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/queue.h"
30 #include "qemu/coroutine.h"
31 #include "qemu/stats64.h"
32 #include "qemu/timer.h"
33 #include "qemu/hbitmap.h"
34 #include "block/snapshot.h"
35 #include "qemu/main-loop.h"
36 #include "qemu/throttle.h"
38 #define BLOCK_FLAG_LAZY_REFCOUNTS 8
40 #define BLOCK_OPT_SIZE "size"
41 #define BLOCK_OPT_ENCRYPT "encryption"
42 #define BLOCK_OPT_ENCRYPT_FORMAT "encrypt.format"
43 #define BLOCK_OPT_COMPAT6 "compat6"
44 #define BLOCK_OPT_HWVERSION "hwversion"
45 #define BLOCK_OPT_BACKING_FILE "backing_file"
46 #define BLOCK_OPT_BACKING_FMT "backing_fmt"
47 #define BLOCK_OPT_CLUSTER_SIZE "cluster_size"
48 #define BLOCK_OPT_TABLE_SIZE "table_size"
49 #define BLOCK_OPT_PREALLOC "preallocation"
50 #define BLOCK_OPT_SUBFMT "subformat"
51 #define BLOCK_OPT_COMPAT_LEVEL "compat"
52 #define BLOCK_OPT_LAZY_REFCOUNTS "lazy_refcounts"
53 #define BLOCK_OPT_ADAPTER_TYPE "adapter_type"
54 #define BLOCK_OPT_REDUNDANCY "redundancy"
55 #define BLOCK_OPT_NOCOW "nocow"
56 #define BLOCK_OPT_OBJECT_SIZE "object_size"
57 #define BLOCK_OPT_REFCOUNT_BITS "refcount_bits"
59 #define BLOCK_PROBE_BUF_SIZE 512
61 enum BdrvTrackedRequestType
{
67 typedef struct BdrvTrackedRequest
{
71 enum BdrvTrackedRequestType type
;
74 int64_t overlap_offset
;
75 unsigned int overlap_bytes
;
77 QLIST_ENTRY(BdrvTrackedRequest
) list
;
78 Coroutine
*co
; /* owner, used for deadlock detection */
79 CoQueue wait_queue
; /* coroutines blocked on this request */
81 struct BdrvTrackedRequest
*waiting_for
;
85 const char *format_name
;
88 /* set to true if the BlockDriver is a block filter. Block filters pass
89 * certain callbacks that refer to data (see block.c) to their bs->file if
90 * the driver doesn't implement them. Drivers that do not wish to forward
91 * must implement them and return -ENOTSUP.
94 /* for snapshots block filter like Quorum can implement the
95 * following recursive callback.
96 * It's purpose is to recurse on the filter children while calling
97 * bdrv_recurse_is_first_non_filter on them.
98 * For a sample implementation look in the future Quorum block filter.
100 bool (*bdrv_recurse_is_first_non_filter
)(BlockDriverState
*bs
,
101 BlockDriverState
*candidate
);
103 int (*bdrv_probe
)(const uint8_t *buf
, int buf_size
, const char *filename
);
104 int (*bdrv_probe_device
)(const char *filename
);
106 /* Any driver implementing this callback is expected to be able to handle
107 * NULL file names in its .bdrv_open() implementation */
108 void (*bdrv_parse_filename
)(const char *filename
, QDict
*options
, Error
**errp
);
109 /* Drivers not implementing bdrv_parse_filename nor bdrv_open should have
110 * this field set to true, except ones that are defined only by their
112 * An example of the last type will be the quorum block driver.
114 bool bdrv_needs_filename
;
116 /* Set if a driver can support backing files */
117 bool supports_backing
;
119 /* For handling image reopen for split or non-split files */
120 int (*bdrv_reopen_prepare
)(BDRVReopenState
*reopen_state
,
121 BlockReopenQueue
*queue
, Error
**errp
);
122 void (*bdrv_reopen_commit
)(BDRVReopenState
*reopen_state
);
123 void (*bdrv_reopen_abort
)(BDRVReopenState
*reopen_state
);
124 void (*bdrv_join_options
)(QDict
*options
, QDict
*old_options
);
126 int (*bdrv_open
)(BlockDriverState
*bs
, QDict
*options
, int flags
,
128 int (*bdrv_file_open
)(BlockDriverState
*bs
, QDict
*options
, int flags
,
130 void (*bdrv_close
)(BlockDriverState
*bs
);
131 int (*bdrv_create
)(const char *filename
, QemuOpts
*opts
, Error
**errp
);
132 int (*bdrv_make_empty
)(BlockDriverState
*bs
);
134 void (*bdrv_refresh_filename
)(BlockDriverState
*bs
, QDict
*options
);
137 BlockAIOCB
*(*bdrv_aio_readv
)(BlockDriverState
*bs
,
138 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
139 BlockCompletionFunc
*cb
, void *opaque
);
140 BlockAIOCB
*(*bdrv_aio_writev
)(BlockDriverState
*bs
,
141 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
142 BlockCompletionFunc
*cb
, void *opaque
);
143 BlockAIOCB
*(*bdrv_aio_flush
)(BlockDriverState
*bs
,
144 BlockCompletionFunc
*cb
, void *opaque
);
145 BlockAIOCB
*(*bdrv_aio_pdiscard
)(BlockDriverState
*bs
,
146 int64_t offset
, int bytes
,
147 BlockCompletionFunc
*cb
, void *opaque
);
149 int coroutine_fn (*bdrv_co_readv
)(BlockDriverState
*bs
,
150 int64_t sector_num
, int nb_sectors
, QEMUIOVector
*qiov
);
153 * @offset: position in bytes to read at
154 * @bytes: number of bytes to read
155 * @qiov: the buffers to fill with read data
156 * @flags: currently unused, always 0
158 * @offset and @bytes will be a multiple of 'request_alignment',
159 * but the length of individual @qiov elements does not have to
162 * @bytes will always equal the total size of @qiov, and will be
163 * no larger than 'max_transfer'.
