xref: /openbmc/qemu/include/block/block_int.h (revision daa76aa416b1e18ab1fac650ff53d966d8f21f68)
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 "qemu/option.h"
30 #include "qemu/queue.h"
31 #include "qemu/coroutine.h"
32 #include "qemu/timer.h"
33 #include "qapi-types.h"
34 #include "qemu/hbitmap.h"
35 #include "block/snapshot.h"
36 #include "qemu/main-loop.h"
37 #include "qemu/throttle.h"
38 
39 #define BLOCK_FLAG_ENCRYPT          1
40 #define BLOCK_FLAG_LAZY_REFCOUNTS   8
41 
42 #define BLOCK_OPT_SIZE              "size"
43 #define BLOCK_OPT_ENCRYPT           "encryption"
44 #define BLOCK_OPT_COMPAT6           "compat6"
45 #define BLOCK_OPT_HWVERSION         "hwversion"
46 #define BLOCK_OPT_BACKING_FILE      "backing_file"
47 #define BLOCK_OPT_BACKING_FMT       "backing_fmt"
48 #define BLOCK_OPT_CLUSTER_SIZE      "cluster_size"
49 #define BLOCK_OPT_TABLE_SIZE        "table_size"
50 #define BLOCK_OPT_PREALLOC          "preallocation"
51 #define BLOCK_OPT_SUBFMT            "subformat"
52 #define BLOCK_OPT_COMPAT_LEVEL      "compat"
53 #define BLOCK_OPT_LAZY_REFCOUNTS    "lazy_refcounts"
54 #define BLOCK_OPT_ADAPTER_TYPE      "adapter_type"
55 #define BLOCK_OPT_REDUNDANCY        "redundancy"
56 #define BLOCK_OPT_NOCOW             "nocow"
57 #define BLOCK_OPT_OBJECT_SIZE       "object_size"
58 #define BLOCK_OPT_REFCOUNT_BITS     "refcount_bits"
59 
60 #define BLOCK_PROBE_BUF_SIZE        512
61 
62 enum BdrvTrackedRequestType {
63     BDRV_TRACKED_READ,
64     BDRV_TRACKED_WRITE,
65     BDRV_TRACKED_FLUSH,
66     BDRV_TRACKED_IOCTL,
67     BDRV_TRACKED_DISCARD,
68 };
69 
70 typedef struct BdrvTrackedRequest {
71     BlockDriverState *bs;
72     int64_t offset;
73     unsigned int bytes;
74     enum BdrvTrackedRequestType type;
75 
76     bool serialising;
77     int64_t overlap_offset;
78     unsigned int overlap_bytes;
79 
80     QLIST_ENTRY(BdrvTrackedRequest) list;
81     Coroutine *co; /* owner, used for deadlock detection */
82     CoQueue wait_queue; /* coroutines blocked on this request */
83 
84     struct BdrvTrackedRequest *waiting_for;
85 } BdrvTrackedRequest;
86 
87 struct BlockDriver {
88     const char *format_name;
89     int instance_size;
90 
91     /* set to true if the BlockDriver is a block filter */
92     bool is_filter;
93     /* for snapshots block filter like Quorum can implement the
94      * following recursive callback.
95      * It's purpose is to recurse on the filter children while calling
96      * bdrv_recurse_is_first_non_filter on them.
97      * For a sample implementation look in the future Quorum block filter.
98      */
99     bool (*bdrv_recurse_is_first_non_filter)(BlockDriverState *bs,
100                                              BlockDriverState *candidate);
101 
102     int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename);
103     int (*bdrv_probe_device)(const char *filename);
104 
105     /* Any driver implementing this callback is expected to be able to handle
106      * NULL file names in its .bdrv_open() implementation */
107     void (*bdrv_parse_filename)(const char *filename, QDict *options, Error **errp);
108     /* Drivers not implementing bdrv_parse_filename nor bdrv_open should have
109      * this field set to true, except ones that are defined only by their
110      * child's bs.
111      * An example of the last type will be the quorum block driver.
