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
);
273 int coroutine_fn (*bdrv_save_vmstate
)(BlockDriverState
*bs
,
276 int coroutine_fn (*bdrv_load_vmstate
)(BlockDriverState
*bs
,
280 int (*bdrv_change_backing_file
)(BlockDriverState
*bs
,
281 const char *backing_file
, const char *backing_fmt
);
283 /* removable device specific */
284 bool (*bdrv_is_inserted
)(BlockDriverState
*bs
);
285 void (*bdrv_eject
)(BlockDriverState
*bs
, bool eject_flag
);
286 void (*bdrv_lock_medium
)(BlockDriverState
*bs
, bool locked
);
288 /* to control generic scsi devices */
289 BlockAIOCB
*(*bdrv_aio_ioctl
)(BlockDriverState
*bs
,
290 unsigned long int req
, void *buf
,
291 BlockCompletionFunc
*cb
, void *opaque
);
292 int coroutine_fn (*bdrv_co_ioctl
)(BlockDriverState
*bs
,
293 unsigned long int req
, void *buf
);
295 /* List of options for creating images, terminated by name == NULL */
296 QemuOptsList
*create_opts
;
299 * Returns 0 for completed check, -errno for internal errors.
300 * The check results are stored in result.
302 int (*bdrv_check
)(BlockDriverState
*bs
, BdrvCheckResult
*result
,
305 int (*bdrv_amend_options
)(BlockDriverState
*bs
, QemuOpts
*opts
,
306 BlockDriverAmendStatusCB
*status_cb
,
309 void (*bdrv_debug_event
)(BlockDriverState
*bs
, BlkdebugEvent event
);
311 /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
312 int (*bdrv_debug_breakpoint
)(BlockDriverState
*bs
, const char *event
,
314 int (*bdrv_debug_remove_breakpoint
)(BlockDriverState
*bs
,
316 int (*bdrv_debug_resume
)(BlockDriverState
*bs
, const char *tag
);
317 bool (*bdrv_debug_is_suspended
)(BlockDriverState
*bs
, const char *tag
);
319 void (*bdrv_refresh_limits
)(BlockDriverState
*bs
, Error
**errp
);
322 * Returns 1 if newly created images are guaranteed to contain only
323 * zeros, 0 otherwise.
325 int (*bdrv_has_zero_init
)(BlockDriverState
*bs
);
327 /* Remove fd handlers, timers, and other event loop callbacks so the event
328 * loop is no longer in use. Called with no in-flight requests and in
329 * depth-first traversal order with parents before child nodes.
331 void (*bdrv_detach_aio_context
)(BlockDriverState
*bs
);
333 /* Add fd handlers, timers, and other event loop callbacks so I/O requests
334 * can be processed again. Called with no in-flight requests and in
335 * depth-first traversal order with child nodes before parent nodes.
337 void (*bdrv_attach_aio_context
)(BlockDriverState
*bs
,
338 AioContext
*new_context
);
340 /* io queue for linux-aio */
341 void (*bdrv_io_plug
)(BlockDriverState
*bs
);
342 void (*bdrv_io_unplug
)(BlockDriverState
*bs
);
345 * Try to get @bs's logical and physical block size.
346 * On success, store them in @bsz and return zero.
347 * On failure, return negative errno.
349 int (*bdrv_probe_blocksizes
)(BlockDriverState
*bs
, BlockSizes
*bsz
);
351 * Try to get @bs's geometry (cyls, heads, sectors)
352 * On success, store them in @geo and return 0.
353 * On failure return -errno.
354 * Only drivers that want to override guest geometry implement this
355 * callback; see hd_geometry_guess().
357 int (*bdrv_probe_geometry
)(BlockDriverState
*bs
, HDGeometry
*geo
);
360 * bdrv_co_drain_begin is called if implemented in the beginning of a
361 * drain operation to drain and stop any internal sources of requests in
363 * bdrv_co_drain_end is called if implemented at the end of the drain.
