xref: /openbmc/qemu/include/block/block_int.h (revision 073d9f2c)
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 
60 #define BLOCK_PROBE_BUF_SIZE        512
61 
62 enum BdrvTrackedRequestType {
63     BDRV_TRACKED_READ,
64     BDRV_TRACKED_WRITE,
65     BDRV_TRACKED_DISCARD,
66     BDRV_TRACKED_TRUNCATE,
67 };
68 
69 typedef struct BdrvTrackedRequest {
70     BlockDriverState *bs;
71     int64_t offset;
72     uint64_t bytes;
73     enum BdrvTrackedRequestType type;
74 
75     bool serialising;
76     int64_t overlap_offset;
77     uint64_t overlap_bytes;
78 
79     QLIST_ENTRY(BdrvTrackedRequest) list;
80     Coroutine *co; /* owner, used for deadlock detection */
81     CoQueue wait_queue; /* coroutines blocked on this request */
82 
83     struct BdrvTrackedRequest *waiting_for;
84 } BdrvTrackedRequest;
85 
86 struct BlockDriver {
87     const char *format_name;
88     int instance_size;
89 
90     /* set to true if the BlockDriver is a block filter. Block filters pass
91      * certain callbacks that refer to data (see block.c) to their bs->file if
92      * the driver doesn't implement them. Drivers that do not wish to forward
93      * must implement them and return -ENOTSUP.
94      */
95     bool is_filter;
96     /* for snapshots block filter like Quorum can implement the
97      * following recursive callback.
98      * It's purpose is to recurse on the filter children while calling
99      * bdrv_recurse_is_first_non_filter on them.
100      * For a sample implementation look in the future Quorum block filter.
101      */
102     bool (*bdrv_recurse_is_first_non_filter)(BlockDriverState *bs,
103                                              BlockDriverState *candidate);
104 
105     int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename);
106     int (*bdrv_probe_device)(const char *filename);
107 
108     /* Any driver implementing this callback is expected to be able to handle
109      * NULL file names in its .bdrv_open() implementation */
110     void (*bdrv_parse_filename)(const char *filename, QDict *options, Error **errp);
111     /* Drivers not implementing bdrv_parse_filename nor bdrv_open should have
112      * this field set to true, except ones that are defined only by their
113      * child's bs.
114      * An example of the last type will be the quorum block driver.
115      */
116     bool bdrv_needs_filename;
117 
118     /* Set if a driver can support backing files */
119     bool supports_backing;
120 
121     /* For handling image reopen for split or non-split files */
122     int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state,
123                                BlockReopenQueue *queue, Error **errp);
124     void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state);
125     void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state);
126     void (*bdrv_join_options)(QDict *options, QDict *old_options);
127 
128     int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags,
129                      Error **errp);
130 
131     /* Protocol drivers should implement this instead of bdrv_open */
132     int (*bdrv_file_open)(BlockDriverState *bs, QDict *options, int flags,
133                           Error **errp);
134     void (*bdrv_close)(BlockDriverState *bs);
135     int coroutine_fn (*bdrv_co_create)(BlockdevCreateOptions *opts,
136                                        Error **errp);
137     int coroutine_fn (*bdrv_co_create_opts)(const char *filename,
138                                             QemuOpts *opts,
139                                             Error **errp);
140     int (*bdrv_make_empty)(BlockDriverState *bs);
141 
142     void (*bdrv_refresh_filename)(BlockDriverState *bs, QDict *options);
143 
144     /* aio */
145     BlockAIOCB *(*bdrv_aio_preadv)(BlockDriverState *bs,
146         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags,
147         BlockCompletionFunc *cb, void *opaque);
148     BlockAIOCB *(*bdrv_aio_pwritev)(BlockDriverState *bs,
149         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags,
150         BlockCompletionFunc *cb, void *opaque);
151     BlockAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs,
152         BlockCompletionFunc *cb, void *opaque);
153     BlockAIOCB *(*bdrv_aio_pdiscard)(BlockDriverState *bs,
154         int64_t offset, int bytes,
155         BlockCompletionFunc *cb, void *opaque);
156 
157     int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs,
158         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
159 
160     /**
161      * @offset: position in bytes to read at
162      * @bytes: number of bytes to read
163      * @qiov: the buffers to fill with read data
164      * @flags: currently unused, always 0
165      *
166      * @offset and @bytes will be a multiple of 'request_alignment',
167      * but the length of individual @qiov elements does not have to
168      * be a multiple.
169      *
170      * @bytes will always equal the total size of @qiov, and will be
171      * no larger than 'max_transfer'.
172      *
173      * The buffer in @qiov may point directly to guest memory.
174      */
175     int coroutine_fn (*bdrv_co_preadv)(BlockDriverState *bs,
176         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags);
177     int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs,
178         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, int flags);
179     /**
180      * @offset: position in bytes to write at
181      * @bytes: number of bytes to write
182      * @qiov: the buffers containing data to write
183      * @flags: zero or more bits allowed by 'supported_write_flags'
184      *
185      * @offset and @bytes will be a multiple of 'request_alignment',
186      * but the length of individual @qiov elements does not have to
187      * be a multiple.
188      *
189      * @bytes will always equal the total size of @qiov, and will be
190      * no larger than 'max_transfer'.
191      *
192      * The buffer in @qiov may point directly to guest memory.
193      */
194     int coroutine_fn (*bdrv_co_pwritev)(BlockDriverState *bs,
195         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags);
196 
197     /*
198      * Efficiently zero a region of the disk image.  Typically an image format
199      * would use a compact metadata representation to implement this.  This
200      * function pointer may be NULL or return -ENOSUP and .bdrv_co_writev()
201      * will be called instead.
