xref: /openbmc/qemu/include/block/block_int.h (revision 2e5b09fd)
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     /* for snapshots block filter like Quorum can implement the
98      * following recursive callback.
99      * It's purpose is to recurse on the filter children while calling
100      * bdrv_recurse_is_first_non_filter on them.
101      * For a sample implementation look in the future Quorum block filter.
102      */
103     bool (*bdrv_recurse_is_first_non_filter)(BlockDriverState *bs,
104                                              BlockDriverState *candidate);
105 
106     int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename);
107     int (*bdrv_probe_device)(const char *filename);
108 
109     /* Any driver implementing this callback is expected to be able to handle
110      * NULL file names in its .bdrv_open() implementation */
111     void (*bdrv_parse_filename)(const char *filename, QDict *options, Error **errp);
112     /* Drivers not implementing bdrv_parse_filename nor bdrv_open should have
113      * this field set to true, except ones that are defined only by their
114      * child's bs.
115      * An example of the last type will be the quorum block driver.
116      */
117     bool bdrv_needs_filename;
118 
119     /* Set if a driver can support backing files */
120     bool supports_backing;
121 
122     /* For handling image reopen for split or non-split files */
123     int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state,
124                                BlockReopenQueue *queue, Error **errp);
125     void (*bdrv_reopen_commit)(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)(const char *filename,
139                                             QemuOpts *opts,
140                                             Error **errp);
141     int (*bdrv_make_empty)(BlockDriverState *bs);
142 
143     /*
144      * Refreshes the bs->exact_filename field. If that is impossible,
145      * bs->exact_filename has to be left empty.
146      */
147     void (*bdrv_refresh_filename)(BlockDriverState *bs);
148 
149     /*
150      * Gathers the open options for all children into @target.
151      * A simple format driver (without backing file support) might
152      * implement this function like this:
153      *
154      *     QINCREF(bs->file->bs->full_open_options);
155      *     qdict_put(target, "file", bs->file->bs->full_open_options);
156      *
157      * If not specified, the generic implementation will simply put
158      * all children's options under their respective name.
159      *
160      * @backing_overridden is true when bs->backing seems not to be
161      * the child that would result from opening bs->backing_file.
162      * Therefore, if it is true, the backing child's options should be
163      * gathered; otherwise, there is no need since the backing child
164      * is the one implied by the image header.
165      *
166      * Note that ideally this function would not be needed.  Every
167      * block driver which implements it is probably doing something
168      * shady regarding its runtime option structure.
169      */
170     void (*bdrv_gather_child_options)(BlockDriverState *bs, QDict *target,
171                                       bool backing_overridden);
172 
173     /*
174      * Returns an allocated string which is the directory name of this BDS: It
175      * will be used to make relative filenames absolute by prepending this
176      * function's return value to them.
177      */
178     char *(*bdrv_dirname)(BlockDriverState *bs, Error **errp);
179 
180     /* aio */
181     BlockAIOCB *(*bdrv_aio_preadv)(BlockDriverState *bs,
182         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags,
183         BlockCompletionFunc *cb, void *opaque);
184     BlockAIOCB *(*bdrv_aio_pwritev)(BlockDriverState *bs,
185         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags,
186         BlockCompletionFunc *cb, void *opaque);
187     BlockAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs,
188         BlockCompletionFunc *cb, void *opaque);
189     BlockAIOCB *(*bdrv_aio_pdiscard)(BlockDriverState *bs,
190         int64_t offset, int bytes,
191         BlockCompletionFunc *cb, void *opaque);
192 
193     int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs,
194         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
195 
196     /**
197      * @offset: position in bytes to read at
198      * @bytes: number of bytes to read
199      * @qiov: the buffers to fill with read data
200      * @flags: currently unused, always 0
201      *
202      * @offset and @bytes will be a multiple of 'request_alignment',
203      * but the length of individual @qiov elements does not have to
204      * be a multiple.
205      *
206      * @bytes will always equal the total size of @qiov, and will be
207      * no larger than 'max_transfer'.
