1 /* 2 * QEMU System Emulator block driver 3 * 4 * Copyright (c) 2003 Fabrice Bellard 5 * 6 * Permission is hereby granted, free of charge, to any person obtaining a copy 7 * of this software and associated documentation files (the "Software"), to deal 8 * in the Software without restriction, including without limitation the rights 9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell 10 * copies of the Software, and to permit persons to whom the Software is 11 * furnished to do so, subject to the following conditions: 12 * 13 * The above copyright notice and this permission notice shall be included in 14 * all copies or substantial portions of the Software. 15 * 16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN 22 * THE SOFTWARE. 23 */ 24 #ifndef BLOCK_INT_H 25 #define BLOCK_INT_H 26 27 #include "block/accounting.h" 28 #include "block/block.h" 29 #include "block/aio-wait.h" 30 #include "qemu/queue.h" 31 #include "qemu/coroutine.h" 32 #include "qemu/stats64.h" 33 #include "qemu/timer.h" 34 #include "qemu/hbitmap.h" 35 #include "block/snapshot.h" 36 #include "qemu/throttle.h" 37 38 #define BLOCK_FLAG_LAZY_REFCOUNTS 8 39 40 #define BLOCK_OPT_SIZE "size" 41 #define BLOCK_OPT_ENCRYPT "encryption" 42 #define BLOCK_OPT_ENCRYPT_FORMAT "encrypt.format" 43 #define BLOCK_OPT_COMPAT6 "compat6" 44 #define BLOCK_OPT_HWVERSION "hwversion" 45 #define BLOCK_OPT_BACKING_FILE "backing_file" 46 #define BLOCK_OPT_BACKING_FMT "backing_fmt" 47 #define BLOCK_OPT_CLUSTER_SIZE "cluster_size" 48 #define BLOCK_OPT_TABLE_SIZE "table_size" 49 #define BLOCK_OPT_PREALLOC "preallocation" 50 #define BLOCK_OPT_SUBFMT "subformat" 51 #define BLOCK_OPT_COMPAT_LEVEL "compat" 52 #define BLOCK_OPT_LAZY_REFCOUNTS "lazy_refcounts" 53 #define BLOCK_OPT_ADAPTER_TYPE "adapter_type" 54 #define BLOCK_OPT_REDUNDANCY "redundancy" 55 #define BLOCK_OPT_NOCOW "nocow" 56 #define BLOCK_OPT_OBJECT_SIZE "object_size" 57 #define BLOCK_OPT_REFCOUNT_BITS "refcount_bits" 58 #define BLOCK_OPT_DATA_FILE "data_file" 59 #define BLOCK_OPT_DATA_FILE_RAW "data_file_raw" 60 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 typedef struct BdrvTrackedRequest { 71 BlockDriverState *bs; 72 int64_t offset; 73 uint64_t bytes; 74 enum BdrvTrackedRequestType type; 75 76 bool serialising; 77 int64_t overlap_offset; 78 uint64_t overlap_bytes; 79 80 QLIST_ENTRY(BdrvTrackedRequest) list; 81 Coroutine *co; /* owner, used for deadlock detection */ 82 CoQueue wait_queue; /* coroutines blocked on this request */ 83 84 struct BdrvTrackedRequest *waiting_for; 85 } BdrvTrackedRequest; 86 87 struct BlockDriver { 88 const char *format_name; 89 int instance_size; 90 91 /* set to true if the BlockDriver is a block filter. Block filters pass 92 * certain callbacks that refer to data (see block.c) to their bs->file if 93 * the driver doesn't implement them. Drivers that do not wish to forward 94 * must implement them and return -ENOTSUP. 95 */ 96 bool is_filter; 97 /* for snapshots block filter like Quorum can implement the 98 * following recursive callback. 99 * It's purpose is to recurse on the filter children while calling 100 * bdrv_recurse_is_first_non_filter on them. 101 * For a sample implementation look in the future Quorum block filter. 102 */ 103 bool (*bdrv_recurse_is_first_non_filter)(BlockDriverState *bs, 104 BlockDriverState *candidate); 105 106 int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename); 107 int (*bdrv_probe_device)(const char *filename); 108 109 /* Any driver implementing this callback is expected to be able to handle 110 * NULL file names in its .bdrv_open() implementation */ 111 void (*bdrv_parse_filename)(const char *filename, QDict *options, Error **errp); 112 /* Drivers not implementing bdrv_parse_filename nor bdrv_open should have 113 * this field set to true, except ones that are defined only by their 114 * child's bs. 115 * An example of the last type will be the quorum block driver. 116 */ 117 bool bdrv_needs_filename; 118 119 /* Set if a driver can support backing files */ 120 bool supports_backing; 121 122 /* For handling image reopen for split or non-split files */ 123 int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state, 124 BlockReopenQueue *queue, Error **errp); 125 void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state); 126 void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state); 127 void (*bdrv_join_options)(QDict *options, QDict *old_options); 128 129 int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags, 130 Error **errp); 131 132 /* Protocol drivers should implement this instead of bdrv_open */ 133 int (*bdrv_file_open)(BlockDriverState *bs, QDict *options, int flags, 134 Error **errp); 135 void (*bdrv_close)(BlockDriverState *bs); 136 int coroutine_fn (*bdrv_co_create)(BlockdevCreateOptions *opts, 137 Error **errp); 138 int coroutine_fn (*bdrv_co_create_opts)(const char *filename, 139 QemuOpts *opts, 140 Error **errp); 141 int (*bdrv_make_empty)(BlockDriverState *bs); 142 143 /* 144 * Refreshes the bs->exact_filename field. If that is impossible, 145 * bs->exact_filename has to be left empty. 146 */ 147 void (*bdrv_refresh_filename)(BlockDriverState *bs); 148 149 /* 150 * Gathers the open options for all children into @target. 151 * A simple format driver (without backing file support) might 152 * implement this function like this: 153 * 154 * QINCREF(bs->file->bs->full_open_options); 155 * qdict_put(target, "file", bs->file->bs->full_open_options); 156 * 157 * If not specified, the generic implementation will simply put 158 * all children's options under their respective name. 159 * 160 * @backing_overridden is true when bs->backing seems not to be 161 * the child that would result from opening bs->backing_file. 162 * Therefore, if it is true, the backing child's options should be 163 * gathered; otherwise, there is no need since the backing child 164 * is the one implied by the image header. 165 * 166 * Note that ideally this function would not be needed. Every 167 * block driver which implements it is probably doing something 168 * shady regarding its runtime option structure. 