112      */
113     bool bdrv_needs_filename;
114 
115     /* Set if a driver can support backing files */
116     bool supports_backing;
117 
118     /* For handling image reopen for split or non-split files */
119     int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state,
120                                BlockReopenQueue *queue, Error **errp);
121     void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state);
122     void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state);
123     void (*bdrv_join_options)(QDict *options, QDict *old_options);
124 
125     int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags,
126                      Error **errp);
127     int (*bdrv_file_open)(BlockDriverState *bs, QDict *options, int flags,
128                           Error **errp);
129     void (*bdrv_close)(BlockDriverState *bs);
130     int (*bdrv_create)(const char *filename, QemuOpts *opts, Error **errp);
131     int (*bdrv_set_key)(BlockDriverState *bs, const char *key);
132     int (*bdrv_make_empty)(BlockDriverState *bs);
133 
134     void (*bdrv_refresh_filename)(BlockDriverState *bs, QDict *options);
135 
136     /* aio */
137     BlockAIOCB *(*bdrv_aio_readv)(BlockDriverState *bs,
138         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
139         BlockCompletionFunc *cb, void *opaque);
140     BlockAIOCB *(*bdrv_aio_writev)(BlockDriverState *bs,
141         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
142         BlockCompletionFunc *cb, void *opaque);
143     BlockAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs,
144         BlockCompletionFunc *cb, void *opaque);
145     BlockAIOCB *(*bdrv_aio_discard)(BlockDriverState *bs,
146         int64_t sector_num, int nb_sectors,
147         BlockCompletionFunc *cb, void *opaque);
148 
149     int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs,
150         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
151     int coroutine_fn (*bdrv_co_preadv)(BlockDriverState *bs,
152         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags);
153     int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs,
154         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
155     int coroutine_fn (*bdrv_co_writev_flags)(BlockDriverState *bs,
156         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, int flags);
157     int coroutine_fn (*bdrv_co_pwritev)(BlockDriverState *bs,
158         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags);
159 
160     /*
161      * Efficiently zero a region of the disk image.  Typically an image format
162      * would use a compact metadata representation to implement this.  This
163      * function pointer may be NULL or return -ENOSUP and .bdrv_co_writev()
164      * will be called instead.
165      */
166     int coroutine_fn (*bdrv_co_pwrite_zeroes)(BlockDriverState *bs,
167         int64_t offset, int count, BdrvRequestFlags flags);
168     int coroutine_fn (*bdrv_co_discard)(BlockDriverState *bs,
169         int64_t sector_num, int nb_sectors);
170     int64_t coroutine_fn (*bdrv_co_get_block_status)(BlockDriverState *bs,
171         int64_t sector_num, int nb_sectors, int *pnum,
172         BlockDriverState **file);
173 
174     /*
175      * Invalidate any cached meta-data.
176      */
177     void (*bdrv_invalidate_cache)(BlockDriverState *bs, Error **errp);
178     int (*bdrv_inactivate)(BlockDriverState *bs);
179 
180     /*
181      * Flushes all data for all layers by calling bdrv_co_flush for underlying
182      * layers, if needed. This function is needed for deterministic
183      * synchronization of the flush finishing callback.
184      */
185     int coroutine_fn (*bdrv_co_flush)(BlockDriverState *bs);
186 
187     /*
188      * Flushes all data that was already written to the OS all the way down to
189      * the disk (for example raw-posix calls fsync()).
190      */
191     int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs);
192 
193     /*
194      * Flushes all internal caches to the OS. The data may still sit in a
195      * writeback cache of the host OS, but it will survive a crash of the qemu
196      * process.