365 * They should be used by the driver to e.g. manage scheduled I/O
366 * requests, or toggle an internal state. After the end of the drain new
367 * requests will continue normally.
369 void coroutine_fn (*bdrv_co_drain_begin
)(BlockDriverState
*bs
);
370 void coroutine_fn (*bdrv_co_drain_end
)(BlockDriverState
*bs
);
372 void (*bdrv_add_child
)(BlockDriverState
*parent
, BlockDriverState
*child
,
374 void (*bdrv_del_child
)(BlockDriverState
*parent
, BdrvChild
*child
,
378 * Informs the block driver that a permission change is intended. The
379 * driver checks whether the change is permissible and may take other
380 * preparations for the change (e.g. get file system locks). This operation
381 * is always followed either by a call to either .bdrv_set_perm or
382 * .bdrv_abort_perm_update.
384 * Checks whether the requested set of cumulative permissions in @perm
385 * can be granted for accessing @bs and whether no other users are using
386 * permissions other than those given in @shared (both arguments take
387 * BLK_PERM_* bitmasks).
389 * If both conditions are met, 0 is returned. Otherwise, -errno is returned
390 * and errp is set to an error describing the conflict.
392 int (*bdrv_check_perm
)(BlockDriverState
*bs
, uint64_t perm
,
393 uint64_t shared
, Error
**errp
);
396 * Called to inform the driver that the set of cumulative set of used
397 * permissions for @bs has changed to @perm, and the set of sharable
398 * permission to @shared. The driver can use this to propagate changes to
399 * its children (i.e. request permissions only if a parent actually needs
402 * This function is only invoked after bdrv_check_perm(), so block drivers
403 * may rely on preparations made in their .bdrv_check_perm implementation.
405 void (*bdrv_set_perm
)(BlockDriverState
*bs
, uint64_t perm
, uint64_t shared
);
408 * Called to inform the driver that after a previous bdrv_check_perm()
409 * call, the permission update is not performed and any preparations made
410 * for it (e.g. taken file locks) need to be undone.
412 * This function can be called even for nodes that never saw a
413 * bdrv_check_perm() call. It is a no-op then.
415 void (*bdrv_abort_perm_update
)(BlockDriverState
*bs
);
418 * Returns in @nperm and @nshared the permissions that the driver for @bs
419 * needs on its child @c, based on the cumulative permissions requested by
420 * the parents in @parent_perm and @parent_shared.
422 * If @c is NULL, return the permissions for attaching a new child for the
425 * If @reopen_queue is non-NULL, don't return the currently needed
426 * permissions, but those that will be needed after applying the
429 void (*bdrv_child_perm
)(BlockDriverState
*bs
, BdrvChild
*c
,
430 const BdrvChildRole
*role
,
431 BlockReopenQueue
*reopen_queue
,
432 uint64_t parent_perm
, uint64_t parent_shared
,
433 uint64_t *nperm
, uint64_t *nshared
);
436 * Bitmaps should be marked as 'IN_USE' in the image on reopening image
437 * as rw. This handler should realize it. It also should unset readonly
438 * field of BlockDirtyBitmap's in case of success.
