xref: /openbmc/qemu/include/block/block_int.h (revision 58a0067a)
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 for all layers by calling bdrv_co_flush for underlying
180      * layers, if needed. This function is needed for deterministic
181      * synchronization of the flush finishing callback.
182      */
183     int coroutine_fn (*bdrv_co_flush)(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 
201     int64_t (*bdrv_getlength)(BlockDriverState *bs);
202     bool has_variable_length;
203     int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
204 
205     int (*bdrv_write_compressed)(BlockDriverState *bs, int64_t sector_num,
206                                  const uint8_t *buf, int nb_sectors);
207 
208     int (*bdrv_snapshot_create)(BlockDriverState *bs,
209                                 QEMUSnapshotInfo *sn_info);
210     int (*bdrv_snapshot_goto)(BlockDriverState *bs,
211                               const char *snapshot_id);
212     int (*bdrv_snapshot_delete)(BlockDriverState *bs,
213                                 const char *snapshot_id,
214                                 const char *name,
215                                 Error **errp);
216     int (*bdrv_snapshot_list)(BlockDriverState *bs,
217                               QEMUSnapshotInfo **psn_info);
218     int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
219                                   const char *snapshot_id,
220                                   const char *name,
221                                   Error **errp);
222     int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
223     ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs);
224 
225     int (*bdrv_save_vmstate)(BlockDriverState *bs, QEMUIOVector *qiov,
226                              int64_t pos);
227     int (*bdrv_load_vmstate)(BlockDriverState *bs, uint8_t *buf,
228                              int64_t pos, int size);
229 
230     int (*bdrv_change_backing_file)(BlockDriverState *bs,
231         const char *backing_file, const char *backing_fmt);
232 
233     /* removable device specific */
234     bool (*bdrv_is_inserted)(BlockDriverState *bs);
235     int (*bdrv_media_changed)(BlockDriverState *bs);
236     void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
237     void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
238 
239     /* to control generic scsi devices */
240     BlockAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
241         unsigned long int req, void *buf,
242         BlockCompletionFunc *cb, void *opaque);
243 
244     /* List of options for creating images, terminated by name == NULL */
245     QemuOptsList *create_opts;
246 
247     /*
248      * Returns 0 for completed check, -errno for internal errors.
249      * The check results are stored in result.
250      */
251     int (*bdrv_check)(BlockDriverState* bs, BdrvCheckResult *result,
252         BdrvCheckMode fix);
253 
254     int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts,
255                               BlockDriverAmendStatusCB *status_cb,
256                               void *cb_opaque);
257 
258     void (*bdrv_debug_event)(BlockDriverState *bs, BlkdebugEvent event);
259 
260     /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
261     int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
262         const char *tag);
263     int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs,
264         const char *tag);
265     int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
266     bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
267 
268     void (*bdrv_refresh_limits)(BlockDriverState *bs, Error **errp);
269 
270     /*
271      * Returns 1 if newly created images are guaranteed to contain only
272      * zeros, 0 otherwise.
273      */
274     int (*bdrv_has_zero_init)(BlockDriverState *bs);
275 
276     /* Remove fd handlers, timers, and other event loop callbacks so the event
277      * loop is no longer in use.  Called with no in-flight requests and in
278      * depth-first traversal order with parents before child nodes.
279      */
280     void (*bdrv_detach_aio_context)(BlockDriverState *bs);
281 
282     /* Add fd handlers, timers, and other event loop callbacks so I/O requests
283      * can be processed again.  Called with no in-flight requests and in
284      * depth-first traversal order with child nodes before parent nodes.
285      */
286     void (*bdrv_attach_aio_context)(BlockDriverState *bs,
287                                     AioContext *new_context);
288 
289     /* io queue for linux-aio */
290     void (*bdrv_io_plug)(BlockDriverState *bs);
291     void (*bdrv_io_unplug)(BlockDriverState *bs);
292     void (*bdrv_flush_io_queue)(BlockDriverState *bs);
293 
294     /**
295      * Try to get @bs's logical and physical block size.
296      * On success, store them in @bsz and return zero.
297      * On failure, return negative errno.