169 */ 170 void (*bdrv_gather_child_options)(BlockDriverState *bs, QDict *target, 171 bool backing_overridden); 172 173 /* 174 * Returns an allocated string which is the directory name of this BDS: It 175 * will be used to make relative filenames absolute by prepending this 176 * function's return value to them. 177 */ 178 char *(*bdrv_dirname)(BlockDriverState *bs, Error **errp); 179 180 /* aio */ 181 BlockAIOCB *(*bdrv_aio_preadv)(BlockDriverState *bs, 182 uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags, 183 BlockCompletionFunc *cb, void *opaque); 184 BlockAIOCB *(*bdrv_aio_pwritev)(BlockDriverState *bs, 185 uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags, 186 BlockCompletionFunc *cb, void *opaque); 187 BlockAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs, 188 BlockCompletionFunc *cb, void *opaque); 189 BlockAIOCB *(*bdrv_aio_pdiscard)(BlockDriverState *bs, 190 int64_t offset, int bytes, 191 BlockCompletionFunc *cb, void *opaque); 192 193 int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs, 194 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov); 195 196 /** 197 * @offset: position in bytes to read at 198 * @bytes: number of bytes to read 199 * @qiov: the buffers to fill with read data 200 * @flags: currently unused, always 0 201 * 202 * @offset and @bytes will be a multiple of 'request_alignment', 203 * but the length of individual @qiov elements does not have to 204 * be a multiple. 205 * 206 * @bytes will always equal the total size of @qiov, and will be 207 * no larger than 'max_transfer'. 208 * 209 * The buffer in @qiov may point directly to guest memory. 210 */ 211 int coroutine_fn (*bdrv_co_preadv)(BlockDriverState *bs, 212 uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags); 213 int coroutine_fn (*bdrv_co_preadv_part)(BlockDriverState *bs, 214 uint64_t offset, uint64_t bytes, 215 QEMUIOVector *qiov, size_t qiov_offset, int flags); 216 int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs, 217 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, int flags); 218 /** 219 * @offset: position in bytes to write at 220 * @bytes: number of bytes to write 221 * @qiov: the buffers containing data to write 222 * @flags: zero or more bits allowed by 'supported_write_flags' 223 * 224 * @offset and @bytes will be a multiple of 'request_alignment', 225 * but the length of individual @qiov elements does not have to 226 * be a multiple. 227 * 228 * @bytes will always equal the total size of @qiov, and will be 229 * no larger than 'max_transfer'. 230 * 231 * The buffer in @qiov may point directly to guest memory. 232 */ 233 int coroutine_fn (*bdrv_co_pwritev)(BlockDriverState *bs, 234 uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags); 235 int coroutine_fn (*bdrv_co_pwritev_part)(BlockDriverState *bs, 236 uint64_t offset, uint64_t bytes, 237 QEMUIOVector *qiov, size_t qiov_offset, int flags); 238 239 /* 240 * Efficiently zero a region of the disk image. Typically an image format 241 * would use a compact metadata representation to implement this. This 242 * function pointer may be NULL or return -ENOSUP and .bdrv_co_writev() 243 * will be called instead. 244 */ 245 int coroutine_fn (*bdrv_co_pwrite_zeroes)(BlockDriverState *bs, 246 int64_t offset, int bytes, BdrvRequestFlags flags); 247 int coroutine_fn (*bdrv_co_pdiscard)(BlockDriverState *bs, 248 int64_t offset, int bytes); 249 250 /* Map [offset, offset + nbytes) range onto a child of @bs to copy from, 251 * and invoke bdrv_co_copy_range_from(child, ...), or invoke 252 * bdrv_co_copy_range_to() if @bs is the leaf child to copy data from. 253 * 254 * See the comment of bdrv_co_copy_range for the parameter and return value 255 * semantics. 256 */ 257 int coroutine_fn (*bdrv_co_copy_range_from)(BlockDriverState *bs, 258 BdrvChild *src, 259 uint64_t offset, 260 BdrvChild *dst, 261 uint64_t dst_offset, 262 uint64_t bytes, 263 BdrvRequestFlags read_flags, 264 BdrvRequestFlags write_flags); 265 266 /* Map [offset, offset + nbytes) range onto a child of bs to copy data to, 267 * and invoke bdrv_co_copy_range_to(child, src, ...), or perform the copy 268 * operation if @bs is the leaf and @src has the same BlockDriver. Return 269 * -ENOTSUP if @bs is the leaf but @src has a different BlockDriver. 270 * 271 * See the comment of bdrv_co_copy_range for the parameter and return value 272 * semantics. 273 */ 274 int coroutine_fn (*bdrv_co_copy_range_to)(BlockDriverState *bs, 275 BdrvChild *src, 276 uint64_t src_offset, 277 BdrvChild *dst, 278 uint64_t dst_offset, 279 uint64_t bytes, 280 BdrvRequestFlags read_flags, 281 BdrvRequestFlags write_flags); 282 283 /* 284 * Building block for bdrv_block_status[_above] and 285 * bdrv_is_allocated[_above]. The driver should answer only 286 * according to the current layer, and should only need to set 287 * BDRV_BLOCK_DATA, BDRV_BLOCK_ZERO, BDRV_BLOCK_OFFSET_VALID, 288 * and/or BDRV_BLOCK_RAW; if the current layer defers to a backing 289 * layer, the result should be 0 (and not BDRV_BLOCK_ZERO). See 290 * block.h for the overall meaning of the bits. As a hint, the 291 * flag want_zero is true if the caller cares more about precise 292 * mappings (favor accurate _OFFSET_VALID/_ZERO) or false for 293 * overall allocation (favor larger *pnum, perhaps by reporting 294 * _DATA instead of _ZERO). The block layer guarantees input 295 * clamped to bdrv_getlength() and aligned to request_alignment, 296 * as well as non-NULL pnum, map, and file; in turn, the driver 297 * must return an error or set pnum to an aligned non-zero value. 298 */ 299 int coroutine_fn (*bdrv_co_block_status)(BlockDriverState *bs, 300 bool want_zero, int64_t offset, int64_t bytes, int64_t *pnum, 301 int64_t *map, BlockDriverState **file); 302 303 /* 304 * Invalidate any cached meta-data. 305 */ 306 void coroutine_fn (*bdrv_co_invalidate_cache)(BlockDriverState *bs, 307 Error **errp); 308 int (*bdrv_inactivate)(BlockDriverState *bs); 309 310 /* 311 * Flushes all data for all layers by calling bdrv_co_flush for underlying 312 * layers, if needed. This function is needed for deterministic 313 * synchronization of the flush finishing callback. 314 */ 315 int coroutine_fn (*bdrv_co_flush)(BlockDriverState *bs); 316 317 /* 318 * Flushes all data that was already written to the OS all the way down to 319 * the disk (for example file-posix.c calls fsync()). 320 */ 321 int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs); 322 323 /* 324 * Flushes all internal caches to the OS. The data may still sit in a 325 * writeback cache of the host OS, but it will survive a crash of the qemu 326 * process. 327 */ 328 int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs); 329 330 /* 331 * Drivers setting this field must be able to work with just a plain 332 * filename with '<protocol_name>:' as a prefix, and no other options. 333 * Options may be extracted from the filename by implementing 334 * bdrv_parse_filename. 