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 "block/aio-wait.h"
30 #include "qemu/queue.h"
31 #include "qemu/coroutine.h"
32 #include "qemu/stats64.h"
33 #include "qemu/timer.h"
34 #include "qemu/hbitmap.h"
35 #include "block/snapshot.h"
36 #include "qemu/main-loop.h"
37 #include "qemu/throttle.h"
39 #define BLOCK_FLAG_LAZY_REFCOUNTS 8
41 #define BLOCK_OPT_SIZE "size"
42 #define BLOCK_OPT_ENCRYPT "encryption"
43 #define BLOCK_OPT_ENCRYPT_FORMAT "encrypt.format"
44 #define BLOCK_OPT_COMPAT6 "compat6"
45 #define BLOCK_OPT_HWVERSION "hwversion"
46 #define BLOCK_OPT_BACKING_FILE "backing_file"
47 #define BLOCK_OPT_BACKING_FMT "backing_fmt"
48 #define BLOCK_OPT_CLUSTER_SIZE "cluster_size"
49 #define BLOCK_OPT_TABLE_SIZE "table_size"
50 #define BLOCK_OPT_PREALLOC "preallocation"
51 #define BLOCK_OPT_SUBFMT "subformat"
52 #define BLOCK_OPT_COMPAT_LEVEL "compat"
53 #define BLOCK_OPT_LAZY_REFCOUNTS "lazy_refcounts"
54 #define BLOCK_OPT_ADAPTER_TYPE "adapter_type"
55 #define BLOCK_OPT_REDUNDANCY "redundancy"
56 #define BLOCK_OPT_NOCOW "nocow"
57 #define BLOCK_OPT_OBJECT_SIZE "object_size"
58 #define BLOCK_OPT_REFCOUNT_BITS "refcount_bits"
60 #define BLOCK_PROBE_BUF_SIZE 512
62 enum BdrvTrackedRequestType
{
68 typedef struct BdrvTrackedRequest
{
72 enum BdrvTrackedRequestType type
;
75 int64_t overlap_offset
;
76 unsigned int overlap_bytes
;
78 QLIST_ENTRY(BdrvTrackedRequest
) list
;
79 Coroutine
*co
; /* owner, used for deadlock detection */
80 CoQueue wait_queue
; /* coroutines blocked on this request */
82 struct BdrvTrackedRequest
*waiting_for
;
86 const char *format_name
;
89 /* set to true if the BlockDriver is a block filter. Block filters pass
90 * certain callbacks that refer to data (see block.c) to their bs->file if
91 * the driver doesn't implement them. Drivers that do not wish to forward
92 * must implement them and return -ENOTSUP.
95 /* for snapshots block filter like Quorum can implement the
96 * following recursive callback.
97 * It's purpose is to recurse on the filter children while calling
98 * bdrv_recurse_is_first_non_filter on them.
99 * For a sample implementation look in the future Quorum block filter.
101 bool (*bdrv_recurse_is_first_non_filter
)(BlockDriverState
*bs
,
102 BlockDriverState
*candidate
);
104 int (*bdrv_probe
)(const uint8_t *buf
, int buf_size
, const char *filename
);
105 int (*bdrv_probe_device
)(const char *filename
);
107 /* Any driver implementing this callback is expected to be able to handle
108 * NULL file names in its .bdrv_open() implementation */
109 void (*bdrv_parse_filename
)(const char *filename
, QDict
*options
, Error
**errp
);
110 /* Drivers not implementing bdrv_parse_filename nor bdrv_open should have
111 * this field set to true, except ones that are defined only by their
113 * An example of the last type will be the quorum block driver.
115 bool bdrv_needs_filename
;
117 /* Set if a driver can support backing files */
118 bool supports_backing
;
120 /* For handling image reopen for split or non-split files */
121 int (*bdrv_reopen_prepare
)(BDRVReopenState
*reopen_state
,
122 BlockReopenQueue
*queue
, Error
**errp
);
123 void (*bdrv_reopen_commit
)(BDRVReopenState
*reopen_state
);
124 void (*bdrv_reopen_abort
)(BDRVReopenState
*reopen_state
);
125 void (*bdrv_join_options
)(QDict
*options
, QDict
*old_options
);
127 int (*bdrv_open
)(BlockDriverState
*bs
, QDict
*options
, int flags
,
130 /* Protocol drivers should implement this instead of bdrv_open */
131 int (*bdrv_file_open
)(BlockDriverState
*bs
, QDict
*options
, int flags
,
133 void (*bdrv_close
)(BlockDriverState
*bs
);
134 int coroutine_fn (*bdrv_co_create
)(BlockdevCreateOptions
*opts
,
136 int coroutine_fn (*bdrv_co_create_opts
)(const char *filename
,
139 int (*bdrv_make_empty
)(BlockDriverState
*bs
);
141 void (*bdrv_refresh_filename
)(BlockDriverState
*bs
, QDict
*options
);
144 BlockAIOCB
*(*bdrv_aio_preadv
)(BlockDriverState
*bs
,
145 uint64_t offset
, uint64_t bytes
, QEMUIOVector
*qiov
, int flags
,
146 BlockCompletionFunc
*cb
, void *opaque
);
147 BlockAIOCB
*(*bdrv_aio_pwritev
)(BlockDriverState
*bs
,
148 uint64_t offset
, uint64_t bytes
, QEMUIOVector
*qiov
, int flags
,
149 BlockCompletionFunc
*cb
, void *opaque
);
150 BlockAIOCB
*(*bdrv_aio_flush
)(BlockDriverState
*bs
,
151 BlockCompletionFunc
*cb
, void *opaque
);
152 BlockAIOCB
*(*bdrv_aio_pdiscard
)(BlockDriverState
*bs
,
153 int64_t offset
, int bytes
,
154 BlockCompletionFunc
*cb
, void *opaque
);
156 int coroutine_fn (*bdrv_co_readv
)(BlockDriverState
*bs
,
157 int64_t sector_num
, int nb_sectors
, QEMUIOVector
*qiov
);
160 * @offset: position in bytes to read at
161 * @bytes: number of bytes to read
162 * @qiov: the buffers to fill with read data
163 * @flags: currently unused, always 0
165 * @offset and @bytes will be a multiple of 'request_alignment',
166 * but the length of individual @qiov elements does not have to
169 * @bytes will always equal the total size of @qiov, and will be
170 * no larger than 'max_transfer'.
