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
28 #include "qemu-option.h"
29 #include "qemu-queue.h"
30 #include "qemu-coroutine.h"
31 #include "qemu-timer.h"
32 #include "qapi-types.h"
34 #define BLOCK_FLAG_ENCRYPT 1
35 #define BLOCK_FLAG_COMPAT6 4
37 #define BLOCK_IO_LIMIT_READ 0
38 #define BLOCK_IO_LIMIT_WRITE 1
39 #define BLOCK_IO_LIMIT_TOTAL 2
41 #define BLOCK_IO_SLICE_TIME 100000000
42 #define NANOSECONDS_PER_SECOND 1000000000.0
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"
55 typedef struct BdrvTrackedRequest BdrvTrackedRequest;
57 typedef struct BlockIOLimit {
62 typedef struct BlockIOBaseValue {
67 typedef struct BlockJob BlockJob;
72 * A class type for block job objects.
74 typedef struct BlockJobType {
75 /** Derived BlockJob struct size */
78 /** String describing the operation, part of query-block-jobs QMP API */
81 /** Optional callback for job types that support setting a speed limit */
82 void (*set_speed)(BlockJob *job, int64_t speed, Error **errp);
88 * Long-running operation on a BlockDriverState.
91 /** The job type, including the job vtable. */
92 const BlockJobType *job_type;
94 /** The block device on which the job is operating. */
98 * The coroutine that executes the job. If not NULL, it is
99 * reentered when busy is false and the job is cancelled.
104 * Set to true if the job should cancel itself. The flag must
105 * always be tested just before toggling the busy flag from false
106 * to true. After a job has been cancelled, it should only yield
107 * if #qemu_aio_wait will ("sooner or later") reenter the coroutine.
112 * Set to false by the job while it is in a quiescent state, where
113 * no I/O is pending and the job has yielded on any condition
114 * that is not detected by #qemu_aio_wait, such as a timer.
118 /** Offset that is published by the query-block-jobs QMP API */
121 /** Length that is published by the query-block-jobs QMP API */
124 /** Speed that was set with @block_job_set_speed. */
127 /** The completion function that will be called when the job completes. */
128 BlockDriverCompletionFunc *cb;
130 /** The opaque value that is passed to the completion function. */
135 const char *format_name;
137 int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename);
138 int (*bdrv_probe_device)(const char *filename);
139 int (*bdrv_open)(BlockDriverState *bs, int flags);
140 int (*bdrv_file_open)(BlockDriverState *bs, const char *filename, int flags);
141 int (*bdrv_read)(BlockDriverState *bs, int64_t sector_num,
142 uint8_t *buf, int nb_sectors);
143 int (*bdrv_write)(BlockDriverState *bs, int64_t sector_num,
144 const uint8_t *buf, int nb_sectors);
145 void (*bdrv_close)(BlockDriverState *bs);
146 void (*bdrv_rebind)(BlockDriverState *bs);
147 int (*bdrv_create)(const char *filename, QEMUOptionParameter *options);
148 int (*bdrv_set_key)(BlockDriverState *bs, const char *key);
149 int (*bdrv_make_empty)(BlockDriverState *bs);
151 BlockDriverAIOCB *(*bdrv_aio_readv)(BlockDriverState *bs,
152 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
153 BlockDriverCompletionFunc *cb, void *opaque);
154 BlockDriverAIOCB *(*bdrv_aio_writev)(BlockDriverState *bs,
155 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
156 BlockDriverCompletionFunc *cb, void *opaque);
157 BlockDriverAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs,
158 BlockDriverCompletionFunc *cb, void *opaque);
159 BlockDriverAIOCB *(*bdrv_aio_discard)(BlockDriverState *bs,
160 int64_t sector_num, int nb_sectors,
161 BlockDriverCompletionFunc *cb, void *opaque);
163 int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs,
164 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
165 int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs,
166 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
168 * Efficiently zero a region of the disk image. Typically an image format
169 * would use a compact metadata representation to implement this. This
170 * function pointer may be NULL and .bdrv_co_writev() will be called
173 int coroutine_fn (*bdrv_co_write_zeroes)(BlockDriverState *bs,
174 int64_t sector_num, int nb_sectors);
175 int coroutine_fn (*bdrv_co_discard)(BlockDriverState *bs,
176 int64_t sector_num, int nb_sectors);
177 int coroutine_fn (*bdrv_co_is_allocated)(BlockDriverState *bs,
178 int64_t sector_num, int nb_sectors, int *pnum);
181 * Invalidate any cached meta-data.
