xref: /openbmc/qemu/include/block/block_int.h (revision da34e65c)
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/accounting.h"
28 #include "block/block.h"
29 #include "block/throttle-groups.h"
30 #include "qemu/option.h"
31 #include "qemu/queue.h"
32 #include "qemu/coroutine.h"
33 #include "qemu/timer.h"
34 #include "qapi-types.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 #define BLOCK_OPT_NOCOW             "nocow"
58 #define BLOCK_OPT_OBJECT_SIZE       "object_size"
59 #define BLOCK_OPT_REFCOUNT_BITS     "refcount_bits"
60 
61 #define BLOCK_PROBE_BUF_SIZE        512
62 
63 enum BdrvTrackedRequestType {
64     BDRV_TRACKED_READ,
65     BDRV_TRACKED_WRITE,
66     BDRV_TRACKED_FLUSH,
67     BDRV_TRACKED_IOCTL,
68     BDRV_TRACKED_DISCARD,
69 };
70 
71 typedef struct BdrvTrackedRequest {
72     BlockDriverState *bs;
73     int64_t offset;
74     unsigned int bytes;
75     enum BdrvTrackedRequestType type;
76 
77     bool serialising;
78     int64_t overlap_offset;
79     unsigned int overlap_bytes;
80 
81     QLIST_ENTRY(BdrvTrackedRequest) list;
82     Coroutine *co; /* owner, used for deadlock detection */
83     CoQueue wait_queue; /* coroutines blocked on this request */
84 
85     struct BdrvTrackedRequest *waiting_for;
86 } BdrvTrackedRequest;
87 
88 struct BlockDriver {
89     const char *format_name;
90     int instance_size;
91 
92     /* set to true if the BlockDriver is a block filter */
93     bool is_filter;
94     /* for snapshots block filter like Quorum can implement the
95      * following recursive callback.
96      * It's purpose is to recurse on the filter children while calling
97      * bdrv_recurse_is_first_non_filter on them.
98      * For a sample implementation look in the future Quorum block filter.
99      */
100     bool (*bdrv_recurse_is_first_non_filter)(BlockDriverState *bs,
101                                              BlockDriverState *candidate);
102 
103     int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename);
104     int (*bdrv_probe_device)(const char *filename);
105 
106     /* Any driver implementing this callback is expected to be able to handle
107      * NULL file names in its .bdrv_open() implementation */
108     void (*bdrv_parse_filename)(const char *filename, QDict *options, Error **errp);
109     /* Drivers not implementing bdrv_parse_filename nor bdrv_open should have
110      * this field set to true, except ones that are defined only by their
111      * child's bs.
112      * An example of the last type will be the quorum block driver.
113      */
114     bool bdrv_needs_filename;
115 
116     /* Set if a driver can support backing files */
117     bool supports_backing;
118 
119     /* For handling image reopen for split or non-split files */
120     int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state,
121                                BlockReopenQueue *queue, Error **errp);
122     void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state);
123     void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state);
124     void (*bdrv_join_options)(QDict *options, QDict *old_options);
125 
126     int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags,
127                      Error **errp);
128     int (*bdrv_file_open)(BlockDriverState *bs, QDict *options, int flags,
129                           Error **errp);
130     int (*bdrv_read)(BlockDriverState *bs, int64_t sector_num,
131                      uint8_t *buf, int nb_sectors);
132     int (*bdrv_write)(BlockDriverState *bs, int64_t sector_num,
133                       const uint8_t *buf, int nb_sectors);
134     void (*bdrv_close)(BlockDriverState *bs);
135     int (*bdrv_create)(const char *filename, QemuOpts *opts, Error **errp);
136     int (*bdrv_set_key)(BlockDriverState *bs, const char *key);
137     int (*bdrv_make_empty)(BlockDriverState *bs);
138 
139     void (*bdrv_refresh_filename)(BlockDriverState *bs, QDict *options);
140 
141     /* aio */
142     BlockAIOCB *(*bdrv_aio_readv)(BlockDriverState *bs,
143         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
144         BlockCompletionFunc *cb, void *opaque);
145     BlockAIOCB *(*bdrv_aio_writev)(BlockDriverState *bs,
146         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
147         BlockCompletionFunc *cb, void *opaque);
148     BlockAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs,
149         BlockCompletionFunc *cb, void *opaque);
150     BlockAIOCB *(*bdrv_aio_discard)(BlockDriverState *bs,
151         int64_t sector_num, int nb_sectors,
152         BlockCompletionFunc *cb, void *opaque);
153 
154     int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs,
155         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
156     int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs,
157         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
158     /*
159      * Efficiently zero a region of the disk image.  Typically an image format
160      * would use a compact metadata representation to implement this.  This
161      * function pointer may be NULL and .bdrv_co_writev() will be called
162      * instead.
