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 only need to set
206 * BDRV_BLOCK_DATA, BDRV_BLOCK_ZERO, BDRV_BLOCK_OFFSET_VALID,
207 * and/or BDRV_BLOCK_RAW; if the current layer defers to a backing
208 * layer, the result should be 0 (and not BDRV_BLOCK_ZERO). See
209 * block.h for the overall meaning of the bits. As a hint, the
210 * flag want_zero is true if the caller cares more about precise
211 * mappings (favor accurate _OFFSET_VALID/_ZERO) or false for
212 * overall allocation (favor larger *pnum, perhaps by reporting
213 * _DATA instead of _ZERO). The block layer guarantees input
214 * clamped to bdrv_getlength() and aligned to request_alignment,
215 * as well as non-NULL pnum, map, and file; in turn, the driver
216 * must return an error or set pnum to an aligned non-zero value.
218 int coroutine_fn (*bdrv_co_block_status
)(BlockDriverState
*bs
,
219 bool want_zero
, int64_t offset
, int64_t bytes
, int64_t *pnum
,
220 int64_t *map
, BlockDriverState
**file
);
223 * Invalidate any cached meta-data.
225 void (*bdrv_invalidate_cache
)(BlockDriverState
*bs
, Error
**errp
);
226 int (*bdrv_inactivate
)(BlockDriverState
*bs
);
229 * Flushes all data for all layers by calling bdrv_co_flush for underlying
230 * layers, if needed. This function is needed for deterministic
231 * synchronization of the flush finishing callback.
233 int coroutine_fn (*bdrv_co_flush
)(BlockDriverState
*bs
);
236 * Flushes all data that was already written to the OS all the way down to
237 * the disk (for example file-posix.c calls fsync()).
239 int coroutine_fn (*bdrv_co_flush_to_disk
)(BlockDriverState
*bs
);
242 * Flushes all internal caches to the OS. The data may still sit in a
243 * writeback cache of the host OS, but it will survive a crash of the qemu
246 int coroutine_fn (*bdrv_co_flush_to_os
)(BlockDriverState
*bs
);
248 const char *protocol_name
;
249 int (*bdrv_truncate
)(BlockDriverState
*bs
, int64_t offset
,
250 PreallocMode prealloc
, Error
**errp
);
252 int64_t (*bdrv_getlength
)(BlockDriverState
*bs
);
253 bool has_variable_length
;
254 int64_t (*bdrv_get_allocated_file_size
)(BlockDriverState
*bs
);
255 BlockMeasureInfo
*(*bdrv_measure
)(QemuOpts
*opts
, BlockDriverState
*in_bs
,
258 int coroutine_fn (*bdrv_co_pwritev_compressed
)(BlockDriverState
*bs
,
259 uint64_t offset
, uint64_t bytes
, QEMUIOVector
*qiov
);
261 int (*bdrv_snapshot_create
)(BlockDriverState
*bs
,
262 QEMUSnapshotInfo
*sn_info
);
263 int (*bdrv_snapshot_goto
)(BlockDriverState
*bs
,
264 const char *snapshot_id
);
265 int (*bdrv_snapshot_delete
)(BlockDriverState
*bs
,
266 const char *snapshot_id
,
269 int (*bdrv_snapshot_list
)(BlockDriverState
*bs
,
270 QEMUSnapshotInfo
**psn_info
);
271 int (*bdrv_snapshot_load_tmp
)(BlockDriverState
*bs
,
272 const char *snapshot_id
,
275 int (*bdrv_get_info
)(BlockDriverState
*bs
, BlockDriverInfo
*bdi
);
276 ImageInfoSpecific
*(*bdrv_get_specific_info
)(BlockDriverState
*bs
);
278 int coroutine_fn (*bdrv_save_vmstate
)(BlockDriverState
*bs
,
281 int coroutine_fn (*bdrv_load_vmstate
)(BlockDriverState
*bs
,
285 int (*bdrv_change_backing_file
)(BlockDriverState
*bs
,
286 const char *backing_file
, const char *backing_fmt
);
288 /* removable device specific */
289 bool (*bdrv_is_inserted
)(BlockDriverState
*bs
);
290 void (*bdrv_eject
)(BlockDriverState
*bs
, bool eject_flag
);
291 void (*bdrv_lock_medium
)(BlockDriverState
*bs
, bool locked
);
293 /* to control generic scsi devices */
294 BlockAIOCB
*(*bdrv_aio_ioctl
)(BlockDriverState
*bs
,
295 unsigned long int req
, void *buf
,
296 BlockCompletionFunc
*cb
, void *opaque
);
297 int coroutine_fn (*bdrv_co_ioctl
)(BlockDriverState
*bs
,
298 unsigned long int req
, void *buf
);
300 /* List of options for creating images, terminated by name == NULL */
301 QemuOptsList
*create_opts
;
304 * Returns 0 for completed check, -errno for internal errors.