335 */ 336 const char *protocol_name; 337 int coroutine_fn (*bdrv_co_truncate)(BlockDriverState *bs, int64_t offset, 338 PreallocMode prealloc, Error **errp); 339 340 int64_t (*bdrv_getlength)(BlockDriverState *bs); 341 bool has_variable_length; 342 int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs); 343 BlockMeasureInfo *(*bdrv_measure)(QemuOpts *opts, BlockDriverState *in_bs, 344 Error **errp); 345 346 int coroutine_fn (*bdrv_co_pwritev_compressed)(BlockDriverState *bs, 347 uint64_t offset, uint64_t bytes, QEMUIOVector *qiov); 348 int coroutine_fn (*bdrv_co_pwritev_compressed_part)(BlockDriverState *bs, 349 uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, 350 size_t qiov_offset); 351 352 int (*bdrv_snapshot_create)(BlockDriverState *bs, 353 QEMUSnapshotInfo *sn_info); 354 int (*bdrv_snapshot_goto)(BlockDriverState *bs, 355 const char *snapshot_id); 356 int (*bdrv_snapshot_delete)(BlockDriverState *bs, 357 const char *snapshot_id, 358 const char *name, 359 Error **errp); 360 int (*bdrv_snapshot_list)(BlockDriverState *bs, 361 QEMUSnapshotInfo **psn_info); 362 int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs, 363 const char *snapshot_id, 364 const char *name, 365 Error **errp); 366 int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi); 367 ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs, 368 Error **errp); 369 BlockStatsSpecific *(*bdrv_get_specific_stats)(BlockDriverState *bs); 370 371 int coroutine_fn (*bdrv_save_vmstate)(BlockDriverState *bs, 372 QEMUIOVector *qiov, 373 int64_t pos); 374 int coroutine_fn (*bdrv_load_vmstate)(BlockDriverState *bs, 375 QEMUIOVector *qiov, 376 int64_t pos); 377 378 int (*bdrv_change_backing_file)(BlockDriverState *bs, 379 const char *backing_file, const char *backing_fmt); 380 381 /* removable device specific */ 382 bool (*bdrv_is_inserted)(BlockDriverState *bs); 383 void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag); 384 void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked); 385 386 /* to control generic scsi devices */ 387 BlockAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs, 388 unsigned long int req, void *buf, 389 BlockCompletionFunc *cb, void *opaque); 390 int coroutine_fn (*bdrv_co_ioctl)(BlockDriverState *bs, 391 unsigned long int req, void *buf); 392 393 /* List of options for creating images, terminated by name == NULL */ 394 QemuOptsList *create_opts; 395 /* 396 * If this driver supports reopening images this contains a 397 * NULL-terminated list of the runtime options that can be 398 * modified. If an option in this list is unspecified during 399 * reopen then it _must_ be reset to its default value or return 400 * an error. 401 */ 402 const char *const *mutable_opts; 403 404 /* 405 * Returns 0 for completed check, -errno for internal errors. 406 * The check results are stored in result. 407 */ 408 int coroutine_fn (*bdrv_co_check)(BlockDriverState *bs, 409 BdrvCheckResult *result, 410 BdrvCheckMode fix); 411 412 int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts, 413 BlockDriverAmendStatusCB *status_cb, 414 void *cb_opaque, 415 Error **errp); 416 417 void (*bdrv_debug_event)(BlockDriverState *bs, BlkdebugEvent event); 418 419 /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */ 420 int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event, 421 const char *tag); 422 int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs, 423 const char *tag); 424 int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag); 425 bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag); 426 427 void (*bdrv_refresh_limits)(BlockDriverState *bs, Error **errp); 428 429 /* 430 * Returns 1 if newly created images are guaranteed to contain only 431 * zeros, 0 otherwise. 432 * Must return 0 if .bdrv_has_zero_init_truncate() returns 0. 433 */ 434 int (*bdrv_has_zero_init)(BlockDriverState *bs); 435 436 /* 437 * Returns 1 if new areas added by growing the image with 438 * PREALLOC_MODE_OFF contain only zeros, 0 otherwise. 439 */ 440 int (*bdrv_has_zero_init_truncate)(BlockDriverState *bs); 441 442 /* Remove fd handlers, timers, and other event loop callbacks so the event 443 * loop is no longer in use. Called with no in-flight requests and in 444 * depth-first traversal order with parents before child nodes. 445 */ 446 void (*bdrv_detach_aio_context)(BlockDriverState *bs); 447 448 /* Add fd handlers, timers, and other event loop callbacks so I/O requests 449 * can be processed again. Called with no in-flight requests and in 450 * depth-first traversal order with child nodes before parent nodes. 451 */ 452 void (*bdrv_attach_aio_context)(BlockDriverState *bs, 453 AioContext *new_context); 454 455 /* io queue for linux-aio */ 456 void (*bdrv_io_plug)(BlockDriverState *bs); 457 void (*bdrv_io_unplug)(BlockDriverState *bs); 458 459 /** 460 * Try to get @bs's logical and physical block size. 461 * On success, store them in @bsz and return zero. 462 * On failure, return negative errno. 463 */ 464 int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz); 465 /** 466 * Try to get @bs's geometry (cyls, heads, sectors) 467 * On success, store them in @geo and return 0. 468 * On failure return -errno. 469 * Only drivers that want to override guest geometry implement this 470 * callback; see hd_geometry_guess(). 471 */ 472 int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo); 473 474 /** 475 * bdrv_co_drain_begin is called if implemented in the beginning of a 476 * drain operation to drain and stop any internal sources of requests in 477 * the driver. 478 * bdrv_co_drain_end is called if implemented at the end of the drain. 479 * 480 * They should be used by the driver to e.g. manage scheduled I/O 481 * requests, or toggle an internal state. After the end of the drain new 482 * requests will continue normally. 483 */ 484 void coroutine_fn (*bdrv_co_drain_begin)(BlockDriverState *bs); 485 void coroutine_fn (*bdrv_co_drain_end)(BlockDriverState *bs); 486 487 void (*bdrv_add_child)(BlockDriverState *parent, BlockDriverState *child, 488 Error **errp); 489 void (*bdrv_del_child)(BlockDriverState *parent, BdrvChild *child, 490 Error **errp); 491 492 /** 493 * Informs the block driver that a permission change is intended. The 494 * driver checks whether the change is permissible and may take other 495 * preparations for the change (e.g. get file system locks). This operation 496 * is always followed either by a call to either .bdrv_set_perm or 497 * .bdrv_abort_perm_update. 498 * 499 * Checks whether the requested set of cumulative permissions in @perm 500 * can be granted for accessing @bs and whether no other users are using 501 * permissions other than those given in @shared (both arguments take 502 * BLK_PERM_* bitmasks). 