xref: /openbmc/qemu/include/block/block_int.h (revision 38472890)
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/throttle.h"
37 
38 #define BLOCK_FLAG_LAZY_REFCOUNTS   8
39 
40 #define BLOCK_OPT_SIZE              "size"
41 #define BLOCK_OPT_ENCRYPT           "encryption"
42 #define BLOCK_OPT_ENCRYPT_FORMAT    "encrypt.format"
43 #define BLOCK_OPT_COMPAT6           "compat6"
44 #define BLOCK_OPT_HWVERSION         "hwversion"
45 #define BLOCK_OPT_BACKING_FILE      "backing_file"
46 #define BLOCK_OPT_BACKING_FMT       "backing_fmt"
47 #define BLOCK_OPT_CLUSTER_SIZE      "cluster_size"
48 #define BLOCK_OPT_TABLE_SIZE        "table_size"
49 #define BLOCK_OPT_PREALLOC          "preallocation"
50 #define BLOCK_OPT_SUBFMT            "subformat"
51 #define BLOCK_OPT_COMPAT_LEVEL      "compat"
52 #define BLOCK_OPT_LAZY_REFCOUNTS    "lazy_refcounts"
53 #define BLOCK_OPT_ADAPTER_TYPE      "adapter_type"
54 #define BLOCK_OPT_REDUNDANCY        "redundancy"
55 #define BLOCK_OPT_NOCOW             "nocow"
56 #define BLOCK_OPT_OBJECT_SIZE       "object_size"
57 #define BLOCK_OPT_REFCOUNT_BITS     "refcount_bits"
58 #define BLOCK_OPT_DATA_FILE         "data_file"
59 #define BLOCK_OPT_DATA_FILE_RAW     "data_file_raw"
60 
61 #define BLOCK_PROBE_BUF_SIZE        512
62 
63 enum BdrvTrackedRequestType {
64     BDRV_TRACKED_READ,
65     BDRV_TRACKED_WRITE,
66     BDRV_TRACKED_DISCARD,
67     BDRV_TRACKED_TRUNCATE,
68 };
69 
70 typedef struct BdrvTrackedRequest {
71     BlockDriverState *bs;
72     int64_t offset;
73     uint64_t bytes;
74     enum BdrvTrackedRequestType type;
75 
76     bool serialising;
77     int64_t overlap_offset;
78     uint64_t overlap_bytes;
79 
80     QLIST_ENTRY(BdrvTrackedRequest) list;
81     Coroutine *co; /* owner, used for deadlock detection */
82     CoQueue wait_queue; /* coroutines blocked on this request */
83 
84     struct BdrvTrackedRequest *waiting_for;
85 } BdrvTrackedRequest;
86 
87 struct BlockDriver {
88     const char *format_name;
89     int instance_size;
90 
91     /* set to true if the BlockDriver is a block filter. Block filters pass
92      * certain callbacks that refer to data (see block.c) to their bs->file if
93      * the driver doesn't implement them. Drivers that do not wish to forward
94      * must implement them and return -ENOTSUP.
95      */
96     bool is_filter;
97     /*
98      * Return true if @to_replace can be replaced by a BDS with the
99      * same data as @bs without it affecting @bs's behavior (that is,
100      * without it being visible to @bs's parents).
101      */
102     bool (*bdrv_recurse_can_replace)(BlockDriverState *bs,
103                                      BlockDriverState *to_replace);
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_commit_post)(BDRVReopenState *reopen_state);
126     void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state);
127     void (*bdrv_join_options)(QDict *options, QDict *old_options);
128 
129     int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags,
130                      Error **errp);
131 
132     /* Protocol drivers should implement this instead of bdrv_open */
133     int (*bdrv_file_open)(BlockDriverState *bs, QDict *options, int flags,
134                           Error **errp);
135     void (*bdrv_close)(BlockDriverState *bs);
136     int coroutine_fn (*bdrv_co_create)(BlockdevCreateOptions *opts,
137                                        Error **errp);
138     int coroutine_fn (*bdrv_co_create_opts)(BlockDriver *drv,
139                                             const char *filename,
140                                             QemuOpts *opts,
141                                             Error **errp);
142     int (*bdrv_make_empty)(BlockDriverState *bs);
143 
144     /*
145      * Refreshes the bs->exact_filename field. If that is impossible,
146      * bs->exact_filename has to be left empty.
147      */
148     void (*bdrv_refresh_filename)(BlockDriverState *bs);
149 
150     /*
151      * Gathers the open options for all children into @target.
152      * A simple format driver (without backing file support) might
153      * implement this function like this:
154      *
155      *     QINCREF(bs->file->bs->full_open_options);
156      *     qdict_put(target, "file", bs->file->bs->full_open_options);
157      *
158      * If not specified, the generic implementation will simply put
159      * all children's options under their respective name.
160      *
161      * @backing_overridden is true when bs->backing seems not to be
162      * the child that would result from opening bs->backing_file.
163      * Therefore, if it is true, the backing child's options should be
164      * gathered; otherwise, there is no need since the backing child
165      * is the one implied by the image header.
166      *
167      * Note that ideally this function would not be needed.  Every
168      * block driver which implements it is probably doing something
169      * shady regarding its runtime option structure.
170      */
171     void (*bdrv_gather_child_options)(BlockDriverState *bs, QDict *target,
172                                       bool backing_overridden);
173 
174     /*
175      * Returns an allocated string which is the directory name of this BDS: It
176      * will be used to make relative filenames absolute by prepending this
177      * function's return value to them.
178      */
179     char *(*bdrv_dirname)(BlockDriverState *bs, Error **errp);
180 
181     /* aio */
182     BlockAIOCB *(*bdrv_aio_preadv)(BlockDriverState *bs,
183         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags,
184         BlockCompletionFunc *cb, void *opaque);
185     BlockAIOCB *(*bdrv_aio_pwritev)(BlockDriverState *bs,
186         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags,
187         BlockCompletionFunc *cb, void *opaque);
188     BlockAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs,
189         BlockCompletionFunc *cb, void *opaque);
190     BlockAIOCB *(*bdrv_aio_pdiscard)(BlockDriverState *bs,
191         int64_t offset, int bytes,
192         BlockCompletionFunc *cb, void *opaque);
193 
194     int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs,
195         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
196 
197     /**
198      * @offset: position in bytes to read at
199      * @bytes: number of bytes to read
200      * @qiov: the buffers to fill with read data
201      * @flags: currently unused, always 0
202      *
203      * @offset and @bytes will be a multiple of 'request_alignment',
204      * but the length of individual @qiov elements does not have to
205      * be a multiple.
206      *
207      * @bytes will always equal the total size of @qiov, and will be
208      * no larger than 'max_transfer'.
209      *
210      * The buffer in @qiov may point directly to guest memory.
