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/block.h" 28 #include "qemu/option.h" 29 #include "qemu/queue.h" 30 #include "block/coroutine.h" 31 #include "qemu/timer.h" 32 #include "qapi-types.h" 33 #include "qapi/qmp/qerror.h" 34 #include "monitor/monitor.h" 35 #include "qemu/hbitmap.h" 36 #include "block/snapshot.h" 37 #include "qemu/main-loop.h" 38 #include "qemu/throttle.h" 39 40 #define BLOCK_FLAG_ENCRYPT 1 41 #define BLOCK_FLAG_COMPAT6 4 42 #define BLOCK_FLAG_LAZY_REFCOUNTS 8 43 44 #define BLOCK_OPT_SIZE "size" 45 #define BLOCK_OPT_ENCRYPT "encryption" 46 #define BLOCK_OPT_COMPAT6 "compat6" 47 #define BLOCK_OPT_BACKING_FILE "backing_file" 48 #define BLOCK_OPT_BACKING_FMT "backing_fmt" 49 #define BLOCK_OPT_CLUSTER_SIZE "cluster_size" 50 #define BLOCK_OPT_TABLE_SIZE "table_size" 51 #define BLOCK_OPT_PREALLOC "preallocation" 52 #define BLOCK_OPT_SUBFMT "subformat" 53 #define BLOCK_OPT_COMPAT_LEVEL "compat" 54 #define BLOCK_OPT_LAZY_REFCOUNTS "lazy_refcounts" 55 #define BLOCK_OPT_ADAPTER_TYPE "adapter_type" 56 #define BLOCK_OPT_REDUNDANCY "redundancy" 57 #define BLOCK_OPT_NOCOW "nocow" 58 59 typedef struct BdrvTrackedRequest { 60 BlockDriverState *bs; 61 int64_t offset; 62 unsigned int bytes; 63 bool is_write; 64 65 bool serialising; 66 int64_t overlap_offset; 67 unsigned int overlap_bytes; 68 69 QLIST_ENTRY(BdrvTrackedRequest) list; 70 Coroutine *co; /* owner, used for deadlock detection */ 71 CoQueue wait_queue; /* coroutines blocked on this request */ 72 73 struct BdrvTrackedRequest *waiting_for; 74 } BdrvTrackedRequest; 75 76 struct BlockDriver { 77 const char *format_name; 78 int instance_size; 79 80 /* set to true if the BlockDriver is a block filter */ 81 bool is_filter; 82 /* for snapshots block filter like Quorum can implement the 83 * following recursive callback. 84 * It's purpose is to recurse on the filter children while calling 85 * bdrv_recurse_is_first_non_filter on them. 86 * For a sample implementation look in the future Quorum block filter. 87 */ 88 bool (*bdrv_recurse_is_first_non_filter)(BlockDriverState *bs, 89 BlockDriverState *candidate); 90 91 int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename); 92 int (*bdrv_probe_device)(const char *filename); 93 94 /* Any driver implementing this callback is expected to be able to handle 95 * NULL file names in its .bdrv_open() implementation */ 96 void (*bdrv_parse_filename)(const char *filename, QDict *options, Error **errp); 97 /* Drivers not implementing bdrv_parse_filename nor bdrv_open should have 98 * this field set to true, except ones that are defined only by their 99 * child's bs. 100 * An example of the last type will be the quorum block driver. 101 */ 102 bool bdrv_needs_filename; 103 104 /* Set if a driver can support backing files */ 105 bool supports_backing; 106 107 /* For handling image reopen for split or non-split files */ 108 int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state, 109 BlockReopenQueue *queue, Error **errp); 110 void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state); 111 void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state); 112 113 int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags, 114 Error **errp); 115 int (*bdrv_file_open)(BlockDriverState *bs, QDict *options, int flags, 116 Error **errp); 117 int (*bdrv_read)(BlockDriverState *bs, int64_t sector_num, 118 uint8_t *buf, int nb_sectors); 119 int (*bdrv_write)(BlockDriverState *bs, int64_t sector_num, 120 const uint8_t *buf, int nb_sectors); 121 void (*bdrv_close)(BlockDriverState *bs); 122 void (*bdrv_rebind)(BlockDriverState *bs); 123 int (*bdrv_create)(const char *filename, QemuOpts *opts, Error **errp); 124 int (*bdrv_set_key)(BlockDriverState *bs, const char *key); 125 int (*bdrv_make_empty)(BlockDriverState *bs); 126 /* aio */ 127 BlockDriverAIOCB *(*bdrv_aio_readv)(BlockDriverState *bs, 128 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, 129 BlockDriverCompletionFunc *cb, void *opaque); 130 BlockDriverAIOCB *(*bdrv_aio_writev)(BlockDriverState *bs, 131 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, 132 BlockDriverCompletionFunc *cb, void *opaque); 133 BlockDriverAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs, 134 BlockDriverCompletionFunc *cb, void *opaque); 135 BlockDriverAIOCB *(*bdrv_aio_discard)(BlockDriverState *bs, 136 int64_t sector_num, int nb_sectors, 137 BlockDriverCompletionFunc *cb, void *opaque); 138 139 int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs, 140 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov); 141 int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs, 142 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov); 143 /* 144 * Efficiently zero a region of the disk image. Typically an image format 145 * would use a compact metadata representation to implement this. This 146 * function pointer may be NULL and .bdrv_co_writev() will be called 147 * instead. 148 */ 149 int coroutine_fn (*bdrv_co_write_zeroes)(BlockDriverState *bs, 150 int64_t sector_num, int nb_sectors, BdrvRequestFlags flags); 151 int coroutine_fn (*bdrv_co_discard)(BlockDriverState *bs, 152 int64_t sector_num, int nb_sectors); 153 int64_t coroutine_fn (*bdrv_co_get_block_status)(BlockDriverState *bs, 154 int64_t sector_num, int nb_sectors, int *pnum); 155 156 /* 157 * Invalidate any cached meta-data. 158 */ 159 void (*bdrv_invalidate_cache)(BlockDriverState *bs, Error **errp); 160 161 /* 162 * Flushes all data that was already written to the OS all the way down to 163 * the disk (for example raw-posix calls fsync()). 164 */ 165 int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs); 166 167 /* 168 * Flushes all internal caches to the OS. The data may still sit in a 169 * writeback cache of the host OS, but it will survive a crash of the qemu 170 * process. 171 */ 172 int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs); 173 174 const char *protocol_name; 175 int (*bdrv_truncate)(BlockDriverState *bs, int64_t offset); 176 177 int64_t (*bdrv_getlength)(BlockDriverState *bs); 178 bool has_variable_length; 179 int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs); 180 181 int (*bdrv_write_compressed)(BlockDriverState *bs, int64_t sector_num, 182 const uint8_t *buf, int nb_sectors); 183 184 int (*bdrv_snapshot_create)(BlockDriverState *bs, 185 QEMUSnapshotInfo *sn_info); 186 int (*bdrv_snapshot_goto)(BlockDriverState *bs, 187 const char *snapshot_id); 188 int (*bdrv_snapshot_delete)(BlockDriverState *bs, 189 const char *snapshot_id, 190 const char *name, 191 Error **errp); 192 int (*bdrv_snapshot_list)(BlockDriverState *bs, 193 QEMUSnapshotInfo **psn_info); 194 int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs, 195 const char *snapshot_id, 196 const char *name, 197 Error **errp); 198 int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi); 199 ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs); 200 201 int (*bdrv_save_vmstate)(BlockDriverState *bs, QEMUIOVector *qiov, 202 int64_t pos); 203 int (*bdrv_load_vmstate)(BlockDriverState *bs, uint8_t *buf, 204 int64_t pos, int size); 205 206 int (*bdrv_change_backing_file)(BlockDriverState *bs, 207 const char *backing_file, const char *backing_fmt); 208 209 /* removable device specific */ 210 int (*bdrv_is_inserted)(BlockDriverState *bs); 211 int (*bdrv_media_changed)(BlockDriverState *bs); 212 void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag); 213 void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked); 214 215 /* to control generic scsi devices */ 216 int (*bdrv_ioctl)(BlockDriverState *bs, unsigned long int req, void *buf); 217 BlockDriverAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs, 218 unsigned long int req, void *buf, 219 BlockDriverCompletionFunc *cb, void *opaque); 220 221 /* List of options for creating images, terminated by name == NULL */ 222 QemuOptsList *create_opts; 223 224 /* 225 * Returns 0 for completed check, -errno for internal errors. 