503 * 504 * If both conditions are met, 0 is returned. Otherwise, -errno is returned 505 * and errp is set to an error describing the conflict. 506 */ 507 int (*bdrv_check_perm)(BlockDriverState *bs, uint64_t perm, 508 uint64_t shared, Error **errp); 509 510 /** 511 * Called to inform the driver that the set of cumulative set of used 512 * permissions for @bs has changed to @perm, and the set of sharable 513 * permission to @shared. The driver can use this to propagate changes to 514 * its children (i.e. request permissions only if a parent actually needs 515 * them). 516 * 517 * This function is only invoked after bdrv_check_perm(), so block drivers 518 * may rely on preparations made in their .bdrv_check_perm implementation. 519 */ 520 void (*bdrv_set_perm)(BlockDriverState *bs, uint64_t perm, uint64_t shared); 521 522 /* 523 * Called to inform the driver that after a previous bdrv_check_perm() 524 * call, the permission update is not performed and any preparations made 525 * for it (e.g. taken file locks) need to be undone. 526 * 527 * This function can be called even for nodes that never saw a 528 * bdrv_check_perm() call. It is a no-op then. 529 */ 530 void (*bdrv_abort_perm_update)(BlockDriverState *bs); 531 532 /** 533 * Returns in @nperm and @nshared the permissions that the driver for @bs 534 * needs on its child @c, based on the cumulative permissions requested by 535 * the parents in @parent_perm and @parent_shared. 536 * 537 * If @c is NULL, return the permissions for attaching a new child for the 538 * given @role. 539 * 540 * If @reopen_queue is non-NULL, don't return the currently needed 541 * permissions, but those that will be needed after applying the 542 * @reopen_queue. 543 */ 544 void (*bdrv_child_perm)(BlockDriverState *bs, BdrvChild *c, 545 const BdrvChildRole *role, 546 BlockReopenQueue *reopen_queue, 547 uint64_t parent_perm, uint64_t parent_shared, 548 uint64_t *nperm, uint64_t *nshared); 549 550 bool (*bdrv_co_can_store_new_dirty_bitmap)(BlockDriverState *bs, 551 const char *name, 552 uint32_t granularity, 553 Error **errp); 554 int (*bdrv_co_remove_persistent_dirty_bitmap)(BlockDriverState *bs, 555 const char *name, 556 Error **errp); 557 558 /** 559 * Register/unregister a buffer for I/O. For example, when the driver is 560 * interested to know the memory areas that will later be used in iovs, so 561 * that it can do IOMMU mapping with VFIO etc., in order to get better 562 * performance. In the case of VFIO drivers, this callback is used to do 563 * DMA mapping for hot buffers. 564 */ 565 void (*bdrv_register_buf)(BlockDriverState *bs, void *host, size_t size); 566 void (*bdrv_unregister_buf)(BlockDriverState *bs, void *host); 567 QLIST_ENTRY(BlockDriver) list; 568 569 /* Pointer to a NULL-terminated array of names of strong options 570 * that can be specified for bdrv_open(). A strong option is one 571 * that changes the data of a BDS. 572 * If this pointer is NULL, the array is considered empty. 573 * "filename" and "driver" are always considered strong. */ 574 const char *const *strong_runtime_opts; 575 }; 576 577 static inline bool block_driver_can_compress(BlockDriver *drv) 578 { 579 return drv->bdrv_co_pwritev_compressed || 580 drv->bdrv_co_pwritev_compressed_part; 581 } 582 583 typedef struct BlockLimits { 584 /* Alignment requirement, in bytes, for offset/length of I/O 585 * requests. Must be a power of 2 less than INT_MAX; defaults to 586 * 1 for drivers with modern byte interfaces, and to 512 587 * otherwise. */ 588 uint32_t request_alignment; 589 590 /* Maximum number of bytes that can be discarded at once (since it 591 * is signed, it must be < 2G, if set). Must be multiple of 592 * pdiscard_alignment, but need not be power of 2. May be 0 if no 593 * inherent 32-bit limit */ 594 int32_t max_pdiscard; 595 596 /* Optimal alignment for discard requests in bytes. A power of 2 597 * is best but not mandatory. Must be a multiple of 598 * bl.request_alignment, and must be less than max_pdiscard if 599 * that is set. May be 0 if bl.request_alignment is good enough */ 600 uint32_t pdiscard_alignment; 601 602 /* Maximum number of bytes that can zeroized at once (since it is 603 * signed, it must be < 2G, if set). Must be multiple of 604 * pwrite_zeroes_alignment. May be 0 if no inherent 32-bit limit */ 605 int32_t max_pwrite_zeroes; 606 607 /* Optimal alignment for write zeroes requests in bytes. A power 608 * of 2 is best but not mandatory. Must be a multiple of 609 * bl.request_alignment, and must be less than max_pwrite_zeroes 610 * if that is set. May be 0 if bl.request_alignment is good 611 * enough */ 612 uint32_t pwrite_zeroes_alignment; 613 614 /* Optimal transfer length in bytes. A power of 2 is best but not 615 * mandatory. Must be a multiple of bl.request_alignment, or 0 if 616 * no preferred size */ 617 uint32_t opt_transfer; 618 619 /* Maximal transfer length in bytes. Need not be power of 2, but 620 * must be multiple of opt_transfer and bl.request_alignment, or 0 621 * for no 32-bit limit. For now, anything larger than INT_MAX is 622 * clamped down. */ 623 uint32_t max_transfer; 624 625 /* memory alignment, in bytes so that no bounce buffer is needed */ 626 size_t min_mem_alignment; 627 628 /* memory alignment, in bytes, for bounce buffer */ 629 size_t opt_mem_alignment; 630 631 /* maximum number of iovec elements */ 632 int max_iov; 633 } BlockLimits; 634 635 typedef struct BdrvOpBlocker BdrvOpBlocker; 636 637 typedef struct BdrvAioNotifier { 638 void (*attached_aio_context)(AioContext *new_context, void *opaque); 639 void (*detach_aio_context)(void *opaque); 640 641 void *opaque; 642 bool deleted; 643 644 QLIST_ENTRY(BdrvAioNotifier) list; 645 } BdrvAioNotifier; 646 647 struct BdrvChildRole { 648 /* If true, bdrv_replace_node() doesn't change the node this BdrvChild 649 * points to. */ 650 bool stay_at_node; 651 652 /* If true, the parent is a BlockDriverState and bdrv_next_all_states() 653 * will return it. This information is used for drain_all, where every node 654 * will be drained separately, so the drain only needs to be propagated to 655 * non-BDS parents. */ 656 bool parent_is_bds; 657 658 void (*inherit_options)(int *child_flags, QDict *child_options, 659 int parent_flags, QDict *parent_options); 660 661 void (*change_media)(BdrvChild *child, bool load); 662 void (*resize)(BdrvChild *child); 663 664 /* Returns a name that is supposedly more useful for human users than the 665 * node name for identifying the node in question (in particular, a BB 666 * name), or NULL if the parent can't provide a better name. */ 667 const char *(*get_name)(BdrvChild *child); 668 669 /* Returns a malloced string that describes the parent of the child for a 670 * human reader. This could be a node-name, BlockBackend name, qdev ID or 671 * QOM path of the device owning the BlockBackend, job type and ID etc. The 672 * caller is responsible for freeing the memory. */ 673 char *(*get_parent_desc)(BdrvChild *child); 674 675 /* 676 * If this pair of functions is implemented, the parent doesn't issue new 677 * requests after returning from .drained_begin() until .drained_end() is 678 * called. 