211      */
212     int coroutine_fn (*bdrv_co_preadv)(BlockDriverState *bs,
213         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags);
214     int coroutine_fn (*bdrv_co_preadv_part)(BlockDriverState *bs,
215         uint64_t offset, uint64_t bytes,
216         QEMUIOVector *qiov, size_t qiov_offset, int flags);
217     int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs,
218         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, int flags);
219     /**
220      * @offset: position in bytes to write at
221      * @bytes: number of bytes to write
222      * @qiov: the buffers containing data to write
223      * @flags: zero or more bits allowed by 'supported_write_flags'
224      *
225      * @offset and @bytes will be a multiple of 'request_alignment',
226      * but the length of individual @qiov elements does not have to
227      * be a multiple.
228      *
229      * @bytes will always equal the total size of @qiov, and will be
230      * no larger than 'max_transfer'.
231      *
232      * The buffer in @qiov may point directly to guest memory.
233      */
234     int coroutine_fn (*bdrv_co_pwritev)(BlockDriverState *bs,
235         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags);
236     int coroutine_fn (*bdrv_co_pwritev_part)(BlockDriverState *bs,
237         uint64_t offset, uint64_t bytes,
238         QEMUIOVector *qiov, size_t qiov_offset, int flags);
239 
240     /*
241      * Efficiently zero a region of the disk image.  Typically an image format
242      * would use a compact metadata representation to implement this.  This
243      * function pointer may be NULL or return -ENOSUP and .bdrv_co_writev()
244      * will be called instead.
245      */
246     int coroutine_fn (*bdrv_co_pwrite_zeroes)(BlockDriverState *bs,
247         int64_t offset, int bytes, BdrvRequestFlags flags);
248     int coroutine_fn (*bdrv_co_pdiscard)(BlockDriverState *bs,
249         int64_t offset, int bytes);
250 
251     /* Map [offset, offset + nbytes) range onto a child of @bs to copy from,
252      * and invoke bdrv_co_copy_range_from(child, ...), or invoke
253      * bdrv_co_copy_range_to() if @bs is the leaf child to copy data from.
254      *
255      * See the comment of bdrv_co_copy_range for the parameter and return value
256      * semantics.
257      */
258     int coroutine_fn (*bdrv_co_copy_range_from)(BlockDriverState *bs,
259                                                 BdrvChild *src,
260                                                 uint64_t offset,
261                                                 BdrvChild *dst,
262                                                 uint64_t dst_offset,
263                                                 uint64_t bytes,
264                                                 BdrvRequestFlags read_flags,
265                                                 BdrvRequestFlags write_flags);
266 
267     /* Map [offset, offset + nbytes) range onto a child of bs to copy data to,
268      * and invoke bdrv_co_copy_range_to(child, src, ...), or perform the copy
269      * operation if @bs is the leaf and @src has the same BlockDriver.  Return
270      * -ENOTSUP if @bs is the leaf but @src has a different BlockDriver.
271      *
272      * See the comment of bdrv_co_copy_range for the parameter and return value
273      * semantics.
274      */
275     int coroutine_fn (*bdrv_co_copy_range_to)(BlockDriverState *bs,
276                                               BdrvChild *src,
277                                               uint64_t src_offset,
278                                               BdrvChild *dst,
279                                               uint64_t dst_offset,
280                                               uint64_t bytes,
281                                               BdrvRequestFlags read_flags,
282                                               BdrvRequestFlags write_flags);
283 
284     /*
285      * Building block for bdrv_block_status[_above] and
286      * bdrv_is_allocated[_above].  The driver should answer only
287      * according to the current layer, and should only need to set
288      * BDRV_BLOCK_DATA, BDRV_BLOCK_ZERO, BDRV_BLOCK_OFFSET_VALID,
289      * and/or BDRV_BLOCK_RAW; if the current layer defers to a backing
290      * layer, the result should be 0 (and not BDRV_BLOCK_ZERO).  See
291      * block.h for the overall meaning of the bits.  As a hint, the
292      * flag want_zero is true if the caller cares more about precise
293      * mappings (favor accurate _OFFSET_VALID/_ZERO) or false for
294      * overall allocation (favor larger *pnum, perhaps by reporting
295      * _DATA instead of _ZERO).  The block layer guarantees input
296      * clamped to bdrv_getlength() and aligned to request_alignment,
297      * as well as non-NULL pnum, map, and file; in turn, the driver
298      * must return an error or set pnum to an aligned non-zero value.
299      */
300     int coroutine_fn (*bdrv_co_block_status)(BlockDriverState *bs,
301         bool want_zero, int64_t offset, int64_t bytes, int64_t *pnum,
302         int64_t *map, BlockDriverState **file);
303 
304     /*
305      * Invalidate any cached meta-data.
306      */
307     void coroutine_fn (*bdrv_co_invalidate_cache)(BlockDriverState *bs,
308                                                   Error **errp);
309     int (*bdrv_inactivate)(BlockDriverState *bs);
310 
311     /*
312      * Flushes all data for all layers by calling bdrv_co_flush for underlying
313      * layers, if needed. This function is needed for deterministic
314      * synchronization of the flush finishing callback.
315      */
316     int coroutine_fn (*bdrv_co_flush)(BlockDriverState *bs);
317 
318     /* Delete a created file. */
319     int coroutine_fn (*bdrv_co_delete_file)(BlockDriverState *bs,
320                                             Error **errp);
321 
322     /*
323      * Flushes all data that was already written to the OS all the way down to
324      * the disk (for example file-posix.c calls fsync()).
325      */
326     int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs);
327 
328     /*
329      * Flushes all internal caches to the OS. The data may still sit in a
330      * writeback cache of the host OS, but it will survive a crash of the qemu
331      * process.
332      */
333     int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs);
334 
335     /*
336      * Drivers setting this field must be able to work with just a plain
337      * filename with '<protocol_name>:' as a prefix, and no other options.
338      * Options may be extracted from the filename by implementing
339      * bdrv_parse_filename.
340      */
341     const char *protocol_name;
342 
343     /*
344      * Truncate @bs to @offset bytes using the given @prealloc mode
345      * when growing.  Modes other than PREALLOC_MODE_OFF should be
346      * rejected when shrinking @bs.
347      *
348      * If @exact is true, @bs must be resized to exactly @offset.
349      * Otherwise, it is sufficient for @bs (if it is a host block
350      * device and thus there is no way to resize it) to be at least
351      * @offset bytes in length.
352      *
353      * If @exact is true and this function fails but would succeed
354      * with @exact = false, it should return -ENOTSUP.
