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/accounting.h"
28 #include "block/block.h"
29 #include "qemu/option.h"
30 #include "qemu/queue.h"
31 #include "qemu/coroutine.h"
32 #include "qemu/stats64.h"
33 #include "qemu/timer.h"
34 #include "qapi-types.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_LAZY_REFCOUNTS 8
42 #define BLOCK_OPT_SIZE "size"
43 #define BLOCK_OPT_ENCRYPT "encryption"
44 #define BLOCK_OPT_ENCRYPT_FORMAT "encrypt.format"
45 #define BLOCK_OPT_COMPAT6 "compat6"
46 #define BLOCK_OPT_HWVERSION "hwversion"
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 #define BLOCK_OPT_OBJECT_SIZE "object_size"
59 #define BLOCK_OPT_REFCOUNT_BITS "refcount_bits"
61 #define BLOCK_PROBE_BUF_SIZE 512
63 enum BdrvTrackedRequestType
{
69 typedef struct BdrvTrackedRequest
{
73 enum BdrvTrackedRequestType type
;
76 int64_t overlap_offset
;
77 unsigned int overlap_bytes
;
79 QLIST_ENTRY(BdrvTrackedRequest
) list
;
80 Coroutine
*co
; /* owner, used for deadlock detection */
81 CoQueue wait_queue
; /* coroutines blocked on this request */
83 struct BdrvTrackedRequest
*waiting_for
;
87 const char *format_name
;
90 /* set to true if the BlockDriver is a block filter. Block filters pass
91 * certain callbacks that refer to data (see block.c) to their bs->file if
92 * the driver doesn't implement them. Drivers that do not wish to forward
93 * must implement them and return -ENOTSUP.
96 /* for snapshots block filter like Quorum can implement the
97 * following recursive callback.
98 * It's purpose is to recurse on the filter children while calling
99 * bdrv_recurse_is_first_non_filter on them.
100 * For a sample implementation look in the future Quorum block filter.
102 bool (*bdrv_recurse_is_first_non_filter
)(BlockDriverState
*bs
,
103 BlockDriverState
*candidate
);
105 int (*bdrv_probe
)(const uint8_t *buf
, int buf_size
, const char *filename
);
106 int (*bdrv_probe_device
)(const char *filename
);
108 /* Any driver implementing this callback is expected to be able to handle
109 * NULL file names in its .bdrv_open() implementation */
110 void (*bdrv_parse_filename
)(const char *filename
, QDict
*options
, Error
**errp
);
111 /* Drivers not implementing bdrv_parse_filename nor bdrv_open should have
112 * this field set to true, except ones that are defined only by their
114 * An example of the last type will be the quorum block driver.
116 bool bdrv_needs_filename
;
118 /* Set if a driver can support backing files */
119 bool supports_backing
;
121 /* For handling image reopen for split or non-split files */
122 int (*bdrv_reopen_prepare
)(BDRVReopenState
*reopen_state
,
123 BlockReopenQueue
*queue
, Error
**errp
);
124 void (*bdrv_reopen_commit
)(BDRVReopenState
*reopen_state
);
125 void (*bdrv_reopen_abort
)(BDRVReopenState
*reopen_state
);
126 void (*bdrv_join_options
)(QDict
*options
, QDict
*old_options
);
128 int (*bdrv_open
)(BlockDriverState
*bs
, QDict
*options
, int flags
,
130 int (*bdrv_file_open
)(BlockDriverState
*bs
, QDict
*options
, int flags
,
132 void (*bdrv_close
)(BlockDriverState
*bs
);
133 int (*bdrv_create
)(const char *filename
, QemuOpts
*opts
, Error
**errp
);
134 int (*bdrv_make_empty
)(BlockDriverState
*bs
);
136 void (*bdrv_refresh_filename
)(BlockDriverState
*bs
, QDict
*options
);
139 BlockAIOCB
*(*bdrv_aio_readv
)(BlockDriverState
*bs
,
140 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
141 BlockCompletionFunc
*cb
, void *opaque
);
142 BlockAIOCB
*(*bdrv_aio_writev
)(BlockDriverState
*bs
,
143 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
144 BlockCompletionFunc
*cb
, void *opaque
);
145 BlockAIOCB
*(*bdrv_aio_flush
)(BlockDriverState
*bs
,
146 BlockCompletionFunc
*cb
, void *opaque
);
147 BlockAIOCB
*(*bdrv_aio_pdiscard
)(BlockDriverState
*bs
,
148 int64_t offset
, int bytes
,
149 BlockCompletionFunc
*cb
, void *opaque
);
151 int coroutine_fn (*bdrv_co_readv
)(BlockDriverState
*bs
,
152 int64_t sector_num
, int nb_sectors
, QEMUIOVector
*qiov
);
155 * @offset: position in bytes to read at
156 * @bytes: number of bytes to read
157 * @qiov: the buffers to fill with read data
158 * @flags: currently unused, always 0
160 * @offset and @bytes will be a multiple of 'request_alignment',
161 * but the length of individual @qiov elements does not have to
164 * @bytes will always equal the total size of @qiov, and will be
165 * no larger than 'max_transfer'.
167 * The buffer in @qiov may point directly to guest memory.
