xref: /openbmc/qemu/include/block/block_int.h (revision 09a274d8)
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                                                  Error **errp);
324 
325     int coroutine_fn (*bdrv_save_vmstate)(BlockDriverState *bs,
326                                           QEMUIOVector *qiov,
327                                           int64_t pos);
328     int coroutine_fn (*bdrv_load_vmstate)(BlockDriverState *bs,
329                                           QEMUIOVector *qiov,
330                                           int64_t pos);
331 
332     int (*bdrv_change_backing_file)(BlockDriverState *bs,
333         const char *backing_file, const char *backing_fmt);
334 
335     /* removable device specific */
336     bool (*bdrv_is_inserted)(BlockDriverState *bs);
337     void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
338     void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
339 
340     /* to control generic scsi devices */
341     BlockAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
342         unsigned long int req, void *buf,
343         BlockCompletionFunc *cb, void *opaque);
344     int coroutine_fn (*bdrv_co_ioctl)(BlockDriverState *bs,
345                                       unsigned long int req, void *buf);
346 
347     /* List of options for creating images, terminated by name == NULL */
348     QemuOptsList *create_opts;
349 
350     /*
351      * Returns 0 for completed check, -errno for internal errors.
352      * The check results are stored in result.
353      */
354     int coroutine_fn (*bdrv_co_check)(BlockDriverState *bs,
355                                       BdrvCheckResult *result,
356                                       BdrvCheckMode fix);
357 
358     int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts,
359                               BlockDriverAmendStatusCB *status_cb,
360                               void *cb_opaque,
361                               Error **errp);
362 
363     void (*bdrv_debug_event)(BlockDriverState *bs, BlkdebugEvent event);
364 
365     /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
366     int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
367         const char *tag);
368     int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs,
369         const char *tag);
370     int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
371     bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
372 
373     void (*bdrv_refresh_limits)(BlockDriverState *bs, Error **errp);
374 
375     /*
376      * Returns 1 if newly created images are guaranteed to contain only
377      * zeros, 0 otherwise.
378      */
379     int (*bdrv_has_zero_init)(BlockDriverState *bs);
380 
381     /* Remove fd handlers, timers, and other event loop callbacks so the event
382      * loop is no longer in use.  Called with no in-flight requests and in
383      * depth-first traversal order with parents before child nodes.
384      */
385     void (*bdrv_detach_aio_context)(BlockDriverState *bs);
386 
387     /* Add fd handlers, timers, and other event loop callbacks so I/O requests
388      * can be processed again.  Called with no in-flight requests and in
389      * depth-first traversal order with child nodes before parent nodes.
390      */
391     void (*bdrv_attach_aio_context)(BlockDriverState *bs,
392                                     AioContext *new_context);
393 
394     /* io queue for linux-aio */
395     void (*bdrv_io_plug)(BlockDriverState *bs);
396     void (*bdrv_io_unplug)(BlockDriverState *bs);
397 
398     /**
399      * Try to get @bs's logical and physical block size.
400      * On success, store them in @bsz and return zero.
401      * On failure, return negative errno.
402      */
403     int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz);
404     /**
405      * Try to get @bs's geometry (cyls, heads, sectors)
406      * On success, store them in @geo and return 0.
407      * On failure return -errno.
408      * Only drivers that want to override guest geometry implement this
409      * callback; see hd_geometry_guess().
410      */
411     int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo);
412 
413     /**
414      * bdrv_co_drain_begin is called if implemented in the beginning of a
415      * drain operation to drain and stop any internal sources of requests in
416      * the driver.
417      * bdrv_co_drain_end is called if implemented at the end of the drain.
418      *
419      * They should be used by the driver to e.g. manage scheduled I/O
420      * requests, or toggle an internal state. After the end of the drain new
421      * requests will continue normally.
422      */
423     void coroutine_fn (*bdrv_co_drain_begin)(BlockDriverState *bs);
424     void coroutine_fn (*bdrv_co_drain_end)(BlockDriverState *bs);
425 
426     void (*bdrv_add_child)(BlockDriverState *parent, BlockDriverState *child,
427                            Error **errp);
428     void (*bdrv_del_child)(BlockDriverState *parent, BdrvChild *child,
429                            Error **errp);
430 
431     /**
432      * Informs the block driver that a permission change is intended. The
433      * driver checks whether the change is permissible and may take other
434      * preparations for the change (e.g. get file system locks). This operation
435      * is always followed either by a call to either .bdrv_set_perm or
436      * .bdrv_abort_perm_update.
