]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * QEMU System Emulator block driver | |
3 | * | |
4 | * Copyright (c) 2003 Fabrice Bellard | |
5 | * | |
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: | |
12 | * | |
13 | * The above copyright notice and this permission notice shall be included in | |
14 | * all copies or substantial portions of the Software. | |
15 | * | |
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 | |
22 | * THE SOFTWARE. | |
23 | */ | |
24 | #ifndef BLOCK_INT_H | |
25 | #define BLOCK_INT_H | |
26 | ||
27 | #include "block.h" | |
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" | |
33 | ||
34 | #define BLOCK_FLAG_ENCRYPT 1 | |
35 | #define BLOCK_FLAG_COMPAT6 4 | |
36 | ||
37 | #define BLOCK_IO_LIMIT_READ 0 | |
38 | #define BLOCK_IO_LIMIT_WRITE 1 | |
39 | #define BLOCK_IO_LIMIT_TOTAL 2 | |
40 | ||
41 | #define BLOCK_IO_SLICE_TIME 100000000 | |
42 | #define NANOSECONDS_PER_SECOND 1000000000.0 | |
43 | ||
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 | ||
55 | typedef struct BdrvTrackedRequest BdrvTrackedRequest; | |
56 | ||
57 | typedef struct BlockIOLimit { | |
58 | int64_t bps[3]; | |
59 | int64_t iops[3]; | |
60 | } BlockIOLimit; | |
61 | ||
62 | typedef struct BlockIOBaseValue { | |
63 | uint64_t bytes[2]; | |
64 | uint64_t ios[2]; | |
65 | } BlockIOBaseValue; | |
66 | ||
67 | typedef struct BlockJob BlockJob; | |
68 | ||
69 | /** | |
70 | * BlockJobType: | |
71 | * | |
72 | * A class type for block job objects. | |
73 | */ | |
74 | typedef struct BlockJobType { | |
75 | /** Derived BlockJob struct size */ | |
76 | size_t instance_size; | |
77 | ||
78 | /** String describing the operation, part of query-block-jobs QMP API */ | |
79 | const char *job_type; | |
80 | ||
81 | /** Optional callback for job types that support setting a speed limit */ | |
82 | void (*set_speed)(BlockJob *job, int64_t speed, Error **errp); | |
83 | } BlockJobType; | |
84 | ||
85 | /** | |
86 | * BlockJob: | |
87 | * | |
88 | * Long-running operation on a BlockDriverState. | |
89 | */ | |
90 | struct BlockJob { | |
91 | /** The job type, including the job vtable. */ | |
92 | const BlockJobType *job_type; | |
93 | ||
94 | /** The block device on which the job is operating. */ | |
95 | BlockDriverState *bs; | |
96 | ||
97 | /** | |
98 | * The coroutine that executes the job. If not NULL, it is | |
99 | * reentered when busy is false and the job is cancelled. | |
100 | */ | |
101 | Coroutine *co; | |
102 | ||
103 | /** | |
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. | |
108 | */ | |
109 | bool cancelled; | |
110 | ||
111 | /** | |
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. | |
115 | */ | |
116 | bool busy; | |
117 | ||
118 | /** Offset that is published by the query-block-jobs QMP API */ | |
119 | int64_t offset; | |
120 | ||
121 | /** Length that is published by the query-block-jobs QMP API */ | |
122 | int64_t len; | |
123 | ||
124 | /** Speed that was set with @block_job_set_speed. */ | |
125 | int64_t speed; | |
126 | ||
127 | /** The completion function that will be called when the job completes. */ | |
128 | BlockDriverCompletionFunc *cb; | |
129 | ||
130 | /** The opaque value that is passed to the completion function. */ | |
131 | void *opaque; | |
132 | }; | |
133 | ||
134 | struct BlockDriver { | |
135 | const char *format_name; | |
136 | int instance_size; | |
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); | |
150 | /* aio */ | |
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); | |
162 | ||
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); | |
167 | /* | |
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 | |
171 | * instead. | |
172 | */ | |
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); | |
179 | ||
180 | /* | |
181 | * Invalidate any cached meta-data. | |
182 | */ | |
183 | void (*bdrv_invalidate_cache)(BlockDriverState *bs); | |
184 | ||
185 | /* | |
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()). | |
188 | */ | |
189 | int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs); | |
190 | ||
191 | /* | |
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 | |
194 | * process. | |
195 | */ | |