355      */
356     int coroutine_fn (*bdrv_co_truncate)(BlockDriverState *bs, int64_t offset,
357                                          bool exact, PreallocMode prealloc,
358                                          Error **errp);
359 
360     int64_t (*bdrv_getlength)(BlockDriverState *bs);
361     bool has_variable_length;
362     int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
363     BlockMeasureInfo *(*bdrv_measure)(QemuOpts *opts, BlockDriverState *in_bs,
364                                       Error **errp);
365 
366     int coroutine_fn (*bdrv_co_pwritev_compressed)(BlockDriverState *bs,
367         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov);
368     int coroutine_fn (*bdrv_co_pwritev_compressed_part)(BlockDriverState *bs,
369         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov,
370         size_t qiov_offset);
371 
372     int (*bdrv_snapshot_create)(BlockDriverState *bs,
373                                 QEMUSnapshotInfo *sn_info);
374     int (*bdrv_snapshot_goto)(BlockDriverState *bs,
375                               const char *snapshot_id);
376     int (*bdrv_snapshot_delete)(BlockDriverState *bs,
377                                 const char *snapshot_id,
378                                 const char *name,
379                                 Error **errp);
380     int (*bdrv_snapshot_list)(BlockDriverState *bs,
381                               QEMUSnapshotInfo **psn_info);
382     int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
383                                   const char *snapshot_id,
384                                   const char *name,
385                                   Error **errp);
386     int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
387     ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs,
388                                                  Error **errp);
389     BlockStatsSpecific *(*bdrv_get_specific_stats)(BlockDriverState *bs);
390 
391     int coroutine_fn (*bdrv_save_vmstate)(BlockDriverState *bs,
392                                           QEMUIOVector *qiov,
393                                           int64_t pos);
394     int coroutine_fn (*bdrv_load_vmstate)(BlockDriverState *bs,
395                                           QEMUIOVector *qiov,
396                                           int64_t pos);
397 
398     int (*bdrv_change_backing_file)(BlockDriverState *bs,
399         const char *backing_file, const char *backing_fmt);
400 
401     /* removable device specific */
402     bool (*bdrv_is_inserted)(BlockDriverState *bs);
403     void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
404     void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
405 
406     /* to control generic scsi devices */
407     BlockAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
408         unsigned long int req, void *buf,
409         BlockCompletionFunc *cb, void *opaque);
410     int coroutine_fn (*bdrv_co_ioctl)(BlockDriverState *bs,
411                                       unsigned long int req, void *buf);
412 
413     /* List of options for creating images, terminated by name == NULL */
414     QemuOptsList *create_opts;
415     /*
416      * If this driver supports reopening images this contains a
417      * NULL-terminated list of the runtime options that can be
418      * modified. If an option in this list is unspecified during
419      * reopen then it _must_ be reset to its default value or return
420      * an error.
421      */
422     const char *const *mutable_opts;
423 
424     /*
425      * Returns 0 for completed check, -errno for internal errors.
426      * The check results are stored in result.
427      */
428     int coroutine_fn (*bdrv_co_check)(BlockDriverState *bs,
429                                       BdrvCheckResult *result,
430                                       BdrvCheckMode fix);
431 
432     int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts,
433                               BlockDriverAmendStatusCB *status_cb,
434                               void *cb_opaque,
435                               Error **errp);
436 
437     void (*bdrv_debug_event)(BlockDriverState *bs, BlkdebugEvent event);
438 
439     /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
440     int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
441         const char *tag);
442     int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs,
443         const char *tag);
444     int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
445     bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
446 
447     void (*bdrv_refresh_limits)(BlockDriverState *bs, Error **errp);
448 
449     /*
450      * Returns 1 if newly created images are guaranteed to contain only
451      * zeros, 0 otherwise.
452      * Must return 0 if .bdrv_has_zero_init_truncate() returns 0.
453      */
454     int (*bdrv_has_zero_init)(BlockDriverState *bs);
455 
456     /*
457      * Returns 1 if new areas added by growing the image with
458      * PREALLOC_MODE_OFF contain only zeros, 0 otherwise.
459      */
460     int (*bdrv_has_zero_init_truncate)(BlockDriverState *bs);
461 
462     /* Remove fd handlers, timers, and other event loop callbacks so the event
463      * loop is no longer in use.  Called with no in-flight requests and in
464      * depth-first traversal order with parents before child nodes.
465      */
466     void (*bdrv_detach_aio_context)(BlockDriverState *bs);
467 
468     /* Add fd handlers, timers, and other event loop callbacks so I/O requests
469      * can be processed again.  Called with no in-flight requests and in
470      * depth-first traversal order with child nodes before parent nodes.
471      */
472     void (*bdrv_attach_aio_context)(BlockDriverState *bs,
473                                     AioContext *new_context);
474 
475     /* io queue for linux-aio */
476     void (*bdrv_io_plug)(BlockDriverState *bs);
477     void (*bdrv_io_unplug)(BlockDriverState *bs);
478 
479     /**
480      * Try to get @bs's logical and physical block size.
481      * On success, store them in @bsz and return zero.
482      * On failure, return negative errno.
483      */
484     int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz);
485     /**
486      * Try to get @bs's geometry (cyls, heads, sectors)
487      * On success, store them in @geo and return 0.
488      * On failure return -errno.
489      * Only drivers that want to override guest geometry implement this
490      * callback; see hd_geometry_guess().
491      */
492     int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo);
493 
494     /**
495      * bdrv_co_drain_begin is called if implemented in the beginning of a
496      * drain operation to drain and stop any internal sources of requests in
497      * the driver.
498      * bdrv_co_drain_end is called if implemented at the end of the drain.
499      *
500      * They should be used by the driver to e.g. manage scheduled I/O
501      * requests, or toggle an internal state. After the end of the drain new
502      * requests will continue normally.
503      */
504     void coroutine_fn (*bdrv_co_drain_begin)(BlockDriverState *bs);
505     void coroutine_fn (*bdrv_co_drain_end)(BlockDriverState *bs);
506 
507     void (*bdrv_add_child)(BlockDriverState *parent, BlockDriverState *child,
508                            Error **errp);
509     void (*bdrv_del_child)(BlockDriverState *parent, BdrvChild *child,
510                            Error **errp);
511 
512     /**
513      * Informs the block driver that a permission change is intended. The
514      * driver checks whether the change is permissible and may take other
515      * preparations for the change (e.g. get file system locks). This operation
516      * is always followed either by a call to either .bdrv_set_perm or
517      * .bdrv_abort_perm_update.
518      *
519      * Checks whether the requested set of cumulative permissions in @perm
520      * can be granted for accessing @bs and whether no other users are using
521      * permissions other than those given in @shared (both arguments take
522      * BLK_PERM_* bitmasks).
523      *
524      * If both conditions are met, 0 is returned. Otherwise, -errno is returned
525      * and errp is set to an error describing the conflict.
526      */
527     int (*bdrv_check_perm)(BlockDriverState *bs, uint64_t perm,
528                            uint64_t shared, Error **errp);
529 
530     /**
531      * Called to inform the driver that the set of cumulative set of used
532      * permissions for @bs has changed to @perm, and the set of sharable
533      * permission to @shared. The driver can use this to propagate changes to
534      * its children (i.e. request permissions only if a parent actually needs
535      * them).
536      *
537      * This function is only invoked after bdrv_check_perm(), so block drivers
538      * may rely on preparations made in their .bdrv_check_perm implementation.
539      */
540     void (*bdrv_set_perm)(BlockDriverState *bs, uint64_t perm, uint64_t shared);
541 
542     /*
543      * Called to inform the driver that after a previous bdrv_check_perm()
544      * call, the permission update is not performed and any preparations made
545      * for it (e.g. taken file locks) need to be undone.
