2 * QEMU System Emulator block driver
4 * Copyright (c) 2003 Fabrice Bellard
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:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
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
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"
40 #define BLOCK_FLAG_ENCRYPT 1
41 #define BLOCK_FLAG_COMPAT6 4
42 #define BLOCK_FLAG_LAZY_REFCOUNTS 8
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"
58 typedef struct BdrvTrackedRequest
{
65 int64_t overlap_offset
;
66 unsigned int overlap_bytes
;
68 QLIST_ENTRY(BdrvTrackedRequest
) list
;
69 Coroutine
*co
; /* owner, used for deadlock detection */
70 CoQueue wait_queue
; /* coroutines blocked on this request */
72 struct BdrvTrackedRequest
*waiting_for
;
76 const char *format_name
;
79 /* this table of boolean contains authorizations for the block operations */
80 bool authorizations
[BS_AUTHORIZATION_COUNT
];
81 /* for snapshots complex block filter like Quorum can implement the
82 * following recursive callback instead of BS_IS_A_FILTER.
83 * It's purpose is to recurse on the filter children while calling
84 * bdrv_recurse_is_first_non_filter on them.
85 * For a sample implementation look in the future Quorum block filter.
87 bool (*bdrv_recurse_is_first_non_filter
)(BlockDriverState
*bs
,
88 BlockDriverState
*candidate
);
90 int (*bdrv_probe
)(const uint8_t *buf
, int buf_size
, const char *filename
);
91 int (*bdrv_probe_device
)(const char *filename
);
93 /* Any driver implementing this callback is expected to be able to handle
94 * NULL file names in its .bdrv_open() implementation */
95 void (*bdrv_parse_filename
)(const char *filename
, QDict
*options
, Error
**errp
);
96 /* Drivers not implementing bdrv_parse_filename nor bdrv_open should have
97 * this field set to true, except ones that are defined only by their
99 * An example of the last type will be the quorum block driver.
101 bool bdrv_needs_filename
;
103 /* For handling image reopen for split or non-split files */
104 int (*bdrv_reopen_prepare
)(BDRVReopenState
*reopen_state
,
105 BlockReopenQueue
*queue
, Error
**errp
);
106 void (*bdrv_reopen_commit
)(BDRVReopenState
*reopen_state
);
107 void (*bdrv_reopen_abort
)(BDRVReopenState
*reopen_state
);
109 int (*bdrv_open
)(BlockDriverState
*bs
, QDict
*options
, int flags
,
111 int (*bdrv_file_open
)(BlockDriverState
*bs
, QDict
*options
, int flags
,
113 int (*bdrv_read
)(BlockDriverState
*bs
, int64_t sector_num
,
114 uint8_t *buf
, int nb_sectors
);
115 int (*bdrv_write
)(BlockDriverState
*bs
, int64_t sector_num
,
116 const uint8_t *buf
, int nb_sectors
);
117 void (*bdrv_close
)(BlockDriverState
*bs
);
118 void (*bdrv_rebind
)(BlockDriverState
*bs
);
119 int (*bdrv_create
)(const char *filename
, QEMUOptionParameter
*options
,
121 int (*bdrv_set_key
)(BlockDriverState
*bs
, const char *key
);
122 int (*bdrv_make_empty
)(BlockDriverState
*bs
);
124 BlockDriverAIOCB
*(*bdrv_aio_readv
)(BlockDriverState
*bs
,
125 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
126 BlockDriverCompletionFunc
*cb
, void *opaque
);
127 BlockDriverAIOCB
*(*bdrv_aio_writev
)(BlockDriverState
*bs
,
128 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
129 BlockDriverCompletionFunc
*cb
, void *opaque
);
130 BlockDriverAIOCB
*(*bdrv_aio_flush
)(BlockDriverState
*bs
,
131 BlockDriverCompletionFunc
*cb
, void *opaque
);
132 BlockDriverAIOCB
*(*bdrv_aio_discard
)(BlockDriverState
*bs
,
133 int64_t sector_num
, int nb_sectors
,
134 BlockDriverCompletionFunc
*cb
, void *opaque
);
136 int coroutine_fn (*bdrv_co_readv
)(BlockDriverState
*bs
,
137 int64_t sector_num
, int nb_sectors
, QEMUIOVector
*qiov
);
138 int coroutine_fn (*bdrv_co_writev
)(BlockDriverState
*bs
,
139 int64_t sector_num
, int nb_sectors
, QEMUIOVector
*qiov
);
141 * Efficiently zero a region of the disk image. Typically an image format
142 * would use a compact metadata representation to implement this. This
143 * function pointer may be NULL and .bdrv_co_writev() will be called
146 int coroutine_fn (*bdrv_co_write_zeroes
)(BlockDriverState
*bs
,
147 int64_t sector_num
, int nb_sectors
, BdrvRequestFlags flags
);
148 int coroutine_fn (*bdrv_co_discard
)(BlockDriverState
*bs
,
149 int64_t sector_num
, int nb_sectors
);
150 int64_t coroutine_fn (*bdrv_co_get_block_status
)(BlockDriverState
*bs
,
151 int64_t sector_num
, int nb_sectors
, int *pnum
);
154 * Invalidate any cached meta-data.