226 * The check results are stored in result. 227 */ 228 int (*bdrv_check)(BlockDriverState* bs, BdrvCheckResult *result, 229 BdrvCheckMode fix); 230 231 int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts); 232 233 void (*bdrv_debug_event)(BlockDriverState *bs, BlkDebugEvent event); 234 235 /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */ 236 int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event, 237 const char *tag); 238 int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs, 239 const char *tag); 240 int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag); 241 bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag); 242 243 int (*bdrv_refresh_limits)(BlockDriverState *bs); 244 245 /* 246 * Returns 1 if newly created images are guaranteed to contain only 247 * zeros, 0 otherwise. 248 */ 249 int (*bdrv_has_zero_init)(BlockDriverState *bs); 250 251 /* Remove fd handlers, timers, and other event loop callbacks so the event 252 * loop is no longer in use. Called with no in-flight requests and in 253 * depth-first traversal order with parents before child nodes. 254 */ 255 void (*bdrv_detach_aio_context)(BlockDriverState *bs); 256 257 /* Add fd handlers, timers, and other event loop callbacks so I/O requests 258 * can be processed again. Called with no in-flight requests and in 259 * depth-first traversal order with child nodes before parent nodes. 260 */ 261 void (*bdrv_attach_aio_context)(BlockDriverState *bs, 262 AioContext *new_context); 263 264 QLIST_ENTRY(BlockDriver) list; 265 }; 266 267 typedef struct BlockLimits { 268 /* maximum number of sectors that can be discarded at once */ 269 int max_discard; 270 271 /* optimal alignment for discard requests in sectors */ 272 int64_t discard_alignment; 273 274 /* maximum number of sectors that can zeroized at once */ 275 int max_write_zeroes; 276 277 /* optimal alignment for write zeroes requests in sectors */ 278 int64_t write_zeroes_alignment; 279 280 /* optimal transfer length in sectors */ 281 int opt_transfer_length; 282 283 /* memory alignment so that no bounce buffer is needed */ 284 size_t opt_mem_alignment; 285 } BlockLimits; 286 287 typedef struct BdrvOpBlocker BdrvOpBlocker; 288 289 /* 290 * Note: the function bdrv_append() copies and swaps contents of 291 * BlockDriverStates, so if you add new fields to this struct, please 292 * inspect bdrv_append() to determine if the new fields need to be 293 * copied as well. 294 */ 295 struct BlockDriverState { 296 int64_t total_sectors; /* if we are reading a disk image, give its 297 size in sectors */ 298 int read_only; /* if true, the media is read only */ 299 int open_flags; /* flags used to open the file, re-used for re-open */ 300 int encrypted; /* if true, the media is encrypted */ 301 int valid_key; /* if true, a valid encryption key has been set */ 302 int sg; /* if true, the device is a /dev/sg* */ 303 int copy_on_read; /* if true, copy read backing sectors into image 304 note this is a reference count */ 305 306 BlockDriver *drv; /* NULL means no media */ 307 void *opaque; 308 309 void *dev; /* attached device model, if any */ 310 /* TODO change to DeviceState when all users are qdevified */ 311 const BlockDevOps *dev_ops; 312 void *dev_opaque; 313 314 AioContext *aio_context; /* event loop used for fd handlers, timers, etc */ 315 316 char filename[1024]; 317 char backing_file[1024]; /* if non zero, the image is a diff of 318 this file image */ 319 char backing_format[16]; /* if non-zero and backing_file exists */ 320 321 BlockDriverState *backing_hd; 322 BlockDriverState *file; 323 324 NotifierList close_notifiers; 325 326 /* Callback before write request is processed */ 327 NotifierWithReturnList before_write_notifiers; 328 329 /* number of in-flight serialising requests */ 330 unsigned int serialising_in_flight; 331 332 /* I/O throttling */ 333 ThrottleState throttle_state; 334 CoQueue throttled_reqs[2]; 335 bool io_limits_enabled; 