546      *
547      * This function can be called even for nodes that never saw a
548      * bdrv_check_perm() call. It is a no-op then.
549      */
550     void (*bdrv_abort_perm_update)(BlockDriverState *bs);
551 
552     /**
553      * Returns in @nperm and @nshared the permissions that the driver for @bs
554      * needs on its child @c, based on the cumulative permissions requested by
555      * the parents in @parent_perm and @parent_shared.
556      *
557      * If @c is NULL, return the permissions for attaching a new child for the
558      * given @role.
559      *
560      * If @reopen_queue is non-NULL, don't return the currently needed
561      * permissions, but those that will be needed after applying the
562      * @reopen_queue.
563      */
564      void (*bdrv_child_perm)(BlockDriverState *bs, BdrvChild *c,
565                              const BdrvChildRole *role,
566                              BlockReopenQueue *reopen_queue,
567                              uint64_t parent_perm, uint64_t parent_shared,
568                              uint64_t *nperm, uint64_t *nshared);
569 
570     bool (*bdrv_co_can_store_new_dirty_bitmap)(BlockDriverState *bs,
571                                                const char *name,
572                                                uint32_t granularity,
573                                                Error **errp);
574     int (*bdrv_co_remove_persistent_dirty_bitmap)(BlockDriverState *bs,
575                                                   const char *name,
576                                                   Error **errp);
577 
578     /**
579      * Register/unregister a buffer for I/O. For example, when the driver is
580      * interested to know the memory areas that will later be used in iovs, so
581      * that it can do IOMMU mapping with VFIO etc., in order to get better
582      * performance. In the case of VFIO drivers, this callback is used to do
583      * DMA mapping for hot buffers.
584      */
585     void (*bdrv_register_buf)(BlockDriverState *bs, void *host, size_t size);
586     void (*bdrv_unregister_buf)(BlockDriverState *bs, void *host);
587     QLIST_ENTRY(BlockDriver) list;
588 
589     /* Pointer to a NULL-terminated array of names of strong options
590      * that can be specified for bdrv_open(). A strong option is one
591      * that changes the data of a BDS.
592      * If this pointer is NULL, the array is considered empty.
593      * "filename" and "driver" are always considered strong. */
594     const char *const *strong_runtime_opts;
595 };
596 
597 static inline bool block_driver_can_compress(BlockDriver *drv)
598 {
599     return drv->bdrv_co_pwritev_compressed ||
600            drv->bdrv_co_pwritev_compressed_part;
601 }
602 
603 typedef struct BlockLimits {
604     /* Alignment requirement, in bytes, for offset/length of I/O
605      * requests. Must be a power of 2 less than INT_MAX; defaults to
606      * 1 for drivers with modern byte interfaces, and to 512
607      * otherwise. */
608     uint32_t request_alignment;
609 
610     /* Maximum number of bytes that can be discarded at once (since it
611      * is signed, it must be < 2G, if set). Must be multiple of
612      * pdiscard_alignment, but need not be power of 2. May be 0 if no
613      * inherent 32-bit limit */
614     int32_t max_pdiscard;
615 
616     /* Optimal alignment for discard requests in bytes. A power of 2
617      * is best but not mandatory.  Must be a multiple of
618      * bl.request_alignment, and must be less than max_pdiscard if
619      * that is set. May be 0 if bl.request_alignment is good enough */
620     uint32_t pdiscard_alignment;
621 
622     /* Maximum number of bytes that can zeroized at once (since it is
623      * signed, it must be < 2G, if set). Must be multiple of
624      * pwrite_zeroes_alignment. May be 0 if no inherent 32-bit limit */
625     int32_t max_pwrite_zeroes;
626 
627     /* Optimal alignment for write zeroes requests in bytes. A power
628      * of 2 is best but not mandatory.  Must be a multiple of
629      * bl.request_alignment, and must be less than max_pwrite_zeroes
630      * if that is set. May be 0 if bl.request_alignment is good
631      * enough */
632     uint32_t pwrite_zeroes_alignment;
633 
634     /* Optimal transfer length in bytes.  A power of 2 is best but not
635      * mandatory.  Must be a multiple of bl.request_alignment, or 0 if
636      * no preferred size */
637     uint32_t opt_transfer;
638 
639     /* Maximal transfer length in bytes.  Need not be power of 2, but
640      * must be multiple of opt_transfer and bl.request_alignment, or 0
641      * for no 32-bit limit.  For now, anything larger than INT_MAX is
642      * clamped down. */
643     uint32_t max_transfer;
644 
645     /* memory alignment, in bytes so that no bounce buffer is needed */
646     size_t min_mem_alignment;
647 
648     /* memory alignment, in bytes, for bounce buffer */
649     size_t opt_mem_alignment;
650 
651     /* maximum number of iovec elements */
652     int max_iov;
653 } BlockLimits;
654 
655 typedef struct BdrvOpBlocker BdrvOpBlocker;
656 
657 typedef struct BdrvAioNotifier {
658     void (*attached_aio_context)(AioContext *new_context, void *opaque);
659     void (*detach_aio_context)(void *opaque);
660 
661     void *opaque;
662     bool deleted;
663 
664     QLIST_ENTRY(BdrvAioNotifier) list;
665 } BdrvAioNotifier;
666 
667 struct BdrvChildRole {
668     /* If true, bdrv_replace_node() doesn't change the node this BdrvChild
669      * points to. */
670     bool stay_at_node;
671 
672     /* If true, the parent is a BlockDriverState and bdrv_next_all_states()
673      * will return it. This information is used for drain_all, where every node
674      * will be drained separately, so the drain only needs to be propagated to
675      * non-BDS parents. */
676     bool parent_is_bds;
677 
678     void (*inherit_options)(int *child_flags, QDict *child_options,
679                             int parent_flags, QDict *parent_options);
680 
681     void (*change_media)(BdrvChild *child, bool load);
682     void (*resize)(BdrvChild *child);
683 
684     /* Returns a name that is supposedly more useful for human users than the
685      * node name for identifying the node in question (in particular, a BB
686      * name), or NULL if the parent can't provide a better name. */
687     const char *(*get_name)(BdrvChild *child);
688 
689     /* Returns a malloced string that describes the parent of the child for a
690      * human reader. This could be a node-name, BlockBackend name, qdev ID or
691      * QOM path of the device owning the BlockBackend, job type and ID etc. The
692      * caller is responsible for freeing the memory. */
693     char *(*get_parent_desc)(BdrvChild *child);
694 
695     /*
696      * If this pair of functions is implemented, the parent doesn't issue new
697      * requests after returning from .drained_begin() until .drained_end() is
698      * called.
699      *
700      * These functions must not change the graph (and therefore also must not
701      * call aio_poll(), which could change the graph indirectly).
702      *
703      * If drained_end() schedules background operations, it must atomically
704      * increment *drained_end_counter for each such operation and atomically
705      * decrement it once the operation has settled.
706      *
707      * Note that this can be nested. If drained_begin() was called twice, new
708      * I/O is allowed only after drained_end() was called twice, too.