305 * The check results are stored in result.
307 int (*bdrv_check
)(BlockDriverState
*bs
, BdrvCheckResult
*result
,
310 int (*bdrv_amend_options
)(BlockDriverState
*bs
, QemuOpts
*opts
,
311 BlockDriverAmendStatusCB
*status_cb
,
314 void (*bdrv_debug_event
)(BlockDriverState
*bs
, BlkdebugEvent event
);
316 /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
317 int (*bdrv_debug_breakpoint
)(BlockDriverState
*bs
, const char *event
,
319 int (*bdrv_debug_remove_breakpoint
)(BlockDriverState
*bs
,
321 int (*bdrv_debug_resume
)(BlockDriverState
*bs
, const char *tag
);
322 bool (*bdrv_debug_is_suspended
)(BlockDriverState
*bs
, const char *tag
);
324 void (*bdrv_refresh_limits
)(BlockDriverState
*bs
, Error
**errp
);
327 * Returns 1 if newly created images are guaranteed to contain only
328 * zeros, 0 otherwise.
330 int (*bdrv_has_zero_init
)(BlockDriverState
*bs
);
332 /* Remove fd handlers, timers, and other event loop callbacks so the event
333 * loop is no longer in use. Called with no in-flight requests and in
334 * depth-first traversal order with parents before child nodes.
336 void (*bdrv_detach_aio_context
)(BlockDriverState
*bs
);
338 /* Add fd handlers, timers, and other event loop callbacks so I/O requests
339 * can be processed again. Called with no in-flight requests and in
340 * depth-first traversal order with child nodes before parent nodes.
342 void (*bdrv_attach_aio_context
)(BlockDriverState
*bs
,
343 AioContext
*new_context
);
345 /* io queue for linux-aio */
346 void (*bdrv_io_plug
)(BlockDriverState
*bs
);
347 void (*bdrv_io_unplug
)(BlockDriverState
*bs
);
350 * Try to get @bs's logical and physical block size.
351 * On success, store them in @bsz and return zero.
352 * On failure, return negative errno.
354 int (*bdrv_probe_blocksizes
)(BlockDriverState
*bs
, BlockSizes
*bsz
);
356 * Try to get @bs's geometry (cyls, heads, sectors)
357 * On success, store them in @geo and return 0.
358 * On failure return -errno.
359 * Only drivers that want to override guest geometry implement this
360 * callback; see hd_geometry_guess().
362 int (*bdrv_probe_geometry
)(BlockDriverState
*bs
, HDGeometry
*geo
);
365 * bdrv_co_drain_begin is called if implemented in the beginning of a
366 * drain operation to drain and stop any internal sources of requests in
368 * bdrv_co_drain_end is called if implemented at the end of the drain.
370 * They should be used by the driver to e.g. manage scheduled I/O
371 * requests, or toggle an internal state. After the end of the drain new
372 * requests will continue normally.
374 void coroutine_fn (*bdrv_co_drain_begin
)(BlockDriverState
*bs
);
375 void coroutine_fn (*bdrv_co_drain_end
)(BlockDriverState
*bs
);
377 void (*bdrv_add_child
)(BlockDriverState
*parent
, BlockDriverState
*child
,
379 void (*bdrv_del_child
)(BlockDriverState
*parent
, BdrvChild
*child
,
383 * Informs the block driver that a permission change is intended. The
384 * driver checks whether the change is permissible and may take other
385 * preparations for the change (e.g. get file system locks). This operation
386 * is always followed either by a call to either .bdrv_set_perm or
387 * .bdrv_abort_perm_update.
389 * Checks whether the requested set of cumulative permissions in @perm
390 * can be granted for accessing @bs and whether no other users are using
391 * permissions other than those given in @shared (both arguments take
392 * BLK_PERM_* bitmasks).
394 * If both conditions are met, 0 is returned. Otherwise, -errno is returned
395 * and errp is set to an error describing the conflict.