437      *
438      * Checks whether the requested set of cumulative permissions in @perm
439      * can be granted for accessing @bs and whether no other users are using
440      * permissions other than those given in @shared (both arguments take
441      * BLK_PERM_* bitmasks).
442      *
443      * If both conditions are met, 0 is returned. Otherwise, -errno is returned
444      * and errp is set to an error describing the conflict.
445      */
446     int (*bdrv_check_perm)(BlockDriverState *bs, uint64_t perm,
447                            uint64_t shared, Error **errp);
448 
449     /**
450      * Called to inform the driver that the set of cumulative set of used
451      * permissions for @bs has changed to @perm, and the set of sharable
452      * permission to @shared. The driver can use this to propagate changes to
453      * its children (i.e. request permissions only if a parent actually needs
454      * them).
455      *
456      * This function is only invoked after bdrv_check_perm(), so block drivers
457      * may rely on preparations made in their .bdrv_check_perm implementation.
458      */
459     void (*bdrv_set_perm)(BlockDriverState *bs, uint64_t perm, uint64_t shared);
460 
461     /*
462      * Called to inform the driver that after a previous bdrv_check_perm()
463      * call, the permission update is not performed and any preparations made
464      * for it (e.g. taken file locks) need to be undone.
465      *
466      * This function can be called even for nodes that never saw a
467      * bdrv_check_perm() call. It is a no-op then.
468      */
469     void (*bdrv_abort_perm_update)(BlockDriverState *bs);
470 
471     /**
472      * Returns in @nperm and @nshared the permissions that the driver for @bs
473      * needs on its child @c, based on the cumulative permissions requested by
474      * the parents in @parent_perm and @parent_shared.
475      *
476      * If @c is NULL, return the permissions for attaching a new child for the
477      * given @role.
478      *
479      * If @reopen_queue is non-NULL, don't return the currently needed
480      * permissions, but those that will be needed after applying the
481      * @reopen_queue.
482      */
483      void (*bdrv_child_perm)(BlockDriverState *bs, BdrvChild *c,
484                              const BdrvChildRole *role,
485                              BlockReopenQueue *reopen_queue,
486                              uint64_t parent_perm, uint64_t parent_shared,
487                              uint64_t *nperm, uint64_t *nshared);
488 
489     /**
490      * Bitmaps should be marked as 'IN_USE' in the image on reopening image
491      * as rw. This handler should realize it. It also should unset readonly
492      * field of BlockDirtyBitmap's in case of success.
493      */
494     int (*bdrv_reopen_bitmaps_rw)(BlockDriverState *bs, Error **errp);
495     bool (*bdrv_can_store_new_dirty_bitmap)(BlockDriverState *bs,
496                                             const char *name,
497                                             uint32_t granularity,
498                                             Error **errp);
499     void (*bdrv_remove_persistent_dirty_bitmap)(BlockDriverState *bs,
500                                                 const char *name,
501                                                 Error **errp);
502 
503     /**
504      * Register/unregister a buffer for I/O. For example, when the driver is
505      * interested to know the memory areas that will later be used in iovs, so
506      * that it can do IOMMU mapping with VFIO etc., in order to get better
507      * performance. In the case of VFIO drivers, this callback is used to do
508      * DMA mapping for hot buffers.
509      */
510     void (*bdrv_register_buf)(BlockDriverState *bs, void *host, size_t size);
511     void (*bdrv_unregister_buf)(BlockDriverState *bs, void *host);
512     QLIST_ENTRY(BlockDriver) list;
513 };
514 
515 typedef struct BlockLimits {
516     /* Alignment requirement, in bytes, for offset/length of I/O
517      * requests. Must be a power of 2 less than INT_MAX; defaults to
518      * 1 for drivers with modern byte interfaces, and to 512
519      * otherwise. */
520     uint32_t request_alignment;
521 
522     /* Maximum number of bytes that can be discarded at once (since it
523      * is signed, it must be < 2G, if set). Must be multiple of
524      * pdiscard_alignment, but need not be power of 2. May be 0 if no
525      * inherent 32-bit limit */
526     int32_t max_pdiscard;
527 
528     /* Optimal alignment for discard requests in bytes. A power of 2
529      * is best but not mandatory.  Must be a multiple of