202      */
203     int coroutine_fn (*bdrv_co_pwrite_zeroes)(BlockDriverState *bs,
204         int64_t offset, int bytes, BdrvRequestFlags flags);
205     int coroutine_fn (*bdrv_co_pdiscard)(BlockDriverState *bs,
206         int64_t offset, int bytes);
207 
208     /* Map [offset, offset + nbytes) range onto a child of @bs to copy from,
209      * and invoke bdrv_co_copy_range_from(child, ...), or invoke
210      * bdrv_co_copy_range_to() if @bs is the leaf child to copy data from.
211      *
212      * See the comment of bdrv_co_copy_range for the parameter and return value
213      * semantics.
214      */
215     int coroutine_fn (*bdrv_co_copy_range_from)(BlockDriverState *bs,
216                                                 BdrvChild *src,
217                                                 uint64_t offset,
218                                                 BdrvChild *dst,
219                                                 uint64_t dst_offset,
220                                                 uint64_t bytes,
221                                                 BdrvRequestFlags read_flags,
222                                                 BdrvRequestFlags write_flags);
223 
224     /* Map [offset, offset + nbytes) range onto a child of bs to copy data to,
225      * and invoke bdrv_co_copy_range_to(child, src, ...), or perform the copy
226      * operation if @bs is the leaf and @src has the same BlockDriver.  Return
227      * -ENOTSUP if @bs is the leaf but @src has a different BlockDriver.
228      *
229      * See the comment of bdrv_co_copy_range for the parameter and return value
230      * semantics.
231      */
232     int coroutine_fn (*bdrv_co_copy_range_to)(BlockDriverState *bs,
233                                               BdrvChild *src,
234                                               uint64_t src_offset,
235                                               BdrvChild *dst,
236                                               uint64_t dst_offset,
237                                               uint64_t bytes,
238                                               BdrvRequestFlags read_flags,
239                                               BdrvRequestFlags write_flags);
240 
241     /*
242      * Building block for bdrv_block_status[_above] and
243      * bdrv_is_allocated[_above].  The driver should answer only
244      * according to the current layer, and should only need to set
245      * BDRV_BLOCK_DATA, BDRV_BLOCK_ZERO, BDRV_BLOCK_OFFSET_VALID,
246      * and/or BDRV_BLOCK_RAW; if the current layer defers to a backing
247      * layer, the result should be 0 (and not BDRV_BLOCK_ZERO).  See
248      * block.h for the overall meaning of the bits.  As a hint, the
249      * flag want_zero is true if the caller cares more about precise
250      * mappings (favor accurate _OFFSET_VALID/_ZERO) or false for
251      * overall allocation (favor larger *pnum, perhaps by reporting
252      * _DATA instead of _ZERO).  The block layer guarantees input
253      * clamped to bdrv_getlength() and aligned to request_alignment,
254      * as well as non-NULL pnum, map, and file; in turn, the driver
255      * must return an error or set pnum to an aligned non-zero value.
256      */
257     int coroutine_fn (*bdrv_co_block_status)(BlockDriverState *bs,
258         bool want_zero, int64_t offset, int64_t bytes, int64_t *pnum,
259         int64_t *map, BlockDriverState **file);
260 
261     /*
262      * Invalidate any cached meta-data.
263      */
264     void coroutine_fn (*bdrv_co_invalidate_cache)(BlockDriverState *bs,
265                                                   Error **errp);
266     int (*bdrv_inactivate)(BlockDriverState *bs);
267 
268     /*
269      * Flushes all data for all layers by calling bdrv_co_flush for underlying
270      * layers, if needed. This function is needed for deterministic
271      * synchronization of the flush finishing callback.
272      */
273     int coroutine_fn (*bdrv_co_flush)(BlockDriverState *bs);
274 
275     /*
276      * Flushes all data that was already written to the OS all the way down to
277      * the disk (for example file-posix.c calls fsync()).
278      */
279     int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs);
280 
281     /*
282      * Flushes all internal caches to the OS. The data may still sit in a
283      * writeback cache of the host OS, but it will survive a crash of the qemu
284      * process.
285      */
286     int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs);
287 
288     /*
289      * Drivers setting this field must be able to work with just a plain
290      * filename with '<protocol_name>:' as a prefix, and no other options.
291      * Options may be extracted from the filename by implementing
292      * bdrv_parse_filename.
293      */
294     const char *protocol_name;
295     int coroutine_fn (*bdrv_co_truncate)(BlockDriverState *bs, int64_t offset,
296                                          PreallocMode prealloc, Error **errp);
297 
298     int64_t (*bdrv_getlength)(BlockDriverState *bs);
299     bool has_variable_length;
300     int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
301     BlockMeasureInfo *(*bdrv_measure)(QemuOpts *opts, BlockDriverState *in_bs,
302                                       Error **errp);
303 
304     int coroutine_fn (*bdrv_co_pwritev_compressed)(BlockDriverState *bs,
305         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov);
306 
307     int (*bdrv_snapshot_create)(BlockDriverState *bs,
308                                 QEMUSnapshotInfo *sn_info);
309     int (*bdrv_snapshot_goto)(BlockDriverState *bs,
310                               const char *snapshot_id);
311     int (*bdrv_snapshot_delete)(BlockDriverState *bs,
312                                 const char *snapshot_id,
313                                 const char *name,
314                                 Error **errp);
315     int (*bdrv_snapshot_list)(BlockDriverState *bs,
316                               QEMUSnapshotInfo **psn_info);
317     int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
318                                   const char *snapshot_id,
319                                   const char *name,
320                                   Error **errp);
321     int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
322     ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs);
323 
324     int coroutine_fn (*bdrv_save_vmstate)(BlockDriverState *bs,
325                                           QEMUIOVector *qiov,
326                                           int64_t pos);
327     int coroutine_fn (*bdrv_load_vmstate)(BlockDriverState *bs,
328                                           QEMUIOVector *qiov,
329                                           int64_t pos);
330 
331     int (*bdrv_change_backing_file)(BlockDriverState *bs,
332         const char *backing_file, const char *backing_fmt);
333 
334     /* removable device specific */
335     bool (*bdrv_is_inserted)(BlockDriverState *bs);
336     void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
337     void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
338 
339     /* to control generic scsi devices */
340     BlockAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
341         unsigned long int req, void *buf,
342         BlockCompletionFunc *cb, void *opaque);
343     int coroutine_fn (*bdrv_co_ioctl)(BlockDriverState *bs,
344                                       unsigned long int req, void *buf);
345 
346     /* List of options for creating images, terminated by name == NULL */
347     QemuOptsList *create_opts;
348 
349     /*
350      * Returns 0 for completed check, -errno for internal errors.