165 * The buffer in @qiov may point directly to guest memory.
167 int coroutine_fn (*bdrv_co_preadv
)(BlockDriverState
*bs
,
168 uint64_t offset
, uint64_t bytes
, QEMUIOVector
*qiov
, int flags
);
169 int coroutine_fn (*bdrv_co_writev
)(BlockDriverState
*bs
,
170 int64_t sector_num
, int nb_sectors
, QEMUIOVector
*qiov
);
171 int coroutine_fn (*bdrv_co_writev_flags
)(BlockDriverState
*bs
,
172 int64_t sector_num
, int nb_sectors
, QEMUIOVector
*qiov
, int flags
);
174 * @offset: position in bytes to write at
175 * @bytes: number of bytes to write
176 * @qiov: the buffers containing data to write
177 * @flags: zero or more bits allowed by 'supported_write_flags'
179 * @offset and @bytes will be a multiple of 'request_alignment',
180 * but the length of individual @qiov elements does not have to
183 * @bytes will always equal the total size of @qiov, and will be
184 * no larger than 'max_transfer'.
186 * The buffer in @qiov may point directly to guest memory.
188 int coroutine_fn (*bdrv_co_pwritev
)(BlockDriverState
*bs
,
189 uint64_t offset
, uint64_t bytes
, QEMUIOVector
*qiov
, int flags
);
192 * Efficiently zero a region of the disk image. Typically an image format
193 * would use a compact metadata representation to implement this. This
194 * function pointer may be NULL or return -ENOSUP and .bdrv_co_writev()
195 * will be called instead.
197 int coroutine_fn (*bdrv_co_pwrite_zeroes
)(BlockDriverState
*bs
,
198 int64_t offset
, int bytes
, BdrvRequestFlags flags
);
199 int coroutine_fn (*bdrv_co_pdiscard
)(BlockDriverState
*bs
,
200 int64_t offset
, int bytes
);
203 * Building block for bdrv_block_status[_above] and
204 * bdrv_is_allocated[_above]. The driver should answer only
205 * according to the current layer, and should not set
206 * BDRV_BLOCK_ALLOCATED, but may set BDRV_BLOCK_RAW. See block.h
207 * for the meaning of _DATA, _ZERO, and _OFFSET_VALID. The block
208 * layer guarantees input aligned to request_alignment, as well as
209 * non-NULL pnum and file.
211 int64_t coroutine_fn (*bdrv_co_get_block_status
)(BlockDriverState
*bs
,
212 int64_t sector_num
, int nb_sectors
, int *pnum
,
213 BlockDriverState
**file
);
216 * Invalidate any cached meta-data.
218 void (*bdrv_invalidate_cache
)(BlockDriverState
*bs
, Error
**errp
);
219 int (*bdrv_inactivate
)(BlockDriverState
*bs
);
222 * Flushes all data for all layers by calling bdrv_co_flush for underlying
223 * layers, if needed. This function is needed for deterministic
224 * synchronization of the flush finishing callback.
226 int coroutine_fn (*bdrv_co_flush
)(BlockDriverState
*bs
);
229 * Flushes all data that was already written to the OS all the way down to
230 * the disk (for example file-posix.c calls fsync()).
232 int coroutine_fn (*bdrv_co_flush_to_disk
)(BlockDriverState
*bs
);
235 * Flushes all internal caches to the OS. The data may still sit in a
236 * writeback cache of the host OS, but it will survive a crash of the qemu
239 int coroutine_fn (*bdrv_co_flush_to_os
)(BlockDriverState
*bs
);
241 const char *protocol_name
;
242 int (*bdrv_truncate
)(BlockDriverState
*bs
, int64_t offset
,
243 PreallocMode prealloc
, Error
**errp
);
245 int64_t (*bdrv_getlength
)(BlockDriverState
*bs
);
246 bool has_variable_length
;
247 int64_t (*bdrv_get_allocated_file_size
)(BlockDriverState
*bs
);
248 BlockMeasureInfo
*(*bdrv_measure
)(QemuOpts
*opts
, BlockDriverState
*in_bs
,
251 int coroutine_fn (*bdrv_co_pwritev_compressed
)(BlockDriverState
*bs
,
252 uint64_t offset
, uint64_t bytes
, QEMUIOVector
*qiov
);
254 int (*bdrv_snapshot_create
)(BlockDriverState
*bs
,
255 QEMUSnapshotInfo
*sn_info
);
256 int (*bdrv_snapshot_goto
)(BlockDriverState
*bs
,
257 const char *snapshot_id
);
258 int (*bdrv_snapshot_delete
)(BlockDriverState
*bs
,
259 const char *snapshot_id
,
262 int (*bdrv_snapshot_list
)(BlockDriverState
*bs
,
263 QEMUSnapshotInfo
**psn_info
);
264 int (*bdrv_snapshot_load_tmp
)(BlockDriverState
*bs
,
265 const char *snapshot_id
,
268 int (*bdrv_get_info
)(BlockDriverState
*bs
, BlockDriverInfo
*bdi
);
269 ImageInfoSpecific
*(*bdrv_get_specific_info
)(BlockDriverState
*bs
);
271 int coroutine_fn (*bdrv_save_vmstate
)(BlockDriverState
*bs
,
274 int coroutine_fn (*bdrv_load_vmstate
)(BlockDriverState
*bs
,
278 int (*bdrv_change_backing_file
)(BlockDriverState
*bs
,
279 const char *backing_file
, const char *backing_fmt
);
281 /* removable device specific */
282 bool (*bdrv_is_inserted
)(BlockDriverState
*bs
);
283 void (*bdrv_eject
)(BlockDriverState
*bs
, bool eject_flag
);
284 void (*bdrv_lock_medium
)(BlockDriverState
*bs
, bool locked
);
286 /* to control generic scsi devices */
287 BlockAIOCB
*(*bdrv_aio_ioctl
)(BlockDriverState
*bs
,
288 unsigned long int req
, void *buf
,
289 BlockCompletionFunc
*cb
, void *opaque
);
290 int coroutine_fn (*bdrv_co_ioctl
)(BlockDriverState
*bs
,
291 unsigned long int req
, void *buf
);
293 /* List of options for creating images, terminated by name == NULL */
294 QemuOptsList
*create_opts
;
297 * Returns 0 for completed check, -errno for internal errors.