530      * bl.request_alignment, and must be less than max_pdiscard if
531      * that is set. May be 0 if bl.request_alignment is good enough */
532     uint32_t pdiscard_alignment;
533 
534     /* Maximum number of bytes that can zeroized at once (since it is
535      * signed, it must be < 2G, if set). Must be multiple of
536      * pwrite_zeroes_alignment. May be 0 if no inherent 32-bit limit */
537     int32_t max_pwrite_zeroes;
538 
539     /* Optimal alignment for write zeroes requests in bytes. A power
540      * of 2 is best but not mandatory.  Must be a multiple of
541      * bl.request_alignment, and must be less than max_pwrite_zeroes
542      * if that is set. May be 0 if bl.request_alignment is good
543      * enough */
544     uint32_t pwrite_zeroes_alignment;
545 
546     /* Optimal transfer length in bytes.  A power of 2 is best but not
547      * mandatory.  Must be a multiple of bl.request_alignment, or 0 if
548      * no preferred size */
549     uint32_t opt_transfer;
550 
551     /* Maximal transfer length in bytes.  Need not be power of 2, but
552      * must be multiple of opt_transfer and bl.request_alignment, or 0
553      * for no 32-bit limit.  For now, anything larger than INT_MAX is
554      * clamped down. */
555     uint32_t max_transfer;
556 
557     /* memory alignment, in bytes so that no bounce buffer is needed */
558     size_t min_mem_alignment;
559 
560     /* memory alignment, in bytes, for bounce buffer */
561     size_t opt_mem_alignment;
562 
563     /* maximum number of iovec elements */
564     int max_iov;
565 } BlockLimits;
566 
567 typedef struct BdrvOpBlocker BdrvOpBlocker;
568 
569 typedef struct BdrvAioNotifier {
570     void (*attached_aio_context)(AioContext *new_context, void *opaque);
571     void (*detach_aio_context)(void *opaque);
572 
573     void *opaque;
574     bool deleted;
575 
576     QLIST_ENTRY(BdrvAioNotifier) list;
577 } BdrvAioNotifier;
578 
579 struct BdrvChildRole {
580     /* If true, bdrv_replace_node() doesn't change the node this BdrvChild
581      * points to. */
582     bool stay_at_node;
583 
584     /* If true, the parent is a BlockDriverState and bdrv_next_all_states()
585      * will return it. This information is used for drain_all, where every node
586      * will be drained separately, so the drain only needs to be propagated to
587      * non-BDS parents. */
588     bool parent_is_bds;
589 
590     void (*inherit_options)(int *child_flags, QDict *child_options,
591                             int parent_flags, QDict *parent_options);
592 
593     void (*change_media)(BdrvChild *child, bool load);
594     void (*resize)(BdrvChild *child);
595 
596     /* Returns a name that is supposedly more useful for human users than the
597      * node name for identifying the node in question (in particular, a BB
598      * name), or NULL if the parent can't provide a better name. */
599     const char *(*get_name)(BdrvChild *child);
600 
601     /* Returns a malloced string that describes the parent of the child for a
602      * human reader. This could be a node-name, BlockBackend name, qdev ID or
603      * QOM path of the device owning the BlockBackend, job type and ID etc. The
604      * caller is responsible for freeing the memory. */
605     char *(*get_parent_desc)(BdrvChild *child);
606 
607     /*
608      * If this pair of functions is implemented, the parent doesn't issue new
609      * requests after returning from .drained_begin() until .drained_end() is
610      * called.
611      *
612      * These functions must not change the graph (and therefore also must not
613      * call aio_poll(), which could change the graph indirectly).
614      *
615      * Note that this can be nested. If drained_begin() was called twice, new
616      * I/O is allowed only after drained_end() was called twice, too.
617      */
618     void (*drained_begin)(BdrvChild *child);
619     void (*drained_end)(BdrvChild *child);
620 
621     /*
622      * Returns whether the parent has pending requests for the child. This
623      * callback is polled after .drained_begin() has been called until all
624      * activity on the child has stopped.
625      */
626     bool (*drained_poll)(BdrvChild *child);
627 
628     /* Notifies the parent that the child has been activated/inactivated (e.g.
629      * when migration is completing) and it can start/stop requesting
630      * permissions and doing I/O on it. */
631     void (*activate)(BdrvChild *child, Error **errp);
632     int (*inactivate)(BdrvChild *child);
633 
634     void (*attach)(BdrvChild *child);
635     void (*detach)(BdrvChild *child);
636 
637     /* Notifies the parent that the filename of its child has changed (e.g.