169 int coroutine_fn (*bdrv_co_preadv
)(BlockDriverState
*bs
,
170 uint64_t offset
, uint64_t bytes
, QEMUIOVector
*qiov
, int flags
);
171 int coroutine_fn (*bdrv_co_writev
)(BlockDriverState
*bs
,
172 int64_t sector_num
, int nb_sectors
, QEMUIOVector
*qiov
);
173 int coroutine_fn (*bdrv_co_writev_flags
)(BlockDriverState
*bs
,
174 int64_t sector_num
, int nb_sectors
, QEMUIOVector
*qiov
, int flags
);
176 * @offset: position in bytes to write at
177 * @bytes: number of bytes to write
178 * @qiov: the buffers containing data to write
179 * @flags: zero or more bits allowed by 'supported_write_flags'
181 * @offset and @bytes will be a multiple of 'request_alignment',
182 * but the length of individual @qiov elements does not have to
185 * @bytes will always equal the total size of @qiov, and will be
186 * no larger than 'max_transfer'.
188 * The buffer in @qiov may point directly to guest memory.
190 int coroutine_fn (*bdrv_co_pwritev
)(BlockDriverState
*bs
,
191 uint64_t offset
, uint64_t bytes
, QEMUIOVector
*qiov
, int flags
);
194 * Efficiently zero a region of the disk image. Typically an image format
195 * would use a compact metadata representation to implement this. This
196 * function pointer may be NULL or return -ENOSUP and .bdrv_co_writev()
197 * will be called instead.
199 int coroutine_fn (*bdrv_co_pwrite_zeroes
)(BlockDriverState
*bs
,
200 int64_t offset
, int bytes
, BdrvRequestFlags flags
);
201 int coroutine_fn (*bdrv_co_pdiscard
)(BlockDriverState
*bs
,
202 int64_t offset
, int bytes
);
205 * Building block for bdrv_block_status[_above] and
206 * bdrv_is_allocated[_above]. The driver should answer only
207 * according to the current layer, and should not set
208 * BDRV_BLOCK_ALLOCATED, but may set BDRV_BLOCK_RAW. See block.h
209 * for the meaning of _DATA, _ZERO, and _OFFSET_VALID. The block
210 * layer guarantees input aligned to request_alignment, as well as
211 * non-NULL pnum and file.
213 int64_t coroutine_fn (*bdrv_co_get_block_status
)(BlockDriverState
*bs
,
214 int64_t sector_num
, int nb_sectors
, int *pnum
,
215 BlockDriverState
**file
);
218 * Invalidate any cached meta-data.
220 void (*bdrv_invalidate_cache
)(BlockDriverState
*bs
, Error
**errp
);
221 int (*bdrv_inactivate
)(BlockDriverState
*bs
);
224 * Flushes all data for all layers by calling bdrv_co_flush for underlying
225 * layers, if needed. This function is needed for deterministic
226 * synchronization of the flush finishing callback.
228 int coroutine_fn (*bdrv_co_flush
)(BlockDriverState
*bs
);
231 * Flushes all data that was already written to the OS all the way down to
232 * the disk (for example file-posix.c calls fsync()).
234 int coroutine_fn (*bdrv_co_flush_to_disk
)(BlockDriverState
*bs
);
237 * Flushes all internal caches to the OS. The data may still sit in a
238 * writeback cache of the host OS, but it will survive a crash of the qemu
241 int coroutine_fn (*bdrv_co_flush_to_os
)(BlockDriverState
*bs
);
243 const char *protocol_name
;
244 int (*bdrv_truncate
)(BlockDriverState
*bs
, int64_t offset
,
245 PreallocMode prealloc
, Error
**errp
);
247 int64_t (*bdrv_getlength
)(BlockDriverState
*bs
);
248 bool has_variable_length
;
249 int64_t (*bdrv_get_allocated_file_size
)(BlockDriverState
*bs
);
250 BlockMeasureInfo
*(*bdrv_measure
)(QemuOpts
*opts
, BlockDriverState
*in_bs
,
253 int coroutine_fn (*bdrv_co_pwritev_compressed
)(BlockDriverState
*bs
,
254 uint64_t offset
, uint64_t bytes
, QEMUIOVector
*qiov
);
256 int (*bdrv_snapshot_create
)(BlockDriverState
*bs
,
257 QEMUSnapshotInfo
*sn_info
);
258 int (*bdrv_snapshot_goto
)(BlockDriverState
*bs
,
259 const char *snapshot_id
);
260 int (*bdrv_snapshot_delete
)(BlockDriverState
*bs
,
261 const char *snapshot_id
,
264 int (*bdrv_snapshot_list
)(BlockDriverState
*bs
,
265 QEMUSnapshotInfo
**psn_info
);
266 int (*bdrv_snapshot_load_tmp
)(BlockDriverState
*bs
,
267 const char *snapshot_id
,
270 int (*bdrv_get_info
)(BlockDriverState
*bs
, BlockDriverInfo
*bdi
);
271 ImageInfoSpecific
*(*bdrv_get_specific_info
)(BlockDriverState
*bs
);
272 int (*bdrv_update
)(BlockDriverState
*bs
, int argc
, char **argv
);
274 int coroutine_fn (*bdrv_save_vmstate
)(BlockDriverState
*bs
,
277 int coroutine_fn (*bdrv_load_vmstate
)(BlockDriverState
*bs
,
281 int (*bdrv_change_backing_file
)(BlockDriverState
*bs
,
282 const char *backing_file
, const char *backing_fmt
);
284 /* removable device specific */
285 bool (*bdrv_is_inserted
)(BlockDriverState
*bs
);
286 void (*bdrv_eject
)(BlockDriverState
*bs
, bool eject_flag
);
287 void (*bdrv_lock_medium
)(BlockDriverState
*bs
, bool locked
);
289 /* to control generic scsi devices */
290 BlockAIOCB
*(*bdrv_aio_ioctl
)(BlockDriverState
*bs
,
291 unsigned long int req
, void *buf
,
292 BlockCompletionFunc
*cb
, void *opaque
);
293 int coroutine_fn (*bdrv_co_ioctl
)(BlockDriverState
*bs
,
294 unsigned long int req
, void *buf
);
296 /* List of options for creating images, terminated by name == NULL */
297 QemuOptsList
*create_opts
;
300 * Returns 0 for completed check, -errno for internal errors.