351      * The check results are stored in result.
352      */
353     int coroutine_fn (*bdrv_co_check)(BlockDriverState *bs,
354                                       BdrvCheckResult *result,
355                                       BdrvCheckMode fix);
356 
357     int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts,
358                               BlockDriverAmendStatusCB *status_cb,
359                               void *cb_opaque,
360                               Error **errp);
361 
362     void (*bdrv_debug_event)(BlockDriverState *bs, BlkdebugEvent event);
363 
364     /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
365     int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
366         const char *tag);
367     int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs,
368         const char *tag);
369     int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
370     bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
371 
372     void (*bdrv_refresh_limits)(BlockDriverState *bs, Error **errp);
373 
374     /*
375      * Returns 1 if newly created images are guaranteed to contain only
376      * zeros, 0 otherwise.
377      */
378     int (*bdrv_has_zero_init)(BlockDriverState *bs);
379 
380     /* Remove fd handlers, timers, and other event loop callbacks so the event
381      * loop is no longer in use.  Called with no in-flight requests and in
382      * depth-first traversal order with parents before child nodes.
383      */
384     void (*bdrv_detach_aio_context)(BlockDriverState *bs);
385 
386     /* Add fd handlers, timers, and other event loop callbacks so I/O requests
387      * can be processed again.  Called with no in-flight requests and in
388      * depth-first traversal order with child nodes before parent nodes.
389      */
390     void (*bdrv_attach_aio_context)(BlockDriverState *bs,
391                                     AioContext *new_context);
392 
393     /* io queue for linux-aio */
394     void (*bdrv_io_plug)(BlockDriverState *bs);
395     void (*bdrv_io_unplug)(BlockDriverState *bs);
396 
397     /**
398      * Try to get @bs's logical and physical block size.
399      * On success, store them in @bsz and return zero.
400      * On failure, return negative errno.
401      */
402     int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz);
403     /**
404      * Try to get @bs's geometry (cyls, heads, sectors)
405      * On success, store them in @geo and return 0.
406      * On failure return -errno.
407      * Only drivers that want to override guest geometry implement this
408      * callback; see hd_geometry_guess().
409      */
410     int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo);
411 
412     /**
413      * bdrv_co_drain_begin is called if implemented in the beginning of a
414      * drain operation to drain and stop any internal sources of requests in
415      * the driver.
416      * bdrv_co_drain_end is called if implemented at the end of the drain.
417      *
418      * They should be used by the driver to e.g. manage scheduled I/O
419      * requests, or toggle an internal state. After the end of the drain new
420      * requests will continue normally.
421      */
422     void coroutine_fn (*bdrv_co_drain_begin)(BlockDriverState *bs);
423     void coroutine_fn (*bdrv_co_drain_end)(BlockDriverState *bs);
424 
425     void (*bdrv_add_child)(BlockDriverState *parent, BlockDriverState *child,
426                            Error **errp);
427     void (*bdrv_del_child)(BlockDriverState *parent, BdrvChild *child,
428                            Error **errp);
429 
430     /**
431      * Informs the block driver that a permission change is intended. The
432      * driver checks whether the change is permissible and may take other
433      * preparations for the change (e.g. get file system locks). This operation
434      * is always followed either by a call to either .bdrv_set_perm or
435      * .bdrv_abort_perm_update.
436      *
437      * Checks whether the requested set of cumulative permissions in @perm
438      * can be granted for accessing @bs and whether no other users are using
439      * permissions other than those given in @shared (both arguments take
440      * BLK_PERM_* bitmasks).
441      *
442      * If both conditions are met, 0 is returned. Otherwise, -errno is returned
443      * and errp is set to an error describing the conflict.
444      */
445     int (*bdrv_check_perm)(BlockDriverState *bs, uint64_t perm,
446                            uint64_t shared, Error **errp);
447 
448     /**
449      * Called to inform the driver that the set of cumulative set of used
450      * permissions for @bs has changed to @perm, and the set of sharable
451      * permission to @shared. The driver can use this to propagate changes to
452      * its children (i.e. request permissions only if a parent actually needs
453      * them).
454      *
455      * This function is only invoked after bdrv_check_perm(), so block drivers
456      * may rely on preparations made in their .bdrv_check_perm implementation.
457      */
458     void (*bdrv_set_perm)(BlockDriverState *bs, uint64_t perm, uint64_t shared);
459 
460     /*
461      * Called to inform the driver that after a previous bdrv_check_perm()
462      * call, the permission update is not performed and any preparations made
463      * for it (e.g. taken file locks) need to be undone.
464      *
465      * This function can be called even for nodes that never saw a
466      * bdrv_check_perm() call. It is a no-op then.
467      */
468     void (*bdrv_abort_perm_update)(BlockDriverState *bs);
469 
470     /**
471      * Returns in @nperm and @nshared the permissions that the driver for @bs
472      * needs on its child @c, based on the cumulative permissions requested by
473      * the parents in @parent_perm and @parent_shared.
474      *
475      * If @c is NULL, return the permissions for attaching a new child for the
476      * given @role.
477      *
478      * If @reopen_queue is non-NULL, don't return the currently needed
479      * permissions, but those that will be needed after applying the
480      * @reopen_queue.
481      */
482      void (*bdrv_child_perm)(BlockDriverState *bs, BdrvChild *c,
483                              const BdrvChildRole *role,
484                              BlockReopenQueue *reopen_queue,
485                              uint64_t parent_perm, uint64_t parent_shared,
486                              uint64_t *nperm, uint64_t *nshared);
487 
488     /**
489      * Bitmaps should be marked as 'IN_USE' in the image on reopening image
490      * as rw. This handler should realize it. It also should unset readonly
491      * field of BlockDirtyBitmap's in case of success.