163      */
164     int coroutine_fn (*bdrv_co_write_zeroes)(BlockDriverState *bs,
165         int64_t sector_num, int nb_sectors, BdrvRequestFlags flags);
166     int coroutine_fn (*bdrv_co_discard)(BlockDriverState *bs,
167         int64_t sector_num, int nb_sectors);
168     int64_t coroutine_fn (*bdrv_co_get_block_status)(BlockDriverState *bs,
169         int64_t sector_num, int nb_sectors, int *pnum,
170         BlockDriverState **file);
171 
172     /*
173      * Invalidate any cached meta-data.
174      */
175     void (*bdrv_invalidate_cache)(BlockDriverState *bs, Error **errp);
176     int (*bdrv_inactivate)(BlockDriverState *bs);
177 
178     /*
179      * Flushes all data that was already written to the OS all the way down to
180      * the disk (for example raw-posix calls fsync()).
181      */
182     int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs);
183 
184     /*
185      * Flushes all internal caches to the OS. The data may still sit in a
186      * writeback cache of the host OS, but it will survive a crash of the qemu
187      * process.
188      */
189     int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs);
190 
191     const char *protocol_name;
192     int (*bdrv_truncate)(BlockDriverState *bs, int64_t offset);
193 
194     int64_t (*bdrv_getlength)(BlockDriverState *bs);
195     bool has_variable_length;
196     int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
197 
198     int (*bdrv_write_compressed)(BlockDriverState *bs, int64_t sector_num,
199                                  const uint8_t *buf, int nb_sectors);
200 
201     int (*bdrv_snapshot_create)(BlockDriverState *bs,
202                                 QEMUSnapshotInfo *sn_info);
203     int (*bdrv_snapshot_goto)(BlockDriverState *bs,
204                               const char *snapshot_id);
205     int (*bdrv_snapshot_delete)(BlockDriverState *bs,
206                                 const char *snapshot_id,
207                                 const char *name,
208                                 Error **errp);
209     int (*bdrv_snapshot_list)(BlockDriverState *bs,
210                               QEMUSnapshotInfo **psn_info);
211     int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
212                                   const char *snapshot_id,
213                                   const char *name,
214                                   Error **errp);
215     int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
216     ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs);
217 
218     int (*bdrv_save_vmstate)(BlockDriverState *bs, QEMUIOVector *qiov,
219                              int64_t pos);
220     int (*bdrv_load_vmstate)(BlockDriverState *bs, uint8_t *buf,
221                              int64_t pos, int size);
222 
223     int (*bdrv_change_backing_file)(BlockDriverState *bs,
224         const char *backing_file, const char *backing_fmt);
225 
226     /* removable device specific */
227     bool (*bdrv_is_inserted)(BlockDriverState *bs);
228     int (*bdrv_media_changed)(BlockDriverState *bs);
229     void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
230     void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
231 
232     /* to control generic scsi devices */
233     BlockAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
234         unsigned long int req, void *buf,
235         BlockCompletionFunc *cb, void *opaque);
236 
237     /* List of options for creating images, terminated by name == NULL */
238     QemuOptsList *create_opts;
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     int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts,
248                               BlockDriverAmendStatusCB *status_cb,
249                               void *cb_opaque);
250 
251     void (*bdrv_debug_event)(BlockDriverState *bs, BlkdebugEvent event);
252 
253     /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
254     int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
255         const char *tag);
256     int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs,
257         const char *tag);
258     int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
259     bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
260 
261     void (*bdrv_refresh_limits)(BlockDriverState *bs, Error **errp);
262 
263     /*
264      * Returns 1 if newly created images are guaranteed to contain only
265      * zeros, 0 otherwise.
266      */
267     int (*bdrv_has_zero_init)(BlockDriverState *bs);
268 
269     /* Remove fd handlers, timers, and other event loop callbacks so the event
270      * loop is no longer in use.  Called with no in-flight requests and in
271      * depth-first traversal order with parents before child nodes.