208      *
209      * The buffer in @qiov may point directly to guest memory.
210      */
211     int coroutine_fn (*bdrv_co_preadv)(BlockDriverState *bs,
212         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags);
213     int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs,
214         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, int flags);
215     /**
216      * @offset: position in bytes to write at
217      * @bytes: number of bytes to write
218      * @qiov: the buffers containing data to write
219      * @flags: zero or more bits allowed by 'supported_write_flags'
220      *
221      * @offset and @bytes will be a multiple of 'request_alignment',
222      * but the length of individual @qiov elements does not have to
223      * be a multiple.
224      *
225      * @bytes will always equal the total size of @qiov, and will be
226      * no larger than 'max_transfer'.
227      *
228      * The buffer in @qiov may point directly to guest memory.
229      */
230     int coroutine_fn (*bdrv_co_pwritev)(BlockDriverState *bs,
231         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags);
232 
233     /*
234      * Efficiently zero a region of the disk image.  Typically an image format
235      * would use a compact metadata representation to implement this.  This
236      * function pointer may be NULL or return -ENOSUP and .bdrv_co_writev()
237      * will be called instead.
238      */
239     int coroutine_fn (*bdrv_co_pwrite_zeroes)(BlockDriverState *bs,
240         int64_t offset, int bytes, BdrvRequestFlags flags);
241     int coroutine_fn (*bdrv_co_pdiscard)(BlockDriverState *bs,
242         int64_t offset, int bytes);
243 
244     /* Map [offset, offset + nbytes) range onto a child of @bs to copy from,
245      * and invoke bdrv_co_copy_range_from(child, ...), or invoke
246      * bdrv_co_copy_range_to() if @bs is the leaf child to copy data from.
247      *
248      * See the comment of bdrv_co_copy_range for the parameter and return value
249      * semantics.
250      */
251     int coroutine_fn (*bdrv_co_copy_range_from)(BlockDriverState *bs,
252                                                 BdrvChild *src,
253                                                 uint64_t offset,
254                                                 BdrvChild *dst,
255                                                 uint64_t dst_offset,
256                                                 uint64_t bytes,
257                                                 BdrvRequestFlags read_flags,
258                                                 BdrvRequestFlags write_flags);
259 
260     /* Map [offset, offset + nbytes) range onto a child of bs to copy data to,
261      * and invoke bdrv_co_copy_range_to(child, src, ...), or perform the copy
262      * operation if @bs is the leaf and @src has the same BlockDriver.  Return
263      * -ENOTSUP if @bs is the leaf but @src has a different BlockDriver.
264      *
265      * See the comment of bdrv_co_copy_range for the parameter and return value
266      * semantics.
267      */
268     int coroutine_fn (*bdrv_co_copy_range_to)(BlockDriverState *bs,
269                                               BdrvChild *src,
270                                               uint64_t src_offset,
271                                               BdrvChild *dst,
272                                               uint64_t dst_offset,
273                                               uint64_t bytes,
274                                               BdrvRequestFlags read_flags,
275                                               BdrvRequestFlags write_flags);
276 
277     /*
278      * Building block for bdrv_block_status[_above] and
279      * bdrv_is_allocated[_above].  The driver should answer only
280      * according to the current layer, and should only need to set
281      * BDRV_BLOCK_DATA, BDRV_BLOCK_ZERO, BDRV_BLOCK_OFFSET_VALID,
282      * and/or BDRV_BLOCK_RAW; if the current layer defers to a backing
283      * layer, the result should be 0 (and not BDRV_BLOCK_ZERO).  See
284      * block.h for the overall meaning of the bits.  As a hint, the
285      * flag want_zero is true if the caller cares more about precise
286      * mappings (favor accurate _OFFSET_VALID/_ZERO) or false for
287      * overall allocation (favor larger *pnum, perhaps by reporting
288      * _DATA instead of _ZERO).  The block layer guarantees input
289      * clamped to bdrv_getlength() and aligned to request_alignment,
290      * as well as non-NULL pnum, map, and file; in turn, the driver
291      * must return an error or set pnum to an aligned non-zero value.