638      * because the direct child was removed from the backing chain), so that it
639      * can update its reference. */
640     int (*update_filename)(BdrvChild *child, BlockDriverState *new_base,
641                            const char *filename, Error **errp);
642 };
643 
644 extern const BdrvChildRole child_file;
645 extern const BdrvChildRole child_format;
646 extern const BdrvChildRole child_backing;
647 
648 struct BdrvChild {
649     BlockDriverState *bs;
650     char *name;
651     const BdrvChildRole *role;
652     void *opaque;
653 
654     /**
655      * Granted permissions for operating on this BdrvChild (BLK_PERM_* bitmask)
656      */
657     uint64_t perm;
658 
659     /**
660      * Permissions that can still be granted to other users of @bs while this
661      * BdrvChild is still attached to it. (BLK_PERM_* bitmask)
662      */
663     uint64_t shared_perm;
664 
665     QLIST_ENTRY(BdrvChild) next;
666     QLIST_ENTRY(BdrvChild) next_parent;
667 };
668 
669 /*
670  * Note: the function bdrv_append() copies and swaps contents of
671  * BlockDriverStates, so if you add new fields to this struct, please
672  * inspect bdrv_append() to determine if the new fields need to be
673  * copied as well.
674  */
675 struct BlockDriverState {
676     /* Protected by big QEMU lock or read-only after opening.  No special
677      * locking needed during I/O...
678      */
679     int open_flags; /* flags used to open the file, re-used for re-open */
680     bool read_only; /* if true, the media is read only */
681     bool encrypted; /* if true, the media is encrypted */
682     bool sg;        /* if true, the device is a /dev/sg* */
683     bool probed;    /* if true, format was probed rather than specified */
684     bool force_share; /* if true, always allow all shared permissions */
685     bool implicit;  /* if true, this filter node was automatically inserted */
686 
687     BlockDriver *drv; /* NULL means no media */
688     void *opaque;
689 
690     AioContext *aio_context; /* event loop used for fd handlers, timers, etc */
691     /* long-running tasks intended to always use the same AioContext as this
692      * BDS may register themselves in this list to be notified of changes
693      * regarding this BDS's context */
694     QLIST_HEAD(, BdrvAioNotifier) aio_notifiers;
695     bool walking_aio_notifiers; /* to make removal during iteration safe */
696 
697     char filename[PATH_MAX];
698     char backing_file[PATH_MAX]; /* if non zero, the image is a diff of
699                                     this file image */
700     char backing_format[16]; /* if non-zero and backing_file exists */
701 
702     QDict *full_open_options;
703     char exact_filename[PATH_MAX];
704 
705     BdrvChild *backing;
706     BdrvChild *file;
707 
708     /* I/O Limits */
709     BlockLimits bl;
710 
711     /* Flags honored during pwrite (so far: BDRV_REQ_FUA,
712      * BDRV_REQ_WRITE_UNCHANGED).
713      * If a driver does not support BDRV_REQ_WRITE_UNCHANGED, those
714      * writes will be issued as normal writes without the flag set.
715      * This is important to note for drivers that do not explicitly
716      * request a WRITE permission for their children and instead take
717      * the same permissions as their parent did (this is commonly what
718      * block filters do).  Such drivers have to be aware that the
719      * parent may have taken a WRITE_UNCHANGED permission only and is
720      * issuing such requests.  Drivers either must make sure that
721      * these requests do not result in plain WRITE accesses (usually
722      * by supporting BDRV_REQ_WRITE_UNCHANGED, and then forwarding
723      * every incoming write request as-is, including potentially that
724      * flag), or they have to explicitly take the WRITE permission for
725      * their children. */
726     unsigned int supported_write_flags;
727     /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA,
728      * BDRV_REQ_MAY_UNMAP, BDRV_REQ_WRITE_UNCHANGED) */
729     unsigned int supported_zero_flags;
730 
731     /* the following member gives a name to every node on the bs graph. */
732     char node_name[32];
733     /* element of the list of named nodes building the graph */
734     QTAILQ_ENTRY(BlockDriverState) node_list;
735     /* element of the list of all BlockDriverStates (all_bdrv_states) */
736     QTAILQ_ENTRY(BlockDriverState) bs_list;
737     /* element of the list of monitor-owned BDS */
738     QTAILQ_ENTRY(BlockDriverState) monitor_list;
739     int refcnt;
740 
741     /* operation blockers */
742     QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX];
743 
744     /* long-running background operation */
745     BlockJob *job;
746 
747     /* The node that this node inherited default options from (and a reopen on
748      * which can affect this node by changing these defaults). This is always a
749      * parent node of this node. */
750     BlockDriverState *inherits_from;
751     QLIST_HEAD(, BdrvChild) children;
752     QLIST_HEAD(, BdrvChild) parents;
753 
754     QDict *options;
755     QDict *explicit_options;
756     BlockdevDetectZeroesOptions detect_zeroes;
757 
758     /* The error object in use for blocking operations on backing_hd */
759     Error *backing_blocker;
760 
761     /* Protected by AioContext lock */
762 
763     /* If we are reading a disk image, give its size in sectors.