397 int (*bdrv_check_perm
)(BlockDriverState
*bs
, uint64_t perm
,
398 uint64_t shared
, Error
**errp
);
401 * Called to inform the driver that the set of cumulative set of used
402 * permissions for @bs has changed to @perm, and the set of sharable
403 * permission to @shared. The driver can use this to propagate changes to
404 * its children (i.e. request permissions only if a parent actually needs
407 * This function is only invoked after bdrv_check_perm(), so block drivers
408 * may rely on preparations made in their .bdrv_check_perm implementation.
410 void (*bdrv_set_perm
)(BlockDriverState
*bs
, uint64_t perm
, uint64_t shared
);
413 * Called to inform the driver that after a previous bdrv_check_perm()
414 * call, the permission update is not performed and any preparations made
415 * for it (e.g. taken file locks) need to be undone.
417 * This function can be called even for nodes that never saw a
418 * bdrv_check_perm() call. It is a no-op then.
420 void (*bdrv_abort_perm_update
)(BlockDriverState
*bs
);
423 * Returns in @nperm and @nshared the permissions that the driver for @bs
424 * needs on its child @c, based on the cumulative permissions requested by
425 * the parents in @parent_perm and @parent_shared.
427 * If @c is NULL, return the permissions for attaching a new child for the
430 * If @reopen_queue is non-NULL, don't return the currently needed
431 * permissions, but those that will be needed after applying the
434 void (*bdrv_child_perm
)(BlockDriverState
*bs
, BdrvChild
*c
,
435 const BdrvChildRole
*role
,
436 BlockReopenQueue
*reopen_queue
,
437 uint64_t parent_perm
, uint64_t parent_shared
,
438 uint64_t *nperm
, uint64_t *nshared
);
441 * Bitmaps should be marked as 'IN_USE' in the image on reopening image
442 * as rw. This handler should realize it. It also should unset readonly
443 * field of BlockDirtyBitmap's in case of success.
445 int (*bdrv_reopen_bitmaps_rw
)(BlockDriverState
*bs
, Error
**errp
);
446 bool (*bdrv_can_store_new_dirty_bitmap
)(BlockDriverState
*bs
,
448 uint32_t granularity
,
450 void (*bdrv_remove_persistent_dirty_bitmap
)(BlockDriverState
*bs
,
455 * Register/unregister a buffer for I/O. For example, when the driver is
456 * interested to know the memory areas that will later be used in iovs, so
457 * that it can do IOMMU mapping with VFIO etc., in order to get better
458 * performance. In the case of VFIO drivers, this callback is used to do
459 * DMA mapping for hot buffers.
461 void (*bdrv_register_buf
)(BlockDriverState
*bs
, void *host
, size_t size
);
462 void (*bdrv_unregister_buf
)(BlockDriverState
*bs
, void *host
);
463 QLIST_ENTRY(BlockDriver
) list
;
466 typedef struct BlockLimits
{
467 /* Alignment requirement, in bytes, for offset/length of I/O
468 * requests. Must be a power of 2 less than INT_MAX; defaults to
469 * 1 for drivers with modern byte interfaces, and to 512
471 uint32_t request_alignment
;
473 /* Maximum number of bytes that can be discarded at once (since it
474 * is signed, it must be < 2G, if set). Must be multiple of
475 * pdiscard_alignment, but need not be power of 2. May be 0 if no
476 * inherent 32-bit limit */
477 int32_t max_pdiscard
;
479 /* Optimal alignment for discard requests in bytes. A power of 2
480 * is best but not mandatory. Must be a multiple of
481 * bl.request_alignment, and must be less than max_pdiscard if
482 * that is set. May be 0 if bl.request_alignment is good enough */
483 uint32_t pdiscard_alignment
;
485 /* Maximum number of bytes that can zeroized at once (since it is
486 * signed, it must be < 2G, if set). Must be multiple of
487 * pwrite_zeroes_alignment. May be 0 if no inherent 32-bit limit */
488 int32_t max_pwrite_zeroes
;
490 /* Optimal alignment for write zeroes requests in bytes. A power
491 * of 2 is best but not mandatory. Must be a multiple of
492 * bl.request_alignment, and must be less than max_pwrite_zeroes
