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 57 typedef struct BdrvTrackedRequest { 58 BlockDriverState *bs; 59 int64_t sector_num; 60 int nb_sectors; 61 bool is_write; 62 QLIST_ENTRY(BdrvTrackedRequest) list; 63 Coroutine *co; /* owner, used for deadlock detection */ 64 CoQueue wait_queue; /* coroutines blocked on this request */ 65 } BdrvTrackedRequest; 66 67 struct BlockDriver { 68 const char *format_name; 69 int instance_size; 70 int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename); 71 int (*bdrv_probe_device)(const char *filename); 72 73 /* Any driver implementing this callback is expected to be able to handle 74 * NULL file names in its .bdrv_open() implementation */ 75 void (*bdrv_parse_filename)(const char *filename, QDict *options, Error **errp); 76 /* Drivers not implementing bdrv_parse_filename nor bdrv_open should have 77 * this field set to true, except ones that are defined only by their 78 * child's bs. 79 * An example of the last type will be the quorum block driver. 80 */ 81 bool bdrv_needs_filename; 82 83 /* For handling image reopen for split or non-split files */ 84 int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state, 85 BlockReopenQueue *queue, Error **errp); 86 void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state); 87 void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state); 88 89 int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags, 90 Error **errp); 91 int (*bdrv_file_open)(BlockDriverState *bs, QDict *options, int flags, 92 Error **errp); 93 int (*bdrv_read)(BlockDriverState *bs, int64_t sector_num, 94 uint8_t *buf, int nb_sectors); 95 int (*bdrv_write)(BlockDriverState *bs, int64_t sector_num, 96 const uint8_t *buf, int nb_sectors); 97 void (*bdrv_close)(BlockDriverState *bs); 98 void (*bdrv_rebind)(BlockDriverState *bs); 99 int (*bdrv_create)(const char *filename, QEMUOptionParameter *options, 100 Error **errp); 101 int (*bdrv_set_key)(BlockDriverState *bs, const char *key); 102 int (*bdrv_make_empty)(BlockDriverState *bs); 103 /* aio */ 104 BlockDriverAIOCB *(*bdrv_aio_readv)(BlockDriverState *bs, 105 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, 106 BlockDriverCompletionFunc *cb, void *opaque); 107 BlockDriverAIOCB *(*bdrv_aio_writev)(BlockDriverState *bs, 108 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, 109 BlockDriverCompletionFunc *cb, void *opaque); 110 BlockDriverAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs, 111 BlockDriverCompletionFunc *cb, void *opaque); 112 BlockDriverAIOCB *(*bdrv_aio_discard)(BlockDriverState *bs, 113 int64_t sector_num, int nb_sectors, 114 BlockDriverCompletionFunc *cb, void *opaque); 115 116 int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs, 117 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov); 118 int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs, 119 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov); 120 /* 121 * Efficiently zero a region of the disk image. Typically an image format 122 * would use a compact metadata representation to implement this. This 123 * function pointer may be NULL and .bdrv_co_writev() will be called 124 * instead. 125 */ 126 int coroutine_fn (*bdrv_co_write_zeroes)(BlockDriverState *bs, 127 int64_t sector_num, int nb_sectors); 128 int coroutine_fn (*bdrv_co_discard)(BlockDriverState *bs, 129 int64_t sector_num, int nb_sectors); 130 int64_t coroutine_fn (*bdrv_co_get_block_status)(BlockDriverState *bs, 131 int64_t sector_num, int nb_sectors, int *pnum); 132 133 /* 134 * Invalidate any cached meta-data. 135 */ 136 void (*bdrv_invalidate_cache)(BlockDriverState *bs); 137 138 /* 139 * Flushes all data that was already written to the OS all the way down to 140 * the disk (for example raw-posix calls fsync()). 141 */ 142 int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs); 143 144 /* 145 * Flushes all internal caches to the OS. The data may still sit in a 146 * writeback cache of the host OS, but it will survive a crash of the qemu 147 * process. 148 */ 149 int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs); 150 151 const char *protocol_name; 152 int (*bdrv_truncate)(BlockDriverState *bs, int64_t offset); 153 int64_t (*bdrv_getlength)(BlockDriverState *bs); 154 int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs); 155 int (*bdrv_write_compressed)(BlockDriverState *bs, int64_t sector_num, 156 const uint8_t *buf, int nb_sectors); 157 158 int (*bdrv_snapshot_create)(BlockDriverState *bs, 159 QEMUSnapshotInfo *sn_info); 160 int (*bdrv_snapshot_goto)(BlockDriverState *bs, 161 const char *snapshot_id); 162 int (*bdrv_snapshot_delete)(BlockDriverState *bs, 163 const char *snapshot_id, 164 const char *name, 165 Error **errp); 166 int (*bdrv_snapshot_list)(BlockDriverState *bs, 167 QEMUSnapshotInfo **psn_info); 168 