301 * The check results are stored in result.
303 int (*bdrv_check
)(BlockDriverState
*bs
, BdrvCheckResult
*result
,
306 int (*bdrv_amend_options
)(BlockDriverState
*bs
, QemuOpts
*opts
,
307 BlockDriverAmendStatusCB
*status_cb
,
310 void (*bdrv_debug_event
)(BlockDriverState
*bs
, BlkdebugEvent event
);
312 /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
313 int (*bdrv_debug_breakpoint
)(BlockDriverState
*bs
, const char *event
,
315 int (*bdrv_debug_remove_breakpoint
)(BlockDriverState
*bs
,
317 int (*bdrv_debug_resume
)(BlockDriverState
*bs
, const char *tag
);
318 bool (*bdrv_debug_is_suspended
)(BlockDriverState
*bs
, const char *tag
);
320 void (*bdrv_refresh_limits
)(BlockDriverState
*bs
, Error
**errp
);
323 * Returns 1 if newly created images are guaranteed to contain only
324 * zeros, 0 otherwise.
326 int (*bdrv_has_zero_init
)(BlockDriverState
*bs
);
328 /* Remove fd handlers, timers, and other event loop callbacks so the event
329 * loop is no longer in use. Called with no in-flight requests and in
330 * depth-first traversal order with parents before child nodes.
332 void (*bdrv_detach_aio_context
)(BlockDriverState
*bs
);
334 /* Add fd handlers, timers, and other event loop callbacks so I/O requests
335 * can be processed again. Called with no in-flight requests and in
336 * depth-first traversal order with child nodes before parent nodes.
338 void (*bdrv_attach_aio_context
)(BlockDriverState
*bs
,
339 AioContext
*new_context
);
341 /* io queue for linux-aio */
342 void (*bdrv_io_plug
)(BlockDriverState
*bs
);
343 void (*bdrv_io_unplug
)(BlockDriverState
*bs
);
346 * Try to get @bs's logical and physical block size.
347 * On success, store them in @bsz and return zero.
348 * On failure, return negative errno.
350 int (*bdrv_probe_blocksizes
)(BlockDriverState
*bs
, BlockSizes
*bsz
);
352 * Try to get @bs's geometry (cyls, heads, sectors)
353 * On success, store them in @geo and return 0.
354 * On failure return -errno.
355 * Only drivers that want to override guest geometry implement this
356 * callback; see hd_geometry_guess().
358 int (*bdrv_probe_geometry
)(BlockDriverState
*bs
, HDGeometry
*geo
);
361 * bdrv_co_drain_begin is called if implemented in the beginning of a
362 * drain operation to drain and stop any internal sources of requests in
364 * bdrv_co_drain_end is called if implemented at the end of the drain.
366 * They should be used by the driver to e.g. manage scheduled I/O
367 * requests, or toggle an internal state. After the end of the drain new
368 * requests will continue normally.
370 void coroutine_fn (*bdrv_co_drain_begin
)(BlockDriverState
*bs
);
371 void coroutine_fn (*bdrv_co_drain_end
)(BlockDriverState
*bs
);
373 void (*bdrv_add_child
)(BlockDriverState
*parent
, BlockDriverState
*child
,
375 void (*bdrv_del_child
)(BlockDriverState
*parent
, BdrvChild
*child
,
379 * Informs the block driver that a permission change is intended. The
380 * driver checks whether the change is permissible and may take other
381 * preparations for the change (e.g. get file system locks). This operation
382 * is always followed either by a call to either .bdrv_set_perm or
383 * .bdrv_abort_perm_update.
385 * Checks whether the requested set of cumulative permissions in @perm
386 * can be granted for accessing @bs and whether no other users are using
387 * permissions other than those given in @shared (both arguments take
388 * BLK_PERM_* bitmasks).
390 * If both conditions are met, 0 is returned. Otherwise, -errno is returned
391 * and errp is set to an error describing the conflict.