440 int (*bdrv_reopen_bitmaps_rw
)(BlockDriverState
*bs
, Error
**errp
);
441 bool (*bdrv_can_store_new_dirty_bitmap
)(BlockDriverState
*bs
,
443 uint32_t granularity
,
445 void (*bdrv_remove_persistent_dirty_bitmap
)(BlockDriverState
*bs
,
449 QLIST_ENTRY(BlockDriver
) list
;
452 typedef struct BlockLimits
{
453 /* Alignment requirement, in bytes, for offset/length of I/O
454 * requests. Must be a power of 2 less than INT_MAX; defaults to
455 * 1 for drivers with modern byte interfaces, and to 512
457 uint32_t request_alignment
;
459 /* Maximum number of bytes that can be discarded at once (since it
460 * is signed, it must be < 2G, if set). Must be multiple of
461 * pdiscard_alignment, but need not be power of 2. May be 0 if no
462 * inherent 32-bit limit */
463 int32_t max_pdiscard
;
465 /* Optimal alignment for discard requests in bytes. A power of 2
466 * is best but not mandatory. Must be a multiple of
467 * bl.request_alignment, and must be less than max_pdiscard if
468 * that is set. May be 0 if bl.request_alignment is good enough */
469 uint32_t pdiscard_alignment
;
471 /* Maximum number of bytes that can zeroized at once (since it is
472 * signed, it must be < 2G, if set). Must be multiple of
473 * pwrite_zeroes_alignment. May be 0 if no inherent 32-bit limit */
474 int32_t max_pwrite_zeroes
;
476 /* Optimal alignment for write zeroes requests in bytes. A power
477 * of 2 is best but not mandatory. Must be a multiple of
478 * bl.request_alignment, and must be less than max_pwrite_zeroes
479 * if that is set. May be 0 if bl.request_alignment is good
481 uint32_t pwrite_zeroes_alignment
;
483 /* Optimal transfer length in bytes. A power of 2 is best but not
484 * mandatory. Must be a multiple of bl.request_alignment, or 0 if
485 * no preferred size */
486 uint32_t opt_transfer
;
488 /* Maximal transfer length in bytes. Need not be power of 2, but
489 * must be multiple of opt_transfer and bl.request_alignment, or 0
490 * for no 32-bit limit. For now, anything larger than INT_MAX is
492 uint32_t max_transfer
;
494 /* memory alignment, in bytes so that no bounce buffer is needed */
495 size_t min_mem_alignment
;
497 /* memory alignment, in bytes, for bounce buffer */
498 size_t opt_mem_alignment
;
500 /* maximum number of iovec elements */
504 typedef struct BdrvOpBlocker BdrvOpBlocker
;
506 typedef struct BdrvAioNotifier
{
507 void (*attached_aio_context
)(AioContext
*new_context
, void *opaque
);
508 void (*detach_aio_context
)(void *opaque
);
513 QLIST_ENTRY(BdrvAioNotifier
) list
;
516 struct BdrvChildRole
{
517 /* If true, bdrv_replace_node() doesn't change the node this BdrvChild
521 void (*inherit_options
)(int *child_flags
, QDict
*child_options
,
522 int parent_flags
, QDict
*parent_options
);
524 void (*change_media
)(BdrvChild
*child
, bool load
);
525 void (*resize
)(BdrvChild
*child
);
527 /* Returns a name that is supposedly more useful for human users than the
528 * node name for identifying the node in question (in particular, a BB
529 * name), or NULL if the parent can't provide a better name. */
530 const char *(*get_name
)(BdrvChild
*child
);
532 /* Returns a malloced string that describes the parent of the child for a
533 * human reader. This could be a node-name, BlockBackend name, qdev ID or
534 * QOM path of the device owning the BlockBackend, job type and ID etc. The
535 * caller is responsible for freeing the memory. */
536 char *(*get_parent_desc
)(BdrvChild
*child
);
539 * If this pair of functions is implemented, the parent doesn't issue new
540 * requests after returning from .drained_begin() until .drained_end() is
543 * Note that this can be nested. If drained_begin() was called twice, new
544 * I/O is allowed only after drained_end() was called twice, too.
546 void (*drained_begin
)(BdrvChild
*child
);
547 void (*drained_end
)(BdrvChild
*child
);
549 /* Notifies the parent that the child has been activated/inactivated (e.g.
550 * when migration is completing) and it can start/stop requesting
551 * permissions and doing I/O on it. */
552 void (*activate
)(BdrvChild
*child
, Error
**errp
);
553 int (*inactivate
)(BdrvChild
*child
);
555 void (*attach
)(BdrvChild
*child
);
556 void (*detach
)(BdrvChild
*child
);
558 /* Notifies the parent that the filename of its child has changed (e.g.