197      */
198     int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs);
199 
200     const char *protocol_name;
201     int (*bdrv_truncate)(BlockDriverState *bs, int64_t offset);
202 
203     int64_t (*bdrv_getlength)(BlockDriverState *bs);
204     bool has_variable_length;
205     int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
206 
207     int (*bdrv_write_compressed)(BlockDriverState *bs, int64_t sector_num,
208                                  const uint8_t *buf, int nb_sectors);
209 
210     int (*bdrv_snapshot_create)(BlockDriverState *bs,
211                                 QEMUSnapshotInfo *sn_info);
212     int (*bdrv_snapshot_goto)(BlockDriverState *bs,
213                               const char *snapshot_id);
214     int (*bdrv_snapshot_delete)(BlockDriverState *bs,
215                                 const char *snapshot_id,
216                                 const char *name,
217                                 Error **errp);
218     int (*bdrv_snapshot_list)(BlockDriverState *bs,
219                               QEMUSnapshotInfo **psn_info);
220     int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
221                                   const char *snapshot_id,
222                                   const char *name,
223                                   Error **errp);
224     int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
225     ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs);
226 
227     int coroutine_fn (*bdrv_save_vmstate)(BlockDriverState *bs,
228                                           QEMUIOVector *qiov,
229                                           int64_t pos);
230     int coroutine_fn (*bdrv_load_vmstate)(BlockDriverState *bs,
231                                           QEMUIOVector *qiov,
232                                           int64_t pos);
233 
234     int (*bdrv_change_backing_file)(BlockDriverState *bs,
235         const char *backing_file, const char *backing_fmt);
236 
237     /* removable device specific */
238     bool (*bdrv_is_inserted)(BlockDriverState *bs);
239     int (*bdrv_media_changed)(BlockDriverState *bs);
240     void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
241     void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
242 
243     /* to control generic scsi devices */
244     BlockAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
245         unsigned long int req, void *buf,
246         BlockCompletionFunc *cb, void *opaque);
247 
248     /* List of options for creating images, terminated by name == NULL */
249     QemuOptsList *create_opts;
250 
251     /*
252      * Returns 0 for completed check, -errno for internal errors.
253      * The check results are stored in result.
254      */
255     int (*bdrv_check)(BlockDriverState* bs, BdrvCheckResult *result,
256         BdrvCheckMode fix);
257 
258     int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts,
259                               BlockDriverAmendStatusCB *status_cb,
260                               void *cb_opaque);
261 
262     void (*bdrv_debug_event)(BlockDriverState *bs, BlkdebugEvent event);
263 
264     /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
265     int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
266         const char *tag);
267     int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs,
268         const char *tag);
269     int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
270     bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
271 
272     void (*bdrv_refresh_limits)(BlockDriverState *bs, Error **errp);
273 
274     /*
275      * Returns 1 if newly created images are guaranteed to contain only
276      * zeros, 0 otherwise.
277      */
278     int (*bdrv_has_zero_init)(BlockDriverState *bs);
279 
280     /* Remove fd handlers, timers, and other event loop callbacks so the event
281      * loop is no longer in use.  Called with no in-flight requests and in
282      * depth-first traversal order with parents before child nodes.
283      */
284     void (*bdrv_detach_aio_context)(BlockDriverState *bs);
285 
286     /* Add fd handlers, timers, and other event loop callbacks so I/O requests
287      * can be processed again.  Called with no in-flight requests and in
288      * depth-first traversal order with child nodes before parent nodes.
289      */
290     void (*bdrv_attach_aio_context)(BlockDriverState *bs,
291                                     AioContext *new_context);
292 
293     /* io queue for linux-aio */
294     void (*bdrv_io_plug)(BlockDriverState *bs);
295     void (*bdrv_io_unplug)(BlockDriverState *bs);
296 
297     /**
298      * Try to get @bs's logical and physical block size.
299      * On success, store them in @bsz and return zero.
300      * On failure, return negative errno.
301      */
302     int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz);
303     /**
304      * Try to get @bs's geometry (cyls, heads, sectors)
305      * On success, store them in @geo and return 0.
306      * On failure return -errno.
307      * Only drivers that want to override guest geometry implement this
308      * callback; see hd_geometry_guess().
309      */
310     int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo);
311 
312     /**
313      * Drain and stop any internal sources of requests in the driver, and
314      * remain so until next I/O callback (e.g. bdrv_co_writev) is called.