393 int (*bdrv_check_perm
)(BlockDriverState
*bs
, uint64_t perm
,
394 uint64_t shared
, Error
**errp
);
397 * Called to inform the driver that the set of cumulative set of used
398 * permissions for @bs has changed to @perm, and the set of sharable
399 * permission to @shared. The driver can use this to propagate changes to
400 * its children (i.e. request permissions only if a parent actually needs
403 * This function is only invoked after bdrv_check_perm(), so block drivers
404 * may rely on preparations made in their .bdrv_check_perm implementation.
406 void (*bdrv_set_perm
)(BlockDriverState
*bs
, uint64_t perm
, uint64_t shared
);
409 * Called to inform the driver that after a previous bdrv_check_perm()
410 * call, the permission update is not performed and any preparations made
411 * for it (e.g. taken file locks) need to be undone.
413 * This function can be called even for nodes that never saw a
414 * bdrv_check_perm() call. It is a no-op then.
416 void (*bdrv_abort_perm_update
)(BlockDriverState
*bs
);
419 * Returns in @nperm and @nshared the permissions that the driver for @bs
420 * needs on its child @c, based on the cumulative permissions requested by
421 * the parents in @parent_perm and @parent_shared.
423 * If @c is NULL, return the permissions for attaching a new child for the
426 * If @reopen_queue is non-NULL, don't return the currently needed
427 * permissions, but those that will be needed after applying the
430 void (*bdrv_child_perm
)(BlockDriverState
*bs
, BdrvChild
*c
,
431 const BdrvChildRole
*role
,
432 BlockReopenQueue
*reopen_queue
,
433 uint64_t parent_perm
, uint64_t parent_shared
,
434 uint64_t *nperm
, uint64_t *nshared
);
437 * Bitmaps should be marked as 'IN_USE' in the image on reopening image
438 * as rw. This handler should realize it. It also should unset readonly
439 * field of BlockDirtyBitmap's in case of success.
441 int (*bdrv_reopen_bitmaps_rw
)(BlockDriverState
*bs
, Error
**errp
);
442 bool (*bdrv_can_store_new_dirty_bitmap
)(BlockDriverState
*bs
,
444 uint32_t granularity
,
446 void (*bdrv_remove_persistent_dirty_bitmap
)(BlockDriverState
*bs
,
450 QLIST_ENTRY(BlockDriver
) list
;
453 typedef struct BlockLimits
{
454 /* Alignment requirement, in bytes, for offset/length of I/O
455 * requests. Must be a power of 2 less than INT_MAX; defaults to
456 * 1 for drivers with modern byte interfaces, and to 512
458 uint32_t request_alignment
;
460 /* Maximum number of bytes that can be discarded at once (since it
461 * is signed, it must be < 2G, if set). Must be multiple of
462 * pdiscard_alignment, but need not be power of 2. May be 0 if no
463 * inherent 32-bit limit */
464 int32_t max_pdiscard
;
466 /* Optimal alignment for discard requests in bytes. A power of 2
467 * is best but not mandatory. Must be a multiple of
468 * bl.request_alignment, and must be less than max_pdiscard if
469 * that is set. May be 0 if bl.request_alignment is good enough */
470 uint32_t pdiscard_alignment
;
472 /* Maximum number of bytes that can zeroized at once (since it is
473 * signed, it must be < 2G, if set). Must be multiple of
474 * pwrite_zeroes_alignment. May be 0 if no inherent 32-bit limit */
475 int32_t max_pwrite_zeroes
;
477 /* Optimal alignment for write zeroes requests in bytes. A power
478 * of 2 is best but not mandatory. Must be a multiple of
479 * bl.request_alignment, and must be less than max_pwrite_zeroes
480 * if that is set. May be 0 if bl.request_alignment is good
482 uint32_t pwrite_zeroes_alignment
;
484 /* Optimal transfer length in bytes. A power of 2 is best but not
485 * mandatory. Must be a multiple of bl.request_alignment, or 0 if
486 * no preferred size */
487 uint32_t opt_transfer
;
489 /* Maximal transfer length in bytes. Need not be power of 2, but
490 * must be multiple of opt_transfer and bl.request_alignment, or 0
491 * for no 32-bit limit. For now, anything larger than INT_MAX is
493 uint32_t max_transfer
;
495 /* memory alignment, in bytes so that no bounce buffer is needed */
496 size_t min_mem_alignment
;
498 /* memory alignment, in bytes, for bounce buffer */
499 size_t opt_mem_alignment
;
501 /* maximum number of iovec elements */
505 typedef struct BdrvOpBlocker BdrvOpBlocker
;
507 typedef struct BdrvAioNotifier
{
508 void (*attached_aio_context
)(AioContext
*new_context
, void *opaque
);
509 void (*detach_aio_context
)(void *opaque
);
514 QLIST_ENTRY(BdrvAioNotifier
) list
;
517 struct BdrvChildRole
{
518 /* If true, bdrv_replace_node() doesn't change the node this BdrvChild
522 void (*inherit_options
)(int *child_flags
, QDict
*child_options
,
523 int parent_flags
, QDict
*parent_options
);
525 void (*change_media
)(BdrvChild
*child
, bool load
);
526 void (*resize
)(BdrvChild
*child
);
528 /* Returns a name that is supposedly more useful for human users than the
529 * node name for identifying the node in question (in particular, a BB
530 * name), or NULL if the parent can't provide a better name. */
531 const char *(*get_name
)(BdrvChild
*child
);
533 /* Returns a malloced string that describes the parent of the child for a
534 * human reader. This could be a node-name, BlockBackend name, qdev ID or
535 * QOM path of the device owning the BlockBackend, job type and ID etc. The
536 * caller is responsible for freeing the memory. */
537 char *(*get_parent_desc
)(BdrvChild
*child
);
540 * If this pair of functions is implemented, the parent doesn't issue new
541 * requests after returning from .drained_begin() until .drained_end() is
544 * Note that this can be nested. If drained_begin() was called twice, new
545 * I/O is allowed only after drained_end() was called twice, too.