int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs, 169 const char *snapshot_name); 170 int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi); 171 ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs); 172 173 int (*bdrv_save_vmstate)(BlockDriverState *bs, QEMUIOVector *qiov, 174 int64_t pos); 175 int (*bdrv_load_vmstate)(BlockDriverState *bs, uint8_t *buf, 176 int64_t pos, int size); 177 178 int (*bdrv_change_backing_file)(BlockDriverState *bs, 179 const char *backing_file, const char *backing_fmt); 180 181 /* removable device specific */ 182 int (*bdrv_is_inserted)(BlockDriverState *bs); 183 int (*bdrv_media_changed)(BlockDriverState *bs); 184 void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag); 185 void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked); 186 187 /* to control generic scsi devices */ 188 int (*bdrv_ioctl)(BlockDriverState *bs, unsigned long int req, void *buf); 189 BlockDriverAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs, 190 unsigned long int req, void *buf, 191 BlockDriverCompletionFunc *cb, void *opaque); 192 193 /* List of options for creating images, terminated by name == NULL */ 194 QEMUOptionParameter *create_options; 195 196 197 /* 198 * Returns 0 for completed check, -errno for internal errors. 199 * The check results are stored in result. 200 */ 201 int (*bdrv_check)(BlockDriverState* bs, BdrvCheckResult *result, 202 BdrvCheckMode fix); 203 204 int (*bdrv_amend_options)(BlockDriverState *bs, 205 QEMUOptionParameter *options); 206 207 void (*bdrv_debug_event)(BlockDriverState *bs, BlkDebugEvent event); 208 209 /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */ 210 int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event, 211 const char *tag); 212 int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag); 213 bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag); 214 215 /* 216 * Returns 1 if newly created images are guaranteed to contain only 217 * zeros, 0 otherwise. 218 */ 219 int (*bdrv_has_zero_init)(BlockDriverState *bs); 220 221 QLIST_ENTRY(BlockDriver) list; 222 }; 223 224 /* 225 * Note: the function bdrv_append() copies and swaps contents of 226 * BlockDriverStates, so if you add new fields to this struct, please 227 * inspect bdrv_append() to determine if the new fields need to be 228 * copied as well. 229 */ 230 struct BlockDriverState { 231 int64_t total_sectors; /* if we are reading a disk image, give its 232 size in sectors */ 233 int read_only; /* if true, the media is read only */ 234 int open_flags; /* flags used to open the file, re-used for re-open */ 235 int encrypted; /* if true, the media is encrypted */ 236 int valid_key; /* if true, a valid encryption key has been set */ 237 int sg; /* if true, the device is a /dev/sg* */ 238 int copy_on_read; /* if true, copy read backing sectors into image 239 note this is a reference count */ 240 241 BlockDriver *drv; /* NULL means no media */ 242 void *opaque; 243 244 void *dev; /* attached device model, if any */ 245 /* TODO change to DeviceState when all users are qdevified */ 246 const BlockDevOps *dev_ops; 247 void *dev_opaque; 248 249 char filename[1024]; 250 char backing_file[1024]; /* if non zero, the image is a diff of 251 this file image */ 252 char backing_format[16]; /* if non-zero and backing_file exists */ 253 int is_temporary; 254 255 BlockDriverState *backing_hd; 256 BlockDriverState *file; 257 258 NotifierList close_notifiers; 259 260 /* Callback before write request is processed */ 261 NotifierWithReturnList before_write_notifiers; 262 263 /* number of in-flight copy-on-read requests */ 264 unsigned int copy_on_read_in_flight; 265 266 /* I/O throttling */ 267 ThrottleState throttle_state; 268 CoQueue throttled_reqs[2]; 269 bool io_limits_enabled; 270 271 /* I/O stats (display with "info blockstats"). */ 272 uint64_t nr_bytes[BDRV_MAX_IOTYPE]; 273 uint64_t nr_ops[BDRV_MAX_IOTYPE]; 274 uint64_t total_time_ns[BDRV_MAX_IOTYPE]; 275 uint64_t wr_highest_sector; 276 277 /* Whether the disk can expand beyond total_sectors */ 278 int growable; 279 280 /* Whether produces zeros when read beyond eof */ 281 bool zero_beyond_eof; 282 283 /* the memory alignment required for the buffers handled by this driver */ 284 int buffer_alignment; 285 286 /* do we need to tell the quest if we have a volatile write cache? */ 287 int enable_write_cache; 288 289 /* NOTE: the following infos are only hints for real hardware 290 drivers. They are not used by the block driver */ 291 BlockdevOnError on_read_error, on_write_error; 292 bool iostatus_enabled; 293 BlockDeviceIoStatus iostatus; 294 char device_name[32]; 295 HBitmap *dirty_bitmap; 296 int refcnt; 297 int in_use; /* users other than guest access, eg. block migration */ 298 QTAILQ_ENTRY(BlockDriverState) list; 299 300 QLIST_HEAD(, BdrvTrackedRequest) tracked_requests; 301 302 /* long-running background operation */ 303 BlockJob *job; 304 305 QDict *options; 306 }; 307 308 int get_tmp_filename(char *filename, int size); 309 310 void bdrv_set_io_limits(BlockDriverState *bs, 311 ThrottleConfig *cfg); 312 313 314 /** 315 * bdrv_add_before_write_notifier: 316 * 317 * Register a callback that is invoked before write requests are processed but 318 * after any throttling or waiting for overlapping requests. 