172 * The buffer in @qiov may point directly to guest memory.
174 int coroutine_fn (*bdrv_co_preadv
)(BlockDriverState
*bs
,
175 uint64_t offset
, uint64_t bytes
, QEMUIOVector
*qiov
, int flags
);
176 int coroutine_fn (*bdrv_co_writev
)(BlockDriverState
*bs
,
177 int64_t sector_num
, int nb_sectors
, QEMUIOVector
*qiov
, int flags
);
179 * @offset: position in bytes to write at
180 * @bytes: number of bytes to write
181 * @qiov: the buffers containing data to write
182 * @flags: zero or more bits allowed by 'supported_write_flags'
184 * @offset and @bytes will be a multiple of 'request_alignment',
185 * but the length of individual @qiov elements does not have to
188 * @bytes will always equal the total size of @qiov, and will be
189 * no larger than 'max_transfer'.
191 * The buffer in @qiov may point directly to guest memory.
193 int coroutine_fn (*bdrv_co_pwritev
)(BlockDriverState
*bs
,
194 uint64_t offset
, uint64_t bytes
, QEMUIOVector
*qiov
, int flags
);
197 * Efficiently zero a region of the disk image. Typically an image format
198 * would use a compact metadata representation to implement this. This
199 * function pointer may be NULL or return -ENOSUP and .bdrv_co_writev()
200 * will be called instead.
202 int coroutine_fn (*bdrv_co_pwrite_zeroes
)(BlockDriverState
*bs
,
203 int64_t offset
, int bytes
, BdrvRequestFlags flags
);
204 int coroutine_fn (*bdrv_co_pdiscard
)(BlockDriverState
*bs
,
205 int64_t offset
, int bytes
);
208 * Building block for bdrv_block_status[_above] and
209 * bdrv_is_allocated[_above]. The driver should answer only
210 * according to the current layer, and should only need to set
211 * BDRV_BLOCK_DATA, BDRV_BLOCK_ZERO, BDRV_BLOCK_OFFSET_VALID,
212 * and/or BDRV_BLOCK_RAW; if the current layer defers to a backing
213 * layer, the result should be 0 (and not BDRV_BLOCK_ZERO). See
214 * block.h for the overall meaning of the bits. As a hint, the
215 * flag want_zero is true if the caller cares more about precise
216 * mappings (favor accurate _OFFSET_VALID/_ZERO) or false for
217 * overall allocation (favor larger *pnum, perhaps by reporting
218 * _DATA instead of _ZERO). The block layer guarantees input
219 * clamped to bdrv_getlength() and aligned to request_alignment,
220 * as well as non-NULL pnum, map, and file; in turn, the driver
221 * must return an error or set pnum to an aligned non-zero value.
223 int coroutine_fn (*bdrv_co_block_status
)(BlockDriverState
*bs
,
224 bool want_zero
, int64_t offset
, int64_t bytes
, int64_t *pnum
,
225 int64_t *map
, BlockDriverState
**file
);
228 * Invalidate any cached meta-data.
230 void coroutine_fn (*bdrv_co_invalidate_cache
)(BlockDriverState
*bs
,
232 int (*bdrv_inactivate
)(BlockDriverState
*bs
);
235 * Flushes all data for all layers by calling bdrv_co_flush for underlying
236 * layers, if needed. This function is needed for deterministic
237 * synchronization of the flush finishing callback.
239 int coroutine_fn (*bdrv_co_flush
)(BlockDriverState
*bs
);
242 * Flushes all data that was already written to the OS all the way down to
243 * the disk (for example file-posix.c calls fsync()).
245 int coroutine_fn (*bdrv_co_flush_to_disk
)(BlockDriverState
*bs
);
248 * Flushes all internal caches to the OS. The data may still sit in a
249 * writeback cache of the host OS, but it will survive a crash of the qemu
252 int coroutine_fn (*bdrv_co_flush_to_os
)(BlockDriverState
*bs
);
255 * Drivers setting this field must be able to work with just a plain
256 * filename with '<protocol_name>:' as a prefix, and no other options.
257 * Options may be extracted from the filename by implementing
258 * bdrv_parse_filename.