709      */
710     void (*drained_begin)(BdrvChild *child);
711     void (*drained_end)(BdrvChild *child, int *drained_end_counter);
712 
713     /*
714      * Returns whether the parent has pending requests for the child. This
715      * callback is polled after .drained_begin() has been called until all
716      * activity on the child has stopped.
717      */
718     bool (*drained_poll)(BdrvChild *child);
719 
720     /* Notifies the parent that the child has been activated/inactivated (e.g.
721      * when migration is completing) and it can start/stop requesting
722      * permissions and doing I/O on it. */
723     void (*activate)(BdrvChild *child, Error **errp);
724     int (*inactivate)(BdrvChild *child);
725 
726     void (*attach)(BdrvChild *child);
727     void (*detach)(BdrvChild *child);
728 
729     /* Notifies the parent that the filename of its child has changed (e.g.
730      * because the direct child was removed from the backing chain), so that it
731      * can update its reference. */
732     int (*update_filename)(BdrvChild *child, BlockDriverState *new_base,
733                            const char *filename, Error **errp);
734 
735     bool (*can_set_aio_ctx)(BdrvChild *child, AioContext *ctx,
736                             GSList **ignore, Error **errp);
737     void (*set_aio_ctx)(BdrvChild *child, AioContext *ctx, GSList **ignore);
738 };
739 
740 extern const BdrvChildRole child_file;
741 extern const BdrvChildRole child_format;
742 extern const BdrvChildRole child_backing;
743 
744 struct BdrvChild {
745     BlockDriverState *bs;
746     char *name;
747     const BdrvChildRole *role;
748     void *opaque;
749 
750     /**
751      * Granted permissions for operating on this BdrvChild (BLK_PERM_* bitmask)
752      */
753     uint64_t perm;
754 
755     /**
756      * Permissions that can still be granted to other users of @bs while this
757      * BdrvChild is still attached to it. (BLK_PERM_* bitmask)
758      */
759     uint64_t shared_perm;
760 
761     /* backup of permissions during permission update procedure */
762     bool has_backup_perm;
763     uint64_t backup_perm;
764     uint64_t backup_shared_perm;
765 
766     /*
767      * This link is frozen: the child can neither be replaced nor
768      * detached from the parent.
769      */
770     bool frozen;
771 
772     /*
773      * How many times the parent of this child has been drained
774      * (through role->drained_*).
775      * Usually, this is equal to bs->quiesce_counter (potentially
776      * reduced by bdrv_drain_all_count).  It may differ while the
777      * child is entering or leaving a drained section.
778      */
779     int parent_quiesce_counter;
780 
781     QLIST_ENTRY(BdrvChild) next;
782     QLIST_ENTRY(BdrvChild) next_parent;
783 };
784 
785 /*
786  * Note: the function bdrv_append() copies and swaps contents of
787  * BlockDriverStates, so if you add new fields to this struct, please
788  * inspect bdrv_append() to determine if the new fields need to be
789  * copied as well.
790  */
791 struct BlockDriverState {
792     /* Protected by big QEMU lock or read-only after opening.  No special
793      * locking needed during I/O...
794      */
795     int open_flags; /* flags used to open the file, re-used for re-open */
796     bool read_only; /* if true, the media is read only */
797     bool encrypted; /* if true, the media is encrypted */
798     bool sg;        /* if true, the device is a /dev/sg* */
799     bool probed;    /* if true, format was probed rather than specified */
800     bool force_share; /* if true, always allow all shared permissions */
801     bool implicit;  /* if true, this filter node was automatically inserted */
802 
803     BlockDriver *drv; /* NULL means no media */
804     void *opaque;
805 
806     AioContext *aio_context; /* event loop used for fd handlers, timers, etc */
807     /* long-running tasks intended to always use the same AioContext as this
808      * BDS may register themselves in this list to be notified of changes
809      * regarding this BDS's context */
810     QLIST_HEAD(, BdrvAioNotifier) aio_notifiers;
811     bool walking_aio_notifiers; /* to make removal during iteration safe */
812 
813     char filename[PATH_MAX];
814     char backing_file[PATH_MAX]; /* if non zero, the image is a diff of
815                                     this file image */
816     /* The backing filename indicated by the image header; if we ever
817      * open this file, then this is replaced by the resulting BDS's
818      * filename (i.e. after a bdrv_refresh_filename() run). */
819     char auto_backing_file[PATH_MAX];
820     char backing_format[16]; /* if non-zero and backing_file exists */
821 
822     QDict *full_open_options;
823     char exact_filename[PATH_MAX];
824 
825     BdrvChild *backing;
826     BdrvChild *file;
827 
828     /* I/O Limits */
829     BlockLimits bl;
830 
831     /* Flags honored during pwrite (so far: BDRV_REQ_FUA,
832      * BDRV_REQ_WRITE_UNCHANGED).
833      * If a driver does not support BDRV_REQ_WRITE_UNCHANGED, those
834      * writes will be issued as normal writes without the flag set.
835      * This is important to note for drivers that do not explicitly
836      * request a WRITE permission for their children and instead take
837      * the same permissions as their parent did (this is commonly what
838      * block filters do).  Such drivers have to be aware that the
839      * parent may have taken a WRITE_UNCHANGED permission only and is
840      * issuing such requests.  Drivers either must make sure that
841      * these requests do not result in plain WRITE accesses (usually
842      * by supporting BDRV_REQ_WRITE_UNCHANGED, and then forwarding
843      * every incoming write request as-is, including potentially that
844      * flag), or they have to explicitly take the WRITE permission for
845      * their children. */
846     unsigned int supported_write_flags;
847     /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA,
848      * BDRV_REQ_MAY_UNMAP, BDRV_REQ_WRITE_UNCHANGED) */
849     unsigned int supported_zero_flags;
850 
851     /* the following member gives a name to every node on the bs graph. */
852     char node_name[32];
853     /* element of the list of named nodes building the graph */
854     QTAILQ_ENTRY(BlockDriverState) node_list;
855     /* element of the list of all BlockDriverStates (all_bdrv_states) */
856     QTAILQ_ENTRY(BlockDriverState) bs_list;
857     /* element of the list of monitor-owned BDS */
858     QTAILQ_ENTRY(BlockDriverState) monitor_list;
859     int refcnt;
860 
861     /* operation blockers */
862     QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX];
863 
864     /* The node that this node inherited default options from (and a reopen on
865      * which can affect this node by changing these defaults). This is always a
866      * parent node of this node. */
867     BlockDriverState *inherits_from;
868     QLIST_HEAD(, BdrvChild) children;
869     QLIST_HEAD(, BdrvChild) parents;
870 
871     QDict *options;
872     QDict *explicit_options;
873     BlockdevDetectZeroesOptions detect_zeroes;
874 
875     /* The error object in use for blocking operations on backing_hd */
876     Error *backing_blocker;
877 
878     /* Protected by AioContext lock */
879 
880     /* If we are reading a disk image, give its size in sectors.