493 * if that is set. May be 0 if bl.request_alignment is good
495 uint32_t pwrite_zeroes_alignment
;
497 /* Optimal transfer length in bytes. A power of 2 is best but not
498 * mandatory. Must be a multiple of bl.request_alignment, or 0 if
499 * no preferred size */
500 uint32_t opt_transfer
;
502 /* Maximal transfer length in bytes. Need not be power of 2, but
503 * must be multiple of opt_transfer and bl.request_alignment, or 0
504 * for no 32-bit limit. For now, anything larger than INT_MAX is
506 uint32_t max_transfer
;
508 /* memory alignment, in bytes so that no bounce buffer is needed */
509 size_t min_mem_alignment
;
511 /* memory alignment, in bytes, for bounce buffer */
512 size_t opt_mem_alignment
;
514 /* maximum number of iovec elements */
518 typedef struct BdrvOpBlocker BdrvOpBlocker
;
520 typedef struct BdrvAioNotifier
{
521 void (*attached_aio_context
)(AioContext
*new_context
, void *opaque
);
522 void (*detach_aio_context
)(void *opaque
);
527 QLIST_ENTRY(BdrvAioNotifier
) list
;
530 struct BdrvChildRole
{
531 /* If true, bdrv_replace_node() doesn't change the node this BdrvChild
535 void (*inherit_options
)(int *child_flags
, QDict
*child_options
,
536 int parent_flags
, QDict
*parent_options
);
538 void (*change_media
)(BdrvChild
*child
, bool load
);
539 void (*resize
)(BdrvChild
*child
);
541 /* Returns a name that is supposedly more useful for human users than the
542 * node name for identifying the node in question (in particular, a BB
543 * name), or NULL if the parent can't provide a better name. */
544 const char *(*get_name
)(BdrvChild
*child
);
546 /* Returns a malloced string that describes the parent of the child for a
547 * human reader. This could be a node-name, BlockBackend name, qdev ID or
548 * QOM path of the device owning the BlockBackend, job type and ID etc. The
549 * caller is responsible for freeing the memory. */
550 char *(*get_parent_desc
)(BdrvChild
*child
);
553 * If this pair of functions is implemented, the parent doesn't issue new
554 * requests after returning from .drained_begin() until .drained_end() is
557 * Note that this can be nested. If drained_begin() was called twice, new
558 * I/O is allowed only after drained_end() was called twice, too.
560 void (*drained_begin
)(BdrvChild
*child
);
561 void (*drained_end
)(BdrvChild
*child
);
563 /* Notifies the parent that the child has been activated/inactivated (e.g.
564 * when migration is completing) and it can start/stop requesting
565 * permissions and doing I/O on it. */
566 void (*activate
)(BdrvChild
*child
, Error
**errp
);
567 int (*inactivate
)(BdrvChild
*child
);
569 void (*attach
)(BdrvChild
*child
);
570 void (*detach
)(BdrvChild
*child
);
572 /* Notifies the parent that the filename of its child has changed (e.g.
573 * because the direct child was removed from the backing chain), so that it
574 * can update its reference. */
575 int (*update_filename
)(BdrvChild
*child
, BlockDriverState
*new_base
,
576 const char *filename
, Error
**errp
);
579 extern const BdrvChildRole child_file
;
580 extern const BdrvChildRole child_format
;
581 extern const BdrvChildRole child_backing
;
584 BlockDriverState
*bs
;
586 const BdrvChildRole
*role
;
590 * Granted permissions for operating on this BdrvChild (BLK_PERM_* bitmask)
595 * Permissions that can still be granted to other users of @bs while this
596 * BdrvChild is still attached to it. (BLK_PERM_* bitmask)
598 uint64_t shared_perm
;
600 QLIST_ENTRY(BdrvChild
) next
;
601 QLIST_ENTRY(BdrvChild
) next_parent
;
605 * Note: the function bdrv_append() copies and swaps contents of
606 * BlockDriverStates, so if you add new fields to this struct, please
607 * inspect bdrv_append() to determine if the new fields need to be
610 struct BlockDriverState
{
611 /* Protected by big QEMU lock or read-only after opening. No special
612 * locking needed during I/O...