492      */
493     int (*bdrv_reopen_bitmaps_rw)(BlockDriverState *bs, Error **errp);
494     bool (*bdrv_can_store_new_dirty_bitmap)(BlockDriverState *bs,
495                                             const char *name,
496                                             uint32_t granularity,
497                                             Error **errp);
498     void (*bdrv_remove_persistent_dirty_bitmap)(BlockDriverState *bs,
499                                                 const char *name,
500                                                 Error **errp);
501 
502     /**
503      * Register/unregister a buffer for I/O. For example, when the driver is
504      * interested to know the memory areas that will later be used in iovs, so
505      * that it can do IOMMU mapping with VFIO etc., in order to get better
506      * performance. In the case of VFIO drivers, this callback is used to do
507      * DMA mapping for hot buffers.
508      */
509     void (*bdrv_register_buf)(BlockDriverState *bs, void *host, size_t size);
510     void (*bdrv_unregister_buf)(BlockDriverState *bs, void *host);
511     QLIST_ENTRY(BlockDriver) list;
512 };
513 
514 typedef struct BlockLimits {
515     /* Alignment requirement, in bytes, for offset/length of I/O
516      * requests. Must be a power of 2 less than INT_MAX; defaults to
517      * 1 for drivers with modern byte interfaces, and to 512
518      * otherwise. */
519     uint32_t request_alignment;
520 
521     /* Maximum number of bytes that can be discarded at once (since it
522      * is signed, it must be < 2G, if set). Must be multiple of
523      * pdiscard_alignment, but need not be power of 2. May be 0 if no
524      * inherent 32-bit limit */
525     int32_t max_pdiscard;
526 
527     /* Optimal alignment for discard requests in bytes. A power of 2
528      * is best but not mandatory.  Must be a multiple of
529      * bl.request_alignment, and must be less than max_pdiscard if
530      * that is set. May be 0 if bl.request_alignment is good enough */
531     uint32_t pdiscard_alignment;
532 
533     /* Maximum number of bytes that can zeroized at once (since it is
534      * signed, it must be < 2G, if set). Must be multiple of
535      * pwrite_zeroes_alignment. May be 0 if no inherent 32-bit limit */
536     int32_t max_pwrite_zeroes;
537 
538     /* Optimal alignment for write zeroes requests in bytes. A power
539      * of 2 is best but not mandatory.  Must be a multiple of
540      * bl.request_alignment, and must be less than max_pwrite_zeroes
541      * if that is set. May be 0 if bl.request_alignment is good
542      * enough */
543     uint32_t pwrite_zeroes_alignment;
544 
545     /* Optimal transfer length in bytes.  A power of 2 is best but not
546      * mandatory.  Must be a multiple of bl.request_alignment, or 0 if
547      * no preferred size */
548     uint32_t opt_transfer;
549 
550     /* Maximal transfer length in bytes.  Need not be power of 2, but
551      * must be multiple of opt_transfer and bl.request_alignment, or 0
552      * for no 32-bit limit.  For now, anything larger than INT_MAX is
553      * clamped down. */
554     uint32_t max_transfer;
555 
556     /* memory alignment, in bytes so that no bounce buffer is needed */
557     size_t min_mem_alignment;
558 
559     /* memory alignment, in bytes, for bounce buffer */
560     size_t opt_mem_alignment;
561 
562     /* maximum number of iovec elements */
563     int max_iov;
564 } BlockLimits;
565 
566 typedef struct BdrvOpBlocker BdrvOpBlocker;
567 
568 typedef struct BdrvAioNotifier {
569     void (*attached_aio_context)(AioContext *new_context, void *opaque);
570     void (*detach_aio_context)(void *opaque);
571 
572     void *opaque;
573     bool deleted;
574 
575     QLIST_ENTRY(BdrvAioNotifier) list;
576 } BdrvAioNotifier;
577 
578 struct BdrvChildRole {
579     /* If true, bdrv_replace_node() doesn't change the node this BdrvChild
580      * points to. */
581     bool stay_at_node;
582 
583     /* If true, the parent is a BlockDriverState and bdrv_next_all_states()
584      * will return it. This information is used for drain_all, where every node
585      * will be drained separately, so the drain only needs to be propagated to
586      * non-BDS parents. */
587     bool parent_is_bds;
588 
589     void (*inherit_options)(int *child_flags, QDict *child_options,
590                             int parent_flags, QDict *parent_options);
591 
592     void (*change_media)(BdrvChild *child, bool load);
593     void (*resize)(BdrvChild *child);
594 
595     /* Returns a name that is supposedly more useful for human users than the
596      * node name for identifying the node in question (in particular, a BB
597      * name), or NULL if the parent can't provide a better name. */
598     const char *(*get_name)(BdrvChild *child);
599 
600     /* Returns a malloced string that describes the parent of the child for a
601      * human reader. This could be a node-name, BlockBackend name, qdev ID or
602      * QOM path of the device owning the BlockBackend, job type and ID etc. The
603      * caller is responsible for freeing the memory. */
604     char *(*get_parent_desc)(BdrvChild *child);
605 
606     /*
607      * If this pair of functions is implemented, the parent doesn't issue new
608      * requests after returning from .drained_begin() until .drained_end() is
609      * called.
610      *
611      * These functions must not change the graph (and therefore also must not
612      * call aio_poll(), which could change the graph indirectly).
613      *
614      * Note that this can be nested. If drained_begin() was called twice, new
615      * I/O is allowed only after drained_end() was called twice, too.
616      */
617     void (*drained_begin)(BdrvChild *child);
618     void (*drained_end)(BdrvChild *child);
619 
620     /*
621      * Returns whether the parent has pending requests for the child. This
622      * callback is polled after .drained_begin() has been called until all
623      * activity on the child has stopped.