315      */
316     void (*bdrv_drain)(BlockDriverState *bs);
317 
318     void (*bdrv_add_child)(BlockDriverState *parent, BlockDriverState *child,
319                            Error **errp);
320     void (*bdrv_del_child)(BlockDriverState *parent, BdrvChild *child,
321                            Error **errp);
322 
323     QLIST_ENTRY(BlockDriver) list;
324 };
325 
326 typedef struct BlockLimits {
327     /* maximum number of sectors that can be discarded at once */
328     int max_discard;
329 
330     /* optimal alignment for discard requests in sectors */
331     int64_t discard_alignment;
332 
333     /* maximum number of bytes that can zeroized at once (since it is
334      * signed, it must be < 2G, if set) */
335     int32_t max_pwrite_zeroes;
336 
337     /* optimal alignment for write zeroes requests in bytes, must be
338      * power of 2, and less than max_pwrite_zeroes if that is set */
339     uint32_t pwrite_zeroes_alignment;
340 
341     /* optimal transfer length in sectors */
342     int opt_transfer_length;
343 
344     /* maximal transfer length in sectors */
345     int max_transfer_length;
346 
347     /* memory alignment so that no bounce buffer is needed */
348     size_t min_mem_alignment;
349 
350     /* memory alignment for bounce buffer */
351     size_t opt_mem_alignment;
352 
353     /* maximum number of iovec elements */
354     int max_iov;
355 } BlockLimits;
356 
357 typedef struct BdrvOpBlocker BdrvOpBlocker;
358 
359 typedef struct BdrvAioNotifier {
360     void (*attached_aio_context)(AioContext *new_context, void *opaque);
361     void (*detach_aio_context)(void *opaque);
362 
363     void *opaque;
364 
365     QLIST_ENTRY(BdrvAioNotifier) list;
366 } BdrvAioNotifier;
367 
368 struct BdrvChildRole {
369     void (*inherit_options)(int *child_flags, QDict *child_options,
370                             int parent_flags, QDict *parent_options);
371 
372     void (*change_media)(BdrvChild *child, bool load);
373     void (*resize)(BdrvChild *child);
374 
375     /* Returns a name that is supposedly more useful for human users than the
376      * node name for identifying the node in question (in particular, a BB
377      * name), or NULL if the parent can't provide a better name. */
378     const char* (*get_name)(BdrvChild *child);
379 
380     /*
381      * If this pair of functions is implemented, the parent doesn't issue new
382      * requests after returning from .drained_begin() until .drained_end() is
383      * called.
384      *
385      * Note that this can be nested. If drained_begin() was called twice, new
386      * I/O is allowed only after drained_end() was called twice, too.
387      */
388     void (*drained_begin)(BdrvChild *child);
389     void (*drained_end)(BdrvChild *child);
390 };
391 
392 extern const BdrvChildRole child_file;
393 extern const BdrvChildRole child_format;
394 
395 struct BdrvChild {
396     BlockDriverState *bs;
397     char *name;
398     const BdrvChildRole *role;
399     void *opaque;
400     QLIST_ENTRY(BdrvChild) next;
401     QLIST_ENTRY(BdrvChild) next_parent;
402 };
403 
404 /*
405  * Note: the function bdrv_append() copies and swaps contents of
406  * BlockDriverStates, so if you add new fields to this struct, please
407  * inspect bdrv_append() to determine if the new fields need to be
408  * copied as well.
409  */
410 struct BlockDriverState {
411     int64_t total_sectors; /* if we are reading a disk image, give its
412                               size in sectors */
413     int read_only; /* if true, the media is read only */
414     int open_flags; /* flags used to open the file, re-used for re-open */
415     int encrypted; /* if true, the media is encrypted */
416     int valid_key; /* if true, a valid encryption key has been set */
417     int sg;        /* if true, the device is a /dev/sg* */
418     int copy_on_read; /* if true, copy read backing sectors into image
419                          note this is a reference count */
420     bool probed;
421 
422     BlockDriver *drv; /* NULL means no media */
423     void *opaque;
424 
425     AioContext *aio_context; /* event loop used for fd handlers, timers, etc */
426     /* long-running tasks intended to always use the same AioContext as this
427      * BDS may register themselves in this list to be notified of changes
428      * regarding this BDS's context */
429     QLIST_HEAD(, BdrvAioNotifier) aio_notifiers;
430 
431     char filename[PATH_MAX];
432     char backing_file[PATH_MAX]; /* if non zero, the image is a diff of
433                                     this file image */
434     char backing_format[16]; /* if non-zero and backing_file exists */
435 