614 int open_flags
; /* flags used to open the file, re-used for re-open */
615 bool read_only
; /* if true, the media is read only */
616 bool encrypted
; /* if true, the media is encrypted */
617 bool sg
; /* if true, the device is a /dev/sg* */
618 bool probed
; /* if true, format was probed rather than specified */
619 bool force_share
; /* if true, always allow all shared permissions */
620 bool implicit
; /* if true, this filter node was automatically inserted */
622 BlockDriver
*drv
; /* NULL means no media */
625 AioContext
*aio_context
; /* event loop used for fd handlers, timers, etc */
626 /* long-running tasks intended to always use the same AioContext as this
627 * BDS may register themselves in this list to be notified of changes
628 * regarding this BDS's context */
629 QLIST_HEAD(, BdrvAioNotifier
) aio_notifiers
;
630 bool walking_aio_notifiers
; /* to make removal during iteration safe */
632 char filename
[PATH_MAX
];
633 char backing_file
[PATH_MAX
]; /* if non zero, the image is a diff of
635 char backing_format
[16]; /* if non-zero and backing_file exists */
637 QDict
*full_open_options
;
638 char exact_filename
[PATH_MAX
];
646 /* Flags honored during pwrite (so far: BDRV_REQ_FUA) */
647 unsigned int supported_write_flags
;
648 /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA,
649 * BDRV_REQ_MAY_UNMAP) */
650 unsigned int supported_zero_flags
;
652 /* the following member gives a name to every node on the bs graph. */
654 /* element of the list of named nodes building the graph */
655 QTAILQ_ENTRY(BlockDriverState
) node_list
;
656 /* element of the list of all BlockDriverStates (all_bdrv_states) */
657 QTAILQ_ENTRY(BlockDriverState
) bs_list
;
658 /* element of the list of monitor-owned BDS */
659 QTAILQ_ENTRY(BlockDriverState
) monitor_list
;
662 /* operation blockers */
663 QLIST_HEAD(, BdrvOpBlocker
) op_blockers
[BLOCK_OP_TYPE_MAX
];
665 /* long-running background operation */
668 /* The node that this node inherited default options from (and a reopen on
669 * which can affect this node by changing these defaults). This is always a
670 * parent node of this node. */
671 BlockDriverState
*inherits_from
;
672 QLIST_HEAD(, BdrvChild
) children
;
673 QLIST_HEAD(, BdrvChild
) parents
;
676 QDict
*explicit_options
;
677 BlockdevDetectZeroesOptions detect_zeroes
;
679 /* The error object in use for blocking operations on backing_hd */
680 Error
*backing_blocker
;
682 /* Protected by AioContext lock */
684 /* If we are reading a disk image, give its size in sectors.
685 * Generally read-only; it is written to by load_snapshot and
686 * save_snaphost, but the block layer is quiescent during those.
688 int64_t total_sectors
;
690 /* Callback before write request is processed */
691 NotifierWithReturnList before_write_notifiers
;
693 /* threshold limit for writes, in bytes. "High water mark". */
694 uint64_t write_threshold_offset
;
695 NotifierWithReturn write_threshold_notifier
;
697 /* Writing to the list requires the BQL _and_ the dirty_bitmap_mutex.
698 * Reading from the list can be done with either the BQL or the
699 * dirty_bitmap_mutex. Modifying a bitmap only requires
700 * dirty_bitmap_mutex. */
701 QemuMutex dirty_bitmap_mutex
;
702 QLIST_HEAD(, BdrvDirtyBitmap
) dirty_bitmaps
;
704 /* Offset after the highest byte written to */
705 Stat64 wr_highest_offset
;
707 /* If true, copy read backing sectors into image. Can be >1 if more
708 * than one client has requested copy-on-read. Accessed with atomic
713 /* number of in-flight requests; overall and serialising.
714 * Accessed with atomic ops.
716 unsigned int in_flight
;
717 unsigned int serialising_in_flight
;
719 /* Internal to BDRV_POLL_WHILE and bdrv_wakeup. Accessed with atomic
724 /* counter for nested bdrv_io_plug.
725 * Accessed with atomic ops.
729 /* do we need to tell the quest if we have a volatile write cache? */
730 int enable_write_cache
;
732 /* Accessed with atomic ops. */
734 int recursive_quiesce_counter
;
736 unsigned int write_gen
; /* Current data generation */
738 /* Protected by reqs_lock. */
740 QLIST_HEAD(, BdrvTrackedRequest
) tracked_requests
;
741 CoQueue flush_queue
; /* Serializing flush queue */
742 bool active_flush_req
; /* Flush request in flight? */
744 /* Only read/written by whoever has set active_flush_req to true. */
745 unsigned int flushed_gen
; /* Flushed write generation */
748 struct BlockBackendRootState
{
751 BlockdevDetectZeroesOptions detect_zeroes
;
754 typedef enum BlockMirrorBackingMode
{
755 /* Reuse the existing backing chain from the source for the target.
756 * - sync=full: Set backing BDS to NULL.
757 * - sync=top: Use source's backing BDS.