547 void (*drained_begin
)(BdrvChild
*child
);
548 void (*drained_end
)(BdrvChild
*child
);
550 /* Notifies the parent that the child has been activated/inactivated (e.g.
551 * when migration is completing) and it can start/stop requesting
552 * permissions and doing I/O on it. */
553 void (*activate
)(BdrvChild
*child
, Error
**errp
);
554 int (*inactivate
)(BdrvChild
*child
);
556 void (*attach
)(BdrvChild
*child
);
557 void (*detach
)(BdrvChild
*child
);
559 /* Notifies the parent that the filename of its child has changed (e.g.
560 * because the direct child was removed from the backing chain), so that it
561 * can update its reference. */
562 int (*update_filename
)(BdrvChild
*child
, BlockDriverState
*new_base
,
563 const char *filename
, Error
**errp
);
566 extern const BdrvChildRole child_file
;
567 extern const BdrvChildRole child_format
;
568 extern const BdrvChildRole child_backing
;
571 BlockDriverState
*bs
;
573 const BdrvChildRole
*role
;
577 * Granted permissions for operating on this BdrvChild (BLK_PERM_* bitmask)
582 * Permissions that can still be granted to other users of @bs while this
583 * BdrvChild is still attached to it. (BLK_PERM_* bitmask)
585 uint64_t shared_perm
;
587 QLIST_ENTRY(BdrvChild
) next
;
588 QLIST_ENTRY(BdrvChild
) next_parent
;
592 * Note: the function bdrv_append() copies and swaps contents of
593 * BlockDriverStates, so if you add new fields to this struct, please
594 * inspect bdrv_append() to determine if the new fields need to be
597 struct BlockDriverState
{
598 /* Protected by big QEMU lock or read-only after opening. No special
599 * locking needed during I/O...
601 int open_flags
; /* flags used to open the file, re-used for re-open */
602 bool read_only
; /* if true, the media is read only */
603 bool encrypted
; /* if true, the media is encrypted */
604 bool sg
; /* if true, the device is a /dev/sg* */
605 bool probed
; /* if true, format was probed rather than specified */
606 bool force_share
; /* if true, always allow all shared permissions */
607 bool implicit
; /* if true, this filter node was automatically inserted */
609 BlockDriver
*drv
; /* NULL means no media */
612 AioContext
*aio_context
; /* event loop used for fd handlers, timers, etc */
613 /* long-running tasks intended to always use the same AioContext as this
614 * BDS may register themselves in this list to be notified of changes
615 * regarding this BDS's context */
616 QLIST_HEAD(, BdrvAioNotifier
) aio_notifiers
;
617 bool walking_aio_notifiers
; /* to make removal during iteration safe */
619 char filename
[PATH_MAX
];
620 char backing_file
[PATH_MAX
]; /* if non zero, the image is a diff of
622 char backing_format
[16]; /* if non-zero and backing_file exists */
624 QDict
*full_open_options
;
625 char exact_filename
[PATH_MAX
];
633 /* Flags honored during pwrite (so far: BDRV_REQ_FUA) */
634 unsigned int supported_write_flags
;
635 /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA,
636 * BDRV_REQ_MAY_UNMAP) */
637 unsigned int supported_zero_flags
;
639 /* the following member gives a name to every node on the bs graph. */
641 /* element of the list of named nodes building the graph */
642 QTAILQ_ENTRY(BlockDriverState
) node_list
;
643 /* element of the list of all BlockDriverStates (all_bdrv_states) */
644 QTAILQ_ENTRY(BlockDriverState
) bs_list
;
645 /* element of the list of monitor-owned BDS */
646 QTAILQ_ENTRY(BlockDriverState
) monitor_list
;
649 /* operation blockers */
650 QLIST_HEAD(, BdrvOpBlocker
) op_blockers
[BLOCK_OP_TYPE_MAX
];
652 /* long-running background operation */
655 /* The node that this node inherited default options from (and a reopen on
656 * which can affect this node by changing these defaults). This is always a
657 * parent node of this node. */
658 BlockDriverState
*inherits_from
;
659 QLIST_HEAD(, BdrvChild
) children
;
660 QLIST_HEAD(, BdrvChild
) parents
;
663 QDict
*explicit_options
;
664 BlockdevDetectZeroesOptions detect_zeroes
;
666 /* The error object in use for blocking operations on backing_hd */
667 Error
*backing_blocker
;
669 /* Protected by AioContext lock */
671 /* If we are reading a disk image, give its size in sectors.
672 * Generally read-only; it is written to by load_snapshot and
673 * save_snaphost, but the block layer is quiescent during those.