679 * 680 * These functions must not change the graph (and therefore also must not 681 * call aio_poll(), which could change the graph indirectly). 682 * 683 * If drained_end() schedules background operations, it must atomically 684 * increment *drained_end_counter for each such operation and atomically 685 * decrement it once the operation has settled. 686 * 687 * Note that this can be nested. If drained_begin() was called twice, new 688 * I/O is allowed only after drained_end() was called twice, too. 689 */ 690 void (*drained_begin)(BdrvChild *child); 691 void (*drained_end)(BdrvChild *child, int *drained_end_counter); 692 693 /* 694 * Returns whether the parent has pending requests for the child. This 695 * callback is polled after .drained_begin() has been called until all 696 * activity on the child has stopped. 697 */ 698 bool (*drained_poll)(BdrvChild *child); 699 700 /* Notifies the parent that the child has been activated/inactivated (e.g. 701 * when migration is completing) and it can start/stop requesting 702 * permissions and doing I/O on it. */ 703 void (*activate)(BdrvChild *child, Error **errp); 704 int (*inactivate)(BdrvChild *child); 705 706 void (*attach)(BdrvChild *child); 707 void (*detach)(BdrvChild *child); 708 709 /* Notifies the parent that the filename of its child has changed (e.g. 710 * because the direct child was removed from the backing chain), so that it 711 * can update its reference. */ 712 int (*update_filename)(BdrvChild *child, BlockDriverState *new_base, 713 const char *filename, Error **errp); 714 715 bool (*can_set_aio_ctx)(BdrvChild *child, AioContext *ctx, 716 GSList **ignore, Error **errp); 717 void (*set_aio_ctx)(BdrvChild *child, AioContext *ctx, GSList **ignore); 718 }; 719 720 extern const BdrvChildRole child_file; 721 extern const BdrvChildRole child_format; 722 extern const BdrvChildRole child_backing; 723 724 struct BdrvChild { 725 BlockDriverState *bs; 726 char *name; 727 const BdrvChildRole *role; 728 void *opaque; 729 730 /** 731 * Granted permissions for operating on this BdrvChild (BLK_PERM_* bitmask) 732 */ 733 uint64_t perm; 734 735 /** 736 * Permissions that can still be granted to other users of @bs while this 737 * BdrvChild is still attached to it. (BLK_PERM_* bitmask) 738 */ 739 uint64_t shared_perm; 740 741 /* backup of permissions during permission update procedure */ 742 bool has_backup_perm; 743 uint64_t backup_perm; 744 uint64_t backup_shared_perm; 745 746 /* 747 * This link is frozen: the child can neither be replaced nor 748 * detached from the parent. 749 */ 750 bool frozen; 751 752 /* 753 * How many times the parent of this child has been drained 754 * (through role->drained_*). 755 * Usually, this is equal to bs->quiesce_counter (potentially 756 * reduced by bdrv_drain_all_count). It may differ while the 757 * child is entering or leaving a drained section. 758 */ 759 int parent_quiesce_counter; 760 761 QLIST_ENTRY(BdrvChild) next; 762 QLIST_ENTRY(BdrvChild) next_parent; 763 }; 764 765 /* 766 * Note: the function bdrv_append() copies and swaps contents of 767 * BlockDriverStates, so if you add new fields to this struct, please 768 * inspect bdrv_append() to determine if the new fields need to be 769 * copied as well. 770 */ 771 struct BlockDriverState { 772 /* Protected by big QEMU lock or read-only after opening. No special 773 * locking needed during I/O... 774 */ 775 int open_flags; /* flags used to open the file, re-used for re-open */ 776 bool read_only; /* if true, the media is read only */ 777 bool encrypted; /* if true, the media is encrypted */ 778 bool sg; /* if true, the device is a /dev/sg* */ 779 bool probed; /* if true, format was probed rather than specified */ 780 bool force_share; /* if true, always allow all shared permissions */ 781 bool implicit; /* if true, this filter node was automatically inserted */ 782 783 BlockDriver *drv; /* NULL means no media */ 784 void *opaque; 785 786 AioContext *aio_context; /* event loop used for fd handlers, timers, etc */ 787 /* long-running tasks intended to always use the same AioContext as this 788 * BDS may register themselves in this list to be notified of changes 789 * regarding this BDS's context */ 790 QLIST_HEAD(, BdrvAioNotifier) aio_notifiers; 791 bool walking_aio_notifiers; /* to make removal during iteration safe */ 792 793 char filename[PATH_MAX]; 794 char backing_file[PATH_MAX]; /* if non zero, the image is a diff of 795 this file image */ 796 /* The backing filename indicated by the image header; if we ever 797 * open this file, then this is replaced by the resulting BDS's 798 * filename (i.e. after a bdrv_refresh_filename() run). */ 799 char auto_backing_file[PATH_MAX]; 800 char backing_format[16]; /* if non-zero and backing_file exists */ 801 802 QDict *full_open_options; 803 char exact_filename[PATH_MAX]; 804 805 BdrvChild *backing; 806 BdrvChild *file; 807 808 /* I/O Limits */ 809 BlockLimits bl; 810 811 /* Flags honored during pwrite (so far: BDRV_REQ_FUA, 812 * BDRV_REQ_WRITE_UNCHANGED). 