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/block.h"
28 #include "qemu/option.h"
29 #include "qemu/queue.h"
30 #include "block/coroutine.h"
31 #include "qemu/timer.h"
32 #include "qapi-types.h"
33 #include "qapi/qmp/qerror.h"
34 #include "monitor/monitor.h"
35 #include "qemu/hbitmap.h"
36 #include "block/snapshot.h"
37 #include "qemu/main-loop.h"
38 #include "qemu/throttle.h"
40 #define BLOCK_FLAG_ENCRYPT 1
41 #define BLOCK_FLAG_COMPAT6 4
42 #define BLOCK_FLAG_LAZY_REFCOUNTS 8
44 #define BLOCK_OPT_SIZE "size"
45 #define BLOCK_OPT_ENCRYPT "encryption"
46 #define BLOCK_OPT_COMPAT6 "compat6"
47 #define BLOCK_OPT_BACKING_FILE "backing_file"
48 #define BLOCK_OPT_BACKING_FMT "backing_fmt"
49 #define BLOCK_OPT_CLUSTER_SIZE "cluster_size"
50 #define BLOCK_OPT_TABLE_SIZE "table_size"
51 #define BLOCK_OPT_PREALLOC "preallocation"
52 #define BLOCK_OPT_SUBFMT "subformat"
53 #define BLOCK_OPT_COMPAT_LEVEL "compat"
54 #define BLOCK_OPT_LAZY_REFCOUNTS "lazy_refcounts"
55 #define BLOCK_OPT_ADAPTER_TYPE "adapter_type"
56 #define BLOCK_OPT_REDUNDANCY "redundancy"
57 #define BLOCK_OPT_NOCOW "nocow"
59 typedef struct BdrvTrackedRequest {
66 int64_t overlap_offset;
67 unsigned int overlap_bytes;
69 QLIST_ENTRY(BdrvTrackedRequest) list;
70 Coroutine *co; /* owner, used for deadlock detection */
71 CoQueue wait_queue; /* coroutines blocked on this request */
73 struct BdrvTrackedRequest *waiting_for;
77 const char *format_name;
80 /* set to true if the BlockDriver is a block filter */
82 /* for snapshots block filter like Quorum can implement the
83 * following recursive callback.
84 * It's purpose is to recurse on the filter children while calling
85 * bdrv_recurse_is_first_non_filter on them.
86 * For a sample implementation look in the future Quorum block filter.
88 bool (*bdrv_recurse_is_first_non_filter)(BlockDriverState *bs,
89 BlockDriverState *candidate);
91 int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename);
92 int (*bdrv_probe_device)(const char *filename);
94 /* Any driver implementing this callback is expected to be able to handle
95 * NULL file names in its .bdrv_open() implementation */
96 void (*bdrv_parse_filename)(const char *filename, QDict *options, Error **errp);
97 /* Drivers not implementing bdrv_parse_filename nor bdrv_open should have
98 * this field set to true, except ones that are defined only by their
100 * An example of the last type will be the quorum block driver.
102 bool bdrv_needs_filename;
104 /* Set if a driver can support backing files */
105 bool supports_backing;
107 /* For handling image reopen for split or non-split files */
108 int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state,
109 BlockReopenQueue *queue, Error **errp);
110 void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state);
111 void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state);
113 int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags,
115 int (*bdrv_file_open)(BlockDriverState *bs, QDict *options, int flags,
117 int (*bdrv_read)(BlockDriverState *bs, int64_t sector_num,
118 uint8_t *buf, int nb_sectors);
119 int (*bdrv_write)(BlockDriverState *bs, int64_t sector_num,
120 const uint8_t *buf, int nb_sectors);
121 void (*bdrv_close)(BlockDriverState *bs);
122 void (*bdrv_rebind)(BlockDriverState *bs);
123 int (*bdrv_create)(const char *filename, QemuOpts *opts, Error **errp);
124 int (*bdrv_set_key)(BlockDriverState *bs, const char *key);
125 int (*bdrv_make_empty)(BlockDriverState *bs);
127 BlockDriverAIOCB *(*bdrv_aio_readv)(BlockDriverState *bs,
128 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
129 BlockDriverCompletionFunc *cb, void *opaque);
130 BlockDriverAIOCB *(*bdrv_aio_writev)(BlockDriverState *bs,
131 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
132 BlockDriverCompletionFunc *cb, void *opaque);
133 BlockDriverAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs,
134 BlockDriverCompletionFunc *cb, void *opaque);
135 BlockDriverAIOCB *(*bdrv_aio_discard)(BlockDriverState *bs,
136 int64_t sector_num, int nb_sectors,
137 BlockDriverCompletionFunc *cb, void *opaque);
139 int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs,
140 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
141 int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs,
142 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
144 * Efficiently zero a region of the disk image. Typically an image format
145 * would use a compact metadata representation to implement this. This
146 * function pointer may be NULL and .bdrv_co_writev() will be called
149 int coroutine_fn (*bdrv_co_write_zeroes)(BlockDriverState *bs,
150 int64_t sector_num, int nb_sectors, BdrvRequestFlags flags);
151 int coroutine_fn (*bdrv_co_discard)(BlockDriverState *bs,
152 int64_t sector_num, int nb_sectors);
153 int64_t coroutine_fn (*bdrv_co_get_block_status)(BlockDriverState *bs,
154 int64_t sector_num, int nb_sectors, int *pnum);
157 * Invalidate any cached meta-data.
