xref: /openbmc/qemu/include/block/block_int.h (revision b097ba37)
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/aio-wait.h"
30 #include "qemu/queue.h"
31 #include "qemu/coroutine.h"
32 #include "qemu/stats64.h"
33 #include "qemu/timer.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_LAZY_REFCOUNTS   8
40 
41 #define BLOCK_OPT_SIZE              "size"
42 #define BLOCK_OPT_ENCRYPT           "encryption"
43 #define BLOCK_OPT_ENCRYPT_FORMAT    "encrypt.format"
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 #define BLOCK_OPT_DATA_FILE         "data_file"
60 #define BLOCK_OPT_DATA_FILE_RAW     "data_file_raw"
61 
62 #define BLOCK_PROBE_BUF_SIZE        512
63 
64 enum BdrvTrackedRequestType {
65     BDRV_TRACKED_READ,
66     BDRV_TRACKED_WRITE,
67     BDRV_TRACKED_DISCARD,
68     BDRV_TRACKED_TRUNCATE,
69 };
70 
71 typedef struct BdrvTrackedRequest {
72     BlockDriverState *bs;
73     int64_t offset;
74     uint64_t bytes;
75     enum BdrvTrackedRequestType type;
76 
77     bool serialising;
78     int64_t overlap_offset;
79     uint64_t 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. Block filters pass
93      * certain callbacks that refer to data (see block.c) to their bs->file if
94      * the driver doesn't implement them. Drivers that do not wish to forward
95      * must implement them and return -ENOTSUP.
96      */
97     bool is_filter;
98     /* for snapshots block filter like Quorum can implement the
99      * following recursive callback.
100      * It's purpose is to recurse on the filter children while calling
101      * bdrv_recurse_is_first_non_filter on them.
102      * For a sample implementation look in the future Quorum block filter.
103      */
104     bool (*bdrv_recurse_is_first_non_filter)(BlockDriverState *bs,
105                                              BlockDriverState *candidate);
106 
107     int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename);
108     int (*bdrv_probe_device)(const char *filename);
109 
110     /* Any driver implementing this callback is expected to be able to handle
111      * NULL file names in its .bdrv_open() implementation */
112     void (*bdrv_parse_filename)(const char *filename, QDict *options, Error **errp);
113     /* Drivers not implementing bdrv_parse_filename nor bdrv_open should have
114      * this field set to true, except ones that are defined only by their
115      * child's bs.
116      * An example of the last type will be the quorum block driver.
117      */
118     bool bdrv_needs_filename;
119 
120     /* Set if a driver can support backing files */
121     bool supports_backing;
122 
123     /* For handling image reopen for split or non-split files */
124     int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state,
125                                BlockReopenQueue *queue, Error **errp);
126     void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state);
127     void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state);
128     void (*bdrv_join_options)(QDict *options, QDict *old_options);
129 
130     int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags,
131                      Error **errp);
132 
133     /* Protocol drivers should implement this instead of bdrv_open */
134     int (*bdrv_file_open)(BlockDriverState *bs, QDict *options, int flags,
135                           Error **errp);
136     void (*bdrv_close)(BlockDriverState *bs);
137     int coroutine_fn (*bdrv_co_create)(BlockdevCreateOptions *opts,
138                                        Error **errp);
139     int coroutine_fn (*bdrv_co_create_opts)(const char *filename,
140                                             QemuOpts *opts,
141                                             Error **errp);
142     int (*bdrv_make_empty)(BlockDriverState *bs);
143 
144     /*
145      * Refreshes the bs->exact_filename field. If that is impossible,
146      * bs->exact_filename has to be left empty.
147      */
148     void (*bdrv_refresh_filename)(BlockDriverState *bs);
149 
150     /*
151      * Gathers the open options for all children into @target.
152      * A simple format driver (without backing file support) might
153      * implement this function like this:
154      *
155      *     QINCREF(bs->file->bs->full_open_options);
156      *     qdict_put(target, "file", bs->file->bs->full_open_options);
157      *
158      * If not specified, the generic implementation will simply put
159      * all children's options under their respective name.
160      *
161      * @backing_overridden is true when bs->backing seems not to be
162      * the child that would result from opening bs->backing_file.
163      * Therefore, if it is true, the backing child's options should be
164      * gathered; otherwise, there is no need since the backing child
165      * is the one implied by the image header.
166      *
167      * Note that ideally this function would not be needed.  Every
168      * block driver which implements it is probably doing something
169      * shady regarding its runtime option structure.
170      */
171     void (*bdrv_gather_child_options)(BlockDriverState *bs, QDict *target,
172                                       bool backing_overridden);
173 
174     /*
175      * Returns an allocated string which is the directory name of this BDS: It
176      * will be used to make relative filenames absolute by prepending this
177      * function's return value to them.
178      */
179     char *(*bdrv_dirname)(BlockDriverState *bs, Error **errp);
180 
181     /* aio */
182     BlockAIOCB *(*bdrv_aio_preadv)(BlockDriverState *bs,
183         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags,
184         BlockCompletionFunc *cb, void *opaque);
185     BlockAIOCB *(*bdrv_aio_pwritev)(BlockDriverState *bs,
186         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags,
187         BlockCompletionFunc *cb, void *opaque);
188     BlockAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs,
189         BlockCompletionFunc *cb, void *opaque);
190     BlockAIOCB *(*bdrv_aio_pdiscard)(BlockDriverState *bs,
191         int64_t offset, int bytes,
192         BlockCompletionFunc *cb, void *opaque);
193 
194     int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs,
195         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
196 
197     /**
198      * @offset: position in bytes to read at
199      * @bytes: number of bytes to read
200      * @qiov: the buffers to fill with read data
201      * @flags: currently unused, always 0
202      *
203      * @offset and @bytes will be a multiple of 'request_alignment',
204      * but the length of individual @qiov elements does not have to
205      * be a multiple.
206      *
207      * @bytes will always equal the total size of @qiov, and will be
208      * no larger than 'max_transfer'.
209      *
210      * The buffer in @qiov may point directly to guest memory.
211      */
212     int coroutine_fn (*bdrv_co_preadv)(BlockDriverState *bs,
213         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags);
214     int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs,
215         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, int flags);
216     /**
217      * @offset: position in bytes to write at
218      * @bytes: number of bytes to write
219      * @qiov: the buffers containing data to write
220      * @flags: zero or more bits allowed by 'supported_write_flags'
221      *
222      * @offset and @bytes will be a multiple of 'request_alignment',
223      * but the length of individual @qiov elements does not have to
224      * be a multiple.