675 int64_t total_sectors
;
677 /* Callback before write request is processed */
678 NotifierWithReturnList before_write_notifiers
;
680 /* threshold limit for writes, in bytes. "High water mark". */
681 uint64_t write_threshold_offset
;
682 NotifierWithReturn write_threshold_notifier
;
684 /* Writing to the list requires the BQL _and_ the dirty_bitmap_mutex.
685 * Reading from the list can be done with either the BQL or the
686 * dirty_bitmap_mutex. Modifying a bitmap only requires
687 * dirty_bitmap_mutex. */
688 QemuMutex dirty_bitmap_mutex
;
689 QLIST_HEAD(, BdrvDirtyBitmap
) dirty_bitmaps
;
691 /* Offset after the highest byte written to */
692 Stat64 wr_highest_offset
;
694 /* If true, copy read backing sectors into image. Can be >1 if more
695 * than one client has requested copy-on-read. Accessed with atomic
700 /* number of in-flight requests; overall and serialising.
701 * Accessed with atomic ops.
703 unsigned int in_flight
;
704 unsigned int serialising_in_flight
;
706 /* Internal to BDRV_POLL_WHILE and bdrv_wakeup. Accessed with atomic
711 /* counter for nested bdrv_io_plug.
712 * Accessed with atomic ops.
716 /* do we need to tell the quest if we have a volatile write cache? */
717 int enable_write_cache
;
719 /* Accessed with atomic ops. */
721 unsigned int write_gen
; /* Current data generation */
723 /* Protected by reqs_lock. */
725 QLIST_HEAD(, BdrvTrackedRequest
) tracked_requests
;
726 CoQueue flush_queue
; /* Serializing flush queue */
727 bool active_flush_req
; /* Flush request in flight? */
729 /* Only read/written by whoever has set active_flush_req to true. */
730 unsigned int flushed_gen
; /* Flushed write generation */
733 struct BlockBackendRootState
{
736 BlockdevDetectZeroesOptions detect_zeroes
;
739 typedef enum BlockMirrorBackingMode
{
740 /* Reuse the existing backing chain from the source for the target.
741 * - sync=full: Set backing BDS to NULL.
742 * - sync=top: Use source's backing BDS.
743 * - sync=none: Use source as the backing BDS. */
744 MIRROR_SOURCE_BACKING_CHAIN
,
746 /* Open the target's backing chain completely anew */
747 MIRROR_OPEN_BACKING_CHAIN
,
749 /* Do not change the target's backing BDS after job completion */
750 MIRROR_LEAVE_BACKING_CHAIN
,
751 } BlockMirrorBackingMode
;
753 static inline BlockDriverState
*backing_bs(BlockDriverState
*bs
)
755 return bs
->backing
? bs
->backing
->bs
: NULL
;
759 /* Essential block drivers which must always be statically linked into qemu, and
760 * which therefore can be accessed without using bdrv_find_format() */
761 extern BlockDriver bdrv_file
;
762 extern BlockDriver bdrv_raw
;
763 extern BlockDriver bdrv_qcow2
;
765 int coroutine_fn
bdrv_co_preadv(BdrvChild
*child
,
766 int64_t offset
, unsigned int bytes
, QEMUIOVector
*qiov
,
767 BdrvRequestFlags flags
);
768 int coroutine_fn
bdrv_co_pwritev(BdrvChild
*child
,
769 int64_t offset
, unsigned int bytes
, QEMUIOVector
*qiov
,
770 BdrvRequestFlags flags
);
772 int get_tmp_filename(char *filename
, int size
);
773 BlockDriver
*bdrv_probe_all(const uint8_t *buf
, int buf_size
,
774 const char *filename
);
776 void bdrv_parse_filename_strip_prefix(const char *filename
, const char *prefix
,
781 * bdrv_add_before_write_notifier:
783 * Register a callback that is invoked before write requests are processed but
784 * after any throttling or waiting for overlapping requests.
786 void bdrv_add_before_write_notifier(BlockDriverState
*bs
,
787 NotifierWithReturn
*notifier
);
790 * bdrv_detach_aio_context:
792 * May be called from .bdrv_detach_aio_context() to detach children from the
793 * current #AioContext. This is only needed by block drivers that manage their
794 * own children. Both ->file and ->backing are automatically handled and
795 * block drivers should not call this function on them explicitly.
797 void bdrv_detach_aio_context(BlockDriverState
*bs
);
800 * bdrv_attach_aio_context:
802 * May be called from .bdrv_attach_aio_context() to attach children to the new
803 * #AioContext. This is only needed by block drivers that manage their own
804 * children. Both ->file and ->backing are automatically handled and block
805 * drivers should not call this function on them explicitly.
807 void bdrv_attach_aio_context(BlockDriverState
*bs
,
808 AioContext
*new_context
);
811 * bdrv_add_aio_context_notifier:
813 * If a long-running job intends to be always run in the same AioContext as a
814 * certain BDS, it may use this function to be notified of changes regarding the
815 * association of the BDS to an AioContext.
817 * attached_aio_context() is called after the target BDS has been attached to a
818 * new AioContext; detach_aio_context() is called before the target BDS is being
819 * detached from its old AioContext.