624      */
625     bool (*drained_poll)(BdrvChild *child);
626 
627     /* Notifies the parent that the child has been activated/inactivated (e.g.
628      * when migration is completing) and it can start/stop requesting
629      * permissions and doing I/O on it. */
630     void (*activate)(BdrvChild *child, Error **errp);
631     int (*inactivate)(BdrvChild *child);
632 
633     void (*attach)(BdrvChild *child);
634     void (*detach)(BdrvChild *child);
635 
636     /* Notifies the parent that the filename of its child has changed (e.g.
637      * because the direct child was removed from the backing chain), so that it
638      * can update its reference. */
639     int (*update_filename)(BdrvChild *child, BlockDriverState *new_base,
640                            const char *filename, Error **errp);
641 };
642 
643 extern const BdrvChildRole child_file;
644 extern const BdrvChildRole child_format;
645 extern const BdrvChildRole child_backing;
646 
647 struct BdrvChild {
648     BlockDriverState *bs;
649     char *name;
650     const BdrvChildRole *role;
651     void *opaque;
652 
653     /**
654      * Granted permissions for operating on this BdrvChild (BLK_PERM_* bitmask)
655      */
656     uint64_t perm;
657 
658     /**
659      * Permissions that can still be granted to other users of @bs while this
660      * BdrvChild is still attached to it. (BLK_PERM_* bitmask)
661      */
662     uint64_t shared_perm;
663 
664     QLIST_ENTRY(BdrvChild) next;
665     QLIST_ENTRY(BdrvChild) next_parent;
666 };
667 
668 /*
669  * Note: the function bdrv_append() copies and swaps contents of
670  * BlockDriverStates, so if you add new fields to this struct, please
671  * inspect bdrv_append() to determine if the new fields need to be
672  * copied as well.
673  */
674 struct BlockDriverState {
675     /* Protected by big QEMU lock or read-only after opening.  No special
676      * locking needed during I/O...
677      */
678     int open_flags; /* flags used to open the file, re-used for re-open */
679     bool read_only; /* if true, the media is read only */
680     bool encrypted; /* if true, the media is encrypted */
681     bool sg;        /* if true, the device is a /dev/sg* */
682     bool probed;    /* if true, format was probed rather than specified */
683     bool force_share; /* if true, always allow all shared permissions */
684     bool implicit;  /* if true, this filter node was automatically inserted */
685 
686     BlockDriver *drv; /* NULL means no media */
687     void *opaque;
688 
689     AioContext *aio_context; /* event loop used for fd handlers, timers, etc */
690     /* long-running tasks intended to always use the same AioContext as this
691      * BDS may register themselves in this list to be notified of changes
692      * regarding this BDS's context */
693     QLIST_HEAD(, BdrvAioNotifier) aio_notifiers;
694     bool walking_aio_notifiers; /* to make removal during iteration safe */
695 
696     char filename[PATH_MAX];
697     char backing_file[PATH_MAX]; /* if non zero, the image is a diff of
698                                     this file image */
699     char backing_format[16]; /* if non-zero and backing_file exists */
700 
701     QDict *full_open_options;
702     char exact_filename[PATH_MAX];
703 
704     BdrvChild *backing;
705     BdrvChild *file;
706 
707     /* I/O Limits */
708     BlockLimits bl;
709 
710     /* Flags honored during pwrite (so far: BDRV_REQ_FUA,
711      * BDRV_REQ_WRITE_UNCHANGED).
712      * If a driver does not support BDRV_REQ_WRITE_UNCHANGED, those
713      * writes will be issued as normal writes without the flag set.
714      * This is important to note for drivers that do not explicitly
715      * request a WRITE permission for their children and instead take
716      * the same permissions as their parent did (this is commonly what
717      * block filters do).  Such drivers have to be aware that the
718      * parent may have taken a WRITE_UNCHANGED permission only and is
719      * issuing such requests.  Drivers either must make sure that
720      * these requests do not result in plain WRITE accesses (usually
721      * by supporting BDRV_REQ_WRITE_UNCHANGED, and then forwarding
722      * every incoming write request as-is, including potentially that
723      * flag), or they have to explicitly take the WRITE permission for
724      * their children. */
725     unsigned int supported_write_flags;
726     /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA,
727      * BDRV_REQ_MAY_UNMAP, BDRV_REQ_WRITE_UNCHANGED) */
728     unsigned int supported_zero_flags;
729 
730     /* the following member gives a name to every node on the bs graph. */
731     char node_name[32];
732     /* element of the list of named nodes building the graph */
733     QTAILQ_ENTRY(BlockDriverState) node_list;
734     /* element of the list of all BlockDriverStates (all_bdrv_states) */
735     QTAILQ_ENTRY(BlockDriverState) bs_list;
736     /* element of the list of monitor-owned BDS */
737     QTAILQ_ENTRY(BlockDriverState) monitor_list;
738     int refcnt;
739 
740     /* operation blockers */
741     QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX];
742 
743     /* long-running background operation */
744     BlockJob *job;
745 
746     /* The node that this node inherited default options from (and a reopen on
747      * which can affect this node by changing these defaults). This is always a
748      * parent node of this node. */
749     BlockDriverState *inherits_from;
750     QLIST_HEAD(, BdrvChild) children;
751     QLIST_HEAD(, BdrvChild) parents;
752 
753     QDict *options;
754     QDict *explicit_options;
755     BlockdevDetectZeroesOptions detect_zeroes;
756 
757     /* The error object in use for blocking operations on backing_hd */
758     Error *backing_blocker;
759 
760     /* Protected by AioContext lock */
761 
762     /* If we are reading a disk image, give its size in sectors.
763      * Generally read-only; it is written to by load_snapshot and
764      * save_snaphost, but the block layer is quiescent during those.
765      */
766     int64_t total_sectors;
767 
768     /* Callback before write request is processed */
769     NotifierWithReturnList before_write_notifiers;
770 
771     /* threshold limit for writes, in bytes. "High water mark". */
772     uint64_t write_threshold_offset;
773     NotifierWithReturn write_threshold_notifier;
774 
775     /* Writing to the list requires the BQL _and_ the dirty_bitmap_mutex.