225      *
226      * @bytes will always equal the total size of @qiov, and will be
227      * no larger than 'max_transfer'.
228      *
229      * The buffer in @qiov may point directly to guest memory.
230      */
231     int coroutine_fn (*bdrv_co_pwritev)(BlockDriverState *bs,
232         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags);
233 
234     /*
235      * Efficiently zero a region of the disk image.  Typically an image format
236      * would use a compact metadata representation to implement this.  This
237      * function pointer may be NULL or return -ENOSUP and .bdrv_co_writev()
238      * will be called instead.
239      */
240     int coroutine_fn (*bdrv_co_pwrite_zeroes)(BlockDriverState *bs,
241         int64_t offset, int bytes, BdrvRequestFlags flags);
242     int coroutine_fn (*bdrv_co_pdiscard)(BlockDriverState *bs,
243         int64_t offset, int bytes);
244 
245     /* Map [offset, offset + nbytes) range onto a child of @bs to copy from,
246      * and invoke bdrv_co_copy_range_from(child, ...), or invoke
247      * bdrv_co_copy_range_to() if @bs is the leaf child to copy data from.
248      *
249      * See the comment of bdrv_co_copy_range for the parameter and return value
250      * semantics.
251      */
252     int coroutine_fn (*bdrv_co_copy_range_from)(BlockDriverState *bs,
253                                                 BdrvChild *src,
254                                                 uint64_t offset,
255                                                 BdrvChild *dst,
256                                                 uint64_t dst_offset,
257                                                 uint64_t bytes,
258                                                 BdrvRequestFlags read_flags,
259                                                 BdrvRequestFlags write_flags);
260 
261     /* Map [offset, offset + nbytes) range onto a child of bs to copy data to,
262      * and invoke bdrv_co_copy_range_to(child, src, ...), or perform the copy
263      * operation if @bs is the leaf and @src has the same BlockDriver.  Return
264      * -ENOTSUP if @bs is the leaf but @src has a different BlockDriver.
265      *
266      * See the comment of bdrv_co_copy_range for the parameter and return value
267      * semantics.
268      */
269     int coroutine_fn (*bdrv_co_copy_range_to)(BlockDriverState *bs,
270                                               BdrvChild *src,
271                                               uint64_t src_offset,
272                                               BdrvChild *dst,
273                                               uint64_t dst_offset,
274                                               uint64_t bytes,
275                                               BdrvRequestFlags read_flags,
276                                               BdrvRequestFlags write_flags);
277 
278     /*
279      * Building block for bdrv_block_status[_above] and
280      * bdrv_is_allocated[_above].  The driver should answer only
281      * according to the current layer, and should only need to set
282      * BDRV_BLOCK_DATA, BDRV_BLOCK_ZERO, BDRV_BLOCK_OFFSET_VALID,
283      * and/or BDRV_BLOCK_RAW; if the current layer defers to a backing
284      * layer, the result should be 0 (and not BDRV_BLOCK_ZERO).  See
285      * block.h for the overall meaning of the bits.  As a hint, the
286      * flag want_zero is true if the caller cares more about precise
287      * mappings (favor accurate _OFFSET_VALID/_ZERO) or false for
288      * overall allocation (favor larger *pnum, perhaps by reporting
289      * _DATA instead of _ZERO).  The block layer guarantees input
290      * clamped to bdrv_getlength() and aligned to request_alignment,
291      * as well as non-NULL pnum, map, and file; in turn, the driver
292      * must return an error or set pnum to an aligned non-zero value.
293      */
294     int coroutine_fn (*bdrv_co_block_status)(BlockDriverState *bs,
295         bool want_zero, int64_t offset, int64_t bytes, int64_t *pnum,
296         int64_t *map, BlockDriverState **file);
297 
298     /*
299      * Invalidate any cached meta-data.
300      */
301     void coroutine_fn (*bdrv_co_invalidate_cache)(BlockDriverState *bs,
302                                                   Error **errp);
303     int (*bdrv_inactivate)(BlockDriverState *bs);
304 
305     /*
306      * Flushes all data for all layers by calling bdrv_co_flush for underlying
307      * layers, if needed. This function is needed for deterministic
308      * synchronization of the flush finishing callback.
309      */
310     int coroutine_fn (*bdrv_co_flush)(BlockDriverState *bs);
311 
312     /*
313      * Flushes all data that was already written to the OS all the way down to
314      * the disk (for example file-posix.c calls fsync()).
315      */
316     int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs);
317 
318     /*
319      * Flushes all internal caches to the OS. The data may still sit in a
320      * writeback cache of the host OS, but it will survive a crash of the qemu
321      * process.
322      */
323     int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs);
324 
325     /*
326      * Drivers setting this field must be able to work with just a plain
327      * filename with '<protocol_name>:' as a prefix, and no other options.
328      * Options may be extracted from the filename by implementing
329      * bdrv_parse_filename.
330      */
331     const char *protocol_name;
332     int coroutine_fn (*bdrv_co_truncate)(BlockDriverState *bs, int64_t offset,
333                                          PreallocMode prealloc, Error **errp);
334 
335     int64_t (*bdrv_getlength)(BlockDriverState *bs);
336     bool has_variable_length;
337     int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
338     BlockMeasureInfo *(*bdrv_measure)(QemuOpts *opts, BlockDriverState *in_bs,
339                                       Error **errp);
340 
341     int coroutine_fn (*bdrv_co_pwritev_compressed)(BlockDriverState *bs,
342         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov);
343 
344     int (*bdrv_snapshot_create)(BlockDriverState *bs,
345                                 QEMUSnapshotInfo *sn_info);
346     int (*bdrv_snapshot_goto)(BlockDriverState *bs,
347                               const char *snapshot_id);
348     int (*bdrv_snapshot_delete)(BlockDriverState *bs,
349                                 const char *snapshot_id,
350                                 const char *name,
351                                 Error **errp);
352     int (*bdrv_snapshot_list)(BlockDriverState *bs,
353                               QEMUSnapshotInfo **psn_info);
354     int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
355                                   const char *snapshot_id,
356                                   const char *name,
357                                   Error **errp);
358     int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
359     ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs,
360                                                  Error **errp);
361 
362     int coroutine_fn (*bdrv_save_vmstate)(BlockDriverState *bs,
363                                           QEMUIOVector *qiov,
364                                           int64_t pos);
365     int coroutine_fn (*bdrv_load_vmstate)(BlockDriverState *bs,
366                                           QEMUIOVector *qiov,
367                                           int64_t pos);
368 
369     int (*bdrv_change_backing_file)(BlockDriverState *bs,
370         const char *backing_file, const char *backing_fmt);
371 
372     /* removable device specific */
373     bool (*bdrv_is_inserted)(BlockDriverState *bs);
374     void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
375     void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
376 
377     /* to control generic scsi devices */
378     BlockAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
379         unsigned long int req, void *buf,
380         BlockCompletionFunc *cb, void *opaque);
381     int coroutine_fn (*bdrv_co_ioctl)(BlockDriverState *bs,
382                                       unsigned long int req, void *buf);
383 
384     /* List of options for creating images, terminated by name == NULL */
385     QemuOptsList *create_opts;
386     /*
387      * If this driver supports reopening images this contains a
388      * NULL-terminated list of the runtime options that can be
389      * modified. If an option in this list is unspecified during
390      * reopen then it _must_ be reset to its default value or return
391      * an error.