159 void (*bdrv_invalidate_cache)(BlockDriverState *bs, Error **errp);
162 * Flushes all data that was already written to the OS all the way down to
163 * the disk (for example raw-posix calls fsync()).
165 int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs);
168 * Flushes all internal caches to the OS. The data may still sit in a
169 * writeback cache of the host OS, but it will survive a crash of the qemu
172 int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs);
174 const char *protocol_name;
175 int (*bdrv_truncate)(BlockDriverState *bs, int64_t offset);
177 int64_t (*bdrv_getlength)(BlockDriverState *bs);
178 bool has_variable_length;
179 int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
181 int (*bdrv_write_compressed)(BlockDriverState *bs, int64_t sector_num,
182 const uint8_t *buf, int nb_sectors);
184 int (*bdrv_snapshot_create)(BlockDriverState *bs,
185 QEMUSnapshotInfo *sn_info);
186 int (*bdrv_snapshot_goto)(BlockDriverState *bs,
187 const char *snapshot_id);
188 int (*bdrv_snapshot_delete)(BlockDriverState *bs,
189 const char *snapshot_id,
192 int (*bdrv_snapshot_list)(BlockDriverState *bs,
193 QEMUSnapshotInfo **psn_info);
194 int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
195 const char *snapshot_id,
198 int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
199 ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs);
201 int (*bdrv_save_vmstate)(BlockDriverState *bs, QEMUIOVector *qiov,
203 int (*bdrv_load_vmstate)(BlockDriverState *bs, uint8_t *buf,
204 int64_t pos, int size);
206 int (*bdrv_change_backing_file)(BlockDriverState *bs,
207 const char *backing_file, const char *backing_fmt);
209 /* removable device specific */
210 int (*bdrv_is_inserted)(BlockDriverState *bs);
211 int (*bdrv_media_changed)(BlockDriverState *bs);
212 void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
213 void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
215 /* to control generic scsi devices */
216 int (*bdrv_ioctl)(BlockDriverState *bs, unsigned long int req, void *buf);
217 BlockDriverAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
218 unsigned long int req, void *buf,
219 BlockDriverCompletionFunc *cb, void *opaque);
221 /* List of options for creating images, terminated by name == NULL */
222 QemuOptsList *create_opts;
225 * Returns 0 for completed check, -errno for internal errors.
226 * The check results are stored in result.
228 int (*bdrv_check)(BlockDriverState* bs, BdrvCheckResult *result,
231 int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts);
233 void (*bdrv_debug_event)(BlockDriverState *bs, BlkDebugEvent event);
235 /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
236 int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
238 int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs,
240 int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
241 bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
243 int (*bdrv_refresh_limits)(BlockDriverState *bs);
246 * Returns 1 if newly created images are guaranteed to contain only
247 * zeros, 0 otherwise.
249 int (*bdrv_has_zero_init)(BlockDriverState *bs);
251 /* Remove fd handlers, timers, and other event loop callbacks so the event
252 * loop is no longer in use. Called with no in-flight requests and in
253 * depth-first traversal order with parents before child nodes.
255 void (*bdrv_detach_aio_context)(BlockDriverState *bs);
257 /* Add fd handlers, timers, and other event loop callbacks so I/O requests
258 * can be processed again. Called with no in-flight requests and in
259 * depth-first traversal order with child nodes before parent nodes.