559 * because the direct child was removed from the backing chain), so that it
560 * can update its reference. */
561 int (*update_filename
)(BdrvChild
*child
, BlockDriverState
*new_base
,
562 const char *filename
, Error
**errp
);
565 extern const BdrvChildRole child_file
;
566 extern const BdrvChildRole child_format
;
567 extern const BdrvChildRole child_backing
;
570 BlockDriverState
*bs
;
572 const BdrvChildRole
*role
;
576 * Granted permissions for operating on this BdrvChild (BLK_PERM_* bitmask)
581 * Permissions that can still be granted to other users of @bs while this
582 * BdrvChild is still attached to it. (BLK_PERM_* bitmask)
584 uint64_t shared_perm
;
586 QLIST_ENTRY(BdrvChild
) next
;
587 QLIST_ENTRY(BdrvChild
) next_parent
;
591 * Note: the function bdrv_append() copies and swaps contents of
592 * BlockDriverStates, so if you add new fields to this struct, please
593 * inspect bdrv_append() to determine if the new fields need to be
596 struct BlockDriverState
{
597 /* Protected by big QEMU lock or read-only after opening. No special
598 * locking needed during I/O...
600 int open_flags
; /* flags used to open the file, re-used for re-open */
601 bool read_only
; /* if true, the media is read only */
602 bool encrypted
; /* if true, the media is encrypted */
603 bool sg
; /* if true, the device is a /dev/sg* */
604 bool probed
; /* if true, format was probed rather than specified */
605 bool force_share
; /* if true, always allow all shared permissions */
606 bool implicit
; /* if true, this filter node was automatically inserted */
608 BlockDriver
*drv
; /* NULL means no media */
611 AioContext
*aio_context
; /* event loop used for fd handlers, timers, etc */
612 /* long-running tasks intended to always use the same AioContext as this
613 * BDS may register themselves in this list to be notified of changes
614 * regarding this BDS's context */
615 QLIST_HEAD(, BdrvAioNotifier
) aio_notifiers
;
616 bool walking_aio_notifiers
; /* to make removal during iteration safe */
618 char filename
[PATH_MAX
];
619 char backing_file
[PATH_MAX
]; /* if non zero, the image is a diff of
621 char backing_format
[16]; /* if non-zero and backing_file exists */
623 QDict
*full_open_options
;
624 char exact_filename
[PATH_MAX
];
632 /* Flags honored during pwrite (so far: BDRV_REQ_FUA) */
633 unsigned int supported_write_flags
;
634 /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA,
635 * BDRV_REQ_MAY_UNMAP) */
636 unsigned int supported_zero_flags
;
638 /* the following member gives a name to every node on the bs graph. */
640 /* element of the list of named nodes building the graph */
641 QTAILQ_ENTRY(BlockDriverState
) node_list
;
642 /* element of the list of all BlockDriverStates (all_bdrv_states) */
643 QTAILQ_ENTRY(BlockDriverState
) bs_list
;
644 /* element of the list of monitor-owned BDS */
645 QTAILQ_ENTRY(BlockDriverState
) monitor_list
;
648 /* operation blockers */
649 QLIST_HEAD(, BdrvOpBlocker
) op_blockers
[BLOCK_OP_TYPE_MAX
];
651 /* long-running background operation */
654 /* The node that this node inherited default options from (and a reopen on
655 * which can affect this node by changing these defaults). This is always a
656 * parent node of this node. */
657 BlockDriverState
*inherits_from
;
658 QLIST_HEAD(, BdrvChild
) children
;
659 QLIST_HEAD(, BdrvChild
) parents
;
662 QDict
*explicit_options
;
663 BlockdevDetectZeroesOptions detect_zeroes
;
665 /* The error object in use for blocking operations on backing_hd */
666 Error
*backing_blocker
;
668 /* Protected by AioContext lock */
670 /* If we are reading a disk image, give its size in sectors.
671 * Generally read-only; it is written to by load_snapshot and
672 * save_snaphost, but the block layer is quiescent during those.