392      */
393     const char *const *mutable_opts;
394 
395     /*
396      * Returns 0 for completed check, -errno for internal errors.
397      * The check results are stored in result.
398      */
399     int coroutine_fn (*bdrv_co_check)(BlockDriverState *bs,
400                                       BdrvCheckResult *result,
401                                       BdrvCheckMode fix);
402 
403     int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts,
404                               BlockDriverAmendStatusCB *status_cb,
405                               void *cb_opaque,
406                               Error **errp);
407 
408     void (*bdrv_debug_event)(BlockDriverState *bs, BlkdebugEvent event);
409 
410     /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
411     int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
412         const char *tag);
413     int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs,
414         const char *tag);
415     int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
416     bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
417 
418     void (*bdrv_refresh_limits)(BlockDriverState *bs, Error **errp);
419 
420     /*
421      * Returns 1 if newly created images are guaranteed to contain only
422      * zeros, 0 otherwise.
423      */
424     int (*bdrv_has_zero_init)(BlockDriverState *bs);
425 
426     /* Remove fd handlers, timers, and other event loop callbacks so the event
427      * loop is no longer in use.  Called with no in-flight requests and in
428      * depth-first traversal order with parents before child nodes.
429      */
430     void (*bdrv_detach_aio_context)(BlockDriverState *bs);
431 
432     /* Add fd handlers, timers, and other event loop callbacks so I/O requests
433      * can be processed again.  Called with no in-flight requests and in
434      * depth-first traversal order with child nodes before parent nodes.
435      */
436     void (*bdrv_attach_aio_context)(BlockDriverState *bs,
437                                     AioContext *new_context);
438 
439     /* io queue for linux-aio */
440     void (*bdrv_io_plug)(BlockDriverState *bs);
441     void (*bdrv_io_unplug)(BlockDriverState *bs);
442 
443     /**
444      * Try to get @bs's logical and physical block size.
445      * On success, store them in @bsz and return zero.
446      * On failure, return negative errno.
447      */
448     int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz);
449     /**
450      * Try to get @bs's geometry (cyls, heads, sectors)
451      * On success, store them in @geo and return 0.
452      * On failure return -errno.
453      * Only drivers that want to override guest geometry implement this
454      * callback; see hd_geometry_guess().
455      */
456     int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo);
457 
458     /**
459      * bdrv_co_drain_begin is called if implemented in the beginning of a
460      * drain operation to drain and stop any internal sources of requests in
461      * the driver.
462      * bdrv_co_drain_end is called if implemented at the end of the drain.
463      *
464      * They should be used by the driver to e.g. manage scheduled I/O
465      * requests, or toggle an internal state. After the end of the drain new
466      * requests will continue normally.
467      */
468     void coroutine_fn (*bdrv_co_drain_begin)(BlockDriverState *bs);
469     void coroutine_fn (*bdrv_co_drain_end)(BlockDriverState *bs);
470 
471     void (*bdrv_add_child)(BlockDriverState *parent, BlockDriverState *child,
472                            Error **errp);
473     void (*bdrv_del_child)(BlockDriverState *parent, BdrvChild *child,
474                            Error **errp);
475 
476     /**
477      * Informs the block driver that a permission change is intended. The
478      * driver checks whether the change is permissible and may take other
479      * preparations for the change (e.g. get file system locks). This operation
480      * is always followed either by a call to either .bdrv_set_perm or
481      * .bdrv_abort_perm_update.
482      *
483      * Checks whether the requested set of cumulative permissions in @perm
484      * can be granted for accessing @bs and whether no other users are using
485      * permissions other than those given in @shared (both arguments take
486      * BLK_PERM_* bitmasks).
487      *
488      * If both conditions are met, 0 is returned. Otherwise, -errno is returned
489      * and errp is set to an error describing the conflict.
490      */
491     int (*bdrv_check_perm)(BlockDriverState *bs, uint64_t perm,
492                            uint64_t shared, Error **errp);
493 
494     /**
495      * Called to inform the driver that the set of cumulative set of used
496      * permissions for @bs has changed to @perm, and the set of sharable
497      * permission to @shared. The driver can use this to propagate changes to
498      * its children (i.e. request permissions only if a parent actually needs
499      * them).
500      *
501      * This function is only invoked after bdrv_check_perm(), so block drivers
502      * may rely on preparations made in their .bdrv_check_perm implementation.
503      */
504     void (*bdrv_set_perm)(BlockDriverState *bs, uint64_t perm, uint64_t shared);
505 
506     /*
507      * Called to inform the driver that after a previous bdrv_check_perm()
508      * call, the permission update is not performed and any preparations made
509      * for it (e.g. taken file locks) need to be undone.
510      *
511      * This function can be called even for nodes that never saw a
512      * bdrv_check_perm() call. It is a no-op then.
513      */
514     void (*bdrv_abort_perm_update)(BlockDriverState *bs);
515 
516     /**
517      * Returns in @nperm and @nshared the permissions that the driver for @bs
518      * needs on its child @c, based on the cumulative permissions requested by
519      * the parents in @parent_perm and @parent_shared.
520      *
521      * If @c is NULL, return the permissions for attaching a new child for the
522      * given @role.
