable_write_cache; 1232 1233 /* Accessed with atomic ops. */ 1234 int quiesce_counter; 1235 1236 unsigned int write_gen; /* Current data generation */ 1237 1238 /* Protected by reqs_lock. */ 1239 QemuMutex reqs_lock; 1240 QLIST_HEAD(, BdrvTrackedRequest) tracked_requests; 1241 CoQueue flush_queue; /* Serializing flush queue */ 1242 bool active_flush_req; /* Flush request in flight? */ 1243 1244 /* Only read/written by whoever has set active_flush_req to true. */ 1245 unsigned int flushed_gen; /* Flushed write generation */ 1246 1247 /* BdrvChild links to this node may never be frozen */ 1248 bool never_freeze; 1249 1250 /* Lock for block-status cache RCU writers */ 1251 CoMutex bsc_modify_lock; 1252 /* Always non-NULL, but must only be dereferenced under an RCU read guard */ 1253 BdrvBlockStatusCache *block_status_cache; 1254 1255 /* array of write pointers' location of each zone in the zoned device. */ 1256 BlockZoneWps *wps; 1257 }; 1258 1259 struct BlockBackendRootState { 1260 int open_flags; 1261 BlockdevDetectZeroesOptions detect_zeroes; 1262 }; 1263 1264 typedef enum BlockMirrorBackingMode { 1265 /* 1266 * Reuse the existing backing chain from the source for the target. 1267 * - sync=full: Set backing BDS to NULL. 1268 * - sync=top: Use source's backing BDS. 1269 * - sync=none: Use source as the backing BDS. 1270 */ 1271 MIRROR_SOURCE_BACKING_CHAIN, 1272 1273 /* Open the target's backing chain completely anew */ 1274 MIRROR_OPEN_BACKING_CHAIN, 1275 1276 /* Do not change the target's backing BDS after job completion */ 1277 MIRROR_LEAVE_BACKING_CHAIN, 1278 } BlockMirrorBackingMode; 1279 1280 1281 /* 1282 * Essential block drivers which must always be statically linked into qemu, and 1283 * which therefore can be accessed without using bdrv_find_format() 1284 */ 1285 extern BlockDriver bdrv_file; 1286 extern BlockDriver bdrv_raw; 1287 extern BlockDriver bdrv_qcow2; 1288 1289 extern unsigned int bdrv_drain_all_count; 1290 extern QemuOptsList bdrv_create_opts_simple; 1291 1292 /* 1293 * Common functions that are neither I/O nor Global State. 1294 * 1295 * See include/block/block-common.h for more information about 1296 * the Common API. 1297 */ 1298 1299 static inline BlockDriverState *child_bs(BdrvChild *child) 1300 { 1301 return child ? child->bs : NULL; 1302 } 1303 1304 int bdrv_check_request(int64_t offset, int64_t bytes, Error **errp); 1305 char *create_tmp_file(Error **errp); 1306 void bdrv_parse_filename_strip_prefix(const char *filename, const char *prefix, 1307 QDict *options); 1308 1309 1310 int bdrv_check_qiov_request(int64_t offset, int64_t bytes, 1311 QEMUIOVector *qiov, size_t qiov_offset, 1312 Error **errp); 1313 1314 #ifdef _WIN32 1315 int is_windows_drive(const char *filename); 1316 #endif 1317 1318 #endif /* BLOCK_INT_COMMON_H */ 1319