881      * Generally read-only; it is written to by load_snapshot and
882      * save_snaphost, but the block layer is quiescent during those.
883      */
884     int64_t total_sectors;
885 
886     /* Callback before write request is processed */
887     NotifierWithReturnList before_write_notifiers;
888 
889     /* threshold limit for writes, in bytes. "High water mark". */
890     uint64_t write_threshold_offset;
891     NotifierWithReturn write_threshold_notifier;
892 
893     /* Writing to the list requires the BQL _and_ the dirty_bitmap_mutex.
894      * Reading from the list can be done with either the BQL or the
895      * dirty_bitmap_mutex.  Modifying a bitmap only requires
896      * dirty_bitmap_mutex.  */
897     QemuMutex dirty_bitmap_mutex;
898     QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps;
899 
900     /* Offset after the highest byte written to */
901     Stat64 wr_highest_offset;
902 
903     /* If true, copy read backing sectors into image.  Can be >1 if more
904      * than one client has requested copy-on-read.  Accessed with atomic
905      * ops.
906      */
907     int copy_on_read;
908 
909     /* number of in-flight requests; overall and serialising.
910      * Accessed with atomic ops.
911      */
912     unsigned int in_flight;
913     unsigned int serialising_in_flight;
914 
915     /* counter for nested bdrv_io_plug.
916      * Accessed with atomic ops.
917     */
918     unsigned io_plugged;
919 
920     /* do we need to tell the quest if we have a volatile write cache? */
921     int enable_write_cache;
922 
923     /* Accessed with atomic ops.  */
924     int quiesce_counter;
925     int recursive_quiesce_counter;
926 
927     unsigned int write_gen;               /* Current data generation */
928 
929     /* Protected by reqs_lock.  */
930     CoMutex reqs_lock;
931     QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
932     CoQueue flush_queue;                  /* Serializing flush queue */
933     bool active_flush_req;                /* Flush request in flight? */
934 
935     /* Only read/written by whoever has set active_flush_req to true.  */
936     unsigned int flushed_gen;             /* Flushed write generation */
937 
938     /* BdrvChild links to this node may never be frozen */
939     bool never_freeze;
940 };
941 
942 struct BlockBackendRootState {
943     int open_flags;
944     bool read_only;
945     BlockdevDetectZeroesOptions detect_zeroes;
946 };
947 
948 typedef enum BlockMirrorBackingMode {
949     /* Reuse the existing backing chain from the source for the target.
950      * - sync=full: Set backing BDS to NULL.
951      * - sync=top:  Use source's backing BDS.
952      * - sync=none: Use source as the backing BDS. */
953     MIRROR_SOURCE_BACKING_CHAIN,
954 
955     /* Open the target's backing chain completely anew */
956     MIRROR_OPEN_BACKING_CHAIN,
957 
958     /* Do not change the target's backing BDS after job completion */
959     MIRROR_LEAVE_BACKING_CHAIN,
960 } BlockMirrorBackingMode;
961 
962 static inline BlockDriverState *backing_bs(BlockDriverState *bs)
963 {
964     return bs->backing ? bs->backing->bs : NULL;
965 }
966 
967 
968 /* Essential block drivers which must always be statically linked into qemu, and
969  * which therefore can be accessed without using bdrv_find_format() */
970 extern BlockDriver bdrv_file;
971 extern BlockDriver bdrv_raw;
972 extern BlockDriver bdrv_qcow2;
973 
974 int coroutine_fn bdrv_co_preadv(BdrvChild *child,
975     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
976     BdrvRequestFlags flags);
977 int coroutine_fn bdrv_co_preadv_part(BdrvChild *child,
978     int64_t offset, unsigned int bytes,
979     QEMUIOVector *qiov, size_t qiov_offset, BdrvRequestFlags flags);
980 int coroutine_fn bdrv_co_pwritev(BdrvChild *child,
981     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
982     BdrvRequestFlags flags);
983 int coroutine_fn bdrv_co_pwritev_part(BdrvChild *child,
984     int64_t offset, unsigned int bytes,
985     QEMUIOVector *qiov, size_t qiov_offset, BdrvRequestFlags flags);
986 
987 static inline int coroutine_fn bdrv_co_pread(BdrvChild *child,
988     int64_t offset, unsigned int bytes, void *buf, BdrvRequestFlags flags)
989 {
990     QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes);
991 
992     return bdrv_co_preadv(child, offset, bytes, &qiov, flags);
993 }
994 
995 static inline int coroutine_fn bdrv_co_pwrite(BdrvChild *child,
996     int64_t offset, unsigned int bytes, void *buf, BdrvRequestFlags flags)
997 {
998     QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes);
999 
1000     return bdrv_co_pwritev(child, offset, bytes, &qiov, flags);
1001 }
1002 
1003 extern unsigned int bdrv_drain_all_count;
1004 void bdrv_apply_subtree_drain(BdrvChild *child, BlockDriverState *new_parent);
1005 void bdrv_unapply_subtree_drain(BdrvChild *child, BlockDriverState *old_parent);
1006 
1007 bool coroutine_fn bdrv_mark_request_serialising(BdrvTrackedRequest *req, uint64_t align);
1008 BdrvTrackedRequest *coroutine_fn bdrv_co_get_self_request(BlockDriverState *bs);
1009 
1010 int get_tmp_filename(char *filename, int size);
1011 BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size,
1012                             const char *filename);
1013 
1014 void bdrv_parse_filename_strip_prefix(const char *filename, const char *prefix,
1015                                       QDict *options);
1016 
1017 
1018 /**
1019  * bdrv_add_before_write_notifier:
1020  *
1021  * Register a callback that is invoked before write requests are processed but
1022  * after any throttling or waiting for overlapping requests.
1023  */
1024 void bdrv_add_before_write_notifier(BlockDriverState *bs,
1025                                     NotifierWithReturn *notifier);
1026 
1027 /**
1028  * bdrv_add_aio_context_notifier:
1029  *
1030  * If a long-running job intends to be always run in the same AioContext as a
1031  * certain BDS, it may use this function to be notified of changes regarding the
1032  * association of the BDS to an AioContext.
1033  *
1034  * attached_aio_context() is called after the target BDS has been attached to a
1035  * new AioContext; detach_aio_context() is called before the target BDS is being
1036  * detached from its old AioContext.
1037  */
1038 void bdrv_add_aio_context_notifier(BlockDriverState *bs,
1039         void (*attached_aio_context)(AioContext *new_context, void *opaque),
1040         void (*detach_aio_context)(void *opaque), void *opaque);
1041 
1042 /**
1043  * bdrv_remove_aio_context_notifier:
1044  *
1045  * Unsubscribe of change notifications regarding the BDS's AioContext. The
1046  * parameters given here have to be the same as those given to
1047  * bdrv_add_aio_context_notifier().
1048  */
1049 void bdrv_remove_aio_context_notifier(BlockDriverState *bs,
1050                                       void (*aio_context_attached)(AioContext *,
1051                                                                    void *),
1052                                       void (*aio_context_detached)(void *),
1053                                       void *opaque);
1054 
1055 /**
1056  * bdrv_wakeup:
1057  * @bs: The BlockDriverState for which an I/O operation has been completed.
