numa: add -numa node,memdev= option
[qemu.git] / include / block / block_int.h
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/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"
39
40 #define BLOCK_FLAG_ENCRYPT 1
41 #define BLOCK_FLAG_COMPAT6 4
42 #define BLOCK_FLAG_LAZY_REFCOUNTS 8
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 #define BLOCK_OPT_LAZY_REFCOUNTS "lazy_refcounts"
55 #define BLOCK_OPT_ADAPTER_TYPE "adapter_type"
56 #define BLOCK_OPT_REDUNDANCY "redundancy"
57
58 typedef struct BdrvTrackedRequest {
59 BlockDriverState *bs;
60 int64_t offset;
61 unsigned int bytes;
62 bool is_write;
63
64 bool serialising;
65 int64_t overlap_offset;
66 unsigned int overlap_bytes;
67
68 QLIST_ENTRY(BdrvTrackedRequest) list;
69 Coroutine *co; /* owner, used for deadlock detection */
70 CoQueue wait_queue; /* coroutines blocked on this request */
71
72 struct BdrvTrackedRequest *waiting_for;
73 } BdrvTrackedRequest;
74
75 struct BlockDriver {
76 const char *format_name;
77 int instance_size;
78
79 /* set to true if the BlockDriver is a block filter */
80 bool is_filter;
81 /* for snapshots block filter like Quorum can implement the
82 * following recursive callback.
83 * It's purpose is to recurse on the filter children while calling
84 * bdrv_recurse_is_first_non_filter on them.
85 * For a sample implementation look in the future Quorum block filter.
86 */
87 bool (*bdrv_recurse_is_first_non_filter)(BlockDriverState *bs,
88 BlockDriverState *candidate);
89
90 int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename);
91 int (*bdrv_probe_device)(const char *filename);
92
93 /* Any driver implementing this callback is expected to be able to handle
94 * NULL file names in its .bdrv_open() implementation */
95 void (*bdrv_parse_filename)(const char *filename, QDict *options, Error **errp);
96 /* Drivers not implementing bdrv_parse_filename nor bdrv_open should have
97 * this field set to true, except ones that are defined only by their
98 * child's bs.
99 * An example of the last type will be the quorum block driver.
100 */
101 bool bdrv_needs_filename;
102
103 /* For handling image reopen for split or non-split files */
104 int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state,
105 BlockReopenQueue *queue, Error **errp);
106 void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state);
107 void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state);
108
109 int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags,
110 Error **errp);
111 int (*bdrv_file_open)(BlockDriverState *bs, QDict *options, int flags,
112 Error **errp);
113 int (*bdrv_read)(BlockDriverState *bs, int64_t sector_num,
114 uint8_t *buf, int nb_sectors);
115 int (*bdrv_write)(BlockDriverState *bs, int64_t sector_num,
116 const uint8_t *buf, int nb_sectors);
117 void (*bdrv_close)(BlockDriverState *bs);
118 void (*bdrv_rebind)(BlockDriverState *bs);
119 int (*bdrv_create)(const char *filename, QemuOpts *opts, Error **errp);
120 int (*bdrv_set_key)(BlockDriverState *bs, const char *key);
121 int (*bdrv_make_empty)(BlockDriverState *bs);
122 /* aio */
123 BlockDriverAIOCB *(*bdrv_aio_readv)(BlockDriverState *bs,
124 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
125 BlockDriverCompletionFunc *cb, void *opaque);
126 BlockDriverAIOCB *(*bdrv_aio_writev)(BlockDriverState *bs,
127 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
128 BlockDriverCompletionFunc *cb, void *opaque);
129 BlockDriverAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs,
130 BlockDriverCompletionFunc *cb, void *opaque);
131 BlockDriverAIOCB *(*bdrv_aio_discard)(BlockDriverState *bs,
132 int64_t sector_num, int nb_sectors,
133 BlockDriverCompletionFunc *cb, void *opaque);
134
135 int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs,
136 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
137 int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs,
138 int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
139 /*
140 * Efficiently zero a region of the disk image. Typically an image format
141 * would use a compact metadata representation to implement this. This
142 * function pointer may be NULL and .bdrv_co_writev() will be called
143 * instead.
144 */
145 int coroutine_fn (*bdrv_co_write_zeroes)(BlockDriverState *bs,
146 int64_t sector_num, int nb_sectors, BdrvRequestFlags flags);
147 int coroutine_fn (*bdrv_co_discard)(BlockDriverState *bs,
148 int64_t sector_num, int nb_sectors);
149 int64_t coroutine_fn (*bdrv_co_get_block_status)(BlockDriverState *bs,
150 int64_t sector_num, int nb_sectors, int *pnum);
151
152 /*
153 * Invalidate any cached meta-data.