764      * Generally read-only; it is written to by load_snapshot and
765      * save_snaphost, but the block layer is quiescent during those.
766      */
767     int64_t total_sectors;
768 
769     /* Callback before write request is processed */
770     NotifierWithReturnList before_write_notifiers;
771 
772     /* threshold limit for writes, in bytes. "High water mark". */
773     uint64_t write_threshold_offset;
774     NotifierWithReturn write_threshold_notifier;
775 
776     /* Writing to the list requires the BQL _and_ the dirty_bitmap_mutex.
777      * Reading from the list can be done with either the BQL or the
778      * dirty_bitmap_mutex.  Modifying a bitmap only requires
779      * dirty_bitmap_mutex.  */
780     QemuMutex dirty_bitmap_mutex;
781     QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps;
782 
783     /* Offset after the highest byte written to */
784     Stat64 wr_highest_offset;
785 
786     /* If true, copy read backing sectors into image.  Can be >1 if more
787      * than one client has requested copy-on-read.  Accessed with atomic
788      * ops.
789      */
790     int copy_on_read;
791 
792     /* number of in-flight requests; overall and serialising.
793      * Accessed with atomic ops.
794      */
795     unsigned int in_flight;
796     unsigned int serialising_in_flight;
797 
798     /* counter for nested bdrv_io_plug.
799      * Accessed with atomic ops.
800     */
801     unsigned io_plugged;
802 
803     /* do we need to tell the quest if we have a volatile write cache? */
804     int enable_write_cache;
805 
806     /* Accessed with atomic ops.  */
807     int quiesce_counter;
808     int recursive_quiesce_counter;
809 
810     unsigned int write_gen;               /* Current data generation */
811 
812     /* Protected by reqs_lock.  */
813     CoMutex reqs_lock;
814     QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
815     CoQueue flush_queue;                  /* Serializing flush queue */
816     bool active_flush_req;                /* Flush request in flight? */
817 
818     /* Only read/written by whoever has set active_flush_req to true.  */
819     unsigned int flushed_gen;             /* Flushed write generation */
820 };
821 
822 struct BlockBackendRootState {
823     int open_flags;
824     bool read_only;
825     BlockdevDetectZeroesOptions detect_zeroes;
826 };
827 
828 typedef enum BlockMirrorBackingMode {
829     /* Reuse the existing backing chain from the source for the target.
830      * - sync=full: Set backing BDS to NULL.
831      * - sync=top:  Use source's backing BDS.
832      * - sync=none: Use source as the backing BDS. */
833     MIRROR_SOURCE_BACKING_CHAIN,
834 
835     /* Open the target's backing chain completely anew */
836     MIRROR_OPEN_BACKING_CHAIN,
837 
838     /* Do not change the target's backing BDS after job completion */
839     MIRROR_LEAVE_BACKING_CHAIN,
840 } BlockMirrorBackingMode;
841 
842 static inline BlockDriverState *backing_bs(BlockDriverState *bs)
843 {
844     return bs->backing ? bs->backing->bs : NULL;
845 }
846 
847 
848 /* Essential block drivers which must always be statically linked into qemu, and
849  * which therefore can be accessed without using bdrv_find_format() */
850 extern BlockDriver bdrv_file;
851 extern BlockDriver bdrv_raw;
852 extern BlockDriver bdrv_qcow2;
853 
854 int coroutine_fn bdrv_co_preadv(BdrvChild *child,
855     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
856     BdrvRequestFlags flags);
857 int coroutine_fn bdrv_co_pwritev(BdrvChild *child,
858     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
859     BdrvRequestFlags flags);
860 
861 extern unsigned int bdrv_drain_all_count;
862 void bdrv_apply_subtree_drain(BdrvChild *child, BlockDriverState *new_parent);
863 void bdrv_unapply_subtree_drain(BdrvChild *child, BlockDriverState *old_parent);
864 
865 int get_tmp_filename(char *filename, int size);
866 BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size,
867                             const char *filename);
868 
869 void bdrv_parse_filename_strip_prefix(const char *filename, const char *prefix,
870                                       QDict *options);
871 
872 
873 /**
874  * bdrv_add_before_write_notifier:
875  *
876  * Register a callback that is invoked before write requests are processed but
877  * after any throttling or waiting for overlapping requests.
878  */
879 void bdrv_add_before_write_notifier(BlockDriverState *bs,
880                                     NotifierWithReturn *notifier);
881 
882 /**
883  * bdrv_detach_aio_context:
884  *
885  * May be called from .bdrv_detach_aio_context() to detach children from the
886  * current #AioContext.  This is only needed by block drivers that manage their
887  * own children.  Both ->file and ->backing are automatically handled and
888  * block drivers should not call this function on them explicitly.