436     QDict *full_open_options;
437     char exact_filename[PATH_MAX];
438 
439     BdrvChild *backing;
440     BdrvChild *file;
441 
442     /* Callback before write request is processed */
443     NotifierWithReturnList before_write_notifiers;
444 
445     /* number of in-flight serialising requests */
446     unsigned int serialising_in_flight;
447 
448     /* Offset after the highest byte written to */
449     uint64_t wr_highest_offset;
450 
451     /* I/O Limits */
452     BlockLimits bl;
453 
454     /* Alignment requirement for offset/length of I/O requests */
455     unsigned int request_alignment;
456     /* Flags honored during pwrite (so far: BDRV_REQ_FUA) */
457     unsigned int supported_write_flags;
458     /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA,
459      * BDRV_REQ_MAY_UNMAP) */
460     unsigned int supported_zero_flags;
461 
462     /* the following member gives a name to every node on the bs graph. */
463     char node_name[32];
464     /* element of the list of named nodes building the graph */
465     QTAILQ_ENTRY(BlockDriverState) node_list;
466     /* element of the list of all BlockDriverStates (all_bdrv_states) */
467     QTAILQ_ENTRY(BlockDriverState) bs_list;
468     /* element of the list of monitor-owned BDS */
469     QTAILQ_ENTRY(BlockDriverState) monitor_list;
470     QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps;
471     int refcnt;
472 
473     QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
474 
475     /* operation blockers */
476     QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX];
477 
478     /* long-running background operation */
479     BlockJob *job;
480 
481     /* The node that this node inherited default options from (and a reopen on
482      * which can affect this node by changing these defaults). This is always a
483      * parent node of this node. */
484     BlockDriverState *inherits_from;
485     QLIST_HEAD(, BdrvChild) children;
486     QLIST_HEAD(, BdrvChild) parents;
487 
488     QDict *options;
489     QDict *explicit_options;
490     BlockdevDetectZeroesOptions detect_zeroes;
491 
492     /* The error object in use for blocking operations on backing_hd */
493     Error *backing_blocker;
494 
495     /* threshold limit for writes, in bytes. "High water mark". */
496     uint64_t write_threshold_offset;
497     NotifierWithReturn write_threshold_notifier;
498 
499     /* counters for nested bdrv_io_plug and bdrv_io_unplugged_begin */
500     unsigned io_plugged;
501     unsigned io_plug_disabled;
502 
503     int quiesce_counter;
504 };
505 
506 struct BlockBackendRootState {
507     int open_flags;
508     bool read_only;
509     BlockdevDetectZeroesOptions detect_zeroes;
510 };
511 
512 typedef enum BlockMirrorBackingMode {
513     /* Reuse the existing backing chain from the source for the target.
514      * - sync=full: Set backing BDS to NULL.
515      * - sync=top:  Use source's backing BDS.
516      * - sync=none: Use source as the backing BDS. */
517     MIRROR_SOURCE_BACKING_CHAIN,
518 
519     /* Open the target's backing chain completely anew */
520     MIRROR_OPEN_BACKING_CHAIN,
521 
522     /* Do not change the target's backing BDS after job completion */
523     MIRROR_LEAVE_BACKING_CHAIN,
524 } BlockMirrorBackingMode;
525 
526 static inline BlockDriverState *backing_bs(BlockDriverState *bs)
527 {
528     return bs->backing ? bs->backing->bs : NULL;
529 }
530 
531 
532 /* Essential block drivers which must always be statically linked into qemu, and
533  * which therefore can be accessed without using bdrv_find_format() */
534 extern BlockDriver bdrv_file;
535 extern BlockDriver bdrv_raw;
536 extern BlockDriver bdrv_qcow2;
537 
538 /**
539  * bdrv_setup_io_funcs:
540  *
541  * Prepare a #BlockDriver for I/O request processing by populating
542  * unimplemented coroutine and AIO interfaces with generic wrapper functions
543  * that fall back to implemented interfaces.
544  */
545 void bdrv_setup_io_funcs(BlockDriver *bdrv);
546 
547 int coroutine_fn bdrv_co_preadv(BlockDriverState *bs,
548     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
549     BdrvRequestFlags flags);
550 int coroutine_fn bdrv_co_pwritev(BlockDriverState *bs,
551     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
552     BdrvRequestFlags flags);
553 
554 int get_tmp_filename(char *filename, int size);
555 BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size,
556                             const char *filename);
557 
558 
559 /**
560  * bdrv_add_before_write_notifier:
561  *
562  * Register a callback that is invoked before write requests are processed but
563  * after any throttling or waiting for overlapping requests.