154 */
155 void (*bdrv_invalidate_cache)(BlockDriverState *bs, Error **errp);
156
157 /*
158 * Flushes all data that was already written to the OS all the way down to
159 * the disk (for example raw-posix calls fsync()).
160 */
161 int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs);
162
163 /*
164 * Flushes all internal caches to the OS. The data may still sit in a
165 * writeback cache of the host OS, but it will survive a crash of the qemu
166 * process.
167 */
168 int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs);
169
170 const char *protocol_name;
171 int (*bdrv_truncate)(BlockDriverState *bs, int64_t offset);
172
173 int64_t (*bdrv_getlength)(BlockDriverState *bs);
174 bool has_variable_length;
175 int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
176
177 int (*bdrv_write_compressed)(BlockDriverState *bs, int64_t sector_num,
178 const uint8_t *buf, int nb_sectors);
179
180 int (*bdrv_snapshot_create)(BlockDriverState *bs,
181 QEMUSnapshotInfo *sn_info);
182 int (*bdrv_snapshot_goto)(BlockDriverState *bs,
183 const char *snapshot_id);
184 int (*bdrv_snapshot_delete)(BlockDriverState *bs,
185 const char *snapshot_id,
186 const char *name,
187 Error **errp);
188 int (*bdrv_snapshot_list)(BlockDriverState *bs,
189 QEMUSnapshotInfo **psn_info);
190 int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
191 const char *snapshot_id,
192 const char *name,
193 Error **errp);
194 int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
195 ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs);
196
197 int (*bdrv_save_vmstate)(BlockDriverState *bs, QEMUIOVector *qiov,
198 int64_t pos);
199 int (*bdrv_load_vmstate)(BlockDriverState *bs, uint8_t *buf,
200 int64_t pos, int size);
201
202 int (*bdrv_change_backing_file)(BlockDriverState *bs,
203 const char *backing_file, const char *backing_fmt);
204
205 /* removable device specific */
206 int (*bdrv_is_inserted)(BlockDriverState *bs);
207 int (*bdrv_media_changed)(BlockDriverState *bs);
208 void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
209 void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
210
211 /* to control generic scsi devices */
212 int (*bdrv_ioctl)(BlockDriverState *bs, unsigned long int req, void *buf);
213 BlockDriverAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
214 unsigned long int req, void *buf,
215 BlockDriverCompletionFunc *cb, void *opaque);
216
217 /* List of options for creating images, terminated by name == NULL */
218 QemuOptsList *create_opts;
219
220 /*
221 * Returns 0 for completed check, -errno for internal errors.
222 * The check results are stored in result.
223 */
224 int (*bdrv_check)(BlockDriverState* bs, BdrvCheckResult *result,
225 BdrvCheckMode fix);
226
227 int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts);
228
229 void (*bdrv_debug_event)(BlockDriverState *bs, BlkDebugEvent event);
230
231 /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
232 int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
233 const char *tag);
234 int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs,
235 const char *tag);
236 int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
237 bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
238
239 int (*bdrv_refresh_limits)(BlockDriverState *bs);
240
241 /*
242 * Returns 1 if newly created images are guaranteed to contain only
243 * zeros, 0 otherwise.
244 */
245 int (*bdrv_has_zero_init)(BlockDriverState *bs);
246
247 /* Remove fd handlers, timers, and other event loop callbacks so the event
248 * loop is no longer in use. Called with no in-flight requests and in
249 * depth-first traversal order with parents before child nodes.
250 */
251 void (*bdrv_detach_aio_context)(BlockDriverState *bs);
252
253 /* Add fd handlers, timers, and other event loop callbacks so I/O requests
254 * can be processed again. Called with no in-flight requests and in
255 * depth-first traversal order with child nodes before parent nodes.
256 */
257 void (*bdrv_attach_aio_context)(BlockDriverState *bs,
258 AioContext *new_context);
259
260 QLIST_ENTRY(BlockDriver) list;
261 };
262
263 typedef struct BlockLimits {
264 /* maximum number of sectors that can be discarded at once */
265 int max_discard;
266
267 /* optimal alignment for discard requests in sectors */
268 int64_t discard_alignment;
269
270 /* maximum number of sectors that can zeroized at once */
271 int max_write_zeroes;
272
273 /* optimal alignment for write zeroes requests in sectors */
274 int64_t write_zeroes_alignment;
275
276 /* optimal transfer length in sectors */
277 int opt_transfer_length;
278
279 /* memory alignment so that no bounce buffer is needed */
280 size_t opt_mem_alignment;
281 } BlockLimits;
282
283 typedef struct BdrvOpBlocker BdrvOpBlocker;
284
285 /*
286 * Note: the function bdrv_append() copies and swaps contents of
287 * BlockDriverStates, so if you add new fields to this struct, please
288 * inspect bdrv_append() to determine if the new fields need to be
289 * copied as well.