776      * Reading from the list can be done with either the BQL or the
777      * dirty_bitmap_mutex.  Modifying a bitmap only requires
778      * dirty_bitmap_mutex.  */
779     QemuMutex dirty_bitmap_mutex;
780     QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps;
781 
782     /* Offset after the highest byte written to */
783     Stat64 wr_highest_offset;
784 
785     /* If true, copy read backing sectors into image.  Can be >1 if more
786      * than one client has requested copy-on-read.  Accessed with atomic
787      * ops.
788      */
789     int copy_on_read;
790 
791     /* number of in-flight requests; overall and serialising.
792      * Accessed with atomic ops.
793      */
794     unsigned int in_flight;
795     unsigned int serialising_in_flight;
796 
797     /* counter for nested bdrv_io_plug.
798      * Accessed with atomic ops.
799     */
800     unsigned io_plugged;
801 
802     /* do we need to tell the quest if we have a volatile write cache? */
803     int enable_write_cache;
804 
805     /* Accessed with atomic ops.  */
806     int quiesce_counter;
807     int recursive_quiesce_counter;
808 
809     unsigned int write_gen;               /* Current data generation */
810 
811     /* Protected by reqs_lock.  */
812     CoMutex reqs_lock;
813     QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
814     CoQueue flush_queue;                  /* Serializing flush queue */
815     bool active_flush_req;                /* Flush request in flight? */
816 
817     /* Only read/written by whoever has set active_flush_req to true.  */
818     unsigned int flushed_gen;             /* Flushed write generation */
819 };
820 
821 struct BlockBackendRootState {
822     int open_flags;
823     bool read_only;
824     BlockdevDetectZeroesOptions detect_zeroes;
825 };
826 
827 typedef enum BlockMirrorBackingMode {
828     /* Reuse the existing backing chain from the source for the target.
829      * - sync=full: Set backing BDS to NULL.
830      * - sync=top:  Use source's backing BDS.
831      * - sync=none: Use source as the backing BDS. */
832     MIRROR_SOURCE_BACKING_CHAIN,
833 
834     /* Open the target's backing chain completely anew */
835     MIRROR_OPEN_BACKING_CHAIN,
836 
837     /* Do not change the target's backing BDS after job completion */
838     MIRROR_LEAVE_BACKING_CHAIN,
839 } BlockMirrorBackingMode;
840 
841 static inline BlockDriverState *backing_bs(BlockDriverState *bs)
842 {
843     return bs->backing ? bs->backing->bs : NULL;
844 }
845 
846 
847 /* Essential block drivers which must always be statically linked into qemu, and
848  * which therefore can be accessed without using bdrv_find_format() */
849 extern BlockDriver bdrv_file;
850 extern BlockDriver bdrv_raw;
851 extern BlockDriver bdrv_qcow2;
852 
853 int coroutine_fn bdrv_co_preadv(BdrvChild *child,
854     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
855     BdrvRequestFlags flags);
856 int coroutine_fn bdrv_co_pwritev(BdrvChild *child,
857     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
858     BdrvRequestFlags flags);
859 
860 extern unsigned int bdrv_drain_all_count;
861 void bdrv_apply_subtree_drain(BdrvChild *child, BlockDriverState *new_parent);
862 void bdrv_unapply_subtree_drain(BdrvChild *child, BlockDriverState *old_parent);
863 
864 int get_tmp_filename(char *filename, int size);
865 BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size,
866                             const char *filename);
867 
868 void bdrv_parse_filename_strip_prefix(const char *filename, const char *prefix,
869                                       QDict *options);
870 
871 
872 /**
873  * bdrv_add_before_write_notifier:
874  *
875  * Register a callback that is invoked before write requests are processed but
876  * after any throttling or waiting for overlapping requests.
877  */
878 void bdrv_add_before_write_notifier(BlockDriverState *bs,
879                                     NotifierWithReturn *notifier);
880 
881 /**
882  * bdrv_detach_aio_context:
883  *
884  * May be called from .bdrv_detach_aio_context() to detach children from the
885  * current #AioContext.  This is only needed by block drivers that manage their
886  * own children.  Both ->file and ->backing are automatically handled and
887  * block drivers should not call this function on them explicitly.
888  */
889 void bdrv_detach_aio_context(BlockDriverState *bs);
890 
891 /**
892  * bdrv_attach_aio_context:
893  *
894  * May be called from .bdrv_attach_aio_context() to attach children to the new
895  * #AioContext.  This is only needed by block drivers that manage their own
896  * children.  Both ->file and ->backing are automatically handled and block
897  * drivers should not call this function on them explicitly.
898  */
899 void bdrv_attach_aio_context(BlockDriverState *bs,
900                              AioContext *new_context);
901 
902 /**
903  * bdrv_add_aio_context_notifier:
904  *
905  * If a long-running job intends to be always run in the same AioContext as a
906  * certain BDS, it may use this function to be notified of changes regarding the
907  * association of the BDS to an AioContext.
908  *
909  * attached_aio_context() is called after the target BDS has been attached to a
910  * new AioContext; detach_aio_context() is called before the target BDS is being
911  * detached from its old AioContext.
912  */
913 void bdrv_add_aio_context_notifier(BlockDriverState *bs,
914         void (*attached_aio_context)(AioContext *new_context, void *opaque),
915         void (*detach_aio_context)(void *opaque), void *opaque);
916 
917 /**
918  * bdrv_remove_aio_context_notifier:
919  *
920  * Unsubscribe of change notifications regarding the BDS's AioContext. The
921  * parameters given here have to be the same as those given to
922  * bdrv_add_aio_context_notifier().