523      *
524      * If @reopen_queue is non-NULL, don't return the currently needed
525      * permissions, but those that will be needed after applying the
526      * @reopen_queue.
527      */
528      void (*bdrv_child_perm)(BlockDriverState *bs, BdrvChild *c,
529                              const BdrvChildRole *role,
530                              BlockReopenQueue *reopen_queue,
531                              uint64_t parent_perm, uint64_t parent_shared,
532                              uint64_t *nperm, uint64_t *nshared);
533 
534     /**
535      * Bitmaps should be marked as 'IN_USE' in the image on reopening image
536      * as rw. This handler should realize it. It also should unset readonly
537      * field of BlockDirtyBitmap's in case of success.
538      */
539     int (*bdrv_reopen_bitmaps_rw)(BlockDriverState *bs, Error **errp);
540     bool (*bdrv_can_store_new_dirty_bitmap)(BlockDriverState *bs,
541                                             const char *name,
542                                             uint32_t granularity,
543                                             Error **errp);
544     void (*bdrv_remove_persistent_dirty_bitmap)(BlockDriverState *bs,
545                                                 const char *name,
546                                                 Error **errp);
547 
548     /**
549      * Register/unregister a buffer for I/O. For example, when the driver is
550      * interested to know the memory areas that will later be used in iovs, so
551      * that it can do IOMMU mapping with VFIO etc., in order to get better
552      * performance. In the case of VFIO drivers, this callback is used to do
553      * DMA mapping for hot buffers.
554      */
555     void (*bdrv_register_buf)(BlockDriverState *bs, void *host, size_t size);
556     void (*bdrv_unregister_buf)(BlockDriverState *bs, void *host);
557     QLIST_ENTRY(BlockDriver) list;
558 
559     /* Pointer to a NULL-terminated array of names of strong options
560      * that can be specified for bdrv_open(). A strong option is one
561      * that changes the data of a BDS.
562      * If this pointer is NULL, the array is considered empty.
563      * "filename" and "driver" are always considered strong. */
564     const char *const *strong_runtime_opts;
565 };
566 
567 typedef struct BlockLimits {
568     /* Alignment requirement, in bytes, for offset/length of I/O
569      * requests. Must be a power of 2 less than INT_MAX; defaults to
570      * 1 for drivers with modern byte interfaces, and to 512
571      * otherwise. */
572     uint32_t request_alignment;
573 
574     /* Maximum number of bytes that can be discarded at once (since it
575      * is signed, it must be < 2G, if set). Must be multiple of
576      * pdiscard_alignment, but need not be power of 2. May be 0 if no
577      * inherent 32-bit limit */
578     int32_t max_pdiscard;
579 
580     /* Optimal alignment for discard requests in bytes. A power of 2
581      * is best but not mandatory.  Must be a multiple of
582      * bl.request_alignment, and must be less than max_pdiscard if
583      * that is set. May be 0 if bl.request_alignment is good enough */
584     uint32_t pdiscard_alignment;
585 
586     /* Maximum number of bytes that can zeroized at once (since it is
587      * signed, it must be < 2G, if set). Must be multiple of
588      * pwrite_zeroes_alignment. May be 0 if no inherent 32-bit limit */
589     int32_t max_pwrite_zeroes;
590 
591     /* Optimal alignment for write zeroes requests in bytes. A power
592      * of 2 is best but not mandatory.  Must be a multiple of
593      * bl.request_alignment, and must be less than max_pwrite_zeroes
594      * if that is set. May be 0 if bl.request_alignment is good
595      * enough */
596     uint32_t pwrite_zeroes_alignment;
597 
598     /* Optimal transfer length in bytes.  A power of 2 is best but not
599      * mandatory.  Must be a multiple of bl.request_alignment, or 0 if
600      * no preferred size */
601     uint32_t opt_transfer;
602 
603     /* Maximal transfer length in bytes.  Need not be power of 2, but
604      * must be multiple of opt_transfer and bl.request_alignment, or 0
605      * for no 32-bit limit.  For now, anything larger than INT_MAX is
606      * clamped down. */
607     uint32_t max_transfer;
608 
609     /* memory alignment, in bytes so that no bounce buffer is needed */
610     size_t min_mem_alignment;
611 
612     /* memory alignment, in bytes, for bounce buffer */
613     size_t opt_mem_alignment;
614 
615     /* maximum number of iovec elements */
616     int max_iov;
617 } BlockLimits;
618 
619 typedef struct BdrvOpBlocker BdrvOpBlocker;
620 
621 typedef struct BdrvAioNotifier {
622     void (*attached_aio_context)(AioContext *new_context, void *opaque);
623     void (*detach_aio_context)(void *opaque);
624 
625     void *opaque;
626     bool deleted;
627 
628     QLIST_ENTRY(BdrvAioNotifier) list;
629 } BdrvAioNotifier;
630 
631 struct BdrvChildRole {
632     /* If true, bdrv_replace_node() doesn't change the node this BdrvChild
633      * points to. */
634     bool stay_at_node;
635 
636     /* If true, the parent is a BlockDriverState and bdrv_next_all_states()
637      * will return it. This information is used for drain_all, where every node
638      * will be drained separately, so the drain only needs to be propagated to
639      * non-BDS parents. */
640     bool parent_is_bds;
641 
642     void (*inherit_options)(int *child_flags, QDict *child_options,
643                             int parent_flags, QDict *parent_options);
644 
645     void (*change_media)(BdrvChild *child, bool load);
646     void (*resize)(BdrvChild *child);
647 
648     /* Returns a name that is supposedly more useful for human users than the
649      * node name for identifying the node in question (in particular, a BB
650      * name), or NULL if the parent can't provide a better name. */
651     const char *(*get_name)(BdrvChild *child);
652 
653     /* Returns a malloced string that describes the parent of the child for a
654      * human reader. This could be a node-name, BlockBackend name, qdev ID or
655      * QOM path of the device owning the BlockBackend, job type and ID etc. The
656      * caller is responsible for freeing the memory. */
657     char *(*get_parent_desc)(BdrvChild *child);
658 
659     /*
660      * If this pair of functions is implemented, the parent doesn't issue new
661      * requests after returning from .drained_begin() until .drained_end() is
662      * called.
663      *
664      * These functions must not change the graph (and therefore also must not
665      * call aio_poll(), which could change the graph indirectly).
666      *
667      * Note that this can be nested. If drained_begin() was called twice, new
668      * I/O is allowed only after drained_end() was called twice, too.