290 */
291 struct BlockDriverState {
292 int64_t total_sectors; /* if we are reading a disk image, give its
293 size in sectors */
294 int read_only; /* if true, the media is read only */
295 int open_flags; /* flags used to open the file, re-used for re-open */
296 int encrypted; /* if true, the media is encrypted */
297 int valid_key; /* if true, a valid encryption key has been set */
298 int sg; /* if true, the device is a /dev/sg* */
299 int copy_on_read; /* if true, copy read backing sectors into image
300 note this is a reference count */
301
302 BlockDriver *drv; /* NULL means no media */
303 void *opaque;
304
305 void *dev; /* attached device model, if any */
306 /* TODO change to DeviceState when all users are qdevified */
307 const BlockDevOps *dev_ops;
308 void *dev_opaque;
309
310 AioContext *aio_context; /* event loop used for fd handlers, timers, etc */
311
312 char filename[1024];
313 char backing_file[1024]; /* if non zero, the image is a diff of
314 this file image */
315 char backing_format[16]; /* if non-zero and backing_file exists */
316
317 BlockDriverState *backing_hd;
318 BlockDriverState *file;
319
320 NotifierList close_notifiers;
321
322 /* Callback before write request is processed */
323 NotifierWithReturnList before_write_notifiers;
324
325 /* number of in-flight serialising requests */
326 unsigned int serialising_in_flight;
327
328 /* I/O throttling */
329 ThrottleState throttle_state;
330 CoQueue throttled_reqs[2];
331 bool io_limits_enabled;
332
333 /* I/O stats (display with "info blockstats"). */
334 uint64_t nr_bytes[BDRV_MAX_IOTYPE];
335 uint64_t nr_ops[BDRV_MAX_IOTYPE];
336 uint64_t total_time_ns[BDRV_MAX_IOTYPE];
337 uint64_t wr_highest_sector;
338
339 /* I/O Limits */
340 BlockLimits bl;
341
342 /* Whether the disk can expand beyond total_sectors */
343 int growable;
344
345 /* Whether produces zeros when read beyond eof */
346 bool zero_beyond_eof;
347
348 /* Alignment requirement for offset/length of I/O requests */
349 unsigned int request_alignment;
350
351 /* the block size for which the guest device expects atomicity */
352 int guest_block_size;
353
354 /* do we need to tell the quest if we have a volatile write cache? */
355 int enable_write_cache;
356
357 /* NOTE: the following infos are only hints for real hardware
358 drivers. They are not used by the block driver */
359 BlockdevOnError on_read_error, on_write_error;
360 bool iostatus_enabled;
361 BlockDeviceIoStatus iostatus;
362
363 /* the following member gives a name to every node on the bs graph. */
364 char node_name[32];
365 /* element of the list of named nodes building the graph */
366 QTAILQ_ENTRY(BlockDriverState) node_list;
367 /* Device name is the name associated with the "drive" the guest sees */
368 char device_name[32];
369 /* element of the list of "drives" the guest sees */
370 QTAILQ_ENTRY(BlockDriverState) device_list;
371 QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps;
372 int refcnt;
373
374 QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
375
376 /* operation blockers */
377 QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX];
378
379 /* long-running background operation */
380 BlockJob *job;
381
382 QDict *options;
383 BlockdevDetectZeroesOptions detect_zeroes;
384
385 /* The error object in use for blocking operations on backing_hd */
386 Error *backing_blocker;
387 };
388
389 int get_tmp_filename(char *filename, int size);
390
391 void bdrv_set_io_limits(BlockDriverState *bs,
392 ThrottleConfig *cfg);
393
394
395 /**
396 * bdrv_add_before_write_notifier:
397 *
398 * Register a callback that is invoked before write requests are processed but
399 * after any throttling or waiting for overlapping requests.
400 */
401 void bdrv_add_before_write_notifier(BlockDriverState *bs,
402 NotifierWithReturn *notifier);
403
404 /**
405 * bdrv_detach_aio_context:
406 *
407 * May be called from .bdrv_detach_aio_context() to detach children from the
408 * current #AioContext. This is only needed by block drivers that manage their
409 * own children. Both ->file and ->backing_hd are automatically handled and
410 * block drivers should not call this function on them explicitly.