298      */
299     int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz);
300     /**
301      * Try to get @bs's geometry (cyls, heads, sectors)
302      * On success, store them in @geo and return 0.
303      * On failure return -errno.
304      * Only drivers that want to override guest geometry implement this
305      * callback; see hd_geometry_guess().
306      */
307     int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo);
308 
309     /**
310      * Drain and stop any internal sources of requests in the driver, and
311      * remain so until next I/O callback (e.g. bdrv_co_writev) is called.
312      */
313     void (*bdrv_drain)(BlockDriverState *bs);
314 
315     QLIST_ENTRY(BlockDriver) list;
316 };
317 
318 typedef struct BlockLimits {
319     /* maximum number of sectors that can be discarded at once */
320     int max_discard;
321 
322     /* optimal alignment for discard requests in sectors */
323     int64_t discard_alignment;
324 
325     /* maximum number of sectors that can zeroized at once */
326     int max_write_zeroes;
327 
328     /* optimal alignment for write zeroes requests in sectors */
329     int64_t write_zeroes_alignment;
330 
331     /* optimal transfer length in sectors */
332     int opt_transfer_length;
333 
334     /* maximal transfer length in sectors */
335     int max_transfer_length;
336 
337     /* memory alignment so that no bounce buffer is needed */
338     size_t min_mem_alignment;
339 
340     /* memory alignment for bounce buffer */
341     size_t opt_mem_alignment;
342 
343     /* maximum number of iovec elements */
344     int max_iov;
345 } BlockLimits;
346 
347 typedef struct BdrvOpBlocker BdrvOpBlocker;
348 
349 typedef struct BdrvAioNotifier {
350     void (*attached_aio_context)(AioContext *new_context, void *opaque);
351     void (*detach_aio_context)(void *opaque);
352 
353     void *opaque;
354 
355     QLIST_ENTRY(BdrvAioNotifier) list;
356 } BdrvAioNotifier;
357 
358 struct BdrvChildRole {
359     void (*inherit_options)(int *child_flags, QDict *child_options,
360                             int parent_flags, QDict *parent_options);
361 };
362 
363 extern const BdrvChildRole child_file;
364 extern const BdrvChildRole child_format;
365 
366 struct BdrvChild {
367     BlockDriverState *bs;
368     char *name;
369     const BdrvChildRole *role;
370     QLIST_ENTRY(BdrvChild) next;
371     QLIST_ENTRY(BdrvChild) next_parent;
372 };
373 
374 /*
375  * Note: the function bdrv_append() copies and swaps contents of
376  * BlockDriverStates, so if you add new fields to this struct, please
377  * inspect bdrv_append() to determine if the new fields need to be
378  * copied as well.
379  */
380 struct BlockDriverState {
381     int64_t total_sectors; /* if we are reading a disk image, give its
382                               size in sectors */
383     int read_only; /* if true, the media is read only */
384     int open_flags; /* flags used to open the file, re-used for re-open */
385     int encrypted; /* if true, the media is encrypted */
386     int valid_key; /* if true, a valid encryption key has been set */
387     int sg;        /* if true, the device is a /dev/sg* */
388     int copy_on_read; /* if true, copy read backing sectors into image
389                          note this is a reference count */
390     bool probed;
391 
392     BlockDriver *drv; /* NULL means no media */
393     void *opaque;
394 
395     BlockBackend *blk;          /* owning backend, if any */
396 
397     AioContext *aio_context; /* event loop used for fd handlers, timers, etc */
398     /* long-running tasks intended to always use the same AioContext as this
399      * BDS may register themselves in this list to be notified of changes
400      * regarding this BDS's context */
401     QLIST_HEAD(, BdrvAioNotifier) aio_notifiers;
402 
403     char filename[PATH_MAX];
404     char backing_file[PATH_MAX]; /* if non zero, the image is a diff of
405                                     this file image */
406     char backing_format[16]; /* if non-zero and backing_file exists */
407 
408     QDict *full_open_options;
409     char exact_filename[PATH_MAX];
410 
411     BdrvChild *backing;
412     BdrvChild *file;
413 
414     /* Callback before write request is processed */
415     NotifierWithReturnList before_write_notifiers;
416 
417     /* number of in-flight serialising requests */
418     unsigned int serialising_in_flight;
419 
420     /* I/O throttling.