758 * - sync=none: Use source as the backing BDS. */
759 MIRROR_SOURCE_BACKING_CHAIN
,
761 /* Open the target's backing chain completely anew */
762 MIRROR_OPEN_BACKING_CHAIN
,
764 /* Do not change the target's backing BDS after job completion */
765 MIRROR_LEAVE_BACKING_CHAIN
,
766 } BlockMirrorBackingMode
;
768 static inline BlockDriverState
*backing_bs(BlockDriverState
*bs
)
770 return bs
->backing
? bs
->backing
->bs
: NULL
;
774 /* Essential block drivers which must always be statically linked into qemu, and
775 * which therefore can be accessed without using bdrv_find_format() */
776 extern BlockDriver bdrv_file
;
777 extern BlockDriver bdrv_raw
;
778 extern BlockDriver bdrv_qcow2
;
780 int coroutine_fn
bdrv_co_preadv(BdrvChild
*child
,
781 int64_t offset
, unsigned int bytes
, QEMUIOVector
*qiov
,
782 BdrvRequestFlags flags
);
783 int coroutine_fn
bdrv_co_pwritev(BdrvChild
*child
,
784 int64_t offset
, unsigned int bytes
, QEMUIOVector
*qiov
,
785 BdrvRequestFlags flags
);
787 void bdrv_apply_subtree_drain(BdrvChild
*child
, BlockDriverState
*new_parent
);
788 void bdrv_unapply_subtree_drain(BdrvChild
*child
, BlockDriverState
*old_parent
);
790 int get_tmp_filename(char *filename
, int size
);
791 BlockDriver
*bdrv_probe_all(const uint8_t *buf
, int buf_size
,
792 const char *filename
);
794 void bdrv_parse_filename_strip_prefix(const char *filename
, const char *prefix
,
799 * bdrv_add_before_write_notifier:
801 * Register a callback that is invoked before write requests are processed but
802 * after any throttling or waiting for overlapping requests.
804 void bdrv_add_before_write_notifier(BlockDriverState
*bs
,
805 NotifierWithReturn
*notifier
);
808 * bdrv_detach_aio_context:
810 * May be called from .bdrv_detach_aio_context() to detach children from the
811 * current #AioContext. This is only needed by block drivers that manage their
812 * own children. Both ->file and ->backing are automatically handled and
813 * block drivers should not call this function on them explicitly.
815 void bdrv_detach_aio_context(BlockDriverState
*bs
);
818 * bdrv_attach_aio_context:
820 * May be called from .bdrv_attach_aio_context() to attach children to the new
821 * #AioContext. This is only needed by block drivers that manage their own
822 * children. Both ->file and ->backing are automatically handled and block
823 * drivers should not call this function on them explicitly.
825 void bdrv_attach_aio_context(BlockDriverState
*bs
,
826 AioContext
*new_context
);
829 * bdrv_add_aio_context_notifier:
831 * If a long-running job intends to be always run in the same AioContext as a
832 * certain BDS, it may use this function to be notified of changes regarding the
833 * association of the BDS to an AioContext.
835 * attached_aio_context() is called after the target BDS has been attached to a
836 * new AioContext; detach_aio_context() is called before the target BDS is being
837 * detached from its old AioContext.
839 void bdrv_add_aio_context_notifier(BlockDriverState
*bs
,
840 void (*attached_aio_context
)(AioContext
*new_context
, void *opaque
),
841 void (*detach_aio_context
)(void *opaque
), void *opaque
);
844 * bdrv_remove_aio_context_notifier:
846 * Unsubscribe of change notifications regarding the BDS's AioContext. The
847 * parameters given here have to be the same as those given to
848 * bdrv_add_aio_context_notifier().
850 void bdrv_remove_aio_context_notifier(BlockDriverState
*bs
,
851 void (*aio_context_attached
)(AioContext
*,
853 void (*aio_context_detached
)(void *),
858 * @bs: The BlockDriverState for which an I/O operation has been completed.
860 * Wake up the main thread if it is waiting on BDRV_POLL_WHILE. During
861 * synchronous I/O on a BlockDriverState that is attached to another
862 * I/O thread, the main thread lets the I/O thread's event loop run,
863 * waiting for the I/O operation to complete. A bdrv_wakeup will wake
864 * up the main thread if necessary.
866 * Manual calls to bdrv_wakeup are rarely necessary, because
867 * bdrv_dec_in_flight already calls it.
869 void bdrv_wakeup(BlockDriverState
*bs
);
872 int is_windows_drive(const char *filename
);
877 * @job_id: The id of the newly-created job, or %NULL to use the
878 * device name of @bs.
879 * @bs: Block device to operate on.