272      */
273     void (*bdrv_detach_aio_context)(BlockDriverState *bs);
274 
275     /* Add fd handlers, timers, and other event loop callbacks so I/O requests
276      * can be processed again.  Called with no in-flight requests and in
277      * depth-first traversal order with child nodes before parent nodes.
278      */
279     void (*bdrv_attach_aio_context)(BlockDriverState *bs,
280                                     AioContext *new_context);
281 
282     /* io queue for linux-aio */
283     void (*bdrv_io_plug)(BlockDriverState *bs);
284     void (*bdrv_io_unplug)(BlockDriverState *bs);
285     void (*bdrv_flush_io_queue)(BlockDriverState *bs);
286 
287     /**
288      * Try to get @bs's logical and physical block size.
289      * On success, store them in @bsz and return zero.
290      * On failure, return negative errno.
291      */
292     int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz);
293     /**
294      * Try to get @bs's geometry (cyls, heads, sectors)
295      * On success, store them in @geo and return 0.
296      * On failure return -errno.
297      * Only drivers that want to override guest geometry implement this
298      * callback; see hd_geometry_guess().
299      */
300     int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo);
301 
302     /**
303      * Drain and stop any internal sources of requests in the driver, and
304      * remain so until next I/O callback (e.g. bdrv_co_writev) is called.
305      */
306     void (*bdrv_drain)(BlockDriverState *bs);
307 
308     QLIST_ENTRY(BlockDriver) list;
309 };
310 
311 typedef struct BlockLimits {
312     /* maximum number of sectors that can be discarded at once */
313     int max_discard;
314 
315     /* optimal alignment for discard requests in sectors */
316     int64_t discard_alignment;
317 
318     /* maximum number of sectors that can zeroized at once */
319     int max_write_zeroes;
320 
321     /* optimal alignment for write zeroes requests in sectors */
322     int64_t write_zeroes_alignment;
323 
324     /* optimal transfer length in sectors */
325     int opt_transfer_length;
326 
327     /* maximal transfer length in sectors */
328     int max_transfer_length;
329 
330     /* memory alignment so that no bounce buffer is needed */
331     size_t min_mem_alignment;
332 
333     /* memory alignment for bounce buffer */
334     size_t opt_mem_alignment;
335 
336     /* maximum number of iovec elements */
337     int max_iov;
338 } BlockLimits;
339 
340 typedef struct BdrvOpBlocker BdrvOpBlocker;
341 
342 typedef struct BdrvAioNotifier {
343     void (*attached_aio_context)(AioContext *new_context, void *opaque);
344     void (*detach_aio_context)(void *opaque);
345 
346     void *opaque;
347 
348     QLIST_ENTRY(BdrvAioNotifier) list;
349 } BdrvAioNotifier;
350 
351 struct BdrvChildRole {
352     void (*inherit_options)(int *child_flags, QDict *child_options,
353                             int parent_flags, QDict *parent_options);
354 };
355 
356 extern const BdrvChildRole child_file;
357 extern const BdrvChildRole child_format;
358 
359 struct BdrvChild {
360     BlockDriverState *bs;
361     char *name;
362     const BdrvChildRole *role;
363     QLIST_ENTRY(BdrvChild) next;
364     QLIST_ENTRY(BdrvChild) next_parent;
365 };
366 
367 /*
368  * Note: the function bdrv_append() copies and swaps contents of
369  * BlockDriverStates, so if you add new fields to this struct, please
370  * inspect bdrv_append() to determine if the new fields need to be
371  * copied as well.
372  */
373 struct BlockDriverState {
374     int64_t total_sectors; /* if we are reading a disk image, give its
375                               size in sectors */
376     int read_only; /* if true, the media is read only */
377     int open_flags; /* flags used to open the file, re-used for re-open */
378     int encrypted; /* if true, the media is encrypted */
379     int valid_key; /* if true, a valid encryption key has been set */
380     int sg;        /* if true, the device is a /dev/sg* */
381     int copy_on_read; /* if true, copy read backing sectors into image
382                          note this is a reference count */
383     bool probed;
384 
385     BlockDriver *drv; /* NULL means no media */
386     void *opaque;
387 
388     BlockBackend *blk;          /* owning backend, if any */
389 
390     AioContext *aio_context; /* event loop used for fd handlers, timers, etc */
391     /* long-running tasks intended to always use the same AioContext as this
392      * BDS may register themselves in this list to be notified of changes
393      * regarding this BDS's context */
394     QLIST_HEAD(, BdrvAioNotifier) aio_notifiers;
395 
396     char filename[PATH_MAX];
397     char backing_file[PATH_MAX]; /* if non zero, the image is a diff of
398                                     this file image */
399     char backing_format[16]; /* if non-zero and backing_file exists */
400 
401     QDict *full_open_options;
402     char exact_filename[PATH_MAX];
403 
404     BdrvChild *backing;
405     BdrvChild *file;
406 
407     /* Callback before write request is processed */
408     NotifierWithReturnList before_write_notifiers;
409 
410     /* number of in-flight serialising requests */
411     unsigned int serialising_in_flight;
412 
413     /* I/O throttling.