813 * If a driver does not support BDRV_REQ_WRITE_UNCHANGED, those 814 * writes will be issued as normal writes without the flag set. 815 * This is important to note for drivers that do not explicitly 816 * request a WRITE permission for their children and instead take 817 * the same permissions as their parent did (this is commonly what 818 * block filters do). Such drivers have to be aware that the 819 * parent may have taken a WRITE_UNCHANGED permission only and is 820 * issuing such requests. Drivers either must make sure that 821 * these requests do not result in plain WRITE accesses (usually 822 * by supporting BDRV_REQ_WRITE_UNCHANGED, and then forwarding 823 * every incoming write request as-is, including potentially that 824 * flag), or they have to explicitly take the WRITE permission for 825 * their children. */ 826 unsigned int supported_write_flags; 827 /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA, 828 * BDRV_REQ_MAY_UNMAP, BDRV_REQ_WRITE_UNCHANGED) */ 829 unsigned int supported_zero_flags; 830 831 /* the following member gives a name to every node on the bs graph. */ 832 char node_name[32]; 833 /* element of the list of named nodes building the graph */ 834 QTAILQ_ENTRY(BlockDriverState) node_list; 835 /* element of the list of all BlockDriverStates (all_bdrv_states) */ 836 QTAILQ_ENTRY(BlockDriverState) bs_list; 837 /* element of the list of monitor-owned BDS */ 838 QTAILQ_ENTRY(BlockDriverState) monitor_list; 839 int refcnt; 840 841 /* operation blockers */ 842 QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX]; 843 844 /* The node that this node inherited default options from (and a reopen on 845 * which can affect this node by changing these defaults). This is always a 846 * parent node of this node. */ 847 BlockDriverState *inherits_from; 848 QLIST_HEAD(, BdrvChild) children; 849 QLIST_HEAD(, BdrvChild) parents; 850 851 QDict *options; 852 QDict *explicit_options; 853 BlockdevDetectZeroesOptions detect_zeroes; 854 855 /* The error object in use for blocking operations on backing_hd */ 856 Error *backing_blocker; 857 858 /* Protected by AioContext lock */ 859 860 /* If we are reading a disk image, give its size in sectors. 861 * Generally read-only; it is written to by load_snapshot and 862 * save_snaphost, but the block layer is quiescent during those. 863 */ 864 int64_t total_sectors; 865 866 /* Callback before write request is processed */ 867 NotifierWithReturnList before_write_notifiers; 868 869 /* threshold limit for writes, in bytes. "High water mark". */ 870 uint64_t write_threshold_offset; 871 NotifierWithReturn write_threshold_notifier; 872 873 /* Writing to the list requires the BQL _and_ the dirty_bitmap_mutex. 874 * Reading from the list can be done with either the BQL or the 875 * dirty_bitmap_mutex. Modifying a bitmap only requires 876 * dirty_bitmap_mutex. */ 877 QemuMutex dirty_bitmap_mutex; 878 QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps; 879 880 /* Offset after the highest byte written to */ 881 Stat64 wr_highest_offset; 882 883 /* If true, copy read backing sectors into image. Can be >1 if more 884 * than one client has requested copy-on-read. Accessed with atomic 885 * ops. 886 */ 887 int copy_on_read; 888 889 /* number of in-flight requests; overall and serialising. 890 * Accessed with atomic ops. 891 */ 892 unsigned int in_flight; 893 unsigned int serialising_in_flight; 894 895 /* counter for nested bdrv_io_plug. 896 * Accessed with atomic ops. 897 */ 898 unsigned io_plugged; 899 900 /* do we need to tell the quest if we have a volatile write cache? */ 901 int enable_write_cache; 902 903 /* Accessed with atomic ops. */ 904 int quiesce_counter; 905 int recursive_quiesce_counter; 906 907 unsigned int write_gen; /* Current data generation */ 908 909 /* Protected by reqs_lock. */ 910 CoMutex reqs_lock; 911 QLIST_HEAD(, BdrvTrackedRequest) tracked_requests; 912 CoQueue flush_queue; /* Serializing flush queue */ 913 bool active_flush_req; /* Flush request in flight? */ 914 915 /* Only read/written by whoever has set active_flush_req to true. */ 916 unsigned int flushed_gen; /* Flushed write generation */ 917 918 /* BdrvChild links to this node may never be frozen */ 919 bool never_freeze; 920 }; 921 922 struct BlockBackendRootState { 923 int open_flags; 924 bool read_only; 925 BlockdevDetectZeroesOptions detect_zeroes; 926 }; 927 928 typedef enum BlockMirrorBackingMode { 929 /* Reuse the existing backing chain from the source for the target. 930 * - sync=full: Set backing BDS to NULL. 931 * - sync=top: Use source's backing BDS. 932 * - sync=none: Use source as the backing BDS. */ 933 MIRROR_SOURCE_BACKING_CHAIN, 934 935 /* Open the target's backing chain completely anew */ 936 MIRROR_OPEN_BACKING_CHAIN, 937 938 /* Do not change the target's backing BDS after job completion */ 939 MIRROR_LEAVE_BACKING_CHAIN, 940 } BlockMirrorBackingMode; 941 942 static inline BlockDriverState *backing_bs(BlockDriverState *bs) 943 { 944 return bs->backing ? bs->backing->bs : NULL; 945 } 946 947 948 /* Essential block drivers which must always be statically linked into qemu, and 949 * which therefore can be accessed without using bdrv_find_format() */ 950 extern BlockDriver bdrv_file; 951 extern BlockDriver bdrv_raw; 952 extern BlockDriver bdrv_qcow2; 953 954 int coroutine_fn bdrv_co_preadv(BdrvChild *child, 955 int64_t offset, unsigned int bytes, QEMUIOVector *qiov, 956 BdrvRequestFlags flags); 957 int coroutine_fn bdrv_co_preadv_part(BdrvChild *child, 958 int64_t offset, unsigned int bytes, 959 QEMUIOVector *qiov, size_t qiov_offset, BdrvRequestFlags flags); 960 int coroutine_fn bdrv_co_pwritev(BdrvChild *child, 961 int64_t offset, unsigned int bytes, QEMUIOVector *qiov, 962 BdrvRequestFlags flags); 963 int coroutine_fn bdrv_co_pwritev_part(BdrvChild *child, 964 int64_t offset, unsigned int bytes, 965 QEMUIOVector *qiov, size_t qiov_offset, BdrvRequestFlags flags); 966 967 static inline int coroutine_fn bdrv_co_pread(BdrvChild *child, 968 int64_t offset, unsigned int bytes, void *buf, BdrvRequestFlags flags) 969 { 970 QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes); 971 972 return bdrv_co_preadv(child, offset, bytes, &qiov, flags); 973 } 974 975 static inline int coroutine_fn bdrv_co_pwrite(BdrvChild *child, 976 int64_t offset, unsigned int bytes, void *buf, BdrvRequestFlags flags) 977 { 978 QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes); 979 980 return bdrv_co_pwritev(child, offset, bytes, &qiov, flags); 981 } 982 983 extern unsigned int bdrv_drain_all_count; 984 void bdrv_apply_subtree_drain(BdrvChild *child, BlockDriverState *new_parent); 985 void bdrv_unapply_subtree_drain(BdrvChild *child, BlockDriverState *old_parent); 986 987 int get_tmp_filename(char *filename, int size); 988 BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size, 989 const char *filename); 990 991 void bdrv_parse_filename_strip_prefix(const char *filename, const char *prefix, 992 QDict *options); 993 994 995 /** 996 * bdrv_add_before_write_notifier: 997 * 998 * Register a callback that is invoked before write requests are processed but 999 * after any throttling or waiting for overlapping requests. 1000 */ 1001 void bdrv_add_before_write_notifier(BlockDriverState *bs, 1002 NotifierWithReturn *notifier); 1003 1004 /** 1005 * bdrv_add_aio_context_notifier: 1006 * 1007 * If a long-running job intends to be always run in the same AioContext as a 1008 * certain BDS, it may use this function to be notified of changes regarding the 1009 * association of the BDS to an AioContext. 