880 * @base: Block device that will become the new base, or %NULL to
881 * flatten the whole backing file chain onto @bs.
882 * @backing_file_str: The file name that will be written to @bs as the
883 * the new backing file if the job completes. Ignored if @base is %NULL.
884 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
885 * @on_error: The action to take upon error.
886 * @errp: Error object.
888 * Start a streaming operation on @bs. Clusters that are unallocated
889 * in @bs, but allocated in any image between @base and @bs (both
890 * exclusive) will be written to @bs. At the end of a successful
891 * streaming job, the backing file of @bs will be changed to
892 * @backing_file_str in the written image and to @base in the live
895 void stream_start(const char *job_id
, BlockDriverState
*bs
,
896 BlockDriverState
*base
, const char *backing_file_str
,
897 int64_t speed
, BlockdevOnError on_error
, Error
**errp
);
901 * @job_id: The id of the newly-created job, or %NULL to use the
902 * device name of @bs.
903 * @bs: Active block device.
904 * @top: Top block device to be committed.
905 * @base: Block device that will be written into, and become the new top.
906 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
907 * @on_error: The action to take upon error.
908 * @backing_file_str: String to use as the backing file in @top's overlay
909 * @filter_node_name: The node name that should be assigned to the filter
910 * driver that the commit job inserts into the graph above @top. NULL means
911 * that a node name should be autogenerated.
912 * @errp: Error object.
915 void commit_start(const char *job_id
, BlockDriverState
*bs
,
916 BlockDriverState
*base
, BlockDriverState
*top
, int64_t speed
,
917 BlockdevOnError on_error
, const char *backing_file_str
,
918 const char *filter_node_name
, Error
**errp
);
920 * commit_active_start:
921 * @job_id: The id of the newly-created job, or %NULL to use the
922 * device name of @bs.
923 * @bs: Active block device to be committed.
924 * @base: Block device that will be written into, and become the new top.
925 * @creation_flags: Flags that control the behavior of the Job lifetime.
926 * See @BlockJobCreateFlags
927 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
928 * @on_error: The action to take upon error.
929 * @filter_node_name: The node name that should be assigned to the filter
930 * driver that the commit job inserts into the graph above @bs. NULL means that
931 * a node name should be autogenerated.
932 * @cb: Completion function for the job.
933 * @opaque: Opaque pointer value passed to @cb.
934 * @auto_complete: Auto complete the job.
935 * @errp: Error object.
938 void commit_active_start(const char *job_id
, BlockDriverState
*bs
,
939 BlockDriverState
*base
, int creation_flags
,
940 int64_t speed
, BlockdevOnError on_error
,
941 const char *filter_node_name
,
942 BlockCompletionFunc
*cb
, void *opaque
,
943 bool auto_complete
, Error
**errp
);
946 * @job_id: The id of the newly-created job, or %NULL to use the
947 * device name of @bs.
948 * @bs: Block device to operate on.
949 * @target: Block device to write to.
950 * @replaces: Block graph node name to replace once the mirror is done. Can
951 * only be used when full mirroring is selected.
952 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
953 * @granularity: The chosen granularity for the dirty bitmap.
954 * @buf_size: The amount of data that can be in flight at one time.
955 * @mode: Whether to collapse all images in the chain to the target.
956 * @backing_mode: How to establish the target's backing chain after completion.
957 * @on_source_error: The action to take upon error reading from the source.
958 * @on_target_error: The action to take upon error writing to the target.
959 * @unmap: Whether to unmap target where source sectors only contain zeroes.
960 * @filter_node_name: The node name that should be assigned to the filter
961 * driver that the mirror job inserts into the graph above @bs. NULL means that
962 * a node name should be autogenerated.
963 * @errp: Error object.
965 * Start a mirroring operation on @bs. Clusters that are allocated
966 * in @bs will be written to @target until the job is cancelled or
967 * manually completed. At the end of a successful mirroring job,
968 * @bs will be switched to read from @target.
970 void mirror_start(const char *job_id
, BlockDriverState
*bs
,
971 BlockDriverState
*target
, const char *replaces
,
972 int64_t speed
, uint32_t granularity
, int64_t buf_size
,
973 MirrorSyncMode mode
, BlockMirrorBackingMode backing_mode
,
974 BlockdevOnError on_source_error
,
975 BlockdevOnError on_target_error
,
976 bool unmap
, const char *filter_node_name
, Error
**errp
);
980 * @job_id: The id of the newly-created job, or %NULL to use the
981 * device name of @bs.