336 337 /* I/O stats (display with "info blockstats"). */ 338 uint64_t nr_bytes[BDRV_MAX_IOTYPE]; 339 uint64_t nr_ops[BDRV_MAX_IOTYPE]; 340 uint64_t total_time_ns[BDRV_MAX_IOTYPE]; 341 uint64_t wr_highest_sector; 342 343 /* I/O Limits */ 344 BlockLimits bl; 345 346 /* Whether the disk can expand beyond total_sectors */ 347 int growable; 348 349 /* Whether produces zeros when read beyond eof */ 350 bool zero_beyond_eof; 351 352 /* Alignment requirement for offset/length of I/O requests */ 353 unsigned int request_alignment; 354 355 /* the block size for which the guest device expects atomicity */ 356 int guest_block_size; 357 358 /* do we need to tell the quest if we have a volatile write cache? */ 359 int enable_write_cache; 360 361 /* NOTE: the following infos are only hints for real hardware 362 drivers. They are not used by the block driver */ 363 BlockdevOnError on_read_error, on_write_error; 364 bool iostatus_enabled; 365 BlockDeviceIoStatus iostatus; 366 367 /* the following member gives a name to every node on the bs graph. */ 368 char node_name[32]; 369 /* element of the list of named nodes building the graph */ 370 QTAILQ_ENTRY(BlockDriverState) node_list; 371 /* Device name is the name associated with the "drive" the guest sees */ 372 char device_name[32]; 373 /* element of the list of "drives" the guest sees */ 374 QTAILQ_ENTRY(BlockDriverState) device_list; 375 QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps; 376 int refcnt; 377 378 QLIST_HEAD(, BdrvTrackedRequest) tracked_requests; 379 380 /* operation blockers */ 381 QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX]; 382 383 /* long-running background operation */ 384 BlockJob *job; 385 386 QDict *options; 387 BlockdevDetectZeroesOptions detect_zeroes; 388 389 /* The error object in use for blocking operations on backing_hd */ 390 Error *backing_blocker; 391 }; 392 393 int get_tmp_filename(char *filename, int size); 394 395 void bdrv_set_io_limits(BlockDriverState *bs, 396 ThrottleConfig *cfg); 397 398 399 /** 400 * bdrv_add_before_write_notifier: 401 * 402 * Register a callback that is invoked before write requests are processed but 403 * after any throttling or waiting for overlapping requests. 404 */ 405 void bdrv_add_before_write_notifier(BlockDriverState *bs, 406 NotifierWithReturn *notifier); 407 408 /** 409 * bdrv_detach_aio_context: 410 * 411 * May be called from .bdrv_detach_aio_context() to detach children from the 412 * current #AioContext. This is only needed by block drivers that manage their 413 * own children. Both ->file and ->backing_hd are automatically handled and 414 * block drivers should not call this function on them explicitly. 415 */ 416 void bdrv_detach_aio_context(BlockDriverState *bs); 417 418 /** 419 * bdrv_attach_aio_context: 420 * 421 * May be called from .bdrv_attach_aio_context() to attach children to the new 422 * #AioContext. This is only needed by block drivers that manage their own 423 * children. Both ->file and ->backing_hd are automatically handled and block 424 * drivers should not call this function on them explicitly. 425 */ 426 void bdrv_attach_aio_context(BlockDriverState *bs, 427 AioContext *new_context); 428 429 #ifdef _WIN32 430 int is_windows_drive(const char *filename); 431 #endif 432 433 /** 434 * stream_start: 435 * @bs: Block device to operate on. 436 * @base: Block device that will become the new base, or %NULL to 437 * flatten the whole backing file chain onto @bs. 438 * @base_id: The file name that will be written to @bs as the new 439 * backing file if the job completes. Ignored if @base is %NULL. 440 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 441 * @on_error: The action to take upon error. 442 * @cb: Completion function for the job. 443 * @opaque: Opaque pointer value passed to @cb. 444 * @errp: Error object. 