260 const char *protocol_name
;
261 int (*bdrv_truncate
)(BlockDriverState
*bs
, int64_t offset
,
262 PreallocMode prealloc
, Error
**errp
);
264 int64_t (*bdrv_getlength
)(BlockDriverState
*bs
);
265 bool has_variable_length
;
266 int64_t (*bdrv_get_allocated_file_size
)(BlockDriverState
*bs
);
267 BlockMeasureInfo
*(*bdrv_measure
)(QemuOpts
*opts
, BlockDriverState
*in_bs
,
270 int coroutine_fn (*bdrv_co_pwritev_compressed
)(BlockDriverState
*bs
,
271 uint64_t offset
, uint64_t bytes
, QEMUIOVector
*qiov
);
273 int (*bdrv_snapshot_create
)(BlockDriverState
*bs
,
274 QEMUSnapshotInfo
*sn_info
);
275 int (*bdrv_snapshot_goto
)(BlockDriverState
*bs
,
276 const char *snapshot_id
);
277 int (*bdrv_snapshot_delete
)(BlockDriverState
*bs
,
278 const char *snapshot_id
,
281 int (*bdrv_snapshot_list
)(BlockDriverState
*bs
,
282 QEMUSnapshotInfo
**psn_info
);
283 int (*bdrv_snapshot_load_tmp
)(BlockDriverState
*bs
,
284 const char *snapshot_id
,
287 int (*bdrv_get_info
)(BlockDriverState
*bs
, BlockDriverInfo
*bdi
);
288 ImageInfoSpecific
*(*bdrv_get_specific_info
)(BlockDriverState
*bs
);
290 int coroutine_fn (*bdrv_save_vmstate
)(BlockDriverState
*bs
,
293 int coroutine_fn (*bdrv_load_vmstate
)(BlockDriverState
*bs
,
297 int (*bdrv_change_backing_file
)(BlockDriverState
*bs
,
298 const char *backing_file
, const char *backing_fmt
);
300 /* removable device specific */
301 bool (*bdrv_is_inserted
)(BlockDriverState
*bs
);
302 void (*bdrv_eject
)(BlockDriverState
*bs
, bool eject_flag
);
303 void (*bdrv_lock_medium
)(BlockDriverState
*bs
, bool locked
);
305 /* to control generic scsi devices */
306 BlockAIOCB
*(*bdrv_aio_ioctl
)(BlockDriverState
*bs
,
307 unsigned long int req
, void *buf
,
308 BlockCompletionFunc
*cb
, void *opaque
);
309 int coroutine_fn (*bdrv_co_ioctl
)(BlockDriverState
*bs
,
310 unsigned long int req
, void *buf
);
312 /* List of options for creating images, terminated by name == NULL */
313 QemuOptsList
*create_opts
;
316 * Returns 0 for completed check, -errno for internal errors.
317 * The check results are stored in result.
319 int coroutine_fn (*bdrv_co_check
)(BlockDriverState
*bs
,
320 BdrvCheckResult
*result
,
323 int (*bdrv_amend_options
)(BlockDriverState
*bs
, QemuOpts
*opts
,
324 BlockDriverAmendStatusCB
*status_cb
,
327 void (*bdrv_debug_event
)(BlockDriverState
*bs
, BlkdebugEvent event
);
329 /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
330 int (*bdrv_debug_breakpoint
)(BlockDriverState
*bs
, const char *event
,
332 int (*bdrv_debug_remove_breakpoint
)(BlockDriverState
*bs
,
334 int (*bdrv_debug_resume
)(BlockDriverState
*bs
, const char *tag
);
335 bool (*bdrv_debug_is_suspended
)(BlockDriverState
*bs
, const char *tag
);
337 void (*bdrv_refresh_limits
)(BlockDriverState
*bs
, Error
**errp
);
340 * Returns 1 if newly created images are guaranteed to contain only
341 * zeros, 0 otherwise.
343 int (*bdrv_has_zero_init
)(BlockDriverState
*bs
);
345 /* Remove fd handlers, timers, and other event loop callbacks so the event
346 * loop is no longer in use. Called with no in-flight requests and in
347 * depth-first traversal order with parents before child nodes.
349 void (*bdrv_detach_aio_context
)(BlockDriverState
*bs
);
351 /* Add fd handlers, timers, and other event loop callbacks so I/O requests
352 * can be processed again. Called with no in-flight requests and in
353 * depth-first traversal order with child nodes before parent nodes.
355 void (*bdrv_attach_aio_context
)(BlockDriverState
*bs
,
356 AioContext
*new_context
);
358 /* io queue for linux-aio */
359 void (*bdrv_io_plug
)(BlockDriverState
*bs
);
360 void (*bdrv_io_unplug
)(BlockDriverState
*bs
);
363 * Try to get @bs's logical and physical block size.
364 * On success, store them in @bsz and return zero.
365 * On failure, return negative errno.
367 int (*bdrv_probe_blocksizes
)(BlockDriverState
*bs
, BlockSizes
*bsz
);
369 * Try to get @bs's geometry (cyls, heads, sectors)
370 * On success, store them in @geo and return 0.
371 * On failure return -errno.
372 * Only drivers that want to override guest geometry implement this
373 * callback; see hd_geometry_guess().
375 int (*bdrv_probe_geometry
)(BlockDriverState
*bs
, HDGeometry
*geo
);
378 * bdrv_co_drain_begin is called if implemented in the beginning of a
379 * drain operation to drain and stop any internal sources of requests in
381 * bdrv_co_drain_end is called if implemented at the end of the drain.
383 * They should be used by the driver to e.g. manage scheduled I/O
384 * requests, or toggle an internal state. After the end of the drain new
385 * requests will continue normally.
387 void coroutine_fn (*bdrv_co_drain_begin
)(BlockDriverState
*bs
);
388 void coroutine_fn (*bdrv_co_drain_end
)(BlockDriverState
*bs
);
390 void (*bdrv_add_child
)(BlockDriverState
*parent
, BlockDriverState
*child
,
392 void (*bdrv_del_child
)(BlockDriverState
*parent
, BdrvChild
*child
,
396 * Informs the block driver that a permission change is intended. The
397 * driver checks whether the change is permissible and may take other
398 * preparations for the change (e.g. get file system locks). This operation
399 * is always followed either by a call to either .bdrv_set_perm or
400 * .bdrv_abort_perm_update.