196 | int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs); | |
197 | ||
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); | |
204 | ||
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); | |
215 | ||
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); | |
220 | ||
221 | int (*bdrv_change_backing_file)(BlockDriverState *bs, | |
222 | const char *backing_file, const char *backing_fmt); | |
223 | ||
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); | |
229 | ||
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); | |
235 | ||
236 | /* List of options for creating images, terminated by name == NULL */ | |
237 | QEMUOptionParameter *create_options; | |
238 | ||
239 | ||
240 | /* | |
241 | * Returns 0 for completed check, -errno for internal errors. | |
242 | * The check results are stored in result. | |
243 | */ | |
244 | int (*bdrv_check)(BlockDriverState* bs, BdrvCheckResult *result, | |
245 | BdrvCheckMode fix); | |
246 | ||
247 | void (*bdrv_debug_event)(BlockDriverState *bs, BlkDebugEvent event); | |
248 | ||
249 | /* | |
250 | * Returns 1 if newly created images are guaranteed to contain only | |
251 | * zeros, 0 otherwise. | |
252 | */ | |
253 | int (*bdrv_has_zero_init)(BlockDriverState *bs); | |
254 | ||
255 | QLIST_ENTRY(BlockDriver) list; | |
256 | }; | |
257 | ||
258 | /* | |
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 | |
262 | * copied as well. | |
263 | */ | |
264 | struct BlockDriverState { | |
265 | int64_t total_sectors; /* if we are reading a disk image, give its | |
266 | size in sectors */ | |
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 */ | |
275 | ||
276 | BlockDriver *drv; /* NULL means no media */ | |
277 | void *opaque; | |
278 | ||
279 | void *dev; /* attached device model, if any */ | |
280 | /* TODO change to DeviceState when all users are qdevified */ | |
281 | const BlockDevOps *dev_ops; | |
282 | void *dev_opaque; | |
283 | ||
284 | char filename[1024]; | |
285 | char backing_file[1024]; /* if non zero, the image is a diff of | |
286 | this file image */ | |
287 | char backing_format[16]; /* if non-zero and backing_file exists */ | |
288 | int is_temporary; | |
289 | ||
290 | BlockDriverState *backing_hd; | |
291 | BlockDriverState *file; | |
292 | ||
293 | /* number of in-flight copy-on-read requests */ | |
294 | unsigned int copy_on_read_in_flight; | |
295 | ||
296 | /* the time for latest disk I/O */ | |
297 | int64_t slice_time; | |
298 | int64_t slice_start; | |
299 | int64_t slice_end; | |
300 | BlockIOLimit io_limits; | |
301 | BlockIOBaseValue io_base; | |
302 | CoQueue throttled_reqs; | |
303 | QEMUTimer *block_timer; | |
304 | bool io_limits_enabled; | |
305 | ||
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; | |
311 | ||
312 | /* Whether the disk can expand beyond total_sectors */ | |
313 | int growable; | |
314 | ||
315 | /* the memory alignment required for the buffers handled by this driver */ | |
316 | int buffer_alignment; | |
317 | ||
318 | /* do we need to tell the quest if we have a volatile write cache? */ | |
319 | int enable_write_cache; | |
320 | ||
321 | /* NOTE: the following infos are only hints for real hardware | |
322 | drivers. They are not used by the block driver */ | |
323 | BlockErrorAction on_read_error, on_write_error; | |
324 | bool iostatus_enabled; | |
325 | BlockDeviceIoStatus iostatus; | |
326 | char device_name[32]; | |
327 | unsigned long *dirty_bitmap; | |
328 | int64_t dirty_count; | |
329 | int in_use; /* users other than guest access, eg. block migration */ | |
330 | QTAILQ_ENTRY(BlockDriverState) list; | |
331 | ||
332 | QLIST_HEAD(, BdrvTrackedRequest) tracked_requests; | |
333 | ||
334 | /* long-running background operation */ | |
335 | BlockJob *job; | |
336 | }; | |
337 | ||
338 | int get_tmp_filename(char *filename, int size); | |
339 | ||
340 | void bdrv_set_io_limits(BlockDriverState *bs, | |
341 | BlockIOLimit *io_limits); | |
342 | ||
343 | #ifdef _WIN32 | |
344 | int is_windows_drive(const char *filename); | |
345 | #endif | |
346 | ||
347 | /** | |
348 | * block_job_create: | |
349 | * @job_type: The class object for the newly-created job. | |
350 | * @bs: The block | |