821 void bdrv_add_aio_context_notifier(BlockDriverState
*bs
,
822 void (*attached_aio_context
)(AioContext
*new_context
, void *opaque
),
823 void (*detach_aio_context
)(void *opaque
), void *opaque
);
826 * bdrv_remove_aio_context_notifier:
828 * Unsubscribe of change notifications regarding the BDS's AioContext. The
829 * parameters given here have to be the same as those given to
830 * bdrv_add_aio_context_notifier().
832 void bdrv_remove_aio_context_notifier(BlockDriverState
*bs
,
833 void (*aio_context_attached
)(AioContext
*,
835 void (*aio_context_detached
)(void *),
840 * @bs: The BlockDriverState for which an I/O operation has been completed.
842 * Wake up the main thread if it is waiting on BDRV_POLL_WHILE. During
843 * synchronous I/O on a BlockDriverState that is attached to another
844 * I/O thread, the main thread lets the I/O thread's event loop run,
845 * waiting for the I/O operation to complete. A bdrv_wakeup will wake
846 * up the main thread if necessary.
848 * Manual calls to bdrv_wakeup are rarely necessary, because
849 * bdrv_dec_in_flight already calls it.
851 void bdrv_wakeup(BlockDriverState
*bs
);
854 int is_windows_drive(const char *filename
);
859 * @job_id: The id of the newly-created job, or %NULL to use the
860 * device name of @bs.
861 * @bs: Block device to operate on.
862 * @base: Block device that will become the new base, or %NULL to
863 * flatten the whole backing file chain onto @bs.
864 * @backing_file_str: The file name that will be written to @bs as the
865 * the new backing file if the job completes. Ignored if @base is %NULL.
866 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
867 * @on_error: The action to take upon error.
868 * @errp: Error object.
870 * Start a streaming operation on @bs. Clusters that are unallocated
871 * in @bs, but allocated in any image between @base and @bs (both
872 * exclusive) will be written to @bs. At the end of a successful
873 * streaming job, the backing file of @bs will be changed to
874 * @backing_file_str in the written image and to @base in the live
877 void stream_start(const char *job_id
, BlockDriverState
*bs
,
878 BlockDriverState
*base
, const char *backing_file_str
,
879 int64_t speed
, BlockdevOnError on_error
, Error
**errp
);
883 * @job_id: The id of the newly-created job, or %NULL to use the
884 * device name of @bs.
885 * @bs: Active block device.
886 * @top: Top block device to be committed.
887 * @base: Block device that will be written into, and become the new top.
888 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
889 * @on_error: The action to take upon error.
890 * @backing_file_str: String to use as the backing file in @top's overlay
891 * @filter_node_name: The node name that should be assigned to the filter
892 * driver that the commit job inserts into the graph above @top. NULL means
893 * that a node name should be autogenerated.
894 * @errp: Error object.
897 void commit_start(const char *job_id
, BlockDriverState
*bs
,
898 BlockDriverState
*base
, BlockDriverState
*top
, int64_t speed
,
899 BlockdevOnError on_error
, const char *backing_file_str
,
900 const char *filter_node_name
, Error
**errp
);
902 * commit_active_start:
903 * @job_id: The id of the newly-created job, or %NULL to use the
904 * device name of @bs.
905 * @bs: Active block device to be committed.
906 * @base: Block device that will be written into, and become the new top.
907 * @creation_flags: Flags that control the behavior of the Job lifetime.
908 * See @BlockJobCreateFlags
909 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
910 * @on_error: The action to take upon error.
911 * @filter_node_name: The node name that should be assigned to the filter
912 * driver that the commit job inserts into the graph above @bs. NULL means that
913 * a node name should be autogenerated.
914 * @cb: Completion function for the job.
915 * @opaque: Opaque pointer value passed to @cb.
916 * @auto_complete: Auto complete the job.
917 * @errp: Error object.
920 void commit_active_start(const char *job_id
, BlockDriverState
*bs
,
921 BlockDriverState
*base
, int creation_flags
,
922 int64_t speed
, BlockdevOnError on_error
,
923 const char *filter_node_name
,
924 BlockCompletionFunc
*cb
, void *opaque
,
925 bool auto_complete
, Error
**errp
);
928 * @job_id: The id of the newly-created job, or %NULL to use the
929 * device name of @bs.
930 * @bs: Block device to operate on.
931 * @target: Block device to write to.
932 * @replaces: Block graph node name to replace once the mirror is done. Can
933 * only be used when full mirroring is selected.
934 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
935 * @granularity: The chosen granularity for the dirty bitmap.
936 * @buf_size: The amount of data that can be in flight at one time.
937 * @mode: Whether to collapse all images in the chain to the target.
938 * @backing_mode: How to establish the target's backing chain after completion.
939 * @on_source_error: The action to take upon error reading from the source.
940 * @on_target_error: The action to take upon error writing to the target.
941 * @unmap: Whether to unmap target where source sectors only contain zeroes.
942 * @filter_node_name: The node name that should be assigned to the filter
943 * driver that the mirror job inserts into the graph above @bs. NULL means that
944 * a node name should be autogenerated.
945 * @errp: Error object.
947 * Start a mirroring operation on @bs. Clusters that are allocated
948 * in @bs will be written to @target until the job is cancelled or
949 * manually completed. At the end of a successful mirroring job,
950 * @bs will be switched to read from @target.