674 int64_t total_sectors
;
676 /* Callback before write request is processed */
677 NotifierWithReturnList before_write_notifiers
;
679 /* threshold limit for writes, in bytes. "High water mark". */
680 uint64_t write_threshold_offset
;
681 NotifierWithReturn write_threshold_notifier
;
683 /* Writing to the list requires the BQL _and_ the dirty_bitmap_mutex.
684 * Reading from the list can be done with either the BQL or the
685 * dirty_bitmap_mutex. Modifying a bitmap only requires
686 * dirty_bitmap_mutex. */
687 QemuMutex dirty_bitmap_mutex
;
688 QLIST_HEAD(, BdrvDirtyBitmap
) dirty_bitmaps
;
690 /* Offset after the highest byte written to */
691 Stat64 wr_highest_offset
;
693 /* If true, copy read backing sectors into image. Can be >1 if more
694 * than one client has requested copy-on-read. Accessed with atomic
699 /* number of in-flight requests; overall and serialising.
700 * Accessed with atomic ops.
702 unsigned int in_flight
;
703 unsigned int serialising_in_flight
;
705 /* Internal to BDRV_POLL_WHILE and bdrv_wakeup. Accessed with atomic
710 /* counter for nested bdrv_io_plug.
711 * Accessed with atomic ops.
715 /* do we need to tell the quest if we have a volatile write cache? */
716 int enable_write_cache
;
718 /* Accessed with atomic ops. */
720 int recursive_quiesce_counter
;
722 unsigned int write_gen
; /* Current data generation */
724 /* Protected by reqs_lock. */
726 QLIST_HEAD(, BdrvTrackedRequest
) tracked_requests
;
727 CoQueue flush_queue
; /* Serializing flush queue */
728 bool active_flush_req
; /* Flush request in flight? */
730 /* Only read/written by whoever has set active_flush_req to true. */
731 unsigned int flushed_gen
; /* Flushed write generation */
734 struct BlockBackendRootState
{
737 BlockdevDetectZeroesOptions detect_zeroes
;
740 typedef enum BlockMirrorBackingMode
{
741 /* Reuse the existing backing chain from the source for the target.
742 * - sync=full: Set backing BDS to NULL.
743 * - sync=top: Use source's backing BDS.
744 * - sync=none: Use source as the backing BDS. */
745 MIRROR_SOURCE_BACKING_CHAIN
,
747 /* Open the target's backing chain completely anew */
748 MIRROR_OPEN_BACKING_CHAIN
,
750 /* Do not change the target's backing BDS after job completion */
751 MIRROR_LEAVE_BACKING_CHAIN
,
752 } BlockMirrorBackingMode
;
754 static inline BlockDriverState
*backing_bs(BlockDriverState
*bs
)
756 return bs
->backing
? bs
->backing
->bs
: NULL
;
760 /* Essential block drivers which must always be statically linked into qemu, and
761 * which therefore can be accessed without using bdrv_find_format() */
762 extern BlockDriver bdrv_file
;
763 extern BlockDriver bdrv_raw
;
764 extern BlockDriver bdrv_qcow2
;
766 int coroutine_fn
bdrv_co_preadv(BdrvChild
*child
,
767 int64_t offset
, unsigned int bytes
, QEMUIOVector
*qiov
,
768 BdrvRequestFlags flags
);
769 int coroutine_fn
bdrv_co_pwritev(BdrvChild
*child
,
770 int64_t offset
, unsigned int bytes
, QEMUIOVector
*qiov
,
771 BdrvRequestFlags flags
);
773 void bdrv_apply_subtree_drain(BdrvChild
*child
, BlockDriverState
*new_parent
);
774 void bdrv_unapply_subtree_drain(BdrvChild
*child
, BlockDriverState
*old_parent
);
776 int get_tmp_filename(char *filename
, int size
);
777 BlockDriver
*bdrv_probe_all(const uint8_t *buf
, int buf_size
,
778 const char *filename
);
780 void bdrv_parse_filename_strip_prefix(const char *filename
, const char *prefix
,
785 * bdrv_add_before_write_notifier:
787 * Register a callback that is invoked before write requests are processed but
788 * after any throttling or waiting for overlapping requests.