1010 * 1011 * attached_aio_context() is called after the target BDS has been attached to a 1012 * new AioContext; detach_aio_context() is called before the target BDS is being 1013 * detached from its old AioContext. 1014 */ 1015 void bdrv_add_aio_context_notifier(BlockDriverState *bs, 1016 void (*attached_aio_context)(AioContext *new_context, void *opaque), 1017 void (*detach_aio_context)(void *opaque), void *opaque); 1018 1019 /** 1020 * bdrv_remove_aio_context_notifier: 1021 * 1022 * Unsubscribe of change notifications regarding the BDS's AioContext. The 1023 * parameters given here have to be the same as those given to 1024 * bdrv_add_aio_context_notifier(). 1025 */ 1026 void bdrv_remove_aio_context_notifier(BlockDriverState *bs, 1027 void (*aio_context_attached)(AioContext *, 1028 void *), 1029 void (*aio_context_detached)(void *), 1030 void *opaque); 1031 1032 /** 1033 * bdrv_wakeup: 1034 * @bs: The BlockDriverState for which an I/O operation has been completed. 1035 * 1036 * Wake up the main thread if it is waiting on BDRV_POLL_WHILE. During 1037 * synchronous I/O on a BlockDriverState that is attached to another 1038 * I/O thread, the main thread lets the I/O thread's event loop run, 1039 * waiting for the I/O operation to complete. A bdrv_wakeup will wake 1040 * up the main thread if necessary. 1041 * 1042 * Manual calls to bdrv_wakeup are rarely necessary, because 1043 * bdrv_dec_in_flight already calls it. 1044 */ 1045 void bdrv_wakeup(BlockDriverState *bs); 1046 1047 #ifdef _WIN32 1048 int is_windows_drive(const char *filename); 1049 #endif 1050 1051 /** 1052 * stream_start: 1053 * @job_id: The id of the newly-created job, or %NULL to use the 1054 * device name of @bs. 1055 * @bs: Block device to operate on. 1056 * @base: Block device that will become the new base, or %NULL to 1057 * flatten the whole backing file chain onto @bs. 1058 * @backing_file_str: The file name that will be written to @bs as the 1059 * the new backing file if the job completes. Ignored if @base is %NULL. 1060 * @creation_flags: Flags that control the behavior of the Job lifetime. 1061 * See @BlockJobCreateFlags 1062 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 1063 * @on_error: The action to take upon error. 1064 * @errp: Error object. 1065 * 1066 * Start a streaming operation on @bs. Clusters that are unallocated 1067 * in @bs, but allocated in any image between @base and @bs (both 1068 * exclusive) will be written to @bs. At the end of a successful 1069 * streaming job, the backing file of @bs will be changed to 1070 * @backing_file_str in the written image and to @base in the live 1071 * BlockDriverState. 1072 */ 1073 void stream_start(const char *job_id, BlockDriverState *bs, 1074 BlockDriverState *base, const char *backing_file_str, 1075 int creation_flags, int64_t speed, 1076 BlockdevOnError on_error, Error **errp); 1077 1078 /** 1079 * commit_start: 1080 * @job_id: The id of the newly-created job, or %NULL to use the 1081 * device name of @bs. 1082 * @bs: Active block device. 1083 * @top: Top block device to be committed. 1084 * @base: Block device that will be written into, and become the new top. 1085 * @creation_flags: Flags that control the behavior of the Job lifetime. 1086 * See @BlockJobCreateFlags 1087 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 1088 * @on_error: The action to take upon error. 1089 * @backing_file_str: String to use as the backing file in @top's overlay 1090 * @filter_node_name: The node name that should be assigned to the filter 1091 * driver that the commit job inserts into the graph above @top. NULL means 1092 * that a node name should be autogenerated. 1093 * @errp: Error object. 1094 * 1095 */ 1096 void commit_start(const char *job_id, BlockDriverState *bs, 1097 BlockDriverState *base, BlockDriverState *top, 1098 int creation_flags, int64_t speed, 1099 BlockdevOnError on_error, const char *backing_file_str, 1100 const char *filter_node_name, Error **errp); 1101 /** 1102 * commit_active_start: 1103 * @job_id: The id of the newly-created job, or %NULL to use the 1104 * device name of @bs. 1105 * @bs: Active block device to be committed. 1106 * @base: Block device that will be written into, and become the new top. 1107 * @creation_flags: Flags that control the behavior of the Job lifetime. 1108 * See @BlockJobCreateFlags 1109 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 1110 * @on_error: The action to take upon error. 1111 * @filter_node_name: The node name that should be assigned to the filter 1112 * driver that the commit job inserts into the graph above @bs. NULL means that 1113 * a node name should be autogenerated. 1114 * @cb: Completion function for the job. 1115 * @opaque: Opaque pointer value passed to @cb. 1116 * @auto_complete: Auto complete the job. 1117 * @errp: Error object. 1118 * 1119 */ 1120 BlockJob *commit_active_start(const char *job_id, BlockDriverState *bs, 1121 BlockDriverState *base, int creation_flags, 1122 int64_t speed, BlockdevOnError on_error, 1123 const char *filter_node_name, 1124 BlockCompletionFunc *cb, void *opaque, 1125 bool auto_complete, Error **errp); 1126 /* 1127 * mirror_start: 1128 * @job_id: The id of the newly-created job, or %NULL to use the 1129 * device name of @bs. 1130 * @bs: Block device to operate on. 1131 * @target: Block device to write to. 1132 * @replaces: Block graph node name to replace once the mirror is done. Can 1133 * only be used when full mirroring is selected. 1134 * @creation_flags: Flags that control the behavior of the Job lifetime. 1135 * See @BlockJobCreateFlags 1136 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 1137 * @granularity: The chosen granularity for the dirty bitmap. 1138 * @buf_size: The amount of data that can be in flight at one time. 1139 * @mode: Whether to collapse all images in the chain to the target. 1140 * @backing_mode: How to establish the target's backing chain after completion. 1141 * @zero_target: Whether the target should be explicitly zero-initialized 1142 * @on_source_error: The action to take upon error reading from the source. 1143 * @on_target_error: The action to take upon error writing to the target. 1144 * @unmap: Whether to unmap target where source sectors only contain zeroes. 1145 * @filter_node_name: The node name that should be assigned to the filter 1146 * driver that the mirror job inserts into the graph above @bs. NULL means that 1147 * a node name should be autogenerated. 