982 * @bs: Block device to operate on.
983 * @target: Block device to write to.
984 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
985 * @sync_mode: What parts of the disk image should be copied to the destination.
986 * @sync_bitmap: The dirty bitmap if sync_mode is MIRROR_SYNC_MODE_INCREMENTAL.
987 * @on_source_error: The action to take upon error reading from the source.
988 * @on_target_error: The action to take upon error writing to the target.
989 * @creation_flags: Flags that control the behavior of the Job lifetime.
990 * See @BlockJobCreateFlags
991 * @cb: Completion function for the job.
992 * @opaque: Opaque pointer value passed to @cb.
993 * @txn: Transaction that this job is part of (may be NULL).
995 * Create a backup operation on @bs. Clusters in @bs are written to @target
996 * until the job is cancelled or manually completed.
998 BlockJob
*backup_job_create(const char *job_id
, BlockDriverState
*bs
,
999 BlockDriverState
*target
, int64_t speed
,
1000 MirrorSyncMode sync_mode
,
1001 BdrvDirtyBitmap
*sync_bitmap
,
1003 BlockdevOnError on_source_error
,
1004 BlockdevOnError on_target_error
,
1006 BlockCompletionFunc
*cb
, void *opaque
,
1007 BlockJobTxn
*txn
, Error
**errp
);
1009 void hmp_drive_add_node(Monitor
*mon
, const char *optstr
);
1011 BdrvChild
*bdrv_root_attach_child(BlockDriverState
*child_bs
,
1012 const char *child_name
,
1013 const BdrvChildRole
*child_role
,
1014 uint64_t perm
, uint64_t shared_perm
,
1015 void *opaque
, Error
**errp
);
1016 void bdrv_root_unref_child(BdrvChild
*child
);
1018 int bdrv_child_try_set_perm(BdrvChild
*c
, uint64_t perm
, uint64_t shared
,
1021 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1022 * block filters: Forward CONSISTENT_READ, WRITE, WRITE_UNCHANGED and RESIZE to
1024 void bdrv_filter_default_perms(BlockDriverState
*bs
, BdrvChild
*c
,
1025 const BdrvChildRole
*role
,
1026 BlockReopenQueue
*reopen_queue
,
1027 uint64_t perm
, uint64_t shared
,
1028 uint64_t *nperm
, uint64_t *nshared
);
1030 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1031 * (non-raw) image formats: Like above for bs->backing, but for bs->file it
1032 * requires WRITE | RESIZE for read-write images, always requires
1033 * CONSISTENT_READ and doesn't share WRITE. */
1034 void bdrv_format_default_perms(BlockDriverState
*bs
, BdrvChild
*c
,
1035 const BdrvChildRole
*role
,
1036 BlockReopenQueue
*reopen_queue
,
1037 uint64_t perm
, uint64_t shared
,
1038 uint64_t *nperm
, uint64_t *nshared
);
1041 * Default implementation for drivers to pass bdrv_co_block_status() to
1044 int coroutine_fn
bdrv_co_block_status_from_file(BlockDriverState
*bs
,
1050 BlockDriverState
**file
);
1052 * Default implementation for drivers to pass bdrv_co_block_status() to
1053 * their backing file.
1055 int coroutine_fn
bdrv_co_block_status_from_backing(BlockDriverState
*bs
,
1061 BlockDriverState
**file
);
1062 const char *bdrv_get_parent_name(const BlockDriverState
*bs
);
1063 void blk_dev_change_media_cb(BlockBackend
*blk
, bool load
, Error
**errp
);
1064 bool blk_dev_has_removable_media(BlockBackend
*blk
);
1065 bool blk_dev_has_tray(BlockBackend
*blk
);
1066 void blk_dev_eject_request(BlockBackend
*blk
, bool force
);
1067 bool blk_dev_is_tray_open(BlockBackend
*blk
);
1068 bool blk_dev_is_medium_locked(BlockBackend
*blk
);
1070 void bdrv_set_dirty(BlockDriverState
*bs
, int64_t offset
, int64_t bytes
);
1072 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap
*bitmap
, HBitmap
**out
);
1073 void bdrv_undo_clear_dirty_bitmap(BdrvDirtyBitmap
*bitmap
, HBitmap
*in
);
1075 void bdrv_inc_in_flight(BlockDriverState
*bs
);
1076 void bdrv_dec_in_flight(BlockDriverState
*bs
);
1078 void blockdev_close_all_bdrv_states(void);
1080 #endif /* BLOCK_INT_H */