1058  *
1059  * Wake up the main thread if it is waiting on BDRV_POLL_WHILE.  During
1060  * synchronous I/O on a BlockDriverState that is attached to another
1061  * I/O thread, the main thread lets the I/O thread's event loop run,
1062  * waiting for the I/O operation to complete.  A bdrv_wakeup will wake
1063  * up the main thread if necessary.
1064  *
1065  * Manual calls to bdrv_wakeup are rarely necessary, because
1066  * bdrv_dec_in_flight already calls it.
1067  */
1068 void bdrv_wakeup(BlockDriverState *bs);
1069 
1070 #ifdef _WIN32
1071 int is_windows_drive(const char *filename);
1072 #endif
1073 
1074 /**
1075  * stream_start:
1076  * @job_id: The id of the newly-created job, or %NULL to use the
1077  * device name of @bs.
1078  * @bs: Block device to operate on.
1079  * @base: Block device that will become the new base, or %NULL to
1080  * flatten the whole backing file chain onto @bs.
1081  * @backing_file_str: The file name that will be written to @bs as the
1082  * the new backing file if the job completes. Ignored if @base is %NULL.
1083  * @creation_flags: Flags that control the behavior of the Job lifetime.
1084  *                  See @BlockJobCreateFlags
1085  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1086  * @on_error: The action to take upon error.
1087  * @errp: Error object.
1088  *
1089  * Start a streaming operation on @bs.  Clusters that are unallocated
1090  * in @bs, but allocated in any image between @base and @bs (both
1091  * exclusive) will be written to @bs.  At the end of a successful
1092  * streaming job, the backing file of @bs will be changed to
1093  * @backing_file_str in the written image and to @base in the live
1094  * BlockDriverState.
1095  */
1096 void stream_start(const char *job_id, BlockDriverState *bs,
1097                   BlockDriverState *base, const char *backing_file_str,
1098                   int creation_flags, int64_t speed,
1099                   BlockdevOnError on_error, Error **errp);
1100 
1101 /**
1102  * commit_start:
1103  * @job_id: The id of the newly-created job, or %NULL to use the
1104  * device name of @bs.
1105  * @bs: Active block device.
1106  * @top: Top block device to be committed.
1107  * @base: Block device that will be written into, and become the new top.
1108  * @creation_flags: Flags that control the behavior of the Job lifetime.
1109  *                  See @BlockJobCreateFlags
1110  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1111  * @on_error: The action to take upon error.
1112  * @backing_file_str: String to use as the backing file in @top's overlay
1113  * @filter_node_name: The node name that should be assigned to the filter
1114  * driver that the commit job inserts into the graph above @top. NULL means
1115  * that a node name should be autogenerated.
1116  * @errp: Error object.
1117  *
1118  */
1119 void commit_start(const char *job_id, BlockDriverState *bs,
1120                   BlockDriverState *base, BlockDriverState *top,
1121                   int creation_flags, int64_t speed,
1122                   BlockdevOnError on_error, const char *backing_file_str,
1123                   const char *filter_node_name, Error **errp);
1124 /**
1125  * commit_active_start:
1126  * @job_id: The id of the newly-created job, or %NULL to use the
1127  * device name of @bs.
1128  * @bs: Active block device to be committed.
1129  * @base: Block device that will be written into, and become the new top.
1130  * @creation_flags: Flags that control the behavior of the Job lifetime.
1131  *                  See @BlockJobCreateFlags
1132  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1133  * @on_error: The action to take upon error.
1134  * @filter_node_name: The node name that should be assigned to the filter
1135  * driver that the commit job inserts into the graph above @bs. NULL means that
1136  * a node name should be autogenerated.
1137  * @cb: Completion function for the job.
1138  * @opaque: Opaque pointer value passed to @cb.
1139  * @auto_complete: Auto complete the job.
1140  * @errp: Error object.
1141  *
1142  */
1143 BlockJob *commit_active_start(const char *job_id, BlockDriverState *bs,
1144                               BlockDriverState *base, int creation_flags,
1145                               int64_t speed, BlockdevOnError on_error,
1146                               const char *filter_node_name,
1147                               BlockCompletionFunc *cb, void *opaque,
1148                               bool auto_complete, Error **errp);
1149 /*
1150  * mirror_start:
1151  * @job_id: The id of the newly-created job, or %NULL to use the
1152  * device name of @bs.
1153  * @bs: Block device to operate on.
1154  * @target: Block device to write to.
1155  * @replaces: Block graph node name to replace once the mirror is done. Can
1156  *            only be used when full mirroring is selected.
1157  * @creation_flags: Flags that control the behavior of the Job lifetime.
1158  *                  See @BlockJobCreateFlags
1159  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1160  * @granularity: The chosen granularity for the dirty bitmap.
1161  * @buf_size: The amount of data that can be in flight at one time.
1162  * @mode: Whether to collapse all images in the chain to the target.
1163  * @backing_mode: How to establish the target's backing chain after completion.
1164  * @zero_target: Whether the target should be explicitly zero-initialized
1165  * @on_source_error: The action to take upon error reading from the source.
1166  * @on_target_error: The action to take upon error writing to the target.
1167  * @unmap: Whether to unmap target where source sectors only contain zeroes.
1168  * @filter_node_name: The node name that should be assigned to the filter
1169  * driver that the mirror job inserts into the graph above @bs. NULL means that
1170  * a node name should be autogenerated.
1171  * @copy_mode: When to trigger writes to the target.
1172  * @errp: Error object.
1173  *
1174  * Start a mirroring operation on @bs.  Clusters that are allocated
1175  * in @bs will be written to @target until the job is cancelled or
1176  * manually completed.  At the end of a successful mirroring job,
1177  * @bs will be switched to read from @target.
1178  */
1179 void mirror_start(const char *job_id, BlockDriverState *bs,
1180                   BlockDriverState *target, const char *replaces,
1181                   int creation_flags, int64_t speed,
1182                   uint32_t granularity, int64_t buf_size,
1183                   MirrorSyncMode mode, BlockMirrorBackingMode backing_mode,
1184                   bool zero_target,
1185                   BlockdevOnError on_source_error,
1186                   BlockdevOnError on_target_error,
1187                   bool unmap, const char *filter_node_name,
1188                   MirrorCopyMode copy_mode, Error **errp);
1189 
1190 /*
1191  * backup_job_create:
1192  * @job_id: The id of the newly-created job, or %NULL to use the
1193  * device name of @bs.
1194  * @bs: Block device to operate on.
1195  * @target: Block device to write to.
1196  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1197  * @sync_mode: What parts of the disk image should be copied to the destination.
1198  * @sync_bitmap: The dirty bitmap if sync_mode is 'bitmap' or 'incremental'
1199  * @bitmap_mode: The bitmap synchronization policy to use.
1200  * @on_source_error: The action to take upon error reading from the source.