298 * The check results are stored in result.
300 int (*bdrv_check
)(BlockDriverState
*bs
, BdrvCheckResult
*result
,
303 int (*bdrv_amend_options
)(BlockDriverState
*bs
, QemuOpts
*opts
,
304 BlockDriverAmendStatusCB
*status_cb
,
307 void (*bdrv_debug_event
)(BlockDriverState
*bs
, BlkdebugEvent event
);
309 /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
310 int (*bdrv_debug_breakpoint
)(BlockDriverState
*bs
, const char *event
,
312 int (*bdrv_debug_remove_breakpoint
)(BlockDriverState
*bs
,
314 int (*bdrv_debug_resume
)(BlockDriverState
*bs
, const char *tag
);
315 bool (*bdrv_debug_is_suspended
)(BlockDriverState
*bs
, const char *tag
);
317 void (*bdrv_refresh_limits
)(BlockDriverState
*bs
, Error
**errp
);
320 * Returns 1 if newly created images are guaranteed to contain only
321 * zeros, 0 otherwise.
323 int (*bdrv_has_zero_init
)(BlockDriverState
*bs
);
325 /* Remove fd handlers, timers, and other event loop callbacks so the event
326 * loop is no longer in use. Called with no in-flight requests and in
327 * depth-first traversal order with parents before child nodes.
329 void (*bdrv_detach_aio_context
)(BlockDriverState
*bs
);
331 /* Add fd handlers, timers, and other event loop callbacks so I/O requests
332 * can be processed again. Called with no in-flight requests and in
333 * depth-first traversal order with child nodes before parent nodes.
335 void (*bdrv_attach_aio_context
)(BlockDriverState
*bs
,
336 AioContext
*new_context
);
338 /* io queue for linux-aio */
339 void (*bdrv_io_plug
)(BlockDriverState
*bs
);
340 void (*bdrv_io_unplug
)(BlockDriverState
*bs
);
343 * Try to get @bs's logical and physical block size.
344 * On success, store them in @bsz and return zero.
345 * On failure, return negative errno.
347 int (*bdrv_probe_blocksizes
)(BlockDriverState
*bs
, BlockSizes
*bsz
);
349 * Try to get @bs's geometry (cyls, heads, sectors)
350 * On success, store them in @geo and return 0.
351 * On failure return -errno.
352 * Only drivers that want to override guest geometry implement this
353 * callback; see hd_geometry_guess().
355 int (*bdrv_probe_geometry
)(BlockDriverState
*bs
, HDGeometry
*geo
);
358 * bdrv_co_drain_begin is called if implemented in the beginning of a
359 * drain operation to drain and stop any internal sources of requests in
361 * bdrv_co_drain_end is called if implemented at the end of the drain.
363 * They should be used by the driver to e.g. manage scheduled I/O
364 * requests, or toggle an internal state. After the end of the drain new
365 * requests will continue normally.
367 void coroutine_fn (*bdrv_co_drain_begin
)(BlockDriverState
*bs
);
368 void coroutine_fn (*bdrv_co_drain_end
)(BlockDriverState
*bs
);
370 void (*bdrv_add_child
)(BlockDriverState
*parent
, BlockDriverState
*child
,
372 void (*bdrv_del_child
)(BlockDriverState
*parent
, BdrvChild
*child
,
376 * Informs the block driver that a permission change is intended. The
377 * driver checks whether the change is permissible and may take other
378 * preparations for the change (e.g. get file system locks). This operation
379 * is always followed either by a call to either .bdrv_set_perm or
380 * .bdrv_abort_perm_update.
382 * Checks whether the requested set of cumulative permissions in @perm
383 * can be granted for accessing @bs and whether no other users are using
384 * permissions other than those given in @shared (both arguments take
385 * BLK_PERM_* bitmasks).
387 * If both conditions are met, 0 is returned. Otherwise, -errno is returned
388 * and errp is set to an error describing the conflict.