790 void bdrv_add_before_write_notifier(BlockDriverState
*bs
,
791 NotifierWithReturn
*notifier
);
794 * bdrv_detach_aio_context:
796 * May be called from .bdrv_detach_aio_context() to detach children from the
797 * current #AioContext. This is only needed by block drivers that manage their
798 * own children. Both ->file and ->backing are automatically handled and
799 * block drivers should not call this function on them explicitly.
801 void bdrv_detach_aio_context(BlockDriverState
*bs
);
804 * bdrv_attach_aio_context:
806 * May be called from .bdrv_attach_aio_context() to attach children to the new
807 * #AioContext. This is only needed by block drivers that manage their own
808 * children. Both ->file and ->backing are automatically handled and block
809 * drivers should not call this function on them explicitly.
811 void bdrv_attach_aio_context(BlockDriverState
*bs
,
812 AioContext
*new_context
);
815 * bdrv_add_aio_context_notifier:
817 * If a long-running job intends to be always run in the same AioContext as a
818 * certain BDS, it may use this function to be notified of changes regarding the
819 * association of the BDS to an AioContext.
821 * attached_aio_context() is called after the target BDS has been attached to a
822 * new AioContext; detach_aio_context() is called before the target BDS is being
823 * detached from its old AioContext.
825 void bdrv_add_aio_context_notifier(BlockDriverState
*bs
,
826 void (*attached_aio_context
)(AioContext
*new_context
, void *opaque
),
827 void (*detach_aio_context
)(void *opaque
), void *opaque
);
830 * bdrv_remove_aio_context_notifier:
832 * Unsubscribe of change notifications regarding the BDS's AioContext. The
833 * parameters given here have to be the same as those given to
834 * bdrv_add_aio_context_notifier().
836 void bdrv_remove_aio_context_notifier(BlockDriverState
*bs
,
837 void (*aio_context_attached
)(AioContext
*,
839 void (*aio_context_detached
)(void *),
844 * @bs: The BlockDriverState for which an I/O operation has been completed.
846 * Wake up the main thread if it is waiting on BDRV_POLL_WHILE. During
847 * synchronous I/O on a BlockDriverState that is attached to another
848 * I/O thread, the main thread lets the I/O thread's event loop run,
849 * waiting for the I/O operation to complete. A bdrv_wakeup will wake
850 * up the main thread if necessary.
852 * Manual calls to bdrv_wakeup are rarely necessary, because
853 * bdrv_dec_in_flight already calls it.
855 void bdrv_wakeup(BlockDriverState
*bs
);
858 int is_windows_drive(const char *filename
);
863 * @job_id: The id of the newly-created job, or %NULL to use the
864 * device name of @bs.
865 * @bs: Block device to operate on.
866 * @base: Block device that will become the new base, or %NULL to
867 * flatten the whole backing file chain onto @bs.
868 * @backing_file_str: The file name that will be written to @bs as the
869 * the new backing file if the job completes. Ignored if @base is %NULL.
870 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
871 * @on_error: The action to take upon error.
872 * @errp: Error object.
874 * Start a streaming operation on @bs. Clusters that are unallocated
875 * in @bs, but allocated in any image between @base and @bs (both
876 * exclusive) will be written to @bs. At the end of a successful
877 * streaming job, the backing file of @bs will be changed to
878 * @backing_file_str in the written image and to @base in the live
881 void stream_start(const char *job_id
, BlockDriverState
*bs
,
882 BlockDriverState
*base
, const char *backing_file_str
,
883 int64_t speed
, BlockdevOnError on_error
, Error
**errp
);
887 * @job_id: The id of the newly-created job, or %NULL to use the
888 * device name of @bs.
889 * @bs: Active block device.
890 * @top: Top block device to be committed.
891 * @base: Block device that will be written into, and become the new top.