1201  * @on_target_error: The action to take upon error writing to the target.
1202  * @creation_flags: Flags that control the behavior of the Job lifetime.
1203  *                  See @BlockJobCreateFlags
1204  * @cb: Completion function for the job.
1205  * @opaque: Opaque pointer value passed to @cb.
1206  * @txn: Transaction that this job is part of (may be NULL).
1207  *
1208  * Create a backup operation on @bs.  Clusters in @bs are written to @target
1209  * until the job is cancelled or manually completed.
1210  */
1211 BlockJob *backup_job_create(const char *job_id, BlockDriverState *bs,
1212                             BlockDriverState *target, int64_t speed,
1213                             MirrorSyncMode sync_mode,
1214                             BdrvDirtyBitmap *sync_bitmap,
1215                             BitmapSyncMode bitmap_mode,
1216                             bool compress,
1217                             const char *filter_node_name,
1218                             BlockdevOnError on_source_error,
1219                             BlockdevOnError on_target_error,
1220                             int creation_flags,
1221                             BlockCompletionFunc *cb, void *opaque,
1222                             JobTxn *txn, Error **errp);
1223 
1224 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
1225                                   const char *child_name,
1226                                   const BdrvChildRole *child_role,
1227                                   AioContext *ctx,
1228                                   uint64_t perm, uint64_t shared_perm,
1229                                   void *opaque, Error **errp);
1230 void bdrv_root_unref_child(BdrvChild *child);
1231 
1232 void bdrv_get_cumulative_perm(BlockDriverState *bs, uint64_t *perm,
1233                               uint64_t *shared_perm);
1234 
1235 /**
1236  * Sets a BdrvChild's permissions.  Avoid if the parent is a BDS; use
1237  * bdrv_child_refresh_perms() instead and make the parent's
1238  * .bdrv_child_perm() implementation return the correct values.
1239  */
1240 int bdrv_child_try_set_perm(BdrvChild *c, uint64_t perm, uint64_t shared,
1241                             Error **errp);
1242 
1243 /**
1244  * Calls bs->drv->bdrv_child_perm() and updates the child's permission
1245  * masks with the result.
1246  * Drivers should invoke this function whenever an event occurs that
1247  * makes their .bdrv_child_perm() implementation return different
1248  * values than before, but which will not result in the block layer
1249  * automatically refreshing the permissions.
1250  */
1251 int bdrv_child_refresh_perms(BlockDriverState *bs, BdrvChild *c, Error **errp);
1252 
1253 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1254  * block filters: Forward CONSISTENT_READ, WRITE, WRITE_UNCHANGED and RESIZE to
1255  * all children */
1256 void bdrv_filter_default_perms(BlockDriverState *bs, BdrvChild *c,
1257                                const BdrvChildRole *role,
1258                                BlockReopenQueue *reopen_queue,
1259                                uint64_t perm, uint64_t shared,
1260                                uint64_t *nperm, uint64_t *nshared);
1261 
1262 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1263  * (non-raw) image formats: Like above for bs->backing, but for bs->file it
1264  * requires WRITE | RESIZE for read-write images, always requires
1265  * CONSISTENT_READ and doesn't share WRITE. */
1266 void bdrv_format_default_perms(BlockDriverState *bs, BdrvChild *c,
1267                                const BdrvChildRole *role,
1268                                BlockReopenQueue *reopen_queue,
1269                                uint64_t perm, uint64_t shared,
1270                                uint64_t *nperm, uint64_t *nshared);
1271 
1272 bool bdrv_recurse_can_replace(BlockDriverState *bs,
1273                               BlockDriverState *to_replace);
1274 
1275 /*
1276  * Default implementation for drivers to pass bdrv_co_block_status() to
1277  * their file.
1278  */
1279 int coroutine_fn bdrv_co_block_status_from_file(BlockDriverState *bs,
1280                                                 bool want_zero,
1281                                                 int64_t offset,
1282                                                 int64_t bytes,
1283                                                 int64_t *pnum,
1284                                                 int64_t *map,
1285                                                 BlockDriverState **file);
1286 /*
1287  * Default implementation for drivers to pass bdrv_co_block_status() to
1288  * their backing file.
1289  */
1290 int coroutine_fn bdrv_co_block_status_from_backing(BlockDriverState *bs,
1291                                                    bool want_zero,
1292                                                    int64_t offset,
1293                                                    int64_t bytes,
1294                                                    int64_t *pnum,
1295                                                    int64_t *map,
1296                                                    BlockDriverState **file);
1297 const char *bdrv_get_parent_name(const BlockDriverState *bs);
1298 void blk_dev_change_media_cb(BlockBackend *blk, bool load, Error **errp);
1299 bool blk_dev_has_removable_media(BlockBackend *blk);
1300 bool blk_dev_has_tray(BlockBackend *blk);
1301 void blk_dev_eject_request(BlockBackend *blk, bool force);
1302 bool blk_dev_is_tray_open(BlockBackend *blk);
1303 bool blk_dev_is_medium_locked(BlockBackend *blk);
1304 
1305 void bdrv_set_dirty(BlockDriverState *bs, int64_t offset, int64_t bytes);
1306 
1307 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap **out);
1308 void bdrv_restore_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap *backup);
1309 bool bdrv_dirty_bitmap_merge_internal(BdrvDirtyBitmap *dest,
1310                                       const BdrvDirtyBitmap *src,
1311                                       HBitmap **backup, bool lock);
1312 
1313 void bdrv_inc_in_flight(BlockDriverState *bs);
1314 void bdrv_dec_in_flight(BlockDriverState *bs);
1315 
1316 void blockdev_close_all_bdrv_states(void);
1317 
1318 int coroutine_fn bdrv_co_copy_range_from(BdrvChild *src, uint64_t src_offset,
1319                                          BdrvChild *dst, uint64_t dst_offset,
1320                                          uint64_t bytes,
1321                                          BdrvRequestFlags read_flags,
1322                                          BdrvRequestFlags write_flags);
1323 int coroutine_fn bdrv_co_copy_range_to(BdrvChild *src, uint64_t src_offset,
1324                                        BdrvChild *dst, uint64_t dst_offset,
1325                                        uint64_t bytes,
1326                                        BdrvRequestFlags read_flags,
1327                                        BdrvRequestFlags write_flags);
1328 
1329 int refresh_total_sectors(BlockDriverState *bs, int64_t hint);
1330 
1331 void bdrv_set_monitor_owned(BlockDriverState *bs);
1332 BlockDriverState *bds_tree_init(QDict *bs_opts, Error **errp);
1333 
1334 /**
1335  * Simple implementation of bdrv_co_create_opts for protocol drivers
1336  * which only support creation via opening a file
1337  * (usually existing raw storage device)
1338  */
1339 int coroutine_fn bdrv_co_create_opts_simple(BlockDriver *drv,
1340                                             const char *filename,
1341                                             QemuOpts *opts,
1342                                             Error **errp);
1343 extern QemuOptsList bdrv_create_opts_simple;
1344 
1345 #endif /* BLOCK_INT_H */
1346