564  */
565 void bdrv_add_before_write_notifier(BlockDriverState *bs,
566                                     NotifierWithReturn *notifier);
567 
568 /**
569  * bdrv_detach_aio_context:
570  *
571  * May be called from .bdrv_detach_aio_context() to detach children from the
572  * current #AioContext.  This is only needed by block drivers that manage their
573  * own children.  Both ->file and ->backing are automatically handled and
574  * block drivers should not call this function on them explicitly.
575  */
576 void bdrv_detach_aio_context(BlockDriverState *bs);
577 
578 /**
579  * bdrv_attach_aio_context:
580  *
581  * May be called from .bdrv_attach_aio_context() to attach children to the new
582  * #AioContext.  This is only needed by block drivers that manage their own
583  * children.  Both ->file and ->backing are automatically handled and block
584  * drivers should not call this function on them explicitly.
585  */
586 void bdrv_attach_aio_context(BlockDriverState *bs,
587                              AioContext *new_context);
588 
589 /**
590  * bdrv_add_aio_context_notifier:
591  *
592  * If a long-running job intends to be always run in the same AioContext as a
593  * certain BDS, it may use this function to be notified of changes regarding the
594  * association of the BDS to an AioContext.
595  *
596  * attached_aio_context() is called after the target BDS has been attached to a
597  * new AioContext; detach_aio_context() is called before the target BDS is being
598  * detached from its old AioContext.
599  */
600 void bdrv_add_aio_context_notifier(BlockDriverState *bs,
601         void (*attached_aio_context)(AioContext *new_context, void *opaque),
602         void (*detach_aio_context)(void *opaque), void *opaque);
603 
604 /**
605  * bdrv_remove_aio_context_notifier:
606  *
607  * Unsubscribe of change notifications regarding the BDS's AioContext. The
608  * parameters given here have to be the same as those given to
609  * bdrv_add_aio_context_notifier().
610  */
611 void bdrv_remove_aio_context_notifier(BlockDriverState *bs,
612                                       void (*aio_context_attached)(AioContext *,
613                                                                    void *),
614                                       void (*aio_context_detached)(void *),
615                                       void *opaque);
616 
617 #ifdef _WIN32
618 int is_windows_drive(const char *filename);
619 #endif
620 
621 /**
622  * stream_start:
623  * @bs: Block device to operate on.
624  * @base: Block device that will become the new base, or %NULL to
625  * flatten the whole backing file chain onto @bs.
626  * @base_id: The file name that will be written to @bs as the new
627  * backing file if the job completes.  Ignored if @base is %NULL.
628  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
629  * @on_error: The action to take upon error.
630  * @cb: Completion function for the job.
631  * @opaque: Opaque pointer value passed to @cb.
632  * @errp: Error object.
633  *
634  * Start a streaming operation on @bs.  Clusters that are unallocated
635  * in @bs, but allocated in any image between @base and @bs (both
636  * exclusive) will be written to @bs.  At the end of a successful
637  * streaming job, the backing file of @bs will be changed to
638  * @base_id in the written image and to @base in the live BlockDriverState.
639  */
640 void stream_start(BlockDriverState *bs, BlockDriverState *base,
641                   const char *base_id, int64_t speed, BlockdevOnError on_error,
642                   BlockCompletionFunc *cb,
643                   void *opaque, Error **errp);
644 
645 /**
646  * commit_start:
647  * @bs: Active block device.
648  * @top: Top block device to be committed.
649  * @base: Block device that will be written into, and become the new top.
650  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
651  * @on_error: The action to take upon error.
652  * @cb: Completion function for the job.
653  * @opaque: Opaque pointer value passed to @cb.
654  * @backing_file_str: String to use as the backing file in @top's overlay
655  * @errp: Error object.
656  *
657  */
658 void commit_start(BlockDriverState *bs, BlockDriverState *base,
659                  BlockDriverState *top, int64_t speed,
660                  BlockdevOnError on_error, BlockCompletionFunc *cb,
661                  void *opaque, const char *backing_file_str, Error **errp);
662 /**
663  * commit_active_start:
664  * @bs: Active block device to be committed.
665  * @base: Block device that will be written into, and become the new top.