390 int (*bdrv_check_perm
)(BlockDriverState
*bs
, uint64_t perm
,
391 uint64_t shared
, Error
**errp
);
394 * Called to inform the driver that the set of cumulative set of used
395 * permissions for @bs has changed to @perm, and the set of sharable
396 * permission to @shared. The driver can use this to propagate changes to
397 * its children (i.e. request permissions only if a parent actually needs
400 * This function is only invoked after bdrv_check_perm(), so block drivers
401 * may rely on preparations made in their .bdrv_check_perm implementation.
403 void (*bdrv_set_perm
)(BlockDriverState
*bs
, uint64_t perm
, uint64_t shared
);
406 * Called to inform the driver that after a previous bdrv_check_perm()
407 * call, the permission update is not performed and any preparations made
408 * for it (e.g. taken file locks) need to be undone.
410 * This function can be called even for nodes that never saw a
411 * bdrv_check_perm() call. It is a no-op then.
413 void (*bdrv_abort_perm_update
)(BlockDriverState
*bs
);
416 * Returns in @nperm and @nshared the permissions that the driver for @bs
417 * needs on its child @c, based on the cumulative permissions requested by
418 * the parents in @parent_perm and @parent_shared.
420 * If @c is NULL, return the permissions for attaching a new child for the
423 * If @reopen_queue is non-NULL, don't return the currently needed
424 * permissions, but those that will be needed after applying the
427 void (*bdrv_child_perm
)(BlockDriverState
*bs
, BdrvChild
*c
,
428 const BdrvChildRole
*role
,
429 BlockReopenQueue
*reopen_queue
,
430 uint64_t parent_perm
, uint64_t parent_shared
,
431 uint64_t *nperm
, uint64_t *nshared
);
434 * Bitmaps should be marked as 'IN_USE' in the image on reopening image
435 * as rw. This handler should realize it. It also should unset readonly
436 * field of BlockDirtyBitmap's in case of success.
438 int (*bdrv_reopen_bitmaps_rw
)(BlockDriverState
*bs
, Error
**errp
);
439 bool (*bdrv_can_store_new_dirty_bitmap
)(BlockDriverState
*bs
,
441 uint32_t granularity
,
443 void (*bdrv_remove_persistent_dirty_bitmap
)(BlockDriverState
*bs
,
448 * Register/unregister a buffer for I/O. For example, when the driver is
449 * interested to know the memory areas that will later be used in iovs, so
450 * that it can do IOMMU mapping with VFIO etc., in order to get better
451 * performance. In the case of VFIO drivers, this callback is used to do
452 * DMA mapping for hot buffers.
454 void (*bdrv_register_buf
)(BlockDriverState
*bs
, void *host
, size_t size
);
455 void (*bdrv_unregister_buf
)(BlockDriverState
*bs
, void *host
);
456 QLIST_ENTRY(BlockDriver
) list
;
459 typedef struct BlockLimits
{
460 /* Alignment requirement, in bytes, for offset/length of I/O
461 * requests. Must be a power of 2 less than INT_MAX; defaults to
462 * 1 for drivers with modern byte interfaces, and to 512
464 uint32_t request_alignment
;
466 /* Maximum number of bytes that can be discarded at once (since it
467 * is signed, it must be < 2G, if set). Must be multiple of
468 * pdiscard_alignment, but need not be power of 2. May be 0 if no
469 * inherent 32-bit limit */
470 int32_t max_pdiscard
;
472 /* Optimal alignment for discard requests in bytes. A power of 2
473 * is best but not mandatory. Must be a multiple of
474 * bl.request_alignment, and must be less than max_pdiscard if
475 * that is set. May be 0 if bl.request_alignment is good enough */
476 uint32_t pdiscard_alignment
;
478 /* Maximum number of bytes that can zeroized at once (since it is
479 * signed, it must be < 2G, if set). Must be multiple of
480 * pwrite_zeroes_alignment. May be 0 if no inherent 32-bit limit */
481 int32_t max_pwrite_zeroes
;
483 /* Optimal alignment for write zeroes requests in bytes. A power
484 * of 2 is best but not mandatory. Must be a multiple of
485 * bl.request_alignment, and must be less than max_pwrite_zeroes
486 * if that is set. May be 0 if bl.request_alignment is good
488 uint32_t pwrite_zeroes_alignment
;
490 /* Optimal transfer length in bytes. A power of 2 is best but not
491 * mandatory. Must be a multiple of bl.request_alignment, or 0 if
492 * no preferred size */
493 uint32_t opt_transfer
;
495 /* Maximal transfer length in bytes. Need not be power of 2, but
496 * must be multiple of opt_transfer and bl.request_alignment, or 0
497 * for no 32-bit limit. For now, anything larger than INT_MAX is
499 uint32_t max_transfer
;
501 /* memory alignment, in bytes so that no bounce buffer is needed */
502 size_t min_mem_alignment
;
504 /* memory alignment, in bytes, for bounce buffer */
505 size_t opt_mem_alignment
;
507 /* maximum number of iovec elements */
511 typedef struct BdrvOpBlocker BdrvOpBlocker
;
513 typedef struct BdrvAioNotifier
{
514 void (*attached_aio_context
)(AioContext
*new_context
, void *opaque
);
515 void (*detach_aio_context
)(void *opaque
);
520 QLIST_ENTRY(BdrvAioNotifier
) list
;
523 struct BdrvChildRole
{
524 /* If true, bdrv_replace_node() doesn't change the node this BdrvChild
528 void (*inherit_options
)(int *child_flags
, QDict
*child_options
,
529 int parent_flags
, QDict
*parent_options
);
531 void (*change_media
)(BdrvChild
*child
, bool load
);
532 void (*resize
)(BdrvChild
*child
);
534 /* Returns a name that is supposedly more useful for human users than the
535 * node name for identifying the node in question (in particular, a BB
536 * name), or NULL if the parent can't provide a better name. */
537 const char *(*get_name
)(BdrvChild
*child
);
539 /* Returns a malloced string that describes the parent of the child for a
540 * human reader. This could be a node-name, BlockBackend name, qdev ID or
541 * QOM path of the device owning the BlockBackend, job type and ID etc. The
542 * caller is responsible for freeing the memory. */
543 char *(*get_parent_desc
)(BdrvChild
*child
);
546 * If this pair of functions is implemented, the parent doesn't issue new
547 * requests after returning from .drained_begin() until .drained_end() is
550 * Note that this can be nested. If drained_begin() was called twice, new
551 * I/O is allowed only after drained_end() was called twice, too.