156 void (*bdrv_invalidate_cache
)(BlockDriverState
*bs
);
159 * Flushes all data that was already written to the OS all the way down to
160 * the disk (for example raw-posix calls fsync()).
162 int coroutine_fn (*bdrv_co_flush_to_disk
)(BlockDriverState
*bs
);
165 * Flushes all internal caches to the OS. The data may still sit in a
166 * writeback cache of the host OS, but it will survive a crash of the qemu
169 int coroutine_fn (*bdrv_co_flush_to_os
)(BlockDriverState
*bs
);
171 const char *protocol_name
;
172 int (*bdrv_truncate
)(BlockDriverState
*bs
, int64_t offset
);
174 int64_t (*bdrv_getlength
)(BlockDriverState
*bs
);
175 bool has_variable_length
;
176 int64_t (*bdrv_get_allocated_file_size
)(BlockDriverState
*bs
);
178 int (*bdrv_write_compressed
)(BlockDriverState
*bs
, int64_t sector_num
,
179 const uint8_t *buf
, int nb_sectors
);
181 int (*bdrv_snapshot_create
)(BlockDriverState
*bs
,
182 QEMUSnapshotInfo
*sn_info
);
183 int (*bdrv_snapshot_goto
)(BlockDriverState
*bs
,
184 const char *snapshot_id
);
185 int (*bdrv_snapshot_delete
)(BlockDriverState
*bs
,
186 const char *snapshot_id
,
189 int (*bdrv_snapshot_list
)(BlockDriverState
*bs
,
190 QEMUSnapshotInfo
**psn_info
);
191 int (*bdrv_snapshot_load_tmp
)(BlockDriverState
*bs
,
192 const char *snapshot_id
,
195 int (*bdrv_get_info
)(BlockDriverState
*bs
, BlockDriverInfo
*bdi
);
196 ImageInfoSpecific
*(*bdrv_get_specific_info
)(BlockDriverState
*bs
);
198 int (*bdrv_save_vmstate
)(BlockDriverState
*bs
, QEMUIOVector
*qiov
,
200 int (*bdrv_load_vmstate
)(BlockDriverState
*bs
, uint8_t *buf
,
201 int64_t pos
, int size
);
203 int (*bdrv_change_backing_file
)(BlockDriverState
*bs
,
204 const char *backing_file
, const char *backing_fmt
);
206 /* removable device specific */
207 int (*bdrv_is_inserted
)(BlockDriverState
*bs
);
208 int (*bdrv_media_changed
)(BlockDriverState
*bs
);
209 void (*bdrv_eject
)(BlockDriverState
*bs
, bool eject_flag
);
210 void (*bdrv_lock_medium
)(BlockDriverState
*bs
, bool locked
);
212 /* to control generic scsi devices */
213 int (*bdrv_ioctl
)(BlockDriverState
*bs
, unsigned long int req
, void *buf
);
214 BlockDriverAIOCB
*(*bdrv_aio_ioctl
)(BlockDriverState
*bs
,
215 unsigned long int req
, void *buf
,
216 BlockDriverCompletionFunc
*cb
, void *opaque
);
218 /* List of options for creating images, terminated by name == NULL */
219 QEMUOptionParameter
*create_options
;
223 * Returns 0 for completed check, -errno for internal errors.
224 * The check results are stored in result.
226 int (*bdrv_check
)(BlockDriverState
* bs
, BdrvCheckResult
*result
,
229 int (*bdrv_amend_options
)(BlockDriverState
*bs
,
230 QEMUOptionParameter
*options
);
232 void (*bdrv_debug_event
)(BlockDriverState
*bs
, BlkDebugEvent event
);
234 /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
235 int (*bdrv_debug_breakpoint
)(BlockDriverState
*bs
, const char *event
,
237 int (*bdrv_debug_remove_breakpoint
)(BlockDriverState
*bs
,
239 int (*bdrv_debug_resume
)(BlockDriverState
*bs
, const char *tag
);
240 bool (*bdrv_debug_is_suspended
)(BlockDriverState
*bs
, const char *tag
);
242 int (*bdrv_refresh_limits
)(BlockDriverState
*bs
);
245 * Returns 1 if newly created images are guaranteed to contain only
246 * zeros, 0 otherwise.