923  */
924 void bdrv_remove_aio_context_notifier(BlockDriverState *bs,
925                                       void (*aio_context_attached)(AioContext *,
926                                                                    void *),
927                                       void (*aio_context_detached)(void *),
928                                       void *opaque);
929 
930 /**
931  * bdrv_wakeup:
932  * @bs: The BlockDriverState for which an I/O operation has been completed.
933  *
934  * Wake up the main thread if it is waiting on BDRV_POLL_WHILE.  During
935  * synchronous I/O on a BlockDriverState that is attached to another
936  * I/O thread, the main thread lets the I/O thread's event loop run,
937  * waiting for the I/O operation to complete.  A bdrv_wakeup will wake
938  * up the main thread if necessary.
939  *
940  * Manual calls to bdrv_wakeup are rarely necessary, because
941  * bdrv_dec_in_flight already calls it.
942  */
943 void bdrv_wakeup(BlockDriverState *bs);
944 
945 #ifdef _WIN32
946 int is_windows_drive(const char *filename);
947 #endif
948 
949 /**
950  * stream_start:
951  * @job_id: The id of the newly-created job, or %NULL to use the
952  * device name of @bs.
953  * @bs: Block device to operate on.
954  * @base: Block device that will become the new base, or %NULL to
955  * flatten the whole backing file chain onto @bs.
956  * @backing_file_str: The file name that will be written to @bs as the
957  * the new backing file if the job completes. Ignored if @base is %NULL.
958  * @creation_flags: Flags that control the behavior of the Job lifetime.
959  *                  See @BlockJobCreateFlags
960  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
961  * @on_error: The action to take upon error.
962  * @errp: Error object.
963  *
964  * Start a streaming operation on @bs.  Clusters that are unallocated
965  * in @bs, but allocated in any image between @base and @bs (both
966  * exclusive) will be written to @bs.  At the end of a successful
967  * streaming job, the backing file of @bs will be changed to
968  * @backing_file_str in the written image and to @base in the live
969  * BlockDriverState.
970  */
971 void stream_start(const char *job_id, BlockDriverState *bs,
972                   BlockDriverState *base, const char *backing_file_str,
973                   int creation_flags, int64_t speed,
974                   BlockdevOnError on_error, Error **errp);
975 
976 /**
977  * commit_start:
978  * @job_id: The id of the newly-created job, or %NULL to use the
979  * device name of @bs.
980  * @bs: Active block device.
981  * @top: Top block device to be committed.
982  * @base: Block device that will be written into, and become the new top.
983  * @creation_flags: Flags that control the behavior of the Job lifetime.
984  *                  See @BlockJobCreateFlags
985  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
986  * @on_error: The action to take upon error.
987  * @backing_file_str: String to use as the backing file in @top's overlay
988  * @filter_node_name: The node name that should be assigned to the filter
989  * driver that the commit job inserts into the graph above @top. NULL means
990  * that a node name should be autogenerated.
991  * @errp: Error object.
992  *
993  */
994 void commit_start(const char *job_id, BlockDriverState *bs,
995                   BlockDriverState *base, BlockDriverState *top,
996                   int creation_flags, int64_t speed,
997                   BlockdevOnError on_error, const char *backing_file_str,
998                   const char *filter_node_name, Error **errp);
999 /**
1000  * commit_active_start:
1001  * @job_id: The id of the newly-created job, or %NULL to use the
1002  * device name of @bs.
1003  * @bs: Active block device to be committed.
1004  * @base: Block device that will be written into, and become the new top.
1005  * @creation_flags: Flags that control the behavior of the Job lifetime.
1006  *                  See @BlockJobCreateFlags
1007  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1008  * @on_error: The action to take upon error.
1009  * @filter_node_name: The node name that should be assigned to the filter
1010  * driver that the commit job inserts into the graph above @bs. NULL means that
1011  * a node name should be autogenerated.
1012  * @cb: Completion function for the job.
1013  * @opaque: Opaque pointer value passed to @cb.
1014  * @auto_complete: Auto complete the job.
1015  * @errp: Error object.
1016  *
1017  */
1018 void commit_active_start(const char *job_id, BlockDriverState *bs,
1019                          BlockDriverState *base, int creation_flags,
1020                          int64_t speed, BlockdevOnError on_error,
1021                          const char *filter_node_name,
1022                          BlockCompletionFunc *cb, void *opaque,
1023                          bool auto_complete, Error **errp);
1024 /*
1025  * mirror_start:
1026  * @job_id: The id of the newly-created job, or %NULL to use the
1027  * device name of @bs.
1028  * @bs: Block device to operate on.
1029  * @target: Block device to write to.
1030  * @replaces: Block graph node name to replace once the mirror is done. Can
1031  *            only be used when full mirroring is selected.
1032  * @creation_flags: Flags that control the behavior of the Job lifetime.
1033  *                  See @BlockJobCreateFlags
1034  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1035  * @granularity: The chosen granularity for the dirty bitmap.
1036  * @buf_size: The amount of data that can be in flight at one time.
1037  * @mode: Whether to collapse all images in the chain to the target.
1038  * @backing_mode: How to establish the target's backing chain after completion.
1039  * @on_source_error: The action to take upon error reading from the source.
1040  * @on_target_error: The action to take upon error writing to the target.
1041  * @unmap: Whether to unmap target where source sectors only contain zeroes.
1042  * @filter_node_name: The node name that should be assigned to the filter
1043  * driver that the mirror job inserts into the graph above @bs. NULL means that
1044  * a node name should be autogenerated.
1045  * @copy_mode: When to trigger writes to the target.
1046  * @errp: Error object.
1047  *
1048  * Start a mirroring operation on @bs.  Clusters that are allocated
1049  * in @bs will be written to @target until the job is cancelled or
1050  * manually completed.  At the end of a successful mirroring job,
1051  * @bs will be switched to read from @target.