666  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
667  * @on_error: The action to take upon error.
668  * @cb: Completion function for the job.
669  * @opaque: Opaque pointer value passed to @cb.
670  * @errp: Error object.
671  *
672  */
673 void commit_active_start(BlockDriverState *bs, BlockDriverState *base,
674                          int64_t speed,
675                          BlockdevOnError on_error,
676                          BlockCompletionFunc *cb,
677                          void *opaque, Error **errp);
678 /*
679  * mirror_start:
680  * @bs: Block device to operate on.
681  * @target: Block device to write to.
682  * @replaces: Block graph node name to replace once the mirror is done. Can
683  *            only be used when full mirroring is selected.
684  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
685  * @granularity: The chosen granularity for the dirty bitmap.
686  * @buf_size: The amount of data that can be in flight at one time.
687  * @mode: Whether to collapse all images in the chain to the target.
688  * @backing_mode: How to establish the target's backing chain after completion.
689  * @on_source_error: The action to take upon error reading from the source.
690  * @on_target_error: The action to take upon error writing to the target.
691  * @unmap: Whether to unmap target where source sectors only contain zeroes.
692  * @cb: Completion function for the job.
693  * @opaque: Opaque pointer value passed to @cb.
694  * @errp: Error object.
695  *
696  * Start a mirroring operation on @bs.  Clusters that are allocated
697  * in @bs will be written to @bs until the job is cancelled or
698  * manually completed.  At the end of a successful mirroring job,
699  * @bs will be switched to read from @target.
700  */
701 void mirror_start(BlockDriverState *bs, BlockDriverState *target,
702                   const char *replaces,
703                   int64_t speed, uint32_t granularity, int64_t buf_size,
704                   MirrorSyncMode mode, BlockMirrorBackingMode backing_mode,
705                   BlockdevOnError on_source_error,
706                   BlockdevOnError on_target_error,
707                   bool unmap,
708                   BlockCompletionFunc *cb,
709                   void *opaque, Error **errp);
710 
711 /*
712  * backup_start:
713  * @bs: Block device to operate on.
714  * @target: Block device to write to.
715  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
716  * @sync_mode: What parts of the disk image should be copied to the destination.
717  * @sync_bitmap: The dirty bitmap if sync_mode is MIRROR_SYNC_MODE_INCREMENTAL.
718  * @on_source_error: The action to take upon error reading from the source.
719  * @on_target_error: The action to take upon error writing to the target.
720  * @cb: Completion function for the job.
721  * @opaque: Opaque pointer value passed to @cb.
722  * @txn: Transaction that this job is part of (may be NULL).
723  *
724  * Start a backup operation on @bs.  Clusters in @bs are written to @target
725  * until the job is cancelled or manually completed.
726  */
727 void backup_start(BlockDriverState *bs, BlockDriverState *target,
728                   int64_t speed, MirrorSyncMode sync_mode,
729                   BdrvDirtyBitmap *sync_bitmap,
730                   BlockdevOnError on_source_error,
731                   BlockdevOnError on_target_error,
732                   BlockCompletionFunc *cb, void *opaque,
733                   BlockJobTxn *txn, Error **errp);
734 
735 void hmp_drive_add_node(Monitor *mon, const char *optstr);
736 
737 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
738                                   const char *child_name,
739                                   const BdrvChildRole *child_role,
740                                   void *opaque);
741 void bdrv_root_unref_child(BdrvChild *child);
742 
743 const char *bdrv_get_parent_name(const BlockDriverState *bs);
744 void blk_dev_change_media_cb(BlockBackend *blk, bool load);
745 bool blk_dev_has_removable_media(BlockBackend *blk);
746 bool blk_dev_has_tray(BlockBackend *blk);
747 void blk_dev_eject_request(BlockBackend *blk, bool force);
748 bool blk_dev_is_tray_open(BlockBackend *blk);
749 bool blk_dev_is_medium_locked(BlockBackend *blk);
750 
751 void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector, int nr_sectors);
752 bool bdrv_requests_pending(BlockDriverState *bs);
753 
754 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap **out);
755 void bdrv_undo_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap *in);
756 
757 void blockdev_close_all_bdrv_states(void);
758 
759 #endif /* BLOCK_INT_H */
760