402 * Checks whether the requested set of cumulative permissions in @perm
403 * can be granted for accessing @bs and whether no other users are using
404 * permissions other than those given in @shared (both arguments take
405 * BLK_PERM_* bitmasks).
407 * If both conditions are met, 0 is returned. Otherwise, -errno is returned
408 * and errp is set to an error describing the conflict.
410 int (*bdrv_check_perm
)(BlockDriverState
*bs
, uint64_t perm
,
411 uint64_t shared
, Error
**errp
);
414 * Called to inform the driver that the set of cumulative set of used
415 * permissions for @bs has changed to @perm, and the set of sharable
416 * permission to @shared. The driver can use this to propagate changes to
417 * its children (i.e. request permissions only if a parent actually needs
420 * This function is only invoked after bdrv_check_perm(), so block drivers
421 * may rely on preparations made in their .bdrv_check_perm implementation.
423 void (*bdrv_set_perm
)(BlockDriverState
*bs
, uint64_t perm
, uint64_t shared
);
426 * Called to inform the driver that after a previous bdrv_check_perm()
427 * call, the permission update is not performed and any preparations made
428 * for it (e.g. taken file locks) need to be undone.
430 * This function can be called even for nodes that never saw a
431 * bdrv_check_perm() call. It is a no-op then.
433 void (*bdrv_abort_perm_update
)(BlockDriverState
*bs
);
436 * Returns in @nperm and @nshared the permissions that the driver for @bs
437 * needs on its child @c, based on the cumulative permissions requested by
438 * the parents in @parent_perm and @parent_shared.
440 * If @c is NULL, return the permissions for attaching a new child for the
443 * If @reopen_queue is non-NULL, don't return the currently needed
444 * permissions, but those that will be needed after applying the
447 void (*bdrv_child_perm
)(BlockDriverState
*bs
, BdrvChild
*c
,
448 const BdrvChildRole
*role
,
449 BlockReopenQueue
*reopen_queue
,
450 uint64_t parent_perm
, uint64_t parent_shared
,
451 uint64_t *nperm
, uint64_t *nshared
);
454 * Bitmaps should be marked as 'IN_USE' in the image on reopening image
455 * as rw. This handler should realize it. It also should unset readonly
456 * field of BlockDirtyBitmap's in case of success.
458 int (*bdrv_reopen_bitmaps_rw
)(BlockDriverState
*bs
, Error
**errp
);
459 bool (*bdrv_can_store_new_dirty_bitmap
)(BlockDriverState
*bs
,
461 uint32_t granularity
,
463 void (*bdrv_remove_persistent_dirty_bitmap
)(BlockDriverState
*bs
,
468 * Register/unregister a buffer for I/O. For example, when the driver is
469 * interested to know the memory areas that will later be used in iovs, so
470 * that it can do IOMMU mapping with VFIO etc., in order to get better
471 * performance. In the case of VFIO drivers, this callback is used to do
472 * DMA mapping for hot buffers.
474 void (*bdrv_register_buf
)(BlockDriverState
*bs
, void *host
, size_t size
);
475 void (*bdrv_unregister_buf
)(BlockDriverState
*bs
, void *host
);
476 QLIST_ENTRY(BlockDriver
) list
;
479 typedef struct BlockLimits
{
480 /* Alignment requirement, in bytes, for offset/length of I/O
481 * requests. Must be a power of 2 less than INT_MAX; defaults to
482 * 1 for drivers with modern byte interfaces, and to 512
484 uint32_t request_alignment
;
486 /* Maximum number of bytes that can be discarded at once (since it
487 * is signed, it must be < 2G, if set). Must be multiple of
488 * pdiscard_alignment, but need not be power of 2. May be 0 if no
489 * inherent 32-bit limit */
490 int32_t max_pdiscard
;
492 /* Optimal alignment for discard requests in bytes. A power of 2
493 * is best but not mandatory. Must be a multiple of
494 * bl.request_alignment, and must be less than max_pdiscard if
495 * that is set. May be 0 if bl.request_alignment is good enough */
496 uint32_t pdiscard_alignment
;
498 /* Maximum number of bytes that can zeroized at once (since it is
499 * signed, it must be < 2G, if set). Must be multiple of
500 * pwrite_zeroes_alignment. May be 0 if no inherent 32-bit limit */
501 int32_t max_pwrite_zeroes
;
503 /* Optimal alignment for write zeroes requests in bytes. A power
504 * of 2 is best but not mandatory. Must be a multiple of
505 * bl.request_alignment, and must be less than max_pwrite_zeroes
506 * if that is set. May be 0 if bl.request_alignment is good
508 uint32_t pwrite_zeroes_alignment
;
510 /* Optimal transfer length in bytes. A power of 2 is best but not
511 * mandatory. Must be a multiple of bl.request_alignment, or 0 if
512 * no preferred size */
513 uint32_t opt_transfer
;
515 /* Maximal transfer length in bytes. Need not be power of 2, but
516 * must be multiple of opt_transfer and bl.request_alignment, or 0
517 * for no 32-bit limit. For now, anything larger than INT_MAX is
519 uint32_t max_transfer
;
521 /* memory alignment, in bytes so that no bounce buffer is needed */
522 size_t min_mem_alignment
;
524 /* memory alignment, in bytes, for bounce buffer */
525 size_t opt_mem_alignment
;
527 /* maximum number of iovec elements */
531 typedef struct BdrvOpBlocker BdrvOpBlocker
;
533 typedef struct BdrvAioNotifier
{
534 void (*attached_aio_context
)(AioContext
*new_context
, void *opaque
);
535 void (*detach_aio_context
)(void *opaque
);
540 QLIST_ENTRY(BdrvAioNotifier
) list
;
543 struct BdrvChildRole
{