421      * throttle_state tells us if this BDS has I/O limits configured.
422      * io_limits_enabled tells us if they are currently being
423      * enforced, but it can be temporarily set to false */
424     CoQueue      throttled_reqs[2];
425     bool         io_limits_enabled;
426     /* The following fields are protected by the ThrottleGroup lock.
427      * See the ThrottleGroup documentation for details. */
428     ThrottleState *throttle_state;
429     ThrottleTimers throttle_timers;
430     unsigned       pending_reqs[2];
431     QLIST_ENTRY(BlockDriverState) round_robin;
432 
433     /* Offset after the highest byte written to */
434     uint64_t wr_highest_offset;
435 
436     /* I/O Limits */
437     BlockLimits bl;
438 
439     /* Whether produces zeros when read beyond eof */
440     bool zero_beyond_eof;
441 
442     /* Alignment requirement for offset/length of I/O requests */
443     unsigned int request_alignment;
444 
445     /* do we need to tell the quest if we have a volatile write cache? */
446     int enable_write_cache;
447 
448     /* the following member gives a name to every node on the bs graph. */
449     char node_name[32];
450     /* element of the list of named nodes building the graph */
451     QTAILQ_ENTRY(BlockDriverState) node_list;
452     /* element of the list of all BlockDriverStates (all_bdrv_states) */
453     QTAILQ_ENTRY(BlockDriverState) bs_list;
454     /* element of the list of monitor-owned BDS */
455     QTAILQ_ENTRY(BlockDriverState) monitor_list;
456     QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps;
457     int refcnt;
458 
459     QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
460 
461     /* operation blockers */
462     QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX];
463 
464     /* long-running background operation */
465     BlockJob *job;
466 
467     /* The node that this node inherited default options from (and a reopen on
468      * which can affect this node by changing these defaults). This is always a
469      * parent node of this node. */
470     BlockDriverState *inherits_from;
471     QLIST_HEAD(, BdrvChild) children;
472     QLIST_HEAD(, BdrvChild) parents;
473 
474     QDict *options;
475     QDict *explicit_options;
476     BlockdevDetectZeroesOptions detect_zeroes;
477 
478     /* The error object in use for blocking operations on backing_hd */
479     Error *backing_blocker;
480 
481     /* threshold limit for writes, in bytes. "High water mark". */
482     uint64_t write_threshold_offset;
483     NotifierWithReturn write_threshold_notifier;
484 
485     int quiesce_counter;
486 };
487 
488 struct BlockBackendRootState {
489     int open_flags;
490     bool read_only;
491     BlockdevDetectZeroesOptions detect_zeroes;
492 
493     char *throttle_group;
494     ThrottleState *throttle_state;
495 };
496 
497 static inline BlockDriverState *backing_bs(BlockDriverState *bs)
498 {
499     return bs->backing ? bs->backing->bs : NULL;
500 }
501 
502 
503 /* Essential block drivers which must always be statically linked into qemu, and
504  * which therefore can be accessed without using bdrv_find_format() */
505 extern BlockDriver bdrv_file;
506 extern BlockDriver bdrv_raw;
507 extern BlockDriver bdrv_qcow2;
508 
509 /**
510  * bdrv_setup_io_funcs:
511  *
512  * Prepare a #BlockDriver for I/O request processing by populating
513  * unimplemented coroutine and AIO interfaces with generic wrapper functions
514  * that fall back to implemented interfaces.
515  */
516 void bdrv_setup_io_funcs(BlockDriver *bdrv);
517 
518 int coroutine_fn bdrv_co_do_preadv(BlockDriverState *bs,
519     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
520     BdrvRequestFlags flags);
521 int coroutine_fn bdrv_co_do_pwritev(BlockDriverState *bs,
522     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
523     BdrvRequestFlags flags);
524 
525 int get_tmp_filename(char *filename, int size);
526 BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size,
527                             const char *filename);
528 
529 void bdrv_set_io_limits(BlockDriverState *bs,
530                         ThrottleConfig *cfg);
531 
532 
533 /**
534  * bdrv_add_before_write_notifier:
535  *
536  * Register a callback that is invoked before write requests are processed but
537  * after any throttling or waiting for overlapping requests.