1052  */
1053 void mirror_start(const char *job_id, BlockDriverState *bs,
1054                   BlockDriverState *target, const char *replaces,
1055                   int creation_flags, int64_t speed,
1056                   uint32_t granularity, int64_t buf_size,
1057                   MirrorSyncMode mode, BlockMirrorBackingMode backing_mode,
1058                   BlockdevOnError on_source_error,
1059                   BlockdevOnError on_target_error,
1060                   bool unmap, const char *filter_node_name,
1061                   MirrorCopyMode copy_mode, Error **errp);
1062 
1063 /*
1064  * backup_job_create:
1065  * @job_id: The id of the newly-created job, or %NULL to use the
1066  * device name of @bs.
1067  * @bs: Block device to operate on.
1068  * @target: Block device to write to.
1069  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1070  * @sync_mode: What parts of the disk image should be copied to the destination.
1071  * @sync_bitmap: The dirty bitmap if sync_mode is MIRROR_SYNC_MODE_INCREMENTAL.
1072  * @on_source_error: The action to take upon error reading from the source.
1073  * @on_target_error: The action to take upon error writing to the target.
1074  * @creation_flags: Flags that control the behavior of the Job lifetime.
1075  *                  See @BlockJobCreateFlags
1076  * @cb: Completion function for the job.
1077  * @opaque: Opaque pointer value passed to @cb.
1078  * @txn: Transaction that this job is part of (may be NULL).
1079  *
1080  * Create a backup operation on @bs.  Clusters in @bs are written to @target
1081  * until the job is cancelled or manually completed.
1082  */
1083 BlockJob *backup_job_create(const char *job_id, BlockDriverState *bs,
1084                             BlockDriverState *target, int64_t speed,
1085                             MirrorSyncMode sync_mode,
1086                             BdrvDirtyBitmap *sync_bitmap,
1087                             bool compress,
1088                             BlockdevOnError on_source_error,
1089                             BlockdevOnError on_target_error,
1090                             int creation_flags,
1091                             BlockCompletionFunc *cb, void *opaque,
1092                             JobTxn *txn, Error **errp);
1093 
1094 void hmp_drive_add_node(Monitor *mon, const char *optstr);
1095 
1096 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
1097                                   const char *child_name,
1098                                   const BdrvChildRole *child_role,
1099                                   uint64_t perm, uint64_t shared_perm,
1100                                   void *opaque, Error **errp);
1101 void bdrv_root_unref_child(BdrvChild *child);
1102 
1103 int bdrv_child_try_set_perm(BdrvChild *c, uint64_t perm, uint64_t shared,
1104                             Error **errp);
1105 
1106 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1107  * block filters: Forward CONSISTENT_READ, WRITE, WRITE_UNCHANGED and RESIZE to
1108  * all children */
1109 void bdrv_filter_default_perms(BlockDriverState *bs, BdrvChild *c,
1110                                const BdrvChildRole *role,
1111                                BlockReopenQueue *reopen_queue,
1112                                uint64_t perm, uint64_t shared,
1113                                uint64_t *nperm, uint64_t *nshared);
1114 
1115 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1116  * (non-raw) image formats: Like above for bs->backing, but for bs->file it
1117  * requires WRITE | RESIZE for read-write images, always requires
1118  * CONSISTENT_READ and doesn't share WRITE. */
1119 void bdrv_format_default_perms(BlockDriverState *bs, BdrvChild *c,
1120                                const BdrvChildRole *role,
1121                                BlockReopenQueue *reopen_queue,
1122                                uint64_t perm, uint64_t shared,
1123                                uint64_t *nperm, uint64_t *nshared);
1124 
1125 /*
1126  * Default implementation for drivers to pass bdrv_co_block_status() to
1127  * their file.
1128  */
1129 int coroutine_fn bdrv_co_block_status_from_file(BlockDriverState *bs,
1130                                                 bool want_zero,
1131                                                 int64_t offset,
1132                                                 int64_t bytes,
1133                                                 int64_t *pnum,
1134                                                 int64_t *map,
1135                                                 BlockDriverState **file);
1136 /*
1137  * Default implementation for drivers to pass bdrv_co_block_status() to
1138  * their backing file.
1139  */
1140 int coroutine_fn bdrv_co_block_status_from_backing(BlockDriverState *bs,
1141                                                    bool want_zero,
1142                                                    int64_t offset,
1143                                                    int64_t bytes,
1144                                                    int64_t *pnum,
1145                                                    int64_t *map,
1146                                                    BlockDriverState **file);
1147 const char *bdrv_get_parent_name(const BlockDriverState *bs);
1148 void blk_dev_change_media_cb(BlockBackend *blk, bool load, Error **errp);
1149 bool blk_dev_has_removable_media(BlockBackend *blk);
1150 bool blk_dev_has_tray(BlockBackend *blk);
1151 void blk_dev_eject_request(BlockBackend *blk, bool force);
1152 bool blk_dev_is_tray_open(BlockBackend *blk);
1153 bool blk_dev_is_medium_locked(BlockBackend *blk);
1154 
1155 void bdrv_set_dirty(BlockDriverState *bs, int64_t offset, int64_t bytes);
1156 
1157 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap **out);
1158 void bdrv_restore_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap *backup);
1159 
1160 void bdrv_inc_in_flight(BlockDriverState *bs);
1161 void bdrv_dec_in_flight(BlockDriverState *bs);
1162 
1163 void blockdev_close_all_bdrv_states(void);
1164 
1165 int coroutine_fn bdrv_co_copy_range_from(BdrvChild *src, uint64_t src_offset,
1166                                          BdrvChild *dst, uint64_t dst_offset,
1167                                          uint64_t bytes,
1168                                          BdrvRequestFlags read_flags,
1169                                          BdrvRequestFlags write_flags);
1170 int coroutine_fn bdrv_co_copy_range_to(BdrvChild *src, uint64_t src_offset,
1171                                        BdrvChild *dst, uint64_t dst_offset,
1172                                        uint64_t bytes,
1173                                        BdrvRequestFlags read_flags,
1174                                        BdrvRequestFlags write_flags);
1175 
1176 int refresh_total_sectors(BlockDriverState *bs, int64_t hint);
1177 
1178 #endif /* BLOCK_INT_H */
1179