544 /* If true, bdrv_replace_node() doesn't change the node this BdrvChild
548 void (*inherit_options
)(int *child_flags
, QDict
*child_options
,
549 int parent_flags
, QDict
*parent_options
);
551 void (*change_media
)(BdrvChild
*child
, bool load
);
552 void (*resize
)(BdrvChild
*child
);
554 /* Returns a name that is supposedly more useful for human users than the
555 * node name for identifying the node in question (in particular, a BB
556 * name), or NULL if the parent can't provide a better name. */
557 const char *(*get_name
)(BdrvChild
*child
);
559 /* Returns a malloced string that describes the parent of the child for a
560 * human reader. This could be a node-name, BlockBackend name, qdev ID or
561 * QOM path of the device owning the BlockBackend, job type and ID etc. The
562 * caller is responsible for freeing the memory. */
563 char *(*get_parent_desc
)(BdrvChild
*child
);
566 * If this pair of functions is implemented, the parent doesn't issue new
567 * requests after returning from .drained_begin() until .drained_end() is
570 * Note that this can be nested. If drained_begin() was called twice, new
571 * I/O is allowed only after drained_end() was called twice, too.
573 void (*drained_begin
)(BdrvChild
*child
);
574 void (*drained_end
)(BdrvChild
*child
);
576 /* Notifies the parent that the child has been activated/inactivated (e.g.
577 * when migration is completing) and it can start/stop requesting
578 * permissions and doing I/O on it. */
579 void (*activate
)(BdrvChild
*child
, Error
**errp
);
580 int (*inactivate
)(BdrvChild
*child
);
582 void (*attach
)(BdrvChild
*child
);
583 void (*detach
)(BdrvChild
*child
);
585 /* Notifies the parent that the filename of its child has changed (e.g.
586 * because the direct child was removed from the backing chain), so that it
587 * can update its reference. */
588 int (*update_filename
)(BdrvChild
*child
, BlockDriverState
*new_base
,
589 const char *filename
, Error
**errp
);
592 extern const BdrvChildRole child_file
;
593 extern const BdrvChildRole child_format
;
594 extern const BdrvChildRole child_backing
;
597 BlockDriverState
*bs
;
599 const BdrvChildRole
*role
;
603 * Granted permissions for operating on this BdrvChild (BLK_PERM_* bitmask)
608 * Permissions that can still be granted to other users of @bs while this
609 * BdrvChild is still attached to it. (BLK_PERM_* bitmask)
611 uint64_t shared_perm
;
613 QLIST_ENTRY(BdrvChild
) next
;
614 QLIST_ENTRY(BdrvChild
) next_parent
;
618 * Note: the function bdrv_append() copies and swaps contents of
619 * BlockDriverStates, so if you add new fields to this struct, please
620 * inspect bdrv_append() to determine if the new fields need to be
623 struct BlockDriverState
{
624 /* Protected by big QEMU lock or read-only after opening. No special
625 * locking needed during I/O...
627 int open_flags
; /* flags used to open the file, re-used for re-open */
628 bool read_only
; /* if true, the media is read only */
629 bool encrypted
; /* if true, the media is encrypted */
630 bool sg
; /* if true, the device is a /dev/sg* */
631 bool probed
; /* if true, format was probed rather than specified */
632 bool force_share
; /* if true, always allow all shared permissions */
633 bool implicit
; /* if true, this filter node was automatically inserted */
635 BlockDriver
*drv
; /* NULL means no media */
638 AioContext
*aio_context
; /* event loop used for fd handlers, timers, etc */
639 /* long-running tasks intended to always use the same AioContext as this
640 * BDS may register themselves in this list to be notified of changes
641 * regarding this BDS's context */
642 QLIST_HEAD(, BdrvAioNotifier
) aio_notifiers
;
643 bool walking_aio_notifiers
; /* to make removal during iteration safe */
645 char filename
[PATH_MAX
];
646 char backing_file
[PATH_MAX
]; /* if non zero, the image is a diff of
648 char backing_format
[16]; /* if non-zero and backing_file exists */
650 QDict
*full_open_options
;
651 char exact_filename
[PATH_MAX
];
659 /* Flags honored during pwrite (so far: BDRV_REQ_FUA) */
660 unsigned int supported_write_flags
;
661 /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA,
662 * BDRV_REQ_MAY_UNMAP) */
663 unsigned int supported_zero_flags
;
665 /* the following member gives a name to every node on the bs graph. */
667 /* element of the list of named nodes building the graph */
668 QTAILQ_ENTRY(BlockDriverState
) node_list
;
669 /* element of the list of all BlockDriverStates (all_bdrv_states) */
670 QTAILQ_ENTRY(BlockDriverState
) bs_list
;
671 /* element of the list of monitor-owned BDS */
672 QTAILQ_ENTRY(BlockDriverState
) monitor_list
;
675 /* operation blockers */
676 QLIST_HEAD(, BdrvOpBlocker
) op_blockers
[BLOCK_OP_TYPE_MAX
];
678 /* long-running background operation */
681 /* The node that this node inherited default options from (and a reopen on
682 * which can affect this node by changing these defaults). This is always a
683 * parent node of this node. */
684 BlockDriverState
*inherits_from
;
685 QLIST_HEAD(, BdrvChild
) children
;
686 QLIST_HEAD(, BdrvChild
) parents
;
689 QDict
*explicit_options
;
690 BlockdevDetectZeroesOptions detect_zeroes
;
692 /* The error object in use for blocking operations on backing_hd */
693 Error
*backing_blocker
;
695 /* Protected by AioContext lock */
697 /* If we are reading a disk image, give its size in sectors.