414      * throttle_state tells us if this BDS has I/O limits configured.
415      * io_limits_enabled tells us if they are currently being
416      * enforced, but it can be temporarily set to false */
417     CoQueue      throttled_reqs[2];
418     bool         io_limits_enabled;
419     /* The following fields are protected by the ThrottleGroup lock.
420      * See the ThrottleGroup documentation for details. */
421     ThrottleState *throttle_state;
422     ThrottleTimers throttle_timers;
423     unsigned       pending_reqs[2];
424     QLIST_ENTRY(BlockDriverState) round_robin;
425 
426     /* Offset after the highest byte written to */
427     uint64_t wr_highest_offset;
428 
429     /* I/O Limits */
430     BlockLimits bl;
431 
432     /* Whether produces zeros when read beyond eof */
433     bool zero_beyond_eof;
434 
435     /* Alignment requirement for offset/length of I/O requests */
436     unsigned int request_alignment;
437 
438     /* do we need to tell the quest if we have a volatile write cache? */
439     int enable_write_cache;
440 
441     /* the following member gives a name to every node on the bs graph. */
442     char node_name[32];
443     /* element of the list of named nodes building the graph */
444     QTAILQ_ENTRY(BlockDriverState) node_list;
445     /* element of the list of all BlockDriverStates (all_bdrv_states) */
446     QTAILQ_ENTRY(BlockDriverState) bs_list;
447     /* element of the list of monitor-owned BDS */
448     QTAILQ_ENTRY(BlockDriverState) monitor_list;
449     QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps;
450     int refcnt;
451 
452     QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
453 
454     /* operation blockers */
455     QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX];
456 
457     /* long-running background operation */
458     BlockJob *job;
459 
460     /* The node that this node inherited default options from (and a reopen on
461      * which can affect this node by changing these defaults). This is always a
462      * parent node of this node. */
463     BlockDriverState *inherits_from;
464     QLIST_HEAD(, BdrvChild) children;
465     QLIST_HEAD(, BdrvChild) parents;
466 
467     QDict *options;
468     QDict *explicit_options;
469     BlockdevDetectZeroesOptions detect_zeroes;
470 
471     /* The error object in use for blocking operations on backing_hd */
472     Error *backing_blocker;
473 
474     /* threshold limit for writes, in bytes. "High water mark". */
475     uint64_t write_threshold_offset;
476     NotifierWithReturn write_threshold_notifier;
477 
478     int quiesce_counter;
479 };
480 
481 struct BlockBackendRootState {
482     int open_flags;
483     bool read_only;
484     BlockdevDetectZeroesOptions detect_zeroes;
485 
486     char *throttle_group;
487     ThrottleState *throttle_state;
488 };
489 
490 static inline BlockDriverState *backing_bs(BlockDriverState *bs)
491 {
492     return bs->backing ? bs->backing->bs : NULL;
493 }
494 
495 
496 /* Essential block drivers which must always be statically linked into qemu, and
497  * which therefore can be accessed without using bdrv_find_format() */
498 extern BlockDriver bdrv_file;
499 extern BlockDriver bdrv_raw;
500 extern BlockDriver bdrv_qcow2;
501 
502 /**
503  * bdrv_setup_io_funcs:
504  *
505  * Prepare a #BlockDriver for I/O request processing by populating
506  * unimplemented coroutine and AIO interfaces with generic wrapper functions
507  * that fall back to implemented interfaces.