889  */
890 void bdrv_detach_aio_context(BlockDriverState *bs);
891 
892 /**
893  * bdrv_attach_aio_context:
894  *
895  * May be called from .bdrv_attach_aio_context() to attach children to the new
896  * #AioContext.  This is only needed by block drivers that manage their own
897  * children.  Both ->file and ->backing are automatically handled and block
898  * drivers should not call this function on them explicitly.
899  */
900 void bdrv_attach_aio_context(BlockDriverState *bs,
901                              AioContext *new_context);
902 
903 /**
904  * bdrv_add_aio_context_notifier:
905  *
906  * If a long-running job intends to be always run in the same AioContext as a
907  * certain BDS, it may use this function to be notified of changes regarding the
908  * association of the BDS to an AioContext.
909  *
910  * attached_aio_context() is called after the target BDS has been attached to a
911  * new AioContext; detach_aio_context() is called before the target BDS is being
912  * detached from its old AioContext.
913  */
914 void bdrv_add_aio_context_notifier(BlockDriverState *bs,
915         void (*attached_aio_context)(AioContext *new_context, void *opaque),
916         void (*detach_aio_context)(void *opaque), void *opaque);
917 
918 /**
919  * bdrv_remove_aio_context_notifier:
920  *
921  * Unsubscribe of change notifications regarding the BDS's AioContext. The
922  * parameters given here have to be the same as those given to
923  * bdrv_add_aio_context_notifier().
924  */
925 void bdrv_remove_aio_context_notifier(BlockDriverState *bs,
926                                       void (*aio_context_attached)(AioContext *,
927                                                                    void *),
928                                       void (*aio_context_detached)(void *),
929                                       void *opaque);
930 
931 /**
932  * bdrv_wakeup:
933  * @bs: The BlockDriverState for which an I/O operation has been completed.
934  *
935  * Wake up the main thread if it is waiting on BDRV_POLL_WHILE.  During
936  * synchronous I/O on a BlockDriverState that is attached to another
937  * I/O thread, the main thread lets the I/O thread's event loop run,
938  * waiting for the I/O operation to complete.  A bdrv_wakeup will wake
939  * up the main thread if necessary.
940  *
941  * Manual calls to bdrv_wakeup are rarely necessary, because
942  * bdrv_dec_in_flight already calls it.
943  */
944 void bdrv_wakeup(BlockDriverState *bs);
945 
946 #ifdef _WIN32
947 int is_windows_drive(const char *filename);
948 #endif
949 
950 /**
951  * stream_start:
952  * @job_id: The id of the newly-created job, or %NULL to use the
953  * device name of @bs.
954  * @bs: Block device to operate on.
955  * @base: Block device that will become the new base, or %NULL to
956  * flatten the whole backing file chain onto @bs.
957  * @backing_file_str: The file name that will be written to @bs as the
958  * the new backing file if the job completes. Ignored if @base is %NULL.
959  * @creation_flags: Flags that control the behavior of the Job lifetime.
960  *                  See @BlockJobCreateFlags
961  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
962  * @on_error: The action to take upon error.
963  * @errp: Error object.
964  *
965  * Start a streaming operation on @bs.  Clusters that are unallocated
966  * in @bs, but allocated in any image between @base and @bs (both
967  * exclusive) will be written to @bs.  At the end of a successful
968  * streaming job, the backing file of @bs will be changed to
969  * @backing_file_str in the written image and to @base in the live
970  * BlockDriverState.
971  */
972 void stream_start(const char *job_id, BlockDriverState *bs,
973                   BlockDriverState *base, const char *backing_file_str,
974                   int creation_flags, int64_t speed,
975                   BlockdevOnError on_error, Error **errp);
976 
977 /**
978  * commit_start:
979  * @job_id: The id of the newly-created job, or %NULL to use the
980  * device name of @bs.
981  * @bs: Active block device.
982  * @top: Top block device to be committed.
983  * @base: Block device that will be written into, and become the new top.
984  * @creation_flags: Flags that control the behavior of the Job lifetime.
985  *                  See @BlockJobCreateFlags
986  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
987  * @on_error: The action to take upon error.
988  * @backing_file_str: String to use as the backing file in @top's overlay
989  * @filter_node_name: The node name that should be assigned to the filter
990  * driver that the commit job inserts into the graph above @top. NULL means
991  * that a node name should be autogenerated.
992  * @errp: Error object.