248 int (*bdrv_has_zero_init
)(BlockDriverState
*bs
);
250 QLIST_ENTRY(BlockDriver
) list
;
253 typedef struct BlockLimits
{
254 /* maximum number of sectors that can be discarded at once */
257 /* optimal alignment for discard requests in sectors */
258 int64_t discard_alignment
;
260 /* maximum number of sectors that can zeroized at once */
261 int max_write_zeroes
;
263 /* optimal alignment for write zeroes requests in sectors */
264 int64_t write_zeroes_alignment
;
266 /* optimal transfer length in sectors */
267 int opt_transfer_length
;
269 /* memory alignment so that no bounce buffer is needed */
270 size_t opt_mem_alignment
;
274 * Note: the function bdrv_append() copies and swaps contents of
275 * BlockDriverStates, so if you add new fields to this struct, please
276 * inspect bdrv_append() to determine if the new fields need to be
279 struct BlockDriverState
{
280 int64_t total_sectors
; /* if we are reading a disk image, give its
282 int read_only
; /* if true, the media is read only */
283 int open_flags
; /* flags used to open the file, re-used for re-open */
284 int encrypted
; /* if true, the media is encrypted */
285 int valid_key
; /* if true, a valid encryption key has been set */
286 int sg
; /* if true, the device is a /dev/sg* */
287 int copy_on_read
; /* if true, copy read backing sectors into image
288 note this is a reference count */
290 BlockDriver
*drv
; /* NULL means no media */
293 void *dev
; /* attached device model, if any */
294 /* TODO change to DeviceState when all users are qdevified */
295 const BlockDevOps
*dev_ops
;
299 char backing_file
[1024]; /* if non zero, the image is a diff of
301 char backing_format
[16]; /* if non-zero and backing_file exists */
304 BlockDriverState
*backing_hd
;
305 BlockDriverState
*file
;
307 NotifierList close_notifiers
;
309 /* Callback before write request is processed */
310 NotifierWithReturnList before_write_notifiers
;
312 /* number of in-flight serialising requests */
313 unsigned int serialising_in_flight
;
316 ThrottleState throttle_state
;
317 CoQueue throttled_reqs
[2];
318 bool io_limits_enabled
;
320 /* I/O stats (display with "info blockstats"). */
321 uint64_t nr_bytes
[BDRV_MAX_IOTYPE
];
322 uint64_t nr_ops
[BDRV_MAX_IOTYPE
];
323 uint64_t total_time_ns
[BDRV_MAX_IOTYPE
];
324 uint64_t wr_highest_sector
;
329 /* Whether the disk can expand beyond total_sectors */
332 /* Whether produces zeros when read beyond eof */
333 bool zero_beyond_eof
;
335 /* Alignment requirement for offset/length of I/O requests */
336 unsigned int request_alignment
;
338 /* the block size for which the guest device expects atomicity */
339 int guest_block_size
;
341 /* do we need to tell the quest if we have a volatile write cache? */
342 int enable_write_cache
;
344 /* NOTE: the following infos are only hints for real hardware
345 drivers. They are not used by the block driver */
346 BlockdevOnError on_read_error
, on_write_error
;
347 bool iostatus_enabled
;
348 BlockDeviceIoStatus iostatus
;
350 /* the following member gives a name to every node on the bs graph. */
352 /* element of the list of named nodes building the graph */
353 QTAILQ_ENTRY(BlockDriverState
) node_list
;
354 /* Device name is the name associated with the "drive" the guest sees */
355 char device_name
[32];
356 /* element of the list of "drives" the guest sees */
357 QTAILQ_ENTRY(BlockDriverState
) device_list
;
358 QLIST_HEAD(, BdrvDirtyBitmap
) dirty_bitmaps
;
360 int in_use
; /* users other than guest access, eg. block migration */
362 QLIST_HEAD(, BdrvTrackedRequest
) tracked_requests
;
364 /* long-running background operation */
370 int get_tmp_filename(char *filename
, int size
);
372 void bdrv_set_io_limits(BlockDriverState
*bs
,
373 ThrottleConfig
*cfg
);
377 * bdrv_add_before_write_notifier:
379 * Register a callback that is invoked before write requests are processed but
380 * after any throttling or waiting for overlapping requests.