698 * Generally read-only; it is written to by load_snapshot and
699 * save_snaphost, but the block layer is quiescent during those.
701 int64_t total_sectors
;
703 /* Callback before write request is processed */
704 NotifierWithReturnList before_write_notifiers
;
706 /* threshold limit for writes, in bytes. "High water mark". */
707 uint64_t write_threshold_offset
;
708 NotifierWithReturn write_threshold_notifier
;
710 /* Writing to the list requires the BQL _and_ the dirty_bitmap_mutex.
711 * Reading from the list can be done with either the BQL or the
712 * dirty_bitmap_mutex. Modifying a bitmap only requires
713 * dirty_bitmap_mutex. */
714 QemuMutex dirty_bitmap_mutex
;
715 QLIST_HEAD(, BdrvDirtyBitmap
) dirty_bitmaps
;
717 /* Offset after the highest byte written to */
718 Stat64 wr_highest_offset
;
720 /* If true, copy read backing sectors into image. Can be >1 if more
721 * than one client has requested copy-on-read. Accessed with atomic
726 /* number of in-flight requests; overall and serialising.
727 * Accessed with atomic ops.
729 unsigned int in_flight
;
730 unsigned int serialising_in_flight
;
732 /* Kicked to signal main loop when a request completes. */
735 /* counter for nested bdrv_io_plug.
736 * Accessed with atomic ops.
740 /* do we need to tell the quest if we have a volatile write cache? */
741 int enable_write_cache
;
743 /* Accessed with atomic ops. */
745 int recursive_quiesce_counter
;
747 unsigned int write_gen
; /* Current data generation */
749 /* Protected by reqs_lock. */
751 QLIST_HEAD(, BdrvTrackedRequest
) tracked_requests
;
752 CoQueue flush_queue
; /* Serializing flush queue */
753 bool active_flush_req
; /* Flush request in flight? */
755 /* Only read/written by whoever has set active_flush_req to true. */
756 unsigned int flushed_gen
; /* Flushed write generation */
759 struct BlockBackendRootState
{
762 BlockdevDetectZeroesOptions detect_zeroes
;
765 typedef enum BlockMirrorBackingMode
{
766 /* Reuse the existing backing chain from the source for the target.
767 * - sync=full: Set backing BDS to NULL.
768 * - sync=top: Use source's backing BDS.
769 * - sync=none: Use source as the backing BDS. */
770 MIRROR_SOURCE_BACKING_CHAIN
,
772 /* Open the target's backing chain completely anew */
773 MIRROR_OPEN_BACKING_CHAIN
,
775 /* Do not change the target's backing BDS after job completion */
776 MIRROR_LEAVE_BACKING_CHAIN
,
777 } BlockMirrorBackingMode
;
779 static inline BlockDriverState
*backing_bs(BlockDriverState
*bs
)
781 return bs
->backing
? bs
->backing
->bs
: NULL
;
785 /* Essential block drivers which must always be statically linked into qemu, and
786 * which therefore can be accessed without using bdrv_find_format() */
787 extern BlockDriver bdrv_file
;
788 extern BlockDriver bdrv_raw
;
789 extern BlockDriver bdrv_qcow2
;
791 int coroutine_fn
bdrv_co_preadv(BdrvChild
*child
,
792 int64_t offset
, unsigned int bytes
, QEMUIOVector
*qiov
,
793 BdrvRequestFlags flags
);
794 int coroutine_fn
bdrv_co_pwritev(BdrvChild
*child
,
795 int64_t offset
, unsigned int bytes
, QEMUIOVector
*qiov
,
796 BdrvRequestFlags flags
);
798 void bdrv_apply_subtree_drain(BdrvChild
*child
, BlockDriverState
*new_parent
);
799 void bdrv_unapply_subtree_drain(BdrvChild
*child
, BlockDriverState
*old_parent
);
801 int get_tmp_filename(char *filename
, int size
);
802 BlockDriver
*bdrv_probe_all(const uint8_t *buf
, int buf_size
,
803 const char *filename
);
805 void bdrv_parse_filename_strip_prefix(const char *filename
, const char *prefix
,
810 * bdrv_add_before_write_notifier:
812 * Register a callback that is invoked before write requests are processed but
813 * after any throttling or waiting for overlapping requests.
815 void bdrv_add_before_write_notifier(BlockDriverState
*bs
,
816 NotifierWithReturn
*notifier
);
819 * bdrv_detach_aio_context:
821 * May be called from .bdrv_detach_aio_context() to detach children from the
822 * current #AioContext. This is only needed by block drivers that manage their
823 * own children. Both ->file and ->backing are automatically handled and
824 * block drivers should not call this function on them explicitly.
826 void bdrv_detach_aio_context(BlockDriverState
*bs
);
829 * bdrv_attach_aio_context:
831 * May be called from .bdrv_attach_aio_context() to attach children to the new
832 * #AioContext. This is only needed by block drivers that manage their own
833 * children. Both ->file and ->backing are automatically handled and block
834 * drivers should not call this function on them explicitly.