292      */
293     int coroutine_fn (*bdrv_co_block_status)(BlockDriverState *bs,
294         bool want_zero, int64_t offset, int64_t bytes, int64_t *pnum,
295         int64_t *map, BlockDriverState **file);
296 
297     /*
298      * Invalidate any cached meta-data.
299      */
300     void coroutine_fn (*bdrv_co_invalidate_cache)(BlockDriverState *bs,
301                                                   Error **errp);
302     int (*bdrv_inactivate)(BlockDriverState *bs);
303 
304     /*
305      * Flushes all data for all layers by calling bdrv_co_flush for underlying
306      * layers, if needed. This function is needed for deterministic
307      * synchronization of the flush finishing callback.
308      */
309     int coroutine_fn (*bdrv_co_flush)(BlockDriverState *bs);
310 
311     /*
312      * Flushes all data that was already written to the OS all the way down to
313      * the disk (for example file-posix.c calls fsync()).
314      */
315     int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs);
316 
317     /*
318      * Flushes all internal caches to the OS. The data may still sit in a
319      * writeback cache of the host OS, but it will survive a crash of the qemu
320      * process.
321      */
322     int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs);
323 
324     /*
325      * Drivers setting this field must be able to work with just a plain
326      * filename with '<protocol_name>:' as a prefix, and no other options.
327      * Options may be extracted from the filename by implementing
328      * bdrv_parse_filename.
329      */
330     const char *protocol_name;
331     int coroutine_fn (*bdrv_co_truncate)(BlockDriverState *bs, int64_t offset,
332                                          PreallocMode prealloc, Error **errp);
333 
334     int64_t (*bdrv_getlength)(BlockDriverState *bs);
335     bool has_variable_length;
336     int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
337     BlockMeasureInfo *(*bdrv_measure)(QemuOpts *opts, BlockDriverState *in_bs,
338                                       Error **errp);
339 
340     int coroutine_fn (*bdrv_co_pwritev_compressed)(BlockDriverState *bs,
341         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov);
342 
343     int (*bdrv_snapshot_create)(BlockDriverState *bs,
344                                 QEMUSnapshotInfo *sn_info);
345     int (*bdrv_snapshot_goto)(BlockDriverState *bs,
346                               const char *snapshot_id);
347     int (*bdrv_snapshot_delete)(BlockDriverState *bs,
348                                 const char *snapshot_id,
349                                 const char *name,
350                                 Error **errp);
351     int (*bdrv_snapshot_list)(BlockDriverState *bs,
352                               QEMUSnapshotInfo **psn_info);
353     int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
354                                   const char *snapshot_id,
355                                   const char *name,
356                                   Error **errp);
357     int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
358     ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs,
359                                                  Error **errp);
360 
361     int coroutine_fn (*bdrv_save_vmstate)(BlockDriverState *bs,
362                                           QEMUIOVector *qiov,
363                                           int64_t pos);
364     int coroutine_fn (*bdrv_load_vmstate)(BlockDriverState *bs,
365                                           QEMUIOVector *qiov,
366                                           int64_t pos);
367 
368     int (*bdrv_change_backing_file)(BlockDriverState *bs,
369         const char *backing_file, const char *backing_fmt);
370 
371     /* removable device specific */
372     bool (*bdrv_is_inserted)(BlockDriverState *bs);
373     void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
374     void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
375 
376     /* to control generic scsi devices */
377     BlockAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
378         unsigned long int req, void *buf,
379         BlockCompletionFunc *cb, void *opaque);
380     int coroutine_fn (*bdrv_co_ioctl)(BlockDriverState *bs,
381                                       unsigned long int req, void *buf);
382 
383     /* List of options for creating images, terminated by name == NULL */
384     QemuOptsList *create_opts;
385     /*
386      * If this driver supports reopening images this contains a
387      * NULL-terminated list of the runtime options that can be
388      * modified. If an option in this list is unspecified during
389      * reopen then it _must_ be reset to its default value or return
390      * an error.