553 void (*drained_begin
)(BdrvChild
*child
);
554 void (*drained_end
)(BdrvChild
*child
);
556 /* Notifies the parent that the child has been activated/inactivated (e.g.
557 * when migration is completing) and it can start/stop requesting
558 * permissions and doing I/O on it. */
559 void (*activate
)(BdrvChild
*child
, Error
**errp
);
560 int (*inactivate
)(BdrvChild
*child
);
562 void (*attach
)(BdrvChild
*child
);
563 void (*detach
)(BdrvChild
*child
);
565 /* Notifies the parent that the filename of its child has changed (e.g.
566 * because the direct child was removed from the backing chain), so that it
567 * can update its reference. */
568 int (*update_filename
)(BdrvChild
*child
, BlockDriverState
*new_base
,
569 const char *filename
, Error
**errp
);
572 extern const BdrvChildRole child_file
;
573 extern const BdrvChildRole child_format
;
574 extern const BdrvChildRole child_backing
;
577 BlockDriverState
*bs
;
579 const BdrvChildRole
*role
;
583 * Granted permissions for operating on this BdrvChild (BLK_PERM_* bitmask)
588 * Permissions that can still be granted to other users of @bs while this
589 * BdrvChild is still attached to it. (BLK_PERM_* bitmask)
591 uint64_t shared_perm
;
593 QLIST_ENTRY(BdrvChild
) next
;
594 QLIST_ENTRY(BdrvChild
) next_parent
;
598 * Note: the function bdrv_append() copies and swaps contents of
599 * BlockDriverStates, so if you add new fields to this struct, please
600 * inspect bdrv_append() to determine if the new fields need to be
603 struct BlockDriverState
{
604 /* Protected by big QEMU lock or read-only after opening. No special
605 * locking needed during I/O...
607 int open_flags
; /* flags used to open the file, re-used for re-open */
608 bool read_only
; /* if true, the media is read only */
609 bool encrypted
; /* if true, the media is encrypted */
610 bool sg
; /* if true, the device is a /dev/sg* */
611 bool probed
; /* if true, format was probed rather than specified */
612 bool force_share
; /* if true, always allow all shared permissions */
613 bool implicit
; /* if true, this filter node was automatically inserted */
615 BlockDriver
*drv
; /* NULL means no media */
618 AioContext
*aio_context
; /* event loop used for fd handlers, timers, etc */
619 /* long-running tasks intended to always use the same AioContext as this
620 * BDS may register themselves in this list to be notified of changes
621 * regarding this BDS's context */
622 QLIST_HEAD(, BdrvAioNotifier
) aio_notifiers
;
623 bool walking_aio_notifiers
; /* to make removal during iteration safe */
625 char filename
[PATH_MAX
];
626 char backing_file
[PATH_MAX
]; /* if non zero, the image is a diff of
628 char backing_format
[16]; /* if non-zero and backing_file exists */
630 QDict
*full_open_options
;
631 char exact_filename
[PATH_MAX
];
639 /* Flags honored during pwrite (so far: BDRV_REQ_FUA) */
640 unsigned int supported_write_flags
;
641 /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA,
642 * BDRV_REQ_MAY_UNMAP) */
643 unsigned int supported_zero_flags
;
645 /* the following member gives a name to every node on the bs graph. */
647 /* element of the list of named nodes building the graph */
648 QTAILQ_ENTRY(BlockDriverState
) node_list
;
649 /* element of the list of all BlockDriverStates (all_bdrv_states) */
650 QTAILQ_ENTRY(BlockDriverState
) bs_list
;
651 /* element of the list of monitor-owned BDS */
652 QTAILQ_ENTRY(BlockDriverState
) monitor_list
;
655 /* operation blockers */
656 QLIST_HEAD(, BdrvOpBlocker
) op_blockers
[BLOCK_OP_TYPE_MAX
];
658 /* long-running background operation */
661 /* The node that this node inherited default options from (and a reopen on
662 * which can affect this node by changing these defaults). This is always a
663 * parent node of this node. */
664 BlockDriverState
*inherits_from
;
665 QLIST_HEAD(, BdrvChild
) children
;
666 QLIST_HEAD(, BdrvChild
) parents
;
669 QDict
*explicit_options
;
670 BlockdevDetectZeroesOptions detect_zeroes
;
672 /* The error object in use for blocking operations on backing_hd */
673 Error
*backing_blocker
;
675 /* Protected by AioContext lock */
677 /* If we are reading a disk image, give its size in sectors.