508  */
509 void bdrv_setup_io_funcs(BlockDriver *bdrv);
510 
511 int coroutine_fn bdrv_co_do_preadv(BlockDriverState *bs,
512     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
513     BdrvRequestFlags flags);
514 int coroutine_fn bdrv_co_do_pwritev(BlockDriverState *bs,
515     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
516     BdrvRequestFlags flags);
517 
518 int get_tmp_filename(char *filename, int size);
519 BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size,
520                             const char *filename);
521 
522 void bdrv_set_io_limits(BlockDriverState *bs,
523                         ThrottleConfig *cfg);
524 
525 
526 /**
527  * bdrv_add_before_write_notifier:
528  *
529  * Register a callback that is invoked before write requests are processed but
530  * after any throttling or waiting for overlapping requests.
531  */
532 void bdrv_add_before_write_notifier(BlockDriverState *bs,
533                                     NotifierWithReturn *notifier);
534 
535 /**
536  * bdrv_detach_aio_context:
537  *
538  * May be called from .bdrv_detach_aio_context() to detach children from the
539  * current #AioContext.  This is only needed by block drivers that manage their
540  * own children.  Both ->file and ->backing are automatically handled and
541  * block drivers should not call this function on them explicitly.
542  */
543 void bdrv_detach_aio_context(BlockDriverState *bs);
544 
545 /**
546  * bdrv_attach_aio_context:
547  *
548  * May be called from .bdrv_attach_aio_context() to attach children to the new
549  * #AioContext.  This is only needed by block drivers that manage their own
550  * children.  Both ->file and ->backing are automatically handled and block
551  * drivers should not call this function on them explicitly.
552  */
553 void bdrv_attach_aio_context(BlockDriverState *bs,
554                              AioContext *new_context);
555 
556 /**
557  * bdrv_add_aio_context_notifier:
558  *
559  * If a long-running job intends to be always run in the same AioContext as a
560  * certain BDS, it may use this function to be notified of changes regarding the
561  * association of the BDS to an AioContext.
562  *
563  * attached_aio_context() is called after the target BDS has been attached to a
564  * new AioContext; detach_aio_context() is called before the target BDS is being
565  * detached from its old AioContext.
566  */
567 void bdrv_add_aio_context_notifier(BlockDriverState *bs,
568         void (*attached_aio_context)(AioContext *new_context, void *opaque),
569         void (*detach_aio_context)(void *opaque), void *opaque);
570 
571 /**
572  * bdrv_remove_aio_context_notifier:
573  *
574  * Unsubscribe of change notifications regarding the BDS's AioContext. The
575  * parameters given here have to be the same as those given to
576  * bdrv_add_aio_context_notifier().
577  */
578 void bdrv_remove_aio_context_notifier(BlockDriverState *bs,
579                                       void (*aio_context_attached)(AioContext *,
580                                                                    void *),
581                                       void (*aio_context_detached)(void *),
582                                       void *opaque);
583 
584 #ifdef _WIN32
585 int is_windows_drive(const char *filename);
586 #endif
587 
588 /**
589  * stream_start:
590  * @bs: Block device to operate on.
591  * @base: Block device that will become the new base, or %NULL to
592  * flatten the whole backing file chain onto @bs.
593  * @base_id: The file name that will be written to @bs as the new
594  * backing file if the job completes.  Ignored if @base is %NULL.
595  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
596  * @on_error: The action to take upon error.
597  * @cb: Completion function for the job.
598  * @opaque: Opaque pointer value passed to @cb.
599  * @errp: Error object.
600  *
601  * Start a streaming operation on @bs.  Clusters that are unallocated
602  * in @bs, but allocated in any image between @base and @bs (both
603  * exclusive) will be written to @bs.  At the end of a successful
604  * streaming job, the backing file of @bs will be changed to
605  * @base_id in the written image and to @base in the live BlockDriverState.
606  */
607 void stream_start(BlockDriverState *bs, BlockDriverState *base,
608                   const char *base_id, int64_t speed, BlockdevOnError on_error,
609                   BlockCompletionFunc *cb,
610                   void *opaque, Error **errp);
611 
612 /**
613  * commit_start:
614  * @bs: Active block device.
615  * @top: Top block device to be committed.
616  * @base: Block device that will be written into, and become the new top.
617  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
618  * @on_error: The action to take upon error.
619  * @cb: Completion function for the job.
620  * @opaque: Opaque pointer value passed to @cb.
621  * @backing_file_str: String to use as the backing file in @top's overlay
622  * @errp: Error object.
623  *
624  */
625 void commit_start(BlockDriverState *bs, BlockDriverState *base,
626                  BlockDriverState *top, int64_t speed,
627                  BlockdevOnError on_error, BlockCompletionFunc *cb,
628                  void *opaque, const char *backing_file_str, Error **errp);
629 /**
630  * commit_active_start:
631  * @bs: Active block device to be committed.