836 void bdrv_attach_aio_context(BlockDriverState
*bs
,
837 AioContext
*new_context
);
840 * bdrv_add_aio_context_notifier:
842 * If a long-running job intends to be always run in the same AioContext as a
843 * certain BDS, it may use this function to be notified of changes regarding the
844 * association of the BDS to an AioContext.
846 * attached_aio_context() is called after the target BDS has been attached to a
847 * new AioContext; detach_aio_context() is called before the target BDS is being
848 * detached from its old AioContext.
850 void bdrv_add_aio_context_notifier(BlockDriverState
*bs
,
851 void (*attached_aio_context
)(AioContext
*new_context
, void *opaque
),
852 void (*detach_aio_context
)(void *opaque
), void *opaque
);
855 * bdrv_remove_aio_context_notifier:
857 * Unsubscribe of change notifications regarding the BDS's AioContext. The
858 * parameters given here have to be the same as those given to
859 * bdrv_add_aio_context_notifier().
861 void bdrv_remove_aio_context_notifier(BlockDriverState
*bs
,
862 void (*aio_context_attached
)(AioContext
*,
864 void (*aio_context_detached
)(void *),
869 * @bs: The BlockDriverState for which an I/O operation has been completed.
871 * Wake up the main thread if it is waiting on BDRV_POLL_WHILE. During
872 * synchronous I/O on a BlockDriverState that is attached to another
873 * I/O thread, the main thread lets the I/O thread's event loop run,
874 * waiting for the I/O operation to complete. A bdrv_wakeup will wake
875 * up the main thread if necessary.
877 * Manual calls to bdrv_wakeup are rarely necessary, because
878 * bdrv_dec_in_flight already calls it.
880 void bdrv_wakeup(BlockDriverState
*bs
);
883 int is_windows_drive(const char *filename
);
888 * @job_id: The id of the newly-created job, or %NULL to use the
889 * device name of @bs.
890 * @bs: Block device to operate on.
891 * @base: Block device that will become the new base, or %NULL to
892 * flatten the whole backing file chain onto @bs.
893 * @backing_file_str: The file name that will be written to @bs as the
894 * the new backing file if the job completes. Ignored if @base is %NULL.
895 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
896 * @on_error: The action to take upon error.
897 * @errp: Error object.
899 * Start a streaming operation on @bs. Clusters that are unallocated
900 * in @bs, but allocated in any image between @base and @bs (both
901 * exclusive) will be written to @bs. At the end of a successful
902 * streaming job, the backing file of @bs will be changed to
903 * @backing_file_str in the written image and to @base in the live
906 void stream_start(const char *job_id
, BlockDriverState
*bs
,
907 BlockDriverState
*base
, const char *backing_file_str
,
908 int64_t speed
, BlockdevOnError on_error
, Error
**errp
);
912 * @job_id: The id of the newly-created job, or %NULL to use the
913 * device name of @bs.
914 * @bs: Active block device.
915 * @top: Top block device to be committed.
916 * @base: Block device that will be written into, and become the new top.
917 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
918 * @on_error: The action to take upon error.
919 * @backing_file_str: String to use as the backing file in @top's overlay
920 * @filter_node_name: The node name that should be assigned to the filter
921 * driver that the commit job inserts into the graph above @top. NULL means
922 * that a node name should be autogenerated.
923 * @errp: Error object.
926 void commit_start(const char *job_id
, BlockDriverState
*bs
,
927 BlockDriverState
*base
, BlockDriverState
*top
, int64_t speed
,
928 BlockdevOnError on_error
, const char *backing_file_str
,
929 const char *filter_node_name
, Error
**errp
);
931 * commit_active_start:
932 * @job_id: The id of the newly-created job, or %NULL to use the
933 * device name of @bs.
934 * @bs: Active block device to be committed.
935 * @base: Block device that will be written into, and become the new top.
936 * @creation_flags: Flags that control the behavior of the Job lifetime.
937 * See @BlockJobCreateFlags
938 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
939 * @on_error: The action to take upon error.
940 * @filter_node_name: The node name that should be assigned to the filter
941 * driver that the commit job inserts into the graph above @bs. NULL means that
942 * a node name should be autogenerated.
943 * @cb: Completion function for the job.
944 * @opaque: Opaque pointer value passed to @cb.
945 * @auto_complete: Auto complete the job.
946 * @errp: Error object.
949 void commit_active_start(const char *job_id
, BlockDriverState
*bs
,
950 BlockDriverState
*base
, int creation_flags
,
951 int64_t speed
, BlockdevOnError on_error
,
952 const char *filter_node_name
,
953 BlockCompletionFunc
*cb
, void *opaque
,
954 bool auto_complete
, Error
**errp
);
957 * @job_id: The id of the newly-created job, or %NULL to use the
958 * device name of @bs.
959 * @bs: Block device to operate on.
960 * @target: Block device to write to.
961 * @replaces: Block graph node name to replace once the mirror is done. Can
962 * only be used when full mirroring is selected.
963 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
964 * @granularity: The chosen granularity for the dirty bitmap.
965 * @buf_size: The amount of data that can be in flight at one time.
966 * @mode: Whether to collapse all images in the chain to the target.
967 * @backing_mode: How to establish the target's backing chain after completion.
968 * @on_source_error: The action to take upon error reading from the source.
969 * @on_target_error: The action to take upon error writing to the target.