538  */
539 void bdrv_add_before_write_notifier(BlockDriverState *bs,
540                                     NotifierWithReturn *notifier);
541 
542 /**
543  * bdrv_detach_aio_context:
544  *
545  * May be called from .bdrv_detach_aio_context() to detach children from the
546  * current #AioContext.  This is only needed by block drivers that manage their
547  * own children.  Both ->file and ->backing are automatically handled and
548  * block drivers should not call this function on them explicitly.
549  */
550 void bdrv_detach_aio_context(BlockDriverState *bs);
551 
552 /**
553  * bdrv_attach_aio_context:
554  *
555  * May be called from .bdrv_attach_aio_context() to attach children to the new
556  * #AioContext.  This is only needed by block drivers that manage their own
557  * children.  Both ->file and ->backing are automatically handled and block
558  * drivers should not call this function on them explicitly.
559  */
560 void bdrv_attach_aio_context(BlockDriverState *bs,
561                              AioContext *new_context);
562 
563 /**
564  * bdrv_add_aio_context_notifier:
565  *
566  * If a long-running job intends to be always run in the same AioContext as a
567  * certain BDS, it may use this function to be notified of changes regarding the
568  * association of the BDS to an AioContext.
569  *
570  * attached_aio_context() is called after the target BDS has been attached to a
571  * new AioContext; detach_aio_context() is called before the target BDS is being
572  * detached from its old AioContext.
573  */
574 void bdrv_add_aio_context_notifier(BlockDriverState *bs,
575         void (*attached_aio_context)(AioContext *new_context, void *opaque),
576         void (*detach_aio_context)(void *opaque), void *opaque);
577 
578 /**
579  * bdrv_remove_aio_context_notifier:
580  *
581  * Unsubscribe of change notifications regarding the BDS's AioContext. The
582  * parameters given here have to be the same as those given to
583  * bdrv_add_aio_context_notifier().
584  */
585 void bdrv_remove_aio_context_notifier(BlockDriverState *bs,
586                                       void (*aio_context_attached)(AioContext *,
587                                                                    void *),
588                                       void (*aio_context_detached)(void *),
589                                       void *opaque);
590 
591 #ifdef _WIN32
592 int is_windows_drive(const char *filename);
593 #endif
594 
595 /**
596  * stream_start:
597  * @bs: Block device to operate on.
598  * @base: Block device that will become the new base, or %NULL to
599  * flatten the whole backing file chain onto @bs.
600  * @base_id: The file name that will be written to @bs as the new
601  * backing file if the job completes.  Ignored if @base is %NULL.
602  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
603  * @on_error: The action to take upon error.
604  * @cb: Completion function for the job.
605  * @opaque: Opaque pointer value passed to @cb.
606  * @errp: Error object.
607  *
608  * Start a streaming operation on @bs.  Clusters that are unallocated
609  * in @bs, but allocated in any image between @base and @bs (both
610  * exclusive) will be written to @bs.  At the end of a successful
611  * streaming job, the backing file of @bs will be changed to
612  * @base_id in the written image and to @base in the live BlockDriverState.
613  */
614 void stream_start(BlockDriverState *bs, BlockDriverState *base,
615                   const char *base_id, int64_t speed, BlockdevOnError on_error,
616                   BlockCompletionFunc *cb,
617                   void *opaque, Error **errp);
618 
619 /**
620  * commit_start:
621  * @bs: Active block device.
622  * @top: Top block device to be committed.
623  * @base: Block device that will be written into, and become the new top.
624  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
625  * @on_error: The action to take upon error.
626  * @cb: Completion function for the job.
627  * @opaque: Opaque pointer value passed to @cb.
628  * @backing_file_str: String to use as the backing file in @top's overlay
629  * @errp: Error object.
630  *
631  */
632 void commit_start(BlockDriverState *bs, BlockDriverState *base,
633                  BlockDriverState *top, int64_t speed,
634                  BlockdevOnError on_error, BlockCompletionFunc *cb,
635                  void *opaque, const char *backing_file_str, Error **errp);
636 /**
637  * commit_active_start:
638  * @bs: Active block device to be committed.