183 void (*bdrv_invalidate_cache)(BlockDriverState *bs);
186 * Flushes all data that was already written to the OS all the way down to
187 * the disk (for example raw-posix calls fsync()).
189 int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs);
192 * Flushes all internal caches to the OS. The data may still sit in a
193 * writeback cache of the host OS, but it will survive a crash of the qemu
196 int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs);
198 const char *protocol_name;
199 int (*bdrv_truncate)(BlockDriverState *bs, int64_t offset);
200 int64_t (*bdrv_getlength)(BlockDriverState *bs);
201 int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
202 int (*bdrv_write_compressed)(BlockDriverState *bs, int64_t sector_num,
203 const uint8_t *buf, int nb_sectors);
205 int (*bdrv_snapshot_create)(BlockDriverState *bs,
206 QEMUSnapshotInfo *sn_info);
207 int (*bdrv_snapshot_goto)(BlockDriverState *bs,
208 const char *snapshot_id);
209 int (*bdrv_snapshot_delete)(BlockDriverState *bs, const char *snapshot_id);
210 int (*bdrv_snapshot_list)(BlockDriverState *bs,
211 QEMUSnapshotInfo **psn_info);
212 int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
213 const char *snapshot_name);
214 int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
216 int (*bdrv_save_vmstate)(BlockDriverState *bs, const uint8_t *buf,
217 int64_t pos, int size);
218 int (*bdrv_load_vmstate)(BlockDriverState *bs, uint8_t *buf,
219 int64_t pos, int size);
221 int (*bdrv_change_backing_file)(BlockDriverState *bs,
222 const char *backing_file, const char *backing_fmt);
224 /* removable device specific */
225 int (*bdrv_is_inserted)(BlockDriverState *bs);
226 int (*bdrv_media_changed)(BlockDriverState *bs);
227 void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
228 void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
230 /* to control generic scsi devices */
231 int (*bdrv_ioctl)(BlockDriverState *bs, unsigned long int req, void *buf);
232 BlockDriverAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
233 unsigned long int req, void *buf,
234 BlockDriverCompletionFunc *cb, void *opaque);
236 /* List of options for creating images, terminated by name == NULL */
237 QEMUOptionParameter *create_options;
241 * Returns 0 for completed check, -errno for internal errors.
242 * The check results are stored in result.
244 int (*bdrv_check)(BlockDriverState* bs, BdrvCheckResult *result,
247 void (*bdrv_debug_event)(BlockDriverState *bs, BlkDebugEvent event);
250 * Returns 1 if newly created images are guaranteed to contain only
251 * zeros, 0 otherwise.
253 int (*bdrv_has_zero_init)(BlockDriverState *bs);
255 QLIST_ENTRY(BlockDriver) list;
259 * Note: the function bdrv_append() copies and swaps contents of
260 * BlockDriverStates, so if you add new fields to this struct, please
261 * inspect bdrv_append() to determine if the new fields need to be
264 struct BlockDriverState {
265 int64_t total_sectors; /* if we are reading a disk image, give its
267 int read_only; /* if true, the media is read only */
268 int keep_read_only; /* if true, the media was requested to stay read only */
269 int open_flags; /* flags used to open the file, re-used for re-open */
270 int encrypted; /* if true, the media is encrypted */
271 int valid_key; /* if true, a valid encryption key has been set */
272 int sg; /* if true, the device is a /dev/sg* */
273 int copy_on_read; /* if true, copy read backing sectors into image
274 note this is a reference count */
276 BlockDriver *drv; /* NULL means no media */
279 void *dev; /* attached device model, if any */
280 /* TODO change to DeviceState when all users are qdevified */
281 const BlockDevOps *dev_ops;
285 char backing_file[1024]; /* if non zero, the image is a diff of
287 char backing_format[16]; /* if non-zero and backing_file exists */
290 BlockDriverState *backing_hd;
291 BlockDriverState *file;
293 /* number of in-flight copy-on-read requests */
294 unsigned int copy_on_read_in_flight;
296 /* the time for latest disk I/O */
300 BlockIOLimit io_limits;
301 BlockIOBaseValue io_base;
302 CoQueue throttled_reqs;
303 QEMUTimer *block_timer;
304 bool io_limits_enabled;
306 /* I/O stats (display with "info blockstats"). */
307 uint64_t nr_bytes[BDRV_MAX_IOTYPE];