261 void (*bdrv_attach_aio_context)(BlockDriverState *bs,
262 AioContext *new_context);
264 QLIST_ENTRY(BlockDriver) list;
267 typedef struct BlockLimits {
268 /* maximum number of sectors that can be discarded at once */
271 /* optimal alignment for discard requests in sectors */
272 int64_t discard_alignment;
274 /* maximum number of sectors that can zeroized at once */
275 int max_write_zeroes;
277 /* optimal alignment for write zeroes requests in sectors */
278 int64_t write_zeroes_alignment;
280 /* optimal transfer length in sectors */
281 int opt_transfer_length;
283 /* memory alignment so that no bounce buffer is needed */
284 size_t opt_mem_alignment;
287 typedef struct BdrvOpBlocker BdrvOpBlocker;
290 * Note: the function bdrv_append() copies and swaps contents of
291 * BlockDriverStates, so if you add new fields to this struct, please
292 * inspect bdrv_append() to determine if the new fields need to be
295 struct BlockDriverState {
296 int64_t total_sectors; /* if we are reading a disk image, give its
298 int read_only; /* if true, the media is read only */
299 int open_flags; /* flags used to open the file, re-used for re-open */
300 int encrypted; /* if true, the media is encrypted */
301 int valid_key; /* if true, a valid encryption key has been set */
302 int sg; /* if true, the device is a /dev/sg* */
303 int copy_on_read; /* if true, copy read backing sectors into image
304 note this is a reference count */
306 BlockDriver *drv; /* NULL means no media */
309 void *dev; /* attached device model, if any */
310 /* TODO change to DeviceState when all users are qdevified */
311 const BlockDevOps *dev_ops;
314 AioContext *aio_context; /* event loop used for fd handlers, timers, etc */
317 char backing_file[1024]; /* if non zero, the image is a diff of
319 char backing_format[16]; /* if non-zero and backing_file exists */
321 BlockDriverState *backing_hd;
322 BlockDriverState *file;
324 NotifierList close_notifiers;
326 /* Callback before write request is processed */
327 NotifierWithReturnList before_write_notifiers;
329 /* number of in-flight serialising requests */
330 unsigned int serialising_in_flight;
333 ThrottleState throttle_state;
334 CoQueue throttled_reqs[2];
335 bool io_limits_enabled;
337 /* I/O stats (display with "info blockstats"). */
338 uint64_t nr_bytes[BDRV_MAX_IOTYPE];
339 uint64_t nr_ops[BDRV_MAX_IOTYPE];
340 uint64_t total_time_ns[BDRV_MAX_IOTYPE];
341 uint64_t wr_highest_sector;
346 /* Whether the disk can expand beyond total_sectors */
349 /* Whether produces zeros when read beyond eof */
350 bool zero_beyond_eof;
352 /* Alignment requirement for offset/length of I/O requests */
353 unsigned int request_alignment;
355 /* the block size for which the guest device expects atomicity */
356 int guest_block_size;
358 /* do we need to tell the quest if we have a volatile write cache? */
359 int enable_write_cache;
361 /* NOTE: the following infos are only hints for real hardware
362 drivers. They are not used by the block driver */
363 BlockdevOnError on_read_error, on_write_error;
364 bool iostatus_enabled;
365 BlockDeviceIoStatus iostatus;
367 /* the following member gives a name to every node on the bs graph. */
369 /* element of the list of named nodes building the graph */
370 QTAILQ_ENTRY(BlockDriverState) node_list;
371 /* Device name is the name associated with the "drive" the guest sees */
372 char device_name[32];
373 /* element of the list of "drives" the guest sees */
374 QTAILQ_ENTRY(BlockDriverState) device_list;
375 QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps;
378 QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
380 /* operation blockers */
381 QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX];
383 /* long-running background operation */
387 BlockdevDetectZeroesOptions detect_zeroes;
389 /* The error object in use for blocking operations on backing_hd */
390 Error *backing_blocker;
393 int get_tmp_filename(char *filename, int size);
395 void bdrv_set_io_limits(BlockDriverState *bs,
396 ThrottleConfig *cfg);
400 * bdrv_add_before_write_notifier:
402 * Register a callback that is invoked before write requests are processed but
403 * after any throttling or waiting for overlapping requests.
405 void bdrv_add_before_write_notifier(BlockDriverState *bs,
406 NotifierWithReturn *notifier);
409 * bdrv_detach_aio_context:
411 * May be called from .bdrv_detach_aio_context() to detach children from the
412 * current #AioContext. This is only needed by block drivers that manage their
413 * own children. Both ->file and ->backing_hd are automatically handled and
414 * block drivers should not call this function on them explicitly.
416 void bdrv_detach_aio_context(BlockDriverState *bs);
419 * bdrv_attach_aio_context:
421 * May be called from .bdrv_attach_aio_context() to attach children to the new
422 * #AioContext. This is only needed by block drivers that manage their own
423 * children. Both ->file and ->backing_hd are automatically handled and block
424 * drivers should not call this function on them explicitly.