892 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
893 * @on_error: The action to take upon error.
894 * @backing_file_str: String to use as the backing file in @top's overlay
895 * @filter_node_name: The node name that should be assigned to the filter
896 * driver that the commit job inserts into the graph above @top. NULL means
897 * that a node name should be autogenerated.
898 * @errp: Error object.
901 void commit_start(const char *job_id
, BlockDriverState
*bs
,
902 BlockDriverState
*base
, BlockDriverState
*top
, int64_t speed
,
903 BlockdevOnError on_error
, const char *backing_file_str
,
904 const char *filter_node_name
, Error
**errp
);
906 * commit_active_start:
907 * @job_id: The id of the newly-created job, or %NULL to use the
908 * device name of @bs.
909 * @bs: Active block device to be committed.
910 * @base: Block device that will be written into, and become the new top.
911 * @creation_flags: Flags that control the behavior of the Job lifetime.
912 * See @BlockJobCreateFlags
913 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
914 * @on_error: The action to take upon error.
915 * @filter_node_name: The node name that should be assigned to the filter
916 * driver that the commit job inserts into the graph above @bs. NULL means that
917 * a node name should be autogenerated.
918 * @cb: Completion function for the job.
919 * @opaque: Opaque pointer value passed to @cb.
920 * @auto_complete: Auto complete the job.
921 * @errp: Error object.
924 void commit_active_start(const char *job_id
, BlockDriverState
*bs
,
925 BlockDriverState
*base
, int creation_flags
,
926 int64_t speed
, BlockdevOnError on_error
,
927 const char *filter_node_name
,
928 BlockCompletionFunc
*cb
, void *opaque
,
929 bool auto_complete
, Error
**errp
);
932 * @job_id: The id of the newly-created job, or %NULL to use the
933 * device name of @bs.
934 * @bs: Block device to operate on.
935 * @target: Block device to write to.
936 * @replaces: Block graph node name to replace once the mirror is done. Can
937 * only be used when full mirroring is selected.
938 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
939 * @granularity: The chosen granularity for the dirty bitmap.
940 * @buf_size: The amount of data that can be in flight at one time.
941 * @mode: Whether to collapse all images in the chain to the target.
942 * @backing_mode: How to establish the target's backing chain after completion.
943 * @on_source_error: The action to take upon error reading from the source.
944 * @on_target_error: The action to take upon error writing to the target.
945 * @unmap: Whether to unmap target where source sectors only contain zeroes.
946 * @filter_node_name: The node name that should be assigned to the filter
947 * driver that the mirror job inserts into the graph above @bs. NULL means that
948 * a node name should be autogenerated.
949 * @errp: Error object.
951 * Start a mirroring operation on @bs. Clusters that are allocated
952 * in @bs will be written to @target until the job is cancelled or
953 * manually completed. At the end of a successful mirroring job,
954 * @bs will be switched to read from @target.
956 void mirror_start(const char *job_id
, BlockDriverState
*bs
,
957 BlockDriverState
*target
, const char *replaces
,
958 int64_t speed
, uint32_t granularity
, int64_t buf_size
,
959 MirrorSyncMode mode
, BlockMirrorBackingMode backing_mode
,
960 BlockdevOnError on_source_error
,
961 BlockdevOnError on_target_error
,
962 bool unmap
, const char *filter_node_name
, Error
**errp
);
966 * @job_id: The id of the newly-created job, or %NULL to use the
967 * device name of @bs.
968 * @bs: Block device to operate on.
969 * @target: Block device to write to.
970 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
971 * @sync_mode: What parts of the disk image should be copied to the destination.
972 * @sync_bitmap: The dirty bitmap if sync_mode is MIRROR_SYNC_MODE_INCREMENTAL.
973 * @on_source_error: The action to take upon error reading from the source.
974 * @on_target_error: The action to take upon error writing to the target.
975 * @creation_flags: Flags that control the behavior of the Job lifetime.
976 * See @BlockJobCreateFlags
977 * @cb: Completion function for the job.