970 * @unmap: Whether to unmap target where source sectors only contain zeroes.
971 * @filter_node_name: The node name that should be assigned to the filter
972 * driver that the mirror job inserts into the graph above @bs. NULL means that
973 * a node name should be autogenerated.
974 * @errp: Error object.
976 * Start a mirroring operation on @bs. Clusters that are allocated
977 * in @bs will be written to @target until the job is cancelled or
978 * manually completed. At the end of a successful mirroring job,
979 * @bs will be switched to read from @target.
981 void mirror_start(const char *job_id
, BlockDriverState
*bs
,
982 BlockDriverState
*target
, const char *replaces
,
983 int64_t speed
, uint32_t granularity
, int64_t buf_size
,
984 MirrorSyncMode mode
, BlockMirrorBackingMode backing_mode
,
985 BlockdevOnError on_source_error
,
986 BlockdevOnError on_target_error
,
987 bool unmap
, const char *filter_node_name
, Error
**errp
);
991 * @job_id: The id of the newly-created job, or %NULL to use the
992 * device name of @bs.
993 * @bs: Block device to operate on.
994 * @target: Block device to write to.
995 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
996 * @sync_mode: What parts of the disk image should be copied to the destination.
997 * @sync_bitmap: The dirty bitmap if sync_mode is MIRROR_SYNC_MODE_INCREMENTAL.
998 * @on_source_error: The action to take upon error reading from the source.
999 * @on_target_error: The action to take upon error writing to the target.
1000 * @creation_flags: Flags that control the behavior of the Job lifetime.
1001 * See @BlockJobCreateFlags
1002 * @cb: Completion function for the job.
1003 * @opaque: Opaque pointer value passed to @cb.
1004 * @txn: Transaction that this job is part of (may be NULL).
1006 * Create a backup operation on @bs. Clusters in @bs are written to @target
1007 * until the job is cancelled or manually completed.
1009 BlockJob
*backup_job_create(const char *job_id
, BlockDriverState
*bs
,
1010 BlockDriverState
*target
, int64_t speed
,
1011 MirrorSyncMode sync_mode
,
1012 BdrvDirtyBitmap
*sync_bitmap
,
1014 BlockdevOnError on_source_error
,
1015 BlockdevOnError on_target_error
,
1017 BlockCompletionFunc
*cb
, void *opaque
,
1018 BlockJobTxn
*txn
, Error
**errp
);
1020 void hmp_drive_add_node(Monitor
*mon
, const char *optstr
);
1022 BdrvChild
*bdrv_root_attach_child(BlockDriverState
*child_bs
,
1023 const char *child_name
,
1024 const BdrvChildRole
*child_role
,
1025 uint64_t perm
, uint64_t shared_perm
,
1026 void *opaque
, Error
**errp
);
1027 void bdrv_root_unref_child(BdrvChild
*child
);
1029 int bdrv_child_try_set_perm(BdrvChild
*c
, uint64_t perm
, uint64_t shared
,
1032 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1033 * block filters: Forward CONSISTENT_READ, WRITE, WRITE_UNCHANGED and RESIZE to
1035 void bdrv_filter_default_perms(BlockDriverState
*bs
, BdrvChild
*c
,
1036 const BdrvChildRole
*role
,
1037 BlockReopenQueue
*reopen_queue
,
1038 uint64_t perm
, uint64_t shared
,
1039 uint64_t *nperm
, uint64_t *nshared
);
1041 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1042 * (non-raw) image formats: Like above for bs->backing, but for bs->file it
1043 * requires WRITE | RESIZE for read-write images, always requires
1044 * CONSISTENT_READ and doesn't share WRITE. */
1045 void bdrv_format_default_perms(BlockDriverState
*bs
, BdrvChild
*c
,
1046 const BdrvChildRole
*role
,
1047 BlockReopenQueue
*reopen_queue
,
1048 uint64_t perm
, uint64_t shared
,
1049 uint64_t *nperm
, uint64_t *nshared
);
1052 * Default implementation for drivers to pass bdrv_co_block_status() to
1055 int coroutine_fn
bdrv_co_block_status_from_file(BlockDriverState
*bs
,
1061 BlockDriverState
**file
);
1063 * Default implementation for drivers to pass bdrv_co_block_status() to
1064 * their backing file.
1066 int coroutine_fn
bdrv_co_block_status_from_backing(BlockDriverState
*bs
,
1072 BlockDriverState
**file
);
1073 const char *bdrv_get_parent_name(const BlockDriverState
*bs
);
1074 void blk_dev_change_media_cb(BlockBackend
*blk
, bool load
, Error
**errp
);
1075 bool blk_dev_has_removable_media(BlockBackend
*blk
);
1076 bool blk_dev_has_tray(BlockBackend
*blk
);
1077 void blk_dev_eject_request(BlockBackend
*blk
, bool force
);
1078 bool blk_dev_is_tray_open(BlockBackend
*blk
);
1079 bool blk_dev_is_medium_locked(BlockBackend
*blk
);
1081 void bdrv_set_dirty(BlockDriverState
*bs
, int64_t offset
, int64_t bytes
);
1083 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap
*bitmap
, HBitmap
**out
);
1084 void bdrv_undo_clear_dirty_bitmap(BdrvDirtyBitmap
*bitmap
, HBitmap
*in
);
1086 void bdrv_inc_in_flight(BlockDriverState
*bs
);
1087 void bdrv_dec_in_flight(BlockDriverState
*bs
);
1089 void blockdev_close_all_bdrv_states(void);
1091 #endif /* BLOCK_INT_H */