678 * Generally read-only; it is written to by load_snapshot and
679 * save_snaphost, but the block layer is quiescent during those.
681 int64_t total_sectors
;
683 /* Callback before write request is processed */
684 NotifierWithReturnList before_write_notifiers
;
686 /* threshold limit for writes, in bytes. "High water mark". */
687 uint64_t write_threshold_offset
;
688 NotifierWithReturn write_threshold_notifier
;
690 /* Writing to the list requires the BQL _and_ the dirty_bitmap_mutex.
691 * Reading from the list can be done with either the BQL or the
692 * dirty_bitmap_mutex. Modifying a bitmap only requires
693 * dirty_bitmap_mutex. */
694 QemuMutex dirty_bitmap_mutex
;
695 QLIST_HEAD(, BdrvDirtyBitmap
) dirty_bitmaps
;
697 /* Offset after the highest byte written to */
698 Stat64 wr_highest_offset
;
700 /* If true, copy read backing sectors into image. Can be >1 if more
701 * than one client has requested copy-on-read. Accessed with atomic
706 /* number of in-flight requests; overall and serialising.
707 * Accessed with atomic ops.
709 unsigned int in_flight
;
710 unsigned int serialising_in_flight
;
712 /* Internal to BDRV_POLL_WHILE and bdrv_wakeup. Accessed with atomic
717 /* counter for nested bdrv_io_plug.
718 * Accessed with atomic ops.
722 /* do we need to tell the quest if we have a volatile write cache? */
723 int enable_write_cache
;
725 /* Accessed with atomic ops. */
727 int recursive_quiesce_counter
;
729 unsigned int write_gen
; /* Current data generation */
731 /* Protected by reqs_lock. */
733 QLIST_HEAD(, BdrvTrackedRequest
) tracked_requests
;
734 CoQueue flush_queue
; /* Serializing flush queue */
735 bool active_flush_req
; /* Flush request in flight? */
737 /* Only read/written by whoever has set active_flush_req to true. */
738 unsigned int flushed_gen
; /* Flushed write generation */
741 struct BlockBackendRootState
{
744 BlockdevDetectZeroesOptions detect_zeroes
;
747 typedef enum BlockMirrorBackingMode
{
748 /* Reuse the existing backing chain from the source for the target.
749 * - sync=full: Set backing BDS to NULL.
750 * - sync=top: Use source's backing BDS.
751 * - sync=none: Use source as the backing BDS. */
752 MIRROR_SOURCE_BACKING_CHAIN
,
754 /* Open the target's backing chain completely anew */
755 MIRROR_OPEN_BACKING_CHAIN
,
757 /* Do not change the target's backing BDS after job completion */
758 MIRROR_LEAVE_BACKING_CHAIN
,
759 } BlockMirrorBackingMode
;
761 static inline BlockDriverState
*backing_bs(BlockDriverState
*bs
)
763 return bs
->backing
? bs
->backing
->bs
: NULL
;
767 /* Essential block drivers which must always be statically linked into qemu, and
768 * which therefore can be accessed without using bdrv_find_format() */
769 extern BlockDriver bdrv_file
;
770 extern BlockDriver bdrv_raw
;
771 extern BlockDriver bdrv_qcow2
;
773 int coroutine_fn
bdrv_co_preadv(BdrvChild
*child
,
774 int64_t offset
, unsigned int bytes
, QEMUIOVector
*qiov
,
775 BdrvRequestFlags flags
);
776 int coroutine_fn
bdrv_co_pwritev(BdrvChild
*child
,
777 int64_t offset
, unsigned int bytes
, QEMUIOVector
*qiov
,
778 BdrvRequestFlags flags
);
780 void bdrv_apply_subtree_drain(BdrvChild
*child
, BlockDriverState
*new_parent
);
781 void bdrv_unapply_subtree_drain(BdrvChild
*child
, BlockDriverState
*old_parent
);
783 int get_tmp_filename(char *filename
, int size
);
784 BlockDriver
*bdrv_probe_all(const uint8_t *buf
, int buf_size
,
785 const char *filename
);
787 void bdrv_parse_filename_strip_prefix(const char *filename
, const char *prefix
,
792 * bdrv_add_before_write_notifier:
794 * Register a callback that is invoked before write requests are processed but
795 * after any throttling or waiting for overlapping requests.
797 void bdrv_add_before_write_notifier(BlockDriverState
*bs
,
798 NotifierWithReturn
*notifier
);
801 * bdrv_detach_aio_context:
803 * May be called from .bdrv_detach_aio_context() to detach children from the
804 * current #AioContext. This is only needed by block drivers that manage their
805 * own children. Both ->file and ->backing are automatically handled and
806 * block drivers should not call this function on them explicitly.
808 void bdrv_detach_aio_context(BlockDriverState
*bs
);
811 * bdrv_attach_aio_context:
813 * May be called from .bdrv_attach_aio_context() to attach children to the new
814 * #AioContext. This is only needed by block drivers that manage their own
815 * children. Both ->file and ->backing are automatically handled and block
816 * drivers should not call this function on them explicitly.