382 void bdrv_add_before_write_notifier(BlockDriverState
*bs
,
383 NotifierWithReturn
*notifier
);
386 * bdrv_get_aio_context:
388 * Returns: the currently bound #AioContext
390 AioContext
*bdrv_get_aio_context(BlockDriverState
*bs
);
393 int is_windows_drive(const char *filename
);
395 void bdrv_emit_qmp_error_event(const BlockDriverState
*bdrv
,
396 enum MonitorEvent ev
,
397 BlockErrorAction action
, bool is_read
);
401 * @bs: Block device to operate on.
402 * @base: Block device that will become the new base, or %NULL to
403 * flatten the whole backing file chain onto @bs.
404 * @base_id: The file name that will be written to @bs as the new
405 * backing file if the job completes. Ignored if @base is %NULL.
406 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
407 * @on_error: The action to take upon error.
408 * @cb: Completion function for the job.
409 * @opaque: Opaque pointer value passed to @cb.
410 * @errp: Error object.
412 * Start a streaming operation on @bs. Clusters that are unallocated
413 * in @bs, but allocated in any image between @base and @bs (both
414 * exclusive) will be written to @bs. At the end of a successful
415 * streaming job, the backing file of @bs will be changed to
416 * @base_id in the written image and to @base in the live BlockDriverState.
418 void stream_start(BlockDriverState
*bs
, BlockDriverState
*base
,
419 const char *base_id
, int64_t speed
, BlockdevOnError on_error
,
420 BlockDriverCompletionFunc
*cb
,
421 void *opaque
, Error
**errp
);
425 * @bs: Active block device.
426 * @top: Top block device to be committed.
427 * @base: Block device that will be written into, and become the new top.
428 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
429 * @on_error: The action to take upon error.
430 * @cb: Completion function for the job.
431 * @opaque: Opaque pointer value passed to @cb.
432 * @errp: Error object.
435 void commit_start(BlockDriverState
*bs
, BlockDriverState
*base
,
436 BlockDriverState
*top
, int64_t speed
,
437 BlockdevOnError on_error
, BlockDriverCompletionFunc
*cb
,
438 void *opaque
, Error
**errp
);
440 * commit_active_start:
441 * @bs: Active block device to be committed.
442 * @base: Block device that will be written into, and become the new top.
443 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
444 * @on_error: The action to take upon error.
445 * @cb: Completion function for the job.
446 * @opaque: Opaque pointer value passed to @cb.
447 * @errp: Error object.
450 void commit_active_start(BlockDriverState
*bs
, BlockDriverState
*base
,
452 BlockdevOnError on_error
,
453 BlockDriverCompletionFunc
*cb
,
454 void *opaque
, Error
**errp
);
457 * @bs: Block device to operate on.
458 * @target: Block device to write to.
459 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
460 * @granularity: The chosen granularity for the dirty bitmap.
461 * @buf_size: The amount of data that can be in flight at one time.
462 * @mode: Whether to collapse all images in the chain to the target.
463 * @on_source_error: The action to take upon error reading from the source.
464 * @on_target_error: The action to take upon error writing to the target.
465 * @cb: Completion function for the job.
466 * @opaque: Opaque pointer value passed to @cb.
467 * @errp: Error object.
469 * Start a mirroring operation on @bs. Clusters that are allocated
470 * in @bs will be written to @bs until the job is cancelled or
471 * manually completed. At the end of a successful mirroring job,
472 * @bs will be switched to read from @target.
474 void mirror_start(BlockDriverState
*bs
, BlockDriverState
*target
,
475 int64_t speed
, int64_t granularity
, int64_t buf_size
,
476 MirrorSyncMode mode
, BlockdevOnError on_source_error
,
477 BlockdevOnError on_target_error
,
478 BlockDriverCompletionFunc
*cb
,
479 void *opaque
, Error
**errp
);
483 * @bs: Block device to operate on.
484 * @target: Block device to write to.
485 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
486 * @sync_mode: What parts of the disk image should be copied to the destination.
487 * @on_source_error: The action to take upon error reading from the source.
488 * @on_target_error: The action to take upon error writing to the target.
489 * @cb: Completion function for the job.
490 * @opaque: Opaque pointer value passed to @cb.
492 * Start a backup operation on @bs. Clusters in @bs are written to @target
493 * until the job is cancelled or manually completed.
495 void backup_start(BlockDriverState
*bs
, BlockDriverState
*target
,
496 int64_t speed
, MirrorSyncMode sync_mode
,
497 BlockdevOnError on_source_error
,
498 BlockdevOnError on_target_error
,
499 BlockDriverCompletionFunc
*cb
, void *opaque
,
502 #endif /* BLOCK_INT_H */