445 * 446 * Start a streaming operation on @bs. Clusters that are unallocated 447 * in @bs, but allocated in any image between @base and @bs (both 448 * exclusive) will be written to @bs. At the end of a successful 449 * streaming job, the backing file of @bs will be changed to 450 * @base_id in the written image and to @base in the live BlockDriverState. 451 */ 452 void stream_start(BlockDriverState *bs, BlockDriverState *base, 453 const char *base_id, int64_t speed, BlockdevOnError on_error, 454 BlockDriverCompletionFunc *cb, 455 void *opaque, Error **errp); 456 457 /** 458 * commit_start: 459 * @bs: Active block device. 460 * @top: Top block device to be committed. 461 * @base: Block device that will be written into, and become the new top. 462 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 463 * @on_error: The action to take upon error. 464 * @cb: Completion function for the job. 465 * @opaque: Opaque pointer value passed to @cb. 466 * @errp: Error object. 467 * 468 */ 469 void commit_start(BlockDriverState *bs, BlockDriverState *base, 470 BlockDriverState *top, int64_t speed, 471 BlockdevOnError on_error, BlockDriverCompletionFunc *cb, 472 void *opaque, Error **errp); 473 /** 474 * commit_active_start: 475 * @bs: Active block device to be committed. 476 * @base: Block device that will be written into, and become the new top. 477 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 478 * @on_error: The action to take upon error. 479 * @cb: Completion function for the job. 480 * @opaque: Opaque pointer value passed to @cb. 481 * @errp: Error object. 482 * 483 */ 484 void commit_active_start(BlockDriverState *bs, BlockDriverState *base, 485 int64_t speed, 486 BlockdevOnError on_error, 487 BlockDriverCompletionFunc *cb, 488 void *opaque, Error **errp); 489 /* 490 * mirror_start: 491 * @bs: Block device to operate on. 492 * @target: Block device to write to. 493 * @replaces: Block graph node name to replace once the mirror is done. Can 494 * only be used when full mirroring is selected. 495 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 496 * @granularity: The chosen granularity for the dirty bitmap. 497 * @buf_size: The amount of data that can be in flight at one time. 498 * @mode: Whether to collapse all images in the chain to the target. 499 * @on_source_error: The action to take upon error reading from the source. 500 * @on_target_error: The action to take upon error writing to the target. 501 * @cb: Completion function for the job. 502 * @opaque: Opaque pointer value passed to @cb. 503 * @errp: Error object. 504 * 505 * Start a mirroring operation on @bs. Clusters that are allocated 506 * in @bs will be written to @bs until the job is cancelled or 507 * manually completed. At the end of a successful mirroring job, 508 * @bs will be switched to read from @target. 509 */ 510 void mirror_start(BlockDriverState *bs, BlockDriverState *target, 511 const char *replaces, 512 int64_t speed, int64_t granularity, int64_t buf_size, 513 MirrorSyncMode mode, BlockdevOnError on_source_error, 514 BlockdevOnError on_target_error, 515 BlockDriverCompletionFunc *cb, 516 void *opaque, Error **errp); 517 518 /* 519 * backup_start: 520 * @bs: Block device to operate on. 521 * @target: Block device to write to. 522 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 523 * @sync_mode: What parts of the disk image should be copied to the destination. 524 * @on_source_error: The action to take upon error reading from the source. 525 * @on_target_error: The action to take upon error writing to the target. 526 * @cb: Completion function for the job. 527 * @opaque: Opaque pointer value passed to @cb. 528 * 529 * Start a backup operation on @bs. Clusters in @bs are written to @target 530 * until the job is cancelled or manually completed. 531 */ 532 void backup_start(BlockDriverState *bs, BlockDriverState *target, 533 int64_t speed, MirrorSyncMode sync_mode, 534 BlockdevOnError on_source_error, 535 BlockdevOnError on_target_error, 536 BlockDriverCompletionFunc *cb, void *opaque, 537 Error **errp); 538 539 #endif /* BLOCK_INT_H */ 540