818 void bdrv_attach_aio_context(BlockDriverState
*bs
,
819 AioContext
*new_context
);
822 * bdrv_add_aio_context_notifier:
824 * If a long-running job intends to be always run in the same AioContext as a
825 * certain BDS, it may use this function to be notified of changes regarding the
826 * association of the BDS to an AioContext.
828 * attached_aio_context() is called after the target BDS has been attached to a
829 * new AioContext; detach_aio_context() is called before the target BDS is being
830 * detached from its old AioContext.
832 void bdrv_add_aio_context_notifier(BlockDriverState
*bs
,
833 void (*attached_aio_context
)(AioContext
*new_context
, void *opaque
),
834 void (*detach_aio_context
)(void *opaque
), void *opaque
);
837 * bdrv_remove_aio_context_notifier:
839 * Unsubscribe of change notifications regarding the BDS's AioContext. The
840 * parameters given here have to be the same as those given to
841 * bdrv_add_aio_context_notifier().
843 void bdrv_remove_aio_context_notifier(BlockDriverState
*bs
,
844 void (*aio_context_attached
)(AioContext
*,
846 void (*aio_context_detached
)(void *),
851 * @bs: The BlockDriverState for which an I/O operation has been completed.
853 * Wake up the main thread if it is waiting on BDRV_POLL_WHILE. During
854 * synchronous I/O on a BlockDriverState that is attached to another
855 * I/O thread, the main thread lets the I/O thread's event loop run,
856 * waiting for the I/O operation to complete. A bdrv_wakeup will wake
857 * up the main thread if necessary.
859 * Manual calls to bdrv_wakeup are rarely necessary, because
860 * bdrv_dec_in_flight already calls it.
862 void bdrv_wakeup(BlockDriverState
*bs
);
865 int is_windows_drive(const char *filename
);
870 * @job_id: The id of the newly-created job, or %NULL to use the
871 * device name of @bs.
872 * @bs: Block device to operate on.
873 * @base: Block device that will become the new base, or %NULL to
874 * flatten the whole backing file chain onto @bs.
875 * @backing_file_str: The file name that will be written to @bs as the
876 * the new backing file if the job completes. Ignored if @base is %NULL.
877 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
878 * @on_error: The action to take upon error.
879 * @errp: Error object.
881 * Start a streaming operation on @bs. Clusters that are unallocated
882 * in @bs, but allocated in any image between @base and @bs (both
883 * exclusive) will be written to @bs. At the end of a successful
884 * streaming job, the backing file of @bs will be changed to
885 * @backing_file_str in the written image and to @base in the live
888 void stream_start(const char *job_id
, BlockDriverState
*bs
,
889 BlockDriverState
*base
, const char *backing_file_str
,
890 int64_t speed
, BlockdevOnError on_error
, Error
**errp
);
894 * @job_id: The id of the newly-created job, or %NULL to use the
895 * device name of @bs.
896 * @bs: Active block device.
897 * @top: Top block device to be committed.
898 * @base: Block device that will be written into, and become the new top.
899 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
900 * @on_error: The action to take upon error.
901 * @backing_file_str: String to use as the backing file in @top's overlay
902 * @filter_node_name: The node name that should be assigned to the filter
903 * driver that the commit job inserts into the graph above @top. NULL means
904 * that a node name should be autogenerated.
905 * @errp: Error object.
908 void commit_start(const char *job_id
, BlockDriverState
*bs
,
909 BlockDriverState
*base
, BlockDriverState
*top
, int64_t speed
,
910 BlockdevOnError on_error
, const char *backing_file_str
,
911 const char *filter_node_name
, Error
**errp
);
913 * commit_active_start:
914 * @job_id: The id of the newly-created job, or %NULL to use the
915 * device name of @bs.
916 * @bs: Active block device to be committed.
917 * @base: Block device that will be written into, and become the new top.
918 * @creation_flags: Flags that control the behavior of the Job lifetime.
919 * See @BlockJobCreateFlags
920 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
921 * @on_error: The action to take upon error.
922 * @filter_node_name: The node name that should be assigned to the filter
923 * driver that the commit job inserts into the graph above @bs. NULL means that
924 * a node name should be autogenerated.
925 * @cb: Completion function for the job.
926 * @opaque: Opaque pointer value passed to @cb.
927 * @auto_complete: Auto complete the job.
928 * @errp: Error object.
931 void commit_active_start(const char *job_id
, BlockDriverState
*bs
,
932 BlockDriverState
*base
, int creation_flags
,
933 int64_t speed
, BlockdevOnError on_error
,
934 const char *filter_node_name
,
935 BlockCompletionFunc
*cb
, void *opaque
,
936 bool auto_complete
, Error
**errp
);
939 * @job_id: The id of the newly-created job, or %NULL to use the
940 * device name of @bs.
941 * @bs: Block device to operate on.
942 * @target: Block device to write to.
943 * @replaces: Block graph node name to replace once the mirror is done. Can
944 * only be used when full mirroring is selected.
945 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
946 * @granularity: The chosen granularity for the dirty bitmap.
947 * @buf_size: The amount of data that can be in flight at one time.
948 * @mode: Whether to collapse all images in the chain to the target.
949 * @backing_mode: How to establish the target's backing chain after completion.
950 * @on_source_error: The action to take upon error reading from the source.
951 * @on_target_error: The action to take upon error writing to the target.
952 * @unmap: Whether to unmap target where source sectors only contain zeroes.