411 */
412 void bdrv_detach_aio_context(BlockDriverState *bs);
413
414 /**
415 * bdrv_attach_aio_context:
416 *
417 * May be called from .bdrv_attach_aio_context() to attach children to the new
418 * #AioContext. This is only needed by block drivers that manage their own
419 * children. Both ->file and ->backing_hd are automatically handled and block
420 * drivers should not call this function on them explicitly.
421 */
422 void bdrv_attach_aio_context(BlockDriverState *bs,
423 AioContext *new_context);
424
425 #ifdef _WIN32
426 int is_windows_drive(const char *filename);
427 #endif
428 void bdrv_emit_qmp_error_event(const BlockDriverState *bdrv,
429 enum MonitorEvent ev,
430 BlockErrorAction action, bool is_read);
431
432 /**
433 * stream_start:
434 * @bs: Block device to operate on.
435 * @base: Block device that will become the new base, or %NULL to
436 * flatten the whole backing file chain onto @bs.
437 * @base_id: The file name that will be written to @bs as the new
438 * backing file if the job completes. Ignored if @base is %NULL.
439 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
440 * @on_error: The action to take upon error.
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, BlockdevOnError on_error,
453 BlockDriverCompletionFunc *cb,
454 void *opaque, Error **errp);
455
456 /**
457 * commit_start:
458 * @bs: Active block device.
459 * @top: Top block device to be committed.
460 * @base: Block device that will be written into, and become the new top.
461 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
462 * @on_error: The action to take upon error.
463 * @cb: Completion function for the job.
464 * @opaque: Opaque pointer value passed to @cb.
465 * @errp: Error object.
466 *
467 */
468 void commit_start(BlockDriverState *bs, BlockDriverState *base,
469 BlockDriverState *top, int64_t speed,
470 BlockdevOnError on_error, BlockDriverCompletionFunc *cb,
471 void *opaque, Error **errp);
472 /**
473 * commit_active_start:
474 * @bs: Active block device to be committed.
475 * @base: Block device that will be written into, and become the new top.
476 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
477 * @on_error: The action to take upon error.
478 * @cb: Completion function for the job.
479 * @opaque: Opaque pointer value passed to @cb.
480 * @errp: Error object.
481 *
482 */
483 void commit_active_start(BlockDriverState *bs, BlockDriverState *base,
484 int64_t speed,
485 BlockdevOnError on_error,
486 BlockDriverCompletionFunc *cb,
487 void *opaque, Error **errp);
488 /*
489 * mirror_start:
490 * @bs: Block device to operate on.
491 * @target: Block device to write to.
492 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
493 * @granularity: The chosen granularity for the dirty bitmap.
494 * @buf_size: The amount of data that can be in flight at one time.
495 * @mode: Whether to collapse all images in the chain to the target.
496 * @on_source_error: The action to take upon error reading from the source.
497 * @on_target_error: The action to take upon error writing to the target.
498 * @cb: Completion function for the job.
499 * @opaque: Opaque pointer value passed to @cb.
500 * @errp: Error object.
501 *
502 * Start a mirroring operation on @bs. Clusters that are allocated
503 * in @bs will be written to @bs until the job is cancelled or
504 * manually completed. At the end of a successful mirroring job,
505 * @bs will be switched to read from @target.
506 */
507 void mirror_start(BlockDriverState *bs, BlockDriverState *target,
508 int64_t speed, int64_t granularity, int64_t buf_size,
509 MirrorSyncMode mode, BlockdevOnError on_source_error,
510 BlockdevOnError on_target_error,
511 BlockDriverCompletionFunc *cb,
512 void *opaque, Error **errp);
513
514 /*
515 * backup_start:
516 * @bs: Block device to operate on.
517 * @target: Block device to write to.
518 * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
519 * @sync_mode: What parts of the disk image should be copied to the destination.
520 * @on_source_error: The action to take upon error reading from the source.
521 * @on_target_error: The action to take upon error writing to the target.
522 * @cb: Completion function for the job.
523 * @opaque: Opaque pointer value passed to @cb.
524 *
525 * Start a backup operation on @bs. Clusters in @bs are written to @target
526 * until the job is cancelled or manually completed.
527 */
528 void backup_start(BlockDriverState *bs, BlockDriverState *target,
529 int64_t speed, MirrorSyncMode sync_mode,
530 BlockdevOnError on_source_error,
531 BlockdevOnError on_target_error,
532 BlockDriverCompletionFunc *cb, void *opaque,
533 Error **errp);
534
535 #endif /* BLOCK_INT_H */