1148 * @copy_mode: When to trigger writes to the target. 1149 * @errp: Error object. 1150 * 1151 * Start a mirroring operation on @bs. Clusters that are allocated 1152 * in @bs will be written to @target until the job is cancelled or 1153 * manually completed. At the end of a successful mirroring job, 1154 * @bs will be switched to read from @target. 1155 */ 1156 void mirror_start(const char *job_id, BlockDriverState *bs, 1157 BlockDriverState *target, const char *replaces, 1158 int creation_flags, int64_t speed, 1159 uint32_t granularity, int64_t buf_size, 1160 MirrorSyncMode mode, BlockMirrorBackingMode backing_mode, 1161 bool zero_target, 1162 BlockdevOnError on_source_error, 1163 BlockdevOnError on_target_error, 1164 bool unmap, const char *filter_node_name, 1165 MirrorCopyMode copy_mode, Error **errp); 1166 1167 /* 1168 * backup_job_create: 1169 * @job_id: The id of the newly-created job, or %NULL to use the 1170 * device name of @bs. 1171 * @bs: Block device to operate on. 1172 * @target: Block device to write to. 1173 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 1174 * @sync_mode: What parts of the disk image should be copied to the destination. 1175 * @sync_bitmap: The dirty bitmap if sync_mode is 'bitmap' or 'incremental' 1176 * @bitmap_mode: The bitmap synchronization policy to use. 1177 * @on_source_error: The action to take upon error reading from the source. 1178 * @on_target_error: The action to take upon error writing to the target. 1179 * @creation_flags: Flags that control the behavior of the Job lifetime. 1180 * See @BlockJobCreateFlags 1181 * @cb: Completion function for the job. 1182 * @opaque: Opaque pointer value passed to @cb. 1183 * @txn: Transaction that this job is part of (may be NULL). 1184 * 1185 * Create a backup operation on @bs. Clusters in @bs are written to @target 1186 * until the job is cancelled or manually completed. 1187 */ 1188 BlockJob *backup_job_create(const char *job_id, BlockDriverState *bs, 1189 BlockDriverState *target, int64_t speed, 1190 MirrorSyncMode sync_mode, 1191 BdrvDirtyBitmap *sync_bitmap, 1192 BitmapSyncMode bitmap_mode, 1193 bool compress, 1194 const char *filter_node_name, 1195 BlockdevOnError on_source_error, 1196 BlockdevOnError on_target_error, 1197 int creation_flags, 1198 BlockCompletionFunc *cb, void *opaque, 1199 JobTxn *txn, Error **errp); 1200 1201 void hmp_drive_add_node(Monitor *mon, const char *optstr); 1202 1203 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs, 1204 const char *child_name, 1205 const BdrvChildRole *child_role, 1206 AioContext *ctx, 1207 uint64_t perm, uint64_t shared_perm, 1208 void *opaque, Error **errp); 1209 void bdrv_root_unref_child(BdrvChild *child); 1210 1211 /** 1212 * Sets a BdrvChild's permissions. Avoid if the parent is a BDS; use 1213 * bdrv_child_refresh_perms() instead and make the parent's 1214 * .bdrv_child_perm() implementation return the correct values. 1215 */ 1216 int bdrv_child_try_set_perm(BdrvChild *c, uint64_t perm, uint64_t shared, 1217 Error **errp); 1218 1219 /** 1220 * Calls bs->drv->bdrv_child_perm() and updates the child's permission 1221 * masks with the result. 1222 * Drivers should invoke this function whenever an event occurs that 1223 * makes their .bdrv_child_perm() implementation return different 1224 * values than before, but which will not result in the block layer 1225 * automatically refreshing the permissions. 1226 */ 1227 int bdrv_child_refresh_perms(BlockDriverState *bs, BdrvChild *c, Error **errp); 1228 1229 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by 1230 * block filters: Forward CONSISTENT_READ, WRITE, WRITE_UNCHANGED and RESIZE to 1231 * all children */ 1232 void bdrv_filter_default_perms(BlockDriverState *bs, BdrvChild *c, 1233 const BdrvChildRole *role, 1234 BlockReopenQueue *reopen_queue, 1235 uint64_t perm, uint64_t shared, 1236 uint64_t *nperm, uint64_t *nshared); 1237 1238 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by 1239 * (non-raw) image formats: Like above for bs->backing, but for bs->file it 1240 * requires WRITE | RESIZE for read-write images, always requires 1241 * CONSISTENT_READ and doesn't share WRITE. */ 1242 void bdrv_format_default_perms(BlockDriverState *bs, BdrvChild *c, 1243 const BdrvChildRole *role, 1244 BlockReopenQueue *reopen_queue, 1245 uint64_t perm, uint64_t shared, 1246 uint64_t *nperm, uint64_t *nshared); 1247 1248 /* 1249 * Default implementation for drivers to pass bdrv_co_block_status() to 1250 * their file. 1251 */ 1252 int coroutine_fn bdrv_co_block_status_from_file(BlockDriverState *bs, 1253 bool want_zero, 1254 int64_t offset, 1255 int64_t bytes, 1256 int64_t *pnum, 1257 int64_t *map, 1258 BlockDriverState **file); 1259 /* 1260 * Default implementation for drivers to pass bdrv_co_block_status() to 1261 * their backing file. 1262 */ 1263 int coroutine_fn bdrv_co_block_status_from_backing(BlockDriverState *bs, 1264 bool want_zero, 1265 int64_t offset, 1266 int64_t bytes, 1267 int64_t *pnum, 1268 int64_t *map, 1269 BlockDriverState **file); 1270 const char *bdrv_get_parent_name(const BlockDriverState *bs); 1271 void blk_dev_change_media_cb(BlockBackend *blk, bool load, Error **errp); 1272 bool blk_dev_has_removable_media(BlockBackend *blk); 1273 bool blk_dev_has_tray(BlockBackend *blk); 1274 void blk_dev_eject_request(BlockBackend *blk, bool force); 1275 bool blk_dev_is_tray_open(BlockBackend *blk); 1276 bool blk_dev_is_medium_locked(BlockBackend *blk); 1277 1278 void bdrv_set_dirty(BlockDriverState *bs, int64_t offset, int64_t bytes); 1279 1280 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap **out); 1281 void bdrv_restore_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap *backup); 1282 bool bdrv_dirty_bitmap_merge_internal(BdrvDirtyBitmap *dest, 1283 const BdrvDirtyBitmap *src, 1284 HBitmap **backup, bool lock); 1285 1286 void bdrv_inc_in_flight(BlockDriverState *bs); 1287 void bdrv_dec_in_flight(BlockDriverState *bs); 1288 1289 void blockdev_close_all_bdrv_states(void); 1290 1291 int coroutine_fn bdrv_co_copy_range_from(BdrvChild *src, uint64_t src_offset, 1292 BdrvChild *dst, uint64_t dst_offset, 1293 uint64_t bytes, 1294 BdrvRequestFlags read_flags, 1295 BdrvRequestFlags write_flags); 1296 int coroutine_fn bdrv_co_copy_range_to(BdrvChild *src, uint64_t src_offset, 1297 BdrvChild *dst, uint64_t dst_offset, 1298 uint64_t bytes, 1299 BdrvRequestFlags read_flags, 1300 BdrvRequestFlags write_flags); 1301 1302 int refresh_total_sectors(BlockDriverState *bs, int64_t hint); 1303 1304 #endif /* BLOCK_INT_H */ 1305