978 * @opaque: Opaque pointer value passed to @cb.
979 * @txn: Transaction that this job is part of (may be NULL).
981 * Create a backup operation on @bs. Clusters in @bs are written to @target
982 * until the job is cancelled or manually completed.
984 BlockJob
*backup_job_create(const char *job_id
, BlockDriverState
*bs
,
985 BlockDriverState
*target
, int64_t speed
,
986 MirrorSyncMode sync_mode
,
987 BdrvDirtyBitmap
*sync_bitmap
,
989 BlockdevOnError on_source_error
,
990 BlockdevOnError on_target_error
,
992 BlockCompletionFunc
*cb
, void *opaque
,
993 BlockJobTxn
*txn
, Error
**errp
);
995 void hmp_drive_add_node(Monitor
*mon
, const char *optstr
);
997 BdrvChild
*bdrv_root_attach_child(BlockDriverState
*child_bs
,
998 const char *child_name
,
999 const BdrvChildRole
*child_role
,
1000 uint64_t perm
, uint64_t shared_perm
,
1001 void *opaque
, Error
**errp
);
1002 void bdrv_root_unref_child(BdrvChild
*child
);
1004 int bdrv_child_try_set_perm(BdrvChild
*c
, uint64_t perm
, uint64_t shared
,
1007 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1008 * block filters: Forward CONSISTENT_READ, WRITE, WRITE_UNCHANGED and RESIZE to
1010 void bdrv_filter_default_perms(BlockDriverState
*bs
, BdrvChild
*c
,
1011 const BdrvChildRole
*role
,
1012 BlockReopenQueue
*reopen_queue
,
1013 uint64_t perm
, uint64_t shared
,
1014 uint64_t *nperm
, uint64_t *nshared
);
1016 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1017 * (non-raw) image formats: Like above for bs->backing, but for bs->file it
1018 * requires WRITE | RESIZE for read-write images, always requires
1019 * CONSISTENT_READ and doesn't share WRITE. */
1020 void bdrv_format_default_perms(BlockDriverState
*bs
, BdrvChild
*c
,
1021 const BdrvChildRole
*role
,
1022 BlockReopenQueue
*reopen_queue
,
1023 uint64_t perm
, uint64_t shared
,
1024 uint64_t *nperm
, uint64_t *nshared
);
1027 * Default implementation for drivers to pass bdrv_co_get_block_status() to
1030 int64_t coroutine_fn
bdrv_co_get_block_status_from_file(BlockDriverState
*bs
,
1034 BlockDriverState
**file
);
1036 * Default implementation for drivers to pass bdrv_co_get_block_status() to
1037 * their backing file.
1039 int64_t coroutine_fn
bdrv_co_get_block_status_from_backing(BlockDriverState
*bs
,
1043 BlockDriverState
**file
);
1044 const char *bdrv_get_parent_name(const BlockDriverState
*bs
);
1045 void blk_dev_change_media_cb(BlockBackend
*blk
, bool load
, Error
**errp
);
1046 bool blk_dev_has_removable_media(BlockBackend
*blk
);
1047 bool blk_dev_has_tray(BlockBackend
*blk
);
1048 void blk_dev_eject_request(BlockBackend
*blk
, bool force
);
1049 bool blk_dev_is_tray_open(BlockBackend
*blk
);
1050 bool blk_dev_is_medium_locked(BlockBackend
*blk
);
1052 void bdrv_set_dirty(BlockDriverState
*bs
, int64_t offset
, int64_t bytes
);
1054 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap
*bitmap
, HBitmap
**out
);
1055 void bdrv_undo_clear_dirty_bitmap(BdrvDirtyBitmap
*bitmap
, HBitmap
*in
);
1057 void bdrv_inc_in_flight(BlockDriverState
*bs
);
1058 void bdrv_dec_in_flight(BlockDriverState
*bs
);
1060 void blockdev_close_all_bdrv_states(void);
1062 #endif /* BLOCK_INT_H */