632  * @base: Block device that will be written into, and become the new top.
633  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
634  * @on_error: The action to take upon error.
635  * @cb: Completion function for the job.
636  * @opaque: Opaque pointer value passed to @cb.
637  * @errp: Error object.
638  *
639  */
640 void commit_active_start(BlockDriverState *bs, BlockDriverState *base,
641                          int64_t speed,
642                          BlockdevOnError on_error,
643                          BlockCompletionFunc *cb,
644                          void *opaque, Error **errp);
645 /*
646  * mirror_start:
647  * @bs: Block device to operate on.
648  * @target: Block device to write to.
649  * @replaces: Block graph node name to replace once the mirror is done. Can
650  *            only be used when full mirroring is selected.
651  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
652  * @granularity: The chosen granularity for the dirty bitmap.
653  * @buf_size: The amount of data that can be in flight at one time.
654  * @mode: Whether to collapse all images in the chain to the target.
655  * @on_source_error: The action to take upon error reading from the source.
656  * @on_target_error: The action to take upon error writing to the target.
657  * @unmap: Whether to unmap target where source sectors only contain zeroes.
658  * @cb: Completion function for the job.
659  * @opaque: Opaque pointer value passed to @cb.
660  * @errp: Error object.
661  *
662  * Start a mirroring operation on @bs.  Clusters that are allocated
663  * in @bs will be written to @bs until the job is cancelled or
664  * manually completed.  At the end of a successful mirroring job,
665  * @bs will be switched to read from @target.
666  */
667 void mirror_start(BlockDriverState *bs, BlockDriverState *target,
668                   const char *replaces,
669                   int64_t speed, uint32_t granularity, int64_t buf_size,
670                   MirrorSyncMode mode, BlockdevOnError on_source_error,
671                   BlockdevOnError on_target_error,
672                   bool unmap,
673                   BlockCompletionFunc *cb,
674                   void *opaque, Error **errp);
675 
676 /*
677  * backup_start:
678  * @bs: Block device to operate on.
679  * @target: Block device to write to.
680  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
681  * @sync_mode: What parts of the disk image should be copied to the destination.
682  * @sync_bitmap: The dirty bitmap if sync_mode is MIRROR_SYNC_MODE_INCREMENTAL.
683  * @on_source_error: The action to take upon error reading from the source.
684  * @on_target_error: The action to take upon error writing to the target.
685  * @cb: Completion function for the job.
686  * @opaque: Opaque pointer value passed to @cb.
687  * @txn: Transaction that this job is part of (may be NULL).
688  *
689  * Start a backup operation on @bs.  Clusters in @bs are written to @target
690  * until the job is cancelled or manually completed.
691  */
692 void backup_start(BlockDriverState *bs, BlockDriverState *target,
693                   int64_t speed, MirrorSyncMode sync_mode,
694                   BdrvDirtyBitmap *sync_bitmap,
695                   BlockdevOnError on_source_error,
696                   BlockdevOnError on_target_error,
697                   BlockCompletionFunc *cb, void *opaque,
698                   BlockJobTxn *txn, Error **errp);
699 
700 void hmp_drive_add_node(Monitor *mon, const char *optstr);
701 
702 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
703                                   const char *child_name,
704                                   const BdrvChildRole *child_role);
705 void bdrv_root_unref_child(BdrvChild *child);
706 
707 void blk_set_bs(BlockBackend *blk, BlockDriverState *bs);
708 
709 void blk_dev_change_media_cb(BlockBackend *blk, bool load);
710 bool blk_dev_has_removable_media(BlockBackend *blk);
711 bool blk_dev_has_tray(BlockBackend *blk);
712 void blk_dev_eject_request(BlockBackend *blk, bool force);
713 bool blk_dev_is_tray_open(BlockBackend *blk);
714 bool blk_dev_is_medium_locked(BlockBackend *blk);
715 void blk_dev_resize_cb(BlockBackend *blk);
716 
717 void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector, int nr_sectors);
718 bool bdrv_requests_pending(BlockDriverState *bs);
719 
720 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap **out);
721 void bdrv_undo_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap *in);
722 
723 void blockdev_close_all_bdrv_states(void);
724 
725 #endif /* BLOCK_INT_H */
726