669      */
670     void (*drained_begin)(BdrvChild *child);
671     void (*drained_end)(BdrvChild *child);
672 
673     /*
674      * Returns whether the parent has pending requests for the child. This
675      * callback is polled after .drained_begin() has been called until all
676      * activity on the child has stopped.
677      */
678     bool (*drained_poll)(BdrvChild *child);
679 
680     /* Notifies the parent that the child has been activated/inactivated (e.g.
681      * when migration is completing) and it can start/stop requesting
682      * permissions and doing I/O on it. */
683     void (*activate)(BdrvChild *child, Error **errp);
684     int (*inactivate)(BdrvChild *child);
685 
686     void (*attach)(BdrvChild *child);
687     void (*detach)(BdrvChild *child);
688 
689     /* Notifies the parent that the filename of its child has changed (e.g.
690      * because the direct child was removed from the backing chain), so that it
691      * can update its reference. */
692     int (*update_filename)(BdrvChild *child, BlockDriverState *new_base,
693                            const char *filename, Error **errp);
694 };
695 
696 extern const BdrvChildRole child_file;
697 extern const BdrvChildRole child_format;
698 extern const BdrvChildRole child_backing;
699 
700 struct BdrvChild {
701     BlockDriverState *bs;
702     char *name;
703     const BdrvChildRole *role;
704     void *opaque;
705 
706     /**
707      * Granted permissions for operating on this BdrvChild (BLK_PERM_* bitmask)
708      */
709     uint64_t perm;
710 
711     /**
712      * Permissions that can still be granted to other users of @bs while this
713      * BdrvChild is still attached to it. (BLK_PERM_* bitmask)
714      */
715     uint64_t shared_perm;
716 
717     /* backup of permissions during permission update procedure */
718     bool has_backup_perm;
719     uint64_t backup_perm;
720     uint64_t backup_shared_perm;
721 
722     /*
723      * This link is frozen: the child can neither be replaced nor
724      * detached from the parent.
725      */
726     bool frozen;
727 
728     QLIST_ENTRY(BdrvChild) next;
729     QLIST_ENTRY(BdrvChild) next_parent;
730 };
731 
732 /*
733  * Note: the function bdrv_append() copies and swaps contents of
734  * BlockDriverStates, so if you add new fields to this struct, please
735  * inspect bdrv_append() to determine if the new fields need to be
736  * copied as well.
737  */
738 struct BlockDriverState {
739     /* Protected by big QEMU lock or read-only after opening.  No special
740      * locking needed during I/O...
741      */
742     int open_flags; /* flags used to open the file, re-used for re-open */
743     bool read_only; /* if true, the media is read only */
744     bool encrypted; /* if true, the media is encrypted */
745     bool sg;        /* if true, the device is a /dev/sg* */
746     bool probed;    /* if true, format was probed rather than specified */
747     bool force_share; /* if true, always allow all shared permissions */
748     bool implicit;  /* if true, this filter node was automatically inserted */
749 
750     BlockDriver *drv; /* NULL means no media */
751     void *opaque;
752 
753     AioContext *aio_context; /* event loop used for fd handlers, timers, etc */
754     /* long-running tasks intended to always use the same AioContext as this
755      * BDS may register themselves in this list to be notified of changes
756      * regarding this BDS's context */
757     QLIST_HEAD(, BdrvAioNotifier) aio_notifiers;
758     bool walking_aio_notifiers; /* to make removal during iteration safe */
759 
760     char filename[PATH_MAX];
761     char backing_file[PATH_MAX]; /* if non zero, the image is a diff of
762                                     this file image */
763     /* The backing filename indicated by the image header; if we ever
764      * open this file, then this is replaced by the resulting BDS's
765      * filename (i.e. after a bdrv_refresh_filename() run). */
766     char auto_backing_file[PATH_MAX];
767     char backing_format[16]; /* if non-zero and backing_file exists */
768 
769     QDict *full_open_options;
770     char exact_filename[PATH_MAX];
771 
772     BdrvChild *backing;
773     BdrvChild *file;
774 
775     /* I/O Limits */
776     BlockLimits bl;
777 
778     /* Flags honored during pwrite (so far: BDRV_REQ_FUA,
779      * BDRV_REQ_WRITE_UNCHANGED).
780      * If a driver does not support BDRV_REQ_WRITE_UNCHANGED, those
781      * writes will be issued as normal writes without the flag set.
782      * This is important to note for drivers that do not explicitly
783      * request a WRITE permission for their children and instead take
784      * the same permissions as their parent did (this is commonly what
785      * block filters do).  Such drivers have to be aware that the
786      * parent may have taken a WRITE_UNCHANGED permission only and is
787      * issuing such requests.  Drivers either must make sure that
788      * these requests do not result in plain WRITE accesses (usually
789      * by supporting BDRV_REQ_WRITE_UNCHANGED, and then forwarding
790      * every incoming write request as-is, including potentially that
791      * flag), or they have to explicitly take the WRITE permission for
792      * their children. */
793     unsigned int supported_write_flags;
794     /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA,
795      * BDRV_REQ_MAY_UNMAP, BDRV_REQ_WRITE_UNCHANGED) */
796     unsigned int supported_zero_flags;
797 
798     /* the following member gives a name to every node on the bs graph. */
799     char node_name[32];
800     /* element of the list of named nodes building the graph */
801     QTAILQ_ENTRY(BlockDriverState) node_list;
802     /* element of the list of all BlockDriverStates (all_bdrv_states) */
803     QTAILQ_ENTRY(BlockDriverState) bs_list;
804     /* element of the list of monitor-owned BDS */
805     QTAILQ_ENTRY(BlockDriverState) monitor_list;
806     int refcnt;
807 
808     /* operation blockers */
809     QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX];
810 
811     /* long-running background operation */
812     BlockJob *job;
813 
814     /* The node that this node inherited default options from (and a reopen on
815      * which can affect this node by changing these defaults). This is always a
816      * parent node of this node. */
817     BlockDriverState *inherits_from;
818     QLIST_HEAD(, BdrvChild) children;
819     QLIST_HEAD(, BdrvChild) parents;
820 
821     QDict *options;
822     QDict *explicit_options;
823     BlockdevDetectZeroesOptions detect_zeroes;
824 
825     /* The error object in use for blocking operations on backing_hd */
826     Error *backing_blocker;
827 
828     /* Protected by AioContext lock */
829 
830     /* If we are reading a disk image, give its size in sectors.