993  *
994  */
995 void commit_start(const char *job_id, BlockDriverState *bs,
996                   BlockDriverState *base, BlockDriverState *top,
997                   int creation_flags, int64_t speed,
998                   BlockdevOnError on_error, const char *backing_file_str,
999                   const char *filter_node_name, Error **errp);
1000 /**
1001  * commit_active_start:
1002  * @job_id: The id of the newly-created job, or %NULL to use the
1003  * device name of @bs.
1004  * @bs: Active block device to be committed.
1005  * @base: Block device that will be written into, and become the new top.
1006  * @creation_flags: Flags that control the behavior of the Job lifetime.
1007  *                  See @BlockJobCreateFlags
1008  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1009  * @on_error: The action to take upon error.
1010  * @filter_node_name: The node name that should be assigned to the filter
1011  * driver that the commit job inserts into the graph above @bs. NULL means that
1012  * a node name should be autogenerated.
1013  * @cb: Completion function for the job.
1014  * @opaque: Opaque pointer value passed to @cb.
1015  * @auto_complete: Auto complete the job.
1016  * @errp: Error object.
1017  *
1018  */
1019 void commit_active_start(const char *job_id, BlockDriverState *bs,
1020                          BlockDriverState *base, int creation_flags,
1021                          int64_t speed, BlockdevOnError on_error,
1022                          const char *filter_node_name,
1023                          BlockCompletionFunc *cb, void *opaque,
1024                          bool auto_complete, Error **errp);
1025 /*
1026  * mirror_start:
1027  * @job_id: The id of the newly-created job, or %NULL to use the
1028  * device name of @bs.
1029  * @bs: Block device to operate on.
1030  * @target: Block device to write to.
1031  * @replaces: Block graph node name to replace once the mirror is done. Can
1032  *            only be used when full mirroring is selected.
1033  * @creation_flags: Flags that control the behavior of the Job lifetime.
1034  *                  See @BlockJobCreateFlags
1035  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1036  * @granularity: The chosen granularity for the dirty bitmap.
1037  * @buf_size: The amount of data that can be in flight at one time.
1038  * @mode: Whether to collapse all images in the chain to the target.
1039  * @backing_mode: How to establish the target's backing chain after completion.
1040  * @on_source_error: The action to take upon error reading from the source.
1041  * @on_target_error: The action to take upon error writing to the target.
1042  * @unmap: Whether to unmap target where source sectors only contain zeroes.
1043  * @filter_node_name: The node name that should be assigned to the filter
1044  * driver that the mirror job inserts into the graph above @bs. NULL means that
1045  * a node name should be autogenerated.
1046  * @copy_mode: When to trigger writes to the target.
1047  * @errp: Error object.
1048  *
1049  * Start a mirroring operation on @bs.  Clusters that are allocated
1050  * in @bs will be written to @target until the job is cancelled or
1051  * manually completed.  At the end of a successful mirroring job,
1052  * @bs will be switched to read from @target.
1053  */
1054 void mirror_start(const char *job_id, BlockDriverState *bs,
1055                   BlockDriverState *target, const char *replaces,
1056                   int creation_flags, int64_t speed,
1057                   uint32_t granularity, int64_t buf_size,
1058                   MirrorSyncMode mode, BlockMirrorBackingMode backing_mode,
1059                   BlockdevOnError on_source_error,
1060                   BlockdevOnError on_target_error,
1061                   bool unmap, const char *filter_node_name,
1062                   MirrorCopyMode copy_mode, Error **errp);
1063 
1064 /*
1065  * backup_job_create:
1066  * @job_id: The id of the newly-created job, or %NULL to use the
1067  * device name of @bs.
1068  * @bs: Block device to operate on.
1069  * @target: Block device to write to.
1070  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1071  * @sync_mode: What parts of the disk image should be copied to the destination.
1072  * @sync_bitmap: The dirty bitmap if sync_mode is MIRROR_SYNC_MODE_INCREMENTAL.
1073  * @on_source_error: The action to take upon error reading from the source.
1074  * @on_target_error: The action to take upon error writing to the target.
1075  * @creation_flags: Flags that control the behavior of the Job lifetime.
1076  *                  See @BlockJobCreateFlags
1077  * @cb: Completion function for the job.
1078  * @opaque: Opaque pointer value passed to @cb.
1079  * @txn: Transaction that this job is part of (may be NULL).
1080  *
1081  * Create a backup operation on @bs.  Clusters in @bs are written to @target
1082  * until the job is cancelled or manually completed.