952 void mirror_start(const char *job_id
, BlockDriverState
*bs
,
953 BlockDriverState
*target
, const char *replaces
,
954 int64_t speed
, uint32_t granularity
, int64_t buf_size
,
955 MirrorSyncMode mode
, BlockMirrorBackingMode backing_mode
,
956 BlockdevOnError on_source_error
,
957 BlockdevOnError on_target_error
,
958 bool unmap
, const char *filter_node_name
, Error
**errp
);
962 * @job_id: The id of the newly-created job, or %NULL to use the
963 * device name of @bs.
964 * @bs: Block device to operate on.
965 * @target: Block device to write to.
966 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
967 * @sync_mode: What parts of the disk image should be copied to the destination.
968 * @sync_bitmap: The dirty bitmap if sync_mode is MIRROR_SYNC_MODE_INCREMENTAL.
969 * @on_source_error: The action to take upon error reading from the source.
970 * @on_target_error: The action to take upon error writing to the target.
971 * @creation_flags: Flags that control the behavior of the Job lifetime.
972 * See @BlockJobCreateFlags
973 * @cb: Completion function for the job.
974 * @opaque: Opaque pointer value passed to @cb.
975 * @txn: Transaction that this job is part of (may be NULL).
977 * Create a backup operation on @bs. Clusters in @bs are written to @target
978 * until the job is cancelled or manually completed.
980 BlockJob
*backup_job_create(const char *job_id
, BlockDriverState
*bs
,
981 BlockDriverState
*target
, int64_t speed
,
982 MirrorSyncMode sync_mode
,
983 BdrvDirtyBitmap
*sync_bitmap
,
985 BlockdevOnError on_source_error
,
986 BlockdevOnError on_target_error
,
988 BlockCompletionFunc
*cb
, void *opaque
,
989 BlockJobTxn
*txn
, Error
**errp
);
991 void hmp_drive_add_node(Monitor
*mon
, const char *optstr
);
993 BdrvChild
*bdrv_root_attach_child(BlockDriverState
*child_bs
,
994 const char *child_name
,
995 const BdrvChildRole
*child_role
,
996 uint64_t perm
, uint64_t shared_perm
,
997 void *opaque
, Error
**errp
);
998 void bdrv_root_unref_child(BdrvChild
*child
);
1000 int bdrv_child_try_set_perm(BdrvChild
*c
, uint64_t perm
, uint64_t shared
,
1003 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1004 * block filters: Forward CONSISTENT_READ, WRITE, WRITE_UNCHANGED and RESIZE to
1006 void bdrv_filter_default_perms(BlockDriverState
*bs
, BdrvChild
*c
,
1007 const BdrvChildRole
*role
,
1008 BlockReopenQueue
*reopen_queue
,
1009 uint64_t perm
, uint64_t shared
,
1010 uint64_t *nperm
, uint64_t *nshared
);
1012 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1013 * (non-raw) image formats: Like above for bs->backing, but for bs->file it
1014 * requires WRITE | RESIZE for read-write images, always requires
1015 * CONSISTENT_READ and doesn't share WRITE. */
1016 void bdrv_format_default_perms(BlockDriverState
*bs
, BdrvChild
*c
,
1017 const BdrvChildRole
*role
,
1018 BlockReopenQueue
*reopen_queue
,
1019 uint64_t perm
, uint64_t shared
,
1020 uint64_t *nperm
, uint64_t *nshared
);
1023 * Default implementation for drivers to pass bdrv_co_get_block_status() to
1026 int64_t coroutine_fn
bdrv_co_get_block_status_from_file(BlockDriverState
*bs
,
1030 BlockDriverState
**file
);
1032 * Default implementation for drivers to pass bdrv_co_get_block_status() to
1033 * their backing file.
1035 int64_t coroutine_fn
bdrv_co_get_block_status_from_backing(BlockDriverState
*bs
,
1039 BlockDriverState
**file
);
1040 const char *bdrv_get_parent_name(const BlockDriverState
*bs
);
1041 void blk_dev_change_media_cb(BlockBackend
*blk
, bool load
, Error
**errp
);
1042 bool blk_dev_has_removable_media(BlockBackend
*blk
);
1043 bool blk_dev_has_tray(BlockBackend
*blk
);
1044 void blk_dev_eject_request(BlockBackend
*blk
, bool force
);
1045 bool blk_dev_is_tray_open(BlockBackend
*blk
);
1046 bool blk_dev_is_medium_locked(BlockBackend
*blk
);
1048 void bdrv_set_dirty(BlockDriverState
*bs
, int64_t offset
, int64_t bytes
);
1049 bool bdrv_requests_pending(BlockDriverState
*bs
);
1051 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap
*bitmap
, HBitmap
**out
);
1052 void bdrv_undo_clear_dirty_bitmap(BdrvDirtyBitmap
*bitmap
, HBitmap
*in
);
1054 void bdrv_inc_in_flight(BlockDriverState
*bs
);
1055 void bdrv_dec_in_flight(BlockDriverState
*bs
);
1057 void blockdev_close_all_bdrv_states(void);
1059 #endif /* BLOCK_INT_H */