319 */ 320 void bdrv_add_before_write_notifier(BlockDriverState *bs, 321 NotifierWithReturn *notifier); 322 323 /** 324 * bdrv_get_aio_context: 325 * 326 * Returns: the currently bound #AioContext 327 */ 328 AioContext *bdrv_get_aio_context(BlockDriverState *bs); 329 330 #ifdef _WIN32 331 int is_windows_drive(const char *filename); 332 #endif 333 void bdrv_emit_qmp_error_event(const BlockDriverState *bdrv, 334 enum MonitorEvent ev, 335 BlockErrorAction action, bool is_read); 336 337 /** 338 * stream_start: 339 * @bs: Block device to operate on. 340 * @base: Block device that will become the new base, or %NULL to 341 * flatten the whole backing file chain onto @bs. 342 * @base_id: The file name that will be written to @bs as the new 343 * backing file if the job completes. Ignored if @base is %NULL. 344 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 345 * @on_error: The action to take upon error. 346 * @cb: Completion function for the job. 347 * @opaque: Opaque pointer value passed to @cb. 348 * @errp: Error object. 349 * 350 * Start a streaming operation on @bs. Clusters that are unallocated 351 * in @bs, but allocated in any image between @base and @bs (both 352 * exclusive) will be written to @bs. At the end of a successful 353 * streaming job, the backing file of @bs will be changed to 354 * @base_id in the written image and to @base in the live BlockDriverState. 355 */ 356 void stream_start(BlockDriverState *bs, BlockDriverState *base, 357 const char *base_id, int64_t speed, BlockdevOnError on_error, 358 BlockDriverCompletionFunc *cb, 359 void *opaque, Error **errp); 360 361 /** 362 * commit_start: 363 * @bs: Top Block device 364 * @base: Block device that will be written into, and become the new top 365 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 366 * @on_error: The action to take upon error. 367 * @cb: Completion function for the job. 368 * @opaque: Opaque pointer value passed to @cb. 369 * @errp: Error object. 370 * 371 */ 372 void commit_start(BlockDriverState *bs, BlockDriverState *base, 373 BlockDriverState *top, int64_t speed, 374 BlockdevOnError on_error, BlockDriverCompletionFunc *cb, 375 void *opaque, Error **errp); 376 377 /* 378 * mirror_start: 379 * @bs: Block device to operate on. 380 * @target: Block device to write to. 381 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 382 * @granularity: The chosen granularity for the dirty bitmap. 383 * @buf_size: The amount of data that can be in flight at one time. 384 * @mode: Whether to collapse all images in the chain to the target. 385 * @on_source_error: The action to take upon error reading from the source. 386 * @on_target_error: The action to take upon error writing to the target. 387 * @cb: Completion function for the job. 388 * @opaque: Opaque pointer value passed to @cb. 389 * @errp: Error object. 390 * 391 * Start a mirroring operation on @bs. Clusters that are allocated 392 * in @bs will be written to @bs until the job is cancelled or 393 * manually completed. At the end of a successful mirroring job, 394 * @bs will be switched to read from @target. 395 */ 396 void mirror_start(BlockDriverState *bs, BlockDriverState *target, 397 int64_t speed, int64_t granularity, int64_t buf_size, 398 MirrorSyncMode mode, BlockdevOnError on_source_error, 399 BlockdevOnError on_target_error, 400 BlockDriverCompletionFunc *cb, 401 void *opaque, Error **errp); 402 403 /* 404 * backup_start: 405 * @bs: Block device to operate on. 406 * @target: Block device to write to. 407 * @speed: The maximum speed, in bytes per second, or 0 for unlimited. 408 * @sync_mode: What parts of the disk image should be copied to the destination. 409 * @on_source_error: The action to take upon error reading from the source. 410 * @on_target_error: The action to take upon error writing to the target. 411 * @cb: Completion function for the job. 412 * @opaque: Opaque pointer value passed to @cb. 413 * 414 * Start a backup operation on @bs. Clusters in @bs are written to @target 415 * until the job is cancelled or manually completed. 416 */ 417 void backup_start(BlockDriverState *bs, BlockDriverState *target, 418 int64_t speed, MirrorSyncMode sync_mode, 419 BlockdevOnError on_source_error, 420 BlockdevOnError on_target_error, 421 BlockDriverCompletionFunc *cb, void *opaque, 422 Error **errp); 423 424 #endif /* BLOCK_INT_H */ 425