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