831      * Generally read-only; it is written to by load_snapshot and
832      * save_snaphost, but the block layer is quiescent during those.
833      */
834     int64_t total_sectors;
835 
836     /* Callback before write request is processed */
837     NotifierWithReturnList before_write_notifiers;
838 
839     /* threshold limit for writes, in bytes. "High water mark". */
840     uint64_t write_threshold_offset;
841     NotifierWithReturn write_threshold_notifier;
842 
843     /* Writing to the list requires the BQL _and_ the dirty_bitmap_mutex.
844      * Reading from the list can be done with either the BQL or the
845      * dirty_bitmap_mutex.  Modifying a bitmap only requires
846      * dirty_bitmap_mutex.  */
847     QemuMutex dirty_bitmap_mutex;
848     QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps;
849 
850     /* Offset after the highest byte written to */
851     Stat64 wr_highest_offset;
852 
853     /* If true, copy read backing sectors into image.  Can be >1 if more
854      * than one client has requested copy-on-read.  Accessed with atomic
855      * ops.
856      */
857     int copy_on_read;
858 
859     /* number of in-flight requests; overall and serialising.
860      * Accessed with atomic ops.
861      */
862     unsigned int in_flight;
863     unsigned int serialising_in_flight;
864 
865     /* counter for nested bdrv_io_plug.
866      * Accessed with atomic ops.
867     */
868     unsigned io_plugged;
869 
870     /* do we need to tell the quest if we have a volatile write cache? */
871     int enable_write_cache;
872 
873     /* Accessed with atomic ops.  */
874     int quiesce_counter;
875     int recursive_quiesce_counter;
876 
877     unsigned int write_gen;               /* Current data generation */
878 
879     /* Protected by reqs_lock.  */
880     CoMutex reqs_lock;
881     QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
882     CoQueue flush_queue;                  /* Serializing flush queue */
883     bool active_flush_req;                /* Flush request in flight? */
884 
885     /* Only read/written by whoever has set active_flush_req to true.  */
886     unsigned int flushed_gen;             /* Flushed write generation */
887 };
888 
889 struct BlockBackendRootState {
890     int open_flags;
891     bool read_only;
892     BlockdevDetectZeroesOptions detect_zeroes;
893 };
894 
895 typedef enum BlockMirrorBackingMode {
896     /* Reuse the existing backing chain from the source for the target.
897      * - sync=full: Set backing BDS to NULL.
898      * - sync=top:  Use source's backing BDS.
899      * - sync=none: Use source as the backing BDS. */
900     MIRROR_SOURCE_BACKING_CHAIN,
901 
902     /* Open the target's backing chain completely anew */
903     MIRROR_OPEN_BACKING_CHAIN,
904 
905     /* Do not change the target's backing BDS after job completion */
906     MIRROR_LEAVE_BACKING_CHAIN,
907 } BlockMirrorBackingMode;
908 
909 static inline BlockDriverState *backing_bs(BlockDriverState *bs)
910 {
911     return bs->backing ? bs->backing->bs : NULL;
912 }
913 
914 
915 /* Essential block drivers which must always be statically linked into qemu, and
916  * which therefore can be accessed without using bdrv_find_format() */
917 extern BlockDriver bdrv_file;
918 extern BlockDriver bdrv_raw;
919 extern BlockDriver bdrv_qcow2;
920 
921 int coroutine_fn bdrv_co_preadv(BdrvChild *child,
922     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
923     BdrvRequestFlags flags);
924 int coroutine_fn bdrv_co_pwritev(BdrvChild *child,
925     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
926     BdrvRequestFlags flags);
927 
928 static inline int coroutine_fn bdrv_co_pread(BdrvChild *child,
929     int64_t offset, unsigned int bytes, void *buf, BdrvRequestFlags flags)
930 {
931     QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes);
932 
933     return bdrv_co_preadv(child, offset, bytes, &qiov, flags);
934 }
935 
936 static inline int coroutine_fn bdrv_co_pwrite(BdrvChild *child,
937     int64_t offset, unsigned int bytes, void *buf, BdrvRequestFlags flags)
938 {
939     QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes);
940 
941     return bdrv_co_pwritev(child, offset, bytes, &qiov, flags);
942 }
943 
944 extern unsigned int bdrv_drain_all_count;
945 void bdrv_apply_subtree_drain(BdrvChild *child, BlockDriverState *new_parent);
946 void bdrv_unapply_subtree_drain(BdrvChild *child, BlockDriverState *old_parent);
947 
948 int get_tmp_filename(char *filename, int size);
949 BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size,
950                             const char *filename);
951 
952 void bdrv_parse_filename_strip_prefix(const char *filename, const char *prefix,
953                                       QDict *options);
954 
955 
956 /**
957  * bdrv_add_before_write_notifier:
958  *
959  * Register a callback that is invoked before write requests are processed but
960  * after any throttling or waiting for overlapping requests.
961  */
962 void bdrv_add_before_write_notifier(BlockDriverState *bs,
963                                     NotifierWithReturn *notifier);
964 
965 /**
966  * bdrv_detach_aio_context:
967  *
968  * May be called from .bdrv_detach_aio_context() to detach children from the
969  * current #AioContext.  This is only needed by block drivers that manage their
970  * own children.  Both ->file and ->backing are automatically handled and
971  * block drivers should not call this function on them explicitly.
972  */
973 void bdrv_detach_aio_context(BlockDriverState *bs);
974 
975 /**
976  * bdrv_attach_aio_context:
977  *
978  * May be called from .bdrv_attach_aio_context() to attach children to the new
979  * #AioContext.  This is only needed by block drivers that manage their own
980  * children.  Both ->file and ->backing are automatically handled and block
981  * drivers should not call this function on them explicitly.
982  */
983 void bdrv_attach_aio_context(BlockDriverState *bs,
984                              AioContext *new_context);
985 
986 /**
987  * bdrv_add_aio_context_notifier:
988  *
989  * If a long-running job intends to be always run in the same AioContext as a
990  * certain BDS, it may use this function to be notified of changes regarding the
991  * association of the BDS to an AioContext.
992  *
993  * attached_aio_context() is called after the target BDS has been attached to a
994  * new AioContext; detach_aio_context() is called before the target BDS is being
995  * detached from its old AioContext.