639  * @base: Block device that will be written into, and become the new top.
640  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
641  * @on_error: The action to take upon error.
642  * @cb: Completion function for the job.
643  * @opaque: Opaque pointer value passed to @cb.
644  * @errp: Error object.
645  *
646  */
647 void commit_active_start(BlockDriverState *bs, BlockDriverState *base,
648                          int64_t speed,
649                          BlockdevOnError on_error,
650                          BlockCompletionFunc *cb,
651                          void *opaque, Error **errp);
652 /*
653  * mirror_start:
654  * @bs: Block device to operate on.
655  * @target: Block device to write to.
656  * @replaces: Block graph node name to replace once the mirror is done. Can
657  *            only be used when full mirroring is selected.
658  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
659  * @granularity: The chosen granularity for the dirty bitmap.
660  * @buf_size: The amount of data that can be in flight at one time.
661  * @mode: Whether to collapse all images in the chain to the target.
662  * @on_source_error: The action to take upon error reading from the source.
663  * @on_target_error: The action to take upon error writing to the target.
664  * @unmap: Whether to unmap target where source sectors only contain zeroes.
665  * @cb: Completion function for the job.
666  * @opaque: Opaque pointer value passed to @cb.
667  * @errp: Error object.
668  *
669  * Start a mirroring operation on @bs.  Clusters that are allocated
670  * in @bs will be written to @bs until the job is cancelled or
671  * manually completed.  At the end of a successful mirroring job,
672  * @bs will be switched to read from @target.
673  */
674 void mirror_start(BlockDriverState *bs, BlockDriverState *target,
675                   const char *replaces,
676                   int64_t speed, uint32_t granularity, int64_t buf_size,
677                   MirrorSyncMode mode, BlockdevOnError on_source_error,
678                   BlockdevOnError on_target_error,
679                   bool unmap,
680                   BlockCompletionFunc *cb,
681                   void *opaque, Error **errp);
682 
683 /*
684  * backup_start:
685  * @bs: Block device to operate on.
686  * @target: Block device to write to.
687  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
688  * @sync_mode: What parts of the disk image should be copied to the destination.
689  * @sync_bitmap: The dirty bitmap if sync_mode is MIRROR_SYNC_MODE_INCREMENTAL.
690  * @on_source_error: The action to take upon error reading from the source.
691  * @on_target_error: The action to take upon error writing to the target.
692  * @cb: Completion function for the job.
693  * @opaque: Opaque pointer value passed to @cb.
694  * @txn: Transaction that this job is part of (may be NULL).
695  *
696  * Start a backup operation on @bs.  Clusters in @bs are written to @target
697  * until the job is cancelled or manually completed.
698  */
699 void backup_start(BlockDriverState *bs, BlockDriverState *target,
700                   int64_t speed, MirrorSyncMode sync_mode,
701                   BdrvDirtyBitmap *sync_bitmap,
702                   BlockdevOnError on_source_error,
703                   BlockdevOnError on_target_error,
704                   BlockCompletionFunc *cb, void *opaque,
705                   BlockJobTxn *txn, Error **errp);
706 
707 void hmp_drive_add_node(Monitor *mon, const char *optstr);
708 
709 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
710                                   const char *child_name,
711                                   const BdrvChildRole *child_role);
712 void bdrv_root_unref_child(BdrvChild *child);
713 
714 void blk_dev_change_media_cb(BlockBackend *blk, bool load);
715 bool blk_dev_has_removable_media(BlockBackend *blk);
716 bool blk_dev_has_tray(BlockBackend *blk);
717 void blk_dev_eject_request(BlockBackend *blk, bool force);
718 bool blk_dev_is_tray_open(BlockBackend *blk);
719 bool blk_dev_is_medium_locked(BlockBackend *blk);
720 void blk_dev_resize_cb(BlockBackend *blk);
721 
722 void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector, int nr_sectors);
723 bool bdrv_requests_pending(BlockDriverState *bs);
724 
725 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap **out);
726 void bdrv_undo_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap *in);
727 
728 void blockdev_close_all_bdrv_states(void);
729 
730 #endif /* BLOCK_INT_H */
731