1083  */
1084 BlockJob *backup_job_create(const char *job_id, BlockDriverState *bs,
1085                             BlockDriverState *target, int64_t speed,
1086                             MirrorSyncMode sync_mode,
1087                             BdrvDirtyBitmap *sync_bitmap,
1088                             bool compress,
1089                             BlockdevOnError on_source_error,
1090                             BlockdevOnError on_target_error,
1091                             int creation_flags,
1092                             BlockCompletionFunc *cb, void *opaque,
1093                             JobTxn *txn, Error **errp);
1094 
1095 void hmp_drive_add_node(Monitor *mon, const char *optstr);
1096 
1097 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
1098                                   const char *child_name,
1099                                   const BdrvChildRole *child_role,
1100                                   uint64_t perm, uint64_t shared_perm,
1101                                   void *opaque, Error **errp);
1102 void bdrv_root_unref_child(BdrvChild *child);
1103 
1104 int bdrv_child_try_set_perm(BdrvChild *c, uint64_t perm, uint64_t shared,
1105                             Error **errp);
1106 
1107 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1108  * block filters: Forward CONSISTENT_READ, WRITE, WRITE_UNCHANGED and RESIZE to
1109  * all children */
1110 void bdrv_filter_default_perms(BlockDriverState *bs, BdrvChild *c,
1111                                const BdrvChildRole *role,
1112                                BlockReopenQueue *reopen_queue,
1113                                uint64_t perm, uint64_t shared,
1114                                uint64_t *nperm, uint64_t *nshared);
1115 
1116 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1117  * (non-raw) image formats: Like above for bs->backing, but for bs->file it
1118  * requires WRITE | RESIZE for read-write images, always requires
1119  * CONSISTENT_READ and doesn't share WRITE. */
1120 void bdrv_format_default_perms(BlockDriverState *bs, BdrvChild *c,
1121                                const BdrvChildRole *role,
1122                                BlockReopenQueue *reopen_queue,
1123                                uint64_t perm, uint64_t shared,
1124                                uint64_t *nperm, uint64_t *nshared);
1125 
1126 /*
1127  * Default implementation for drivers to pass bdrv_co_block_status() to
1128  * their file.
1129  */
1130 int coroutine_fn bdrv_co_block_status_from_file(BlockDriverState *bs,
1131                                                 bool want_zero,
1132                                                 int64_t offset,
1133                                                 int64_t bytes,
1134                                                 int64_t *pnum,
1135                                                 int64_t *map,
1136                                                 BlockDriverState **file);
1137 /*
1138  * Default implementation for drivers to pass bdrv_co_block_status() to
1139  * their backing file.
1140  */
1141 int coroutine_fn bdrv_co_block_status_from_backing(BlockDriverState *bs,
1142                                                    bool want_zero,
1143                                                    int64_t offset,
1144                                                    int64_t bytes,
1145                                                    int64_t *pnum,
1146                                                    int64_t *map,
1147                                                    BlockDriverState **file);
1148 const char *bdrv_get_parent_name(const BlockDriverState *bs);
1149 void blk_dev_change_media_cb(BlockBackend *blk, bool load, Error **errp);
1150 bool blk_dev_has_removable_media(BlockBackend *blk);
1151 bool blk_dev_has_tray(BlockBackend *blk);
1152 void blk_dev_eject_request(BlockBackend *blk, bool force);
1153 bool blk_dev_is_tray_open(BlockBackend *blk);
1154 bool blk_dev_is_medium_locked(BlockBackend *blk);
1155 
1156 void bdrv_set_dirty(BlockDriverState *bs, int64_t offset, int64_t bytes);
1157 
1158 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap **out);
1159 void bdrv_restore_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap *backup);
1160 
1161 void bdrv_inc_in_flight(BlockDriverState *bs);
1162 void bdrv_dec_in_flight(BlockDriverState *bs);
1163 
1164 void blockdev_close_all_bdrv_states(void);
1165 
1166 int coroutine_fn bdrv_co_copy_range_from(BdrvChild *src, uint64_t src_offset,
1167                                          BdrvChild *dst, uint64_t dst_offset,
1168                                          uint64_t bytes,
1169                                          BdrvRequestFlags read_flags,
1170                                          BdrvRequestFlags write_flags);
1171 int coroutine_fn bdrv_co_copy_range_to(BdrvChild *src, uint64_t src_offset,
1172                                        BdrvChild *dst, uint64_t dst_offset,
1173                                        uint64_t bytes,
1174                                        BdrvRequestFlags read_flags,
1175                                        BdrvRequestFlags write_flags);
1176 
1177 int refresh_total_sectors(BlockDriverState *bs, int64_t hint);
1178 
1179 #endif /* BLOCK_INT_H */
1180