351 | * @speed: The maximum speed, in bytes per second, or 0 for unlimited. | |
352 | * @cb: Completion function for the job. | |
353 | * @opaque: Opaque pointer value passed to @cb. | |
354 | * @errp: Error object. | |
355 | * | |
356 | * Create a new long-running block device job and return it. The job | |
357 | * will call @cb asynchronously when the job completes. Note that | |
358 | * @bs may have been closed at the time the @cb it is called. If | |
359 | * this is the case, the job may be reported as either cancelled or | |
360 | * completed. | |
361 | * | |
362 | * This function is not part of the public job interface; it should be | |
363 | * called from a wrapper that is specific to the job type. | |
364 | */ | |
365 | void *block_job_create(const BlockJobType *job_type, BlockDriverState *bs, | |
366 | int64_t speed, BlockDriverCompletionFunc *cb, | |
367 | void *opaque, Error **errp); | |
368 | ||
369 | /** | |
370 | * block_job_sleep_ns: | |
371 | * @job: The job that calls the function. | |
372 | * @clock: The clock to sleep on. | |
373 | * @ns: How many nanoseconds to stop for. | |
374 | * | |
375 | * Put the job to sleep (assuming that it wasn't canceled) for @ns | |
376 | * nanoseconds. Canceling the job will interrupt the wait immediately. | |
377 | */ | |
378 | void block_job_sleep_ns(BlockJob *job, QEMUClock *clock, int64_t ns); | |
379 | ||
380 | /** | |
381 | * block_job_complete: | |
382 | * @job: The job being completed. | |
383 | * @ret: The status code. | |
384 | * | |
385 | * Call the completion function that was registered at creation time, and | |
386 | * free @job. | |
387 | */ | |
388 | void block_job_complete(BlockJob *job, int ret); | |
389 | ||
390 | /** | |
391 | * block_job_set_speed: | |
392 | * @job: The job to set the speed for. | |
393 | * @speed: The new value | |
394 | * @errp: Error object. | |
395 | * | |
396 | * Set a rate-limiting parameter for the job; the actual meaning may | |
397 | * vary depending on the job type. | |
398 | */ | |
399 | void block_job_set_speed(BlockJob *job, int64_t speed, Error **errp); | |
400 | ||
401 | /** | |
402 | * block_job_cancel: | |
403 | * @job: The job to be canceled. | |
404 | * | |
405 | * Asynchronously cancel the specified job. | |
406 | */ | |
407 | void block_job_cancel(BlockJob *job); | |
408 | ||
409 | /** | |
410 | * block_job_is_cancelled: | |
411 | * @job: The job being queried. | |
412 | * | |
413 | * Returns whether the job is scheduled for cancellation. | |
414 | */ | |
415 | bool block_job_is_cancelled(BlockJob *job); | |
416 | ||
417 | /** | |
418 | * block_job_cancel: | |
419 | * @job: The job to be canceled. | |
420 | * | |
421 | * Asynchronously cancel the job and wait for it to reach a quiescent | |
422 | * state. Note that the completion callback will still be called | |
423 | * asynchronously, hence it is *not* valid to call #bdrv_delete | |
424 | * immediately after #block_job_cancel_sync. Users of block jobs | |
425 | * will usually protect the BlockDriverState objects with a reference | |
426 | * count, should this be a concern. | |
427 | * | |
428 | * Returns the return value from the job if the job actually completed | |
429 | * during the call, or -ECANCELED if it was canceled. | |
430 | */ | |
431 | int block_job_cancel_sync(BlockJob *job); | |
432 | ||
433 | /** | |
434 | * stream_start: | |
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 | * @cb: Completion function for the job. | |
442 | * @opaque: Opaque pointer value passed to @cb. | |
443 | * @errp: Error object. | |
444 | * | |
445 | * Start a streaming operation on @bs. Clusters that are unallocated | |
446 | * in @bs, but allocated in any image between @base and @bs (both | |
447 | * exclusive) will be written to @bs. At the end of a successful | |
448 | * streaming job, the backing file of @bs will be changed to | |
449 | * @base_id in the written image and to @base in the live BlockDriverState. | |
450 | */ | |
451 | void stream_start(BlockDriverState *bs, BlockDriverState *base, | |
452 | const char *base_id, int64_t speed, | |
453 | BlockDriverCompletionFunc *cb, | |
454 | void *opaque, Error **errp); | |
455 | ||
456 | #endif /* BLOCK_INT_H */ |