308 uint64_t nr_ops[BDRV_MAX_IOTYPE];
309 uint64_t total_time_ns[BDRV_MAX_IOTYPE];
310 uint64_t wr_highest_sector;
312 /* Whether the disk can expand beyond total_sectors */
315 /* the memory alignment required for the buffers handled by this driver */
316 int buffer_alignment;
318 /* do we need to tell the quest if we have a volatile write cache? */
319 int enable_write_cache;
321 /* NOTE: the following infos are only hints for real hardware
322 drivers. They are not used by the block driver */
323 int cyls, heads, secs, translation;
324 BlockErrorAction on_read_error, on_write_error;
325 bool iostatus_enabled;
326 BlockDeviceIoStatus iostatus;
327 char device_name[32];
328 unsigned long *dirty_bitmap;
330 int in_use; /* users other than guest access, eg. block migration */
331 QTAILQ_ENTRY(BlockDriverState) list;
333 QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
335 /* long-running background operation */
339 int get_tmp_filename(char *filename, int size);
341 void bdrv_set_io_limits(BlockDriverState *bs,
342 BlockIOLimit *io_limits);
345 int is_windows_drive(const char *filename);
350 * @job_type: The class object for the newly-created job.
352 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
353 * @cb: Completion function for the job.
354 * @opaque: Opaque pointer value passed to @cb.
355 * @errp: Error object.
357 * Create a new long-running block device job and return it. The job
358 * will call @cb asynchronously when the job completes. Note that
359 * @bs may have been closed at the time the @cb it is called. If
360 * this is the case, the job may be reported as either cancelled or
363 * This function is not part of the public job interface; it should be
364 * called from a wrapper that is specific to the job type.
366 void *block_job_create(const BlockJobType *job_type, BlockDriverState *bs,
367 int64_t speed, BlockDriverCompletionFunc *cb,
368 void *opaque, Error **errp);
371 * block_job_sleep_ns:
372 * @job: The job that calls the function.
373 * @clock: The clock to sleep on.
374 * @ns: How many nanoseconds to stop for.
376 * Put the job to sleep (assuming that it wasn't canceled) for @ns
377 * nanoseconds. Canceling the job will interrupt the wait immediately.
379 void block_job_sleep_ns(BlockJob *job, QEMUClock *clock, int64_t ns);
382 * block_job_complete:
383 * @job: The job being completed.
384 * @ret: The status code.
386 * Call the completion function that was registered at creation time, and
389 void block_job_complete(BlockJob *job, int ret);
392 * block_job_set_speed:
393 * @job: The job to set the speed for.
394 * @speed: The new value
395 * @errp: Error object.
397 * Set a rate-limiting parameter for the job; the actual meaning may
398 * vary depending on the job type.
400 void block_job_set_speed(BlockJob *job, int64_t speed, Error **errp);
404 * @job: The job to be canceled.
406 * Asynchronously cancel the specified job.
408 void block_job_cancel(BlockJob *job);
411 * block_job_is_cancelled:
412 * @job: The job being queried.
414 * Returns whether the job is scheduled for cancellation.
416 bool block_job_is_cancelled(BlockJob *job);
420 * @job: The job to be canceled.
422 * Asynchronously cancel the job and wait for it to reach a quiescent
423 * state. Note that the completion callback will still be called
424 * asynchronously, hence it is *not* valid to call #bdrv_delete
425 * immediately after #block_job_cancel_sync. Users of block jobs
426 * will usually protect the BlockDriverState objects with a reference
427 * count, should this be a concern.
429 * Returns the return value from the job if the job actually completed
430 * during the call, or -ECANCELED if it was canceled.
432 int block_job_cancel_sync(BlockJob *job);
436 * @bs: Block device to operate on.
437 * @base: Block device that will become the new base, or %NULL to
438 * flatten the whole backing file chain onto @bs.
439 * @base_id: The file name that will be written to @bs as the new
440 * backing file if the job completes. Ignored if @base is %NULL.
441 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
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,
454 BlockDriverCompletionFunc *cb,
455 void *opaque, Error **errp);
457 #endif /* BLOCK_INT_H */