426 void bdrv_attach_aio_context(BlockDriverState *bs,
427 AioContext *new_context);
430 int is_windows_drive(const char *filename);
435 * @bs: Block device to operate on.
436 * @base: Block device that will become the new base, or %NULL to
437 * flatten the whole backing file chain onto @bs.
438 * @base_id: The file name that will be written to @bs as the new
439 * backing file if the job completes. Ignored if @base is %NULL.
440 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
441 * @on_error: The action to take upon error.
442 * @cb: Completion function for the job.
443 * @opaque: Opaque pointer value passed to @cb.
444 * @errp: Error object.
446 * Start a streaming operation on @bs. Clusters that are unallocated
447 * in @bs, but allocated in any image between @base and @bs (both
448 * exclusive) will be written to @bs. At the end of a successful
449 * streaming job, the backing file of @bs will be changed to
450 * @base_id in the written image and to @base in the live BlockDriverState.
452 void stream_start(BlockDriverState *bs, BlockDriverState *base,
453 const char *base_id, int64_t speed, BlockdevOnError on_error,
454 BlockDriverCompletionFunc *cb,
455 void *opaque, Error **errp);
459 * @bs: Active block device.
460 * @top: Top block device to be committed.
461 * @base: Block device that will be written into, and become the new top.
462 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
463 * @on_error: The action to take upon error.
464 * @cb: Completion function for the job.
465 * @opaque: Opaque pointer value passed to @cb.
466 * @backing_file_str: String to use as the backing file in @top's overlay
467 * @errp: Error object.
470 void commit_start(BlockDriverState *bs, BlockDriverState *base,
471 BlockDriverState *top, int64_t speed,
472 BlockdevOnError on_error, BlockDriverCompletionFunc *cb,
473 void *opaque, const char *backing_file_str, Error **errp);
475 * commit_active_start:
476 * @bs: Active block device to be committed.
477 * @base: Block device that will be written into, and become the new top.
478 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
479 * @on_error: The action to take upon error.
480 * @cb: Completion function for the job.
481 * @opaque: Opaque pointer value passed to @cb.
482 * @errp: Error object.
485 void commit_active_start(BlockDriverState *bs, BlockDriverState *base,
487 BlockdevOnError on_error,
488 BlockDriverCompletionFunc *cb,
489 void *opaque, Error **errp);
492 * @bs: Block device to operate on.
493 * @target: Block device to write to.
494 * @replaces: Block graph node name to replace once the mirror is done. Can
495 * only be used when full mirroring is selected.
496 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
497 * @granularity: The chosen granularity for the dirty bitmap.
498 * @buf_size: The amount of data that can be in flight at one time.
499 * @mode: Whether to collapse all images in the chain to the target.
500 * @on_source_error: The action to take upon error reading from the source.
501 * @on_target_error: The action to take upon error writing to the target.
502 * @cb: Completion function for the job.
503 * @opaque: Opaque pointer value passed to @cb.
504 * @errp: Error object.
506 * Start a mirroring operation on @bs. Clusters that are allocated
507 * in @bs will be written to @bs until the job is cancelled or
508 * manually completed. At the end of a successful mirroring job,
509 * @bs will be switched to read from @target.
511 void mirror_start(BlockDriverState *bs, BlockDriverState *target,
512 const char *replaces,
513 int64_t speed, int64_t granularity, int64_t buf_size,
514 MirrorSyncMode mode, BlockdevOnError on_source_error,
515 BlockdevOnError on_target_error,
516 BlockDriverCompletionFunc *cb,
517 void *opaque, Error **errp);
521 * @bs: Block device to operate on.
522 * @target: Block device to write to.
523 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
524 * @sync_mode: What parts of the disk image should be copied to the destination.
525 * @on_source_error: The action to take upon error reading from the source.
526 * @on_target_error: The action to take upon error writing to the target.
527 * @cb: Completion function for the job.
528 * @opaque: Opaque pointer value passed to @cb.
530 * Start a backup operation on @bs. Clusters in @bs are written to @target
531 * until the job is cancelled or manually completed.
533 void backup_start(BlockDriverState *bs, BlockDriverState *target,
534 int64_t speed, MirrorSyncMode sync_mode,
535 BlockdevOnError on_source_error,
536 BlockdevOnError on_target_error,
537 BlockDriverCompletionFunc *cb, void *opaque,
540 #endif /* BLOCK_INT_H */