996  */
997 void bdrv_add_aio_context_notifier(BlockDriverState *bs,
998         void (*attached_aio_context)(AioContext *new_context, void *opaque),
999         void (*detach_aio_context)(void *opaque), void *opaque);
1000 
1001 /**
1002  * bdrv_remove_aio_context_notifier:
1003  *
1004  * Unsubscribe of change notifications regarding the BDS's AioContext. The
1005  * parameters given here have to be the same as those given to
1006  * bdrv_add_aio_context_notifier().
1007  */
1008 void bdrv_remove_aio_context_notifier(BlockDriverState *bs,
1009                                       void (*aio_context_attached)(AioContext *,
1010                                                                    void *),
1011                                       void (*aio_context_detached)(void *),
1012                                       void *opaque);
1013 
1014 /**
1015  * bdrv_wakeup:
1016  * @bs: The BlockDriverState for which an I/O operation has been completed.
1017  *
1018  * Wake up the main thread if it is waiting on BDRV_POLL_WHILE.  During
1019  * synchronous I/O on a BlockDriverState that is attached to another
1020  * I/O thread, the main thread lets the I/O thread's event loop run,
1021  * waiting for the I/O operation to complete.  A bdrv_wakeup will wake
1022  * up the main thread if necessary.
1023  *
1024  * Manual calls to bdrv_wakeup are rarely necessary, because
1025  * bdrv_dec_in_flight already calls it.
1026  */
1027 void bdrv_wakeup(BlockDriverState *bs);
1028 
1029 #ifdef _WIN32
1030 int is_windows_drive(const char *filename);
1031 #endif
1032 
1033 /**
1034  * stream_start:
1035  * @job_id: The id of the newly-created job, or %NULL to use the
1036  * device name of @bs.
1037  * @bs: Block device to operate on.
1038  * @base: Block device that will become the new base, or %NULL to
1039  * flatten the whole backing file chain onto @bs.
1040  * @backing_file_str: The file name that will be written to @bs as the
1041  * the new backing file if the job completes. Ignored if @base is %NULL.
1042  * @creation_flags: Flags that control the behavior of the Job lifetime.
1043  *                  See @BlockJobCreateFlags
1044  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1045  * @on_error: The action to take upon error.
1046  * @errp: Error object.
1047  *
1048  * Start a streaming operation on @bs.  Clusters that are unallocated
1049  * in @bs, but allocated in any image between @base and @bs (both
1050  * exclusive) will be written to @bs.  At the end of a successful
1051  * streaming job, the backing file of @bs will be changed to
1052  * @backing_file_str in the written image and to @base in the live
1053  * BlockDriverState.
1054  */
1055 void stream_start(const char *job_id, BlockDriverState *bs,
1056                   BlockDriverState *base, const char *backing_file_str,
1057                   int creation_flags, int64_t speed,
1058                   BlockdevOnError on_error, Error **errp);
1059 
1060 /**
1061  * commit_start:
1062  * @job_id: The id of the newly-created job, or %NULL to use the
1063  * device name of @bs.
1064  * @bs: Active block device.
1065  * @top: Top block device to be committed.
1066  * @base: Block device that will be written into, and become the new top.
1067  * @creation_flags: Flags that control the behavior of the Job lifetime.
1068  *                  See @BlockJobCreateFlags
1069  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1070  * @on_error: The action to take upon error.
1071  * @backing_file_str: String to use as the backing file in @top's overlay
1072  * @filter_node_name: The node name that should be assigned to the filter
1073  * driver that the commit job inserts into the graph above @top. NULL means
1074  * that a node name should be autogenerated.
1075  * @errp: Error object.
1076  *
1077  */
1078 void commit_start(const char *job_id, BlockDriverState *bs,
1079                   BlockDriverState *base, BlockDriverState *top,
1080                   int creation_flags, int64_t speed,
1081                   BlockdevOnError on_error, const char *backing_file_str,
1082                   const char *filter_node_name, Error **errp);
1083 /**
1084  * commit_active_start:
1085  * @job_id: The id of the newly-created job, or %NULL to use the
1086  * device name of @bs.
1087  * @bs: Active block device to be committed.
1088  * @base: Block device that will be written into, and become the new top.
1089  * @creation_flags: Flags that control the behavior of the Job lifetime.
1090  *                  See @BlockJobCreateFlags
1091  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1092  * @on_error: The action to take upon error.
1093  * @filter_node_name: The node name that should be assigned to the filter
1094  * driver that the commit job inserts into the graph above @bs. NULL means that
1095  * a node name should be autogenerated.
1096  * @cb: Completion function for the job.
1097  * @opaque: Opaque pointer value passed to @cb.
1098  * @auto_complete: Auto complete the job.
1099  * @errp: Error object.
1100  *
1101  */
1102 void commit_active_start(const char *job_id, BlockDriverState *bs,
1103                          BlockDriverState *base, int creation_flags,
1104                          int64_t speed, BlockdevOnError on_error,
1105                          const char *filter_node_name,
1106                          BlockCompletionFunc *cb, void *opaque,
1107                          bool auto_complete, Error **errp);
1108 /*
1109  * mirror_start:
1110  * @job_id: The id of the newly-created job, or %NULL to use the
1111  * device name of @bs.
1112  * @bs: Block device to operate on.
1113  * @target: Block device to write to.
1114  * @replaces: Block graph node name to replace once the mirror is done. Can
1115  *            only be used when full mirroring is selected.
1116  * @creation_flags: Flags that control the behavior of the Job lifetime.
1117  *                  See @BlockJobCreateFlags
1118  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1119  * @granularity: The chosen granularity for the dirty bitmap.
1120  * @buf_size: The amount of data that can be in flight at one time.
1121  * @mode: Whether to collapse all images in the chain to the target.
1122  * @backing_mode: How to establish the target's backing chain after completion.
1123  * @on_source_error: The action to take upon error reading from the source.
1124  * @on_target_error: The action to take upon error writing to the target.
1125  * @unmap: Whether to unmap target where source sectors only contain zeroes.
1126  * @filter_node_name: The node name that should be assigned to the filter
1127  * driver that the mirror job inserts into the graph above @bs. NULL means that
1128  * a node name should be autogenerated.
1129  * @copy_mode: When to trigger writes to the target.
1130  * @errp: Error object.