953 * @filter_node_name: The node name that should be assigned to the filter
954 * driver that the mirror job inserts into the graph above @bs. NULL means that
955 * a node name should be autogenerated.
956 * @errp: Error object.
958 * Start a mirroring operation on @bs. Clusters that are allocated
959 * in @bs will be written to @target until the job is cancelled or
960 * manually completed. At the end of a successful mirroring job,
961 * @bs will be switched to read from @target.
963 void mirror_start(const char *job_id
, BlockDriverState
*bs
,
964 BlockDriverState
*target
, const char *replaces
,
965 int64_t speed
, uint32_t granularity
, int64_t buf_size
,
966 MirrorSyncMode mode
, BlockMirrorBackingMode backing_mode
,
967 BlockdevOnError on_source_error
,
968 BlockdevOnError on_target_error
,
969 bool unmap
, const char *filter_node_name
, Error
**errp
);
973 * @job_id: The id of the newly-created job, or %NULL to use the
974 * device name of @bs.
975 * @bs: Block device to operate on.
976 * @target: Block device to write to.
977 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
978 * @sync_mode: What parts of the disk image should be copied to the destination.
979 * @sync_bitmap: The dirty bitmap if sync_mode is MIRROR_SYNC_MODE_INCREMENTAL.
980 * @on_source_error: The action to take upon error reading from the source.
981 * @on_target_error: The action to take upon error writing to the target.
982 * @creation_flags: Flags that control the behavior of the Job lifetime.
983 * See @BlockJobCreateFlags
984 * @cb: Completion function for the job.
985 * @opaque: Opaque pointer value passed to @cb.
986 * @txn: Transaction that this job is part of (may be NULL).
988 * Create a backup operation on @bs. Clusters in @bs are written to @target
989 * until the job is cancelled or manually completed.
991 BlockJob
*backup_job_create(const char *job_id
, BlockDriverState
*bs
,
992 BlockDriverState
*target
, int64_t speed
,
993 MirrorSyncMode sync_mode
,
994 BdrvDirtyBitmap
*sync_bitmap
,
996 BlockdevOnError on_source_error
,
997 BlockdevOnError on_target_error
,
999 BlockCompletionFunc
*cb
, void *opaque
,
1000 BlockJobTxn
*txn
, Error
**errp
);
1002 void hmp_drive_add_node(Monitor
*mon
, const char *optstr
);
1004 BdrvChild
*bdrv_root_attach_child(BlockDriverState
*child_bs
,
1005 const char *child_name
,
1006 const BdrvChildRole
*child_role
,
1007 uint64_t perm
, uint64_t shared_perm
,
1008 void *opaque
, Error
**errp
);
1009 void bdrv_root_unref_child(BdrvChild
*child
);
1011 int bdrv_child_try_set_perm(BdrvChild
*c
, uint64_t perm
, uint64_t shared
,
1014 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1015 * block filters: Forward CONSISTENT_READ, WRITE, WRITE_UNCHANGED and RESIZE to
1017 void bdrv_filter_default_perms(BlockDriverState
*bs
, BdrvChild
*c
,
1018 const BdrvChildRole
*role
,
1019 BlockReopenQueue
*reopen_queue
,
1020 uint64_t perm
, uint64_t shared
,
1021 uint64_t *nperm
, uint64_t *nshared
);
1023 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1024 * (non-raw) image formats: Like above for bs->backing, but for bs->file it
1025 * requires WRITE | RESIZE for read-write images, always requires
1026 * CONSISTENT_READ and doesn't share WRITE. */
1027 void bdrv_format_default_perms(BlockDriverState
*bs
, BdrvChild
*c
,
1028 const BdrvChildRole
*role
,
1029 BlockReopenQueue
*reopen_queue
,
1030 uint64_t perm
, uint64_t shared
,
1031 uint64_t *nperm
, uint64_t *nshared
);
1034 * Default implementation for drivers to pass bdrv_co_get_block_status() to
1037 int64_t coroutine_fn
bdrv_co_get_block_status_from_file(BlockDriverState
*bs
,
1041 BlockDriverState
**file
);
1043 * Default implementation for drivers to pass bdrv_co_get_block_status() to
1044 * their backing file.
1046 int64_t coroutine_fn
bdrv_co_get_block_status_from_backing(BlockDriverState
*bs
,
1050 BlockDriverState
**file
);
1051 const char *bdrv_get_parent_name(const BlockDriverState
*bs
);
1052 void blk_dev_change_media_cb(BlockBackend
*blk
, bool load
, Error
**errp
);
1053 bool blk_dev_has_removable_media(BlockBackend
*blk
);
1054 bool blk_dev_has_tray(BlockBackend
*blk
);
1055 void blk_dev_eject_request(BlockBackend
*blk
, bool force
);
1056 bool blk_dev_is_tray_open(BlockBackend
*blk
);
1057 bool blk_dev_is_medium_locked(BlockBackend
*blk
);
1059 void bdrv_set_dirty(BlockDriverState
*bs
, int64_t offset
, int64_t bytes
);
1061 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap
*bitmap
, HBitmap
**out
);
1062 void bdrv_undo_clear_dirty_bitmap(BdrvDirtyBitmap
*bitmap
, HBitmap
*in
);
1064 void bdrv_inc_in_flight(BlockDriverState
*bs
);
1065 void bdrv_dec_in_flight(BlockDriverState
*bs
);
1067 void blockdev_close_all_bdrv_states(void);
1069 #endif /* BLOCK_INT_H */