1131  *
1132  * Start a mirroring operation on @bs.  Clusters that are allocated
1133  * in @bs will be written to @target until the job is cancelled or
1134  * manually completed.  At the end of a successful mirroring job,
1135  * @bs will be switched to read from @target.
1136  */
1137 void mirror_start(const char *job_id, BlockDriverState *bs,
1138                   BlockDriverState *target, const char *replaces,
1139                   int creation_flags, int64_t speed,
1140                   uint32_t granularity, int64_t buf_size,
1141                   MirrorSyncMode mode, BlockMirrorBackingMode backing_mode,
1142                   BlockdevOnError on_source_error,
1143                   BlockdevOnError on_target_error,
1144                   bool unmap, const char *filter_node_name,
1145                   MirrorCopyMode copy_mode, Error **errp);
1146 
1147 /*
1148  * backup_job_create:
1149  * @job_id: The id of the newly-created job, or %NULL to use the
1150  * device name of @bs.
1151  * @bs: Block device to operate on.
1152  * @target: Block device to write to.
1153  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1154  * @sync_mode: What parts of the disk image should be copied to the destination.
1155  * @sync_bitmap: The dirty bitmap if sync_mode is MIRROR_SYNC_MODE_INCREMENTAL.
1156  * @on_source_error: The action to take upon error reading from the source.
1157  * @on_target_error: The action to take upon error writing to the target.
1158  * @creation_flags: Flags that control the behavior of the Job lifetime.
1159  *                  See @BlockJobCreateFlags
1160  * @cb: Completion function for the job.
1161  * @opaque: Opaque pointer value passed to @cb.
1162  * @txn: Transaction that this job is part of (may be NULL).
1163  *
1164  * Create a backup operation on @bs.  Clusters in @bs are written to @target
1165  * until the job is cancelled or manually completed.
1166  */
1167 BlockJob *backup_job_create(const char *job_id, BlockDriverState *bs,
1168                             BlockDriverState *target, int64_t speed,
1169                             MirrorSyncMode sync_mode,
1170                             BdrvDirtyBitmap *sync_bitmap,
1171                             bool compress,
1172                             BlockdevOnError on_source_error,
1173                             BlockdevOnError on_target_error,
1174                             int creation_flags,
1175                             BlockCompletionFunc *cb, void *opaque,
1176                             JobTxn *txn, Error **errp);
1177 
1178 void hmp_drive_add_node(Monitor *mon, const char *optstr);
1179 
1180 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
1181                                   const char *child_name,
1182                                   const BdrvChildRole *child_role,
1183                                   uint64_t perm, uint64_t shared_perm,
1184                                   void *opaque, Error **errp);
1185 void bdrv_root_unref_child(BdrvChild *child);
1186 
1187 int bdrv_child_try_set_perm(BdrvChild *c, uint64_t perm, uint64_t shared,
1188                             Error **errp);
1189 
1190 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1191  * block filters: Forward CONSISTENT_READ, WRITE, WRITE_UNCHANGED and RESIZE to
1192  * all children */
1193 void bdrv_filter_default_perms(BlockDriverState *bs, BdrvChild *c,
1194                                const BdrvChildRole *role,
1195                                BlockReopenQueue *reopen_queue,
1196                                uint64_t perm, uint64_t shared,
1197                                uint64_t *nperm, uint64_t *nshared);
1198 
1199 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1200  * (non-raw) image formats: Like above for bs->backing, but for bs->file it
1201  * requires WRITE | RESIZE for read-write images, always requires
1202  * CONSISTENT_READ and doesn't share WRITE. */
1203 void bdrv_format_default_perms(BlockDriverState *bs, BdrvChild *c,
1204                                const BdrvChildRole *role,
1205                                BlockReopenQueue *reopen_queue,
1206                                uint64_t perm, uint64_t shared,
1207                                uint64_t *nperm, uint64_t *nshared);
1208 
1209 /*
1210  * Default implementation for drivers to pass bdrv_co_block_status() to
1211  * their file.
1212  */
1213 int coroutine_fn bdrv_co_block_status_from_file(BlockDriverState *bs,
1214                                                 bool want_zero,
1215                                                 int64_t offset,
1216                                                 int64_t bytes,
1217                                                 int64_t *pnum,
1218                                                 int64_t *map,
1219                                                 BlockDriverState **file);
1220 /*
1221  * Default implementation for drivers to pass bdrv_co_block_status() to
1222  * their backing file.
1223  */
1224 int coroutine_fn bdrv_co_block_status_from_backing(BlockDriverState *bs,
1225                                                    bool want_zero,
1226                                                    int64_t offset,
1227                                                    int64_t bytes,
1228                                                    int64_t *pnum,
1229                                                    int64_t *map,
1230                                                    BlockDriverState **file);
1231 const char *bdrv_get_parent_name(const BlockDriverState *bs);
1232 void blk_dev_change_media_cb(BlockBackend *blk, bool load, Error **errp);
1233 bool blk_dev_has_removable_media(BlockBackend *blk);
1234 bool blk_dev_has_tray(BlockBackend *blk);
1235 void blk_dev_eject_request(BlockBackend *blk, bool force);
1236 bool blk_dev_is_tray_open(BlockBackend *blk);
1237 bool blk_dev_is_medium_locked(BlockBackend *blk);
1238 
1239 void bdrv_set_dirty(BlockDriverState *bs, int64_t offset, int64_t bytes);
1240 
1241 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap **out);
1242 void bdrv_restore_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap *backup);
1243 
1244 void bdrv_inc_in_flight(BlockDriverState *bs);
1245 void bdrv_dec_in_flight(BlockDriverState *bs);
1246 
1247 void blockdev_close_all_bdrv_states(void);
1248 
1249 int coroutine_fn bdrv_co_copy_range_from(BdrvChild *src, uint64_t src_offset,
1250                                          BdrvChild *dst, uint64_t dst_offset,
1251                                          uint64_t bytes,
1252                                          BdrvRequestFlags read_flags,
1253                                          BdrvRequestFlags write_flags);
1254 int coroutine_fn bdrv_co_copy_range_to(BdrvChild *src, uint64_t src_offset,
1255                                        BdrvChild *dst, uint64_t dst_offset,
1256                                        uint64_t bytes,
1257                                        BdrvRequestFlags read_flags,
1258                                        BdrvRequestFlags write_flags);
1259 
1260 int refresh_total_sectors(BlockDriverState *bs, int64_t hint);
1261 
1262 #endif /* BLOCK_INT_H */
1263