391      */
392     const char *const *mutable_opts;
393 
394     /*
395      * Returns 0 for completed check, -errno for internal errors.
396      * The check results are stored in result.
397      */
398     int coroutine_fn (*bdrv_co_check)(BlockDriverState *bs,
399                                       BdrvCheckResult *result,
400                                       BdrvCheckMode fix);
401 
402     int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts,
403                               BlockDriverAmendStatusCB *status_cb,
404                               void *cb_opaque,
405                               Error **errp);
406 
407     void (*bdrv_debug_event)(BlockDriverState *bs, BlkdebugEvent event);
408 
409     /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
410     int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
411         const char *tag);
412     int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs,
413         const char *tag);
414     int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
415     bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
416 
417     void (*bdrv_refresh_limits)(BlockDriverState *bs, Error **errp);
418 
419     /*
420      * Returns 1 if newly created images are guaranteed to contain only
421      * zeros, 0 otherwise.
422      * Must return 0 if .bdrv_has_zero_init_truncate() returns 0.
423      */
424     int (*bdrv_has_zero_init)(BlockDriverState *bs);
425 
426     /*
427      * Returns 1 if new areas added by growing the image with
428      * PREALLOC_MODE_OFF contain only zeros, 0 otherwise.
429      */
430     int (*bdrv_has_zero_init_truncate)(BlockDriverState *bs);
431 
432     /* Remove fd handlers, timers, and other event loop callbacks so the event
433      * loop is no longer in use.  Called with no in-flight requests and in
434      * depth-first traversal order with parents before child nodes.
435      */
436     void (*bdrv_detach_aio_context)(BlockDriverState *bs);
437 
438     /* Add fd handlers, timers, and other event loop callbacks so I/O requests
439      * can be processed again.  Called with no in-flight requests and in
440      * depth-first traversal order with child nodes before parent nodes.
441      */
442     void (*bdrv_attach_aio_context)(BlockDriverState *bs,
443                                     AioContext *new_context);
444 
445     /* io queue for linux-aio */
446     void (*bdrv_io_plug)(BlockDriverState *bs);
447     void (*bdrv_io_unplug)(BlockDriverState *bs);
448 
449     /**
450      * Try to get @bs's logical and physical block size.
451      * On success, store them in @bsz and return zero.
452      * On failure, return negative errno.
453      */
454     int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz);
455     /**
456      * Try to get @bs's geometry (cyls, heads, sectors)
457      * On success, store them in @geo and return 0.
458      * On failure return -errno.
459      * Only drivers that want to override guest geometry implement this
460      * callback; see hd_geometry_guess().
461      */
462     int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo);
463 
464     /**
465      * bdrv_co_drain_begin is called if implemented in the beginning of a
466      * drain operation to drain and stop any internal sources of requests in
467      * the driver.
468      * bdrv_co_drain_end is called if implemented at the end of the drain.
469      *
470      * They should be used by the driver to e.g. manage scheduled I/O
471      * requests, or toggle an internal state. After the end of the drain new
472      * requests will continue normally.
473      */
474     void coroutine_fn (*bdrv_co_drain_begin)(BlockDriverState *bs);
475     void coroutine_fn (*bdrv_co_drain_end)(BlockDriverState *bs);
476 
477     void (*bdrv_add_child)(BlockDriverState *parent, BlockDriverState *child,
478                            Error **errp);
479     void (*bdrv_del_child)(BlockDriverState *parent, BdrvChild *child,
480                            Error **errp);
481 
482     /**
483      * Informs the block driver that a permission change is intended. The
484      * driver checks whether the change is permissible and may take other
485      * preparations for the change (e.g. get file system locks). This operation
486      * is always followed either by a call to either .bdrv_set_perm or
487      * .bdrv_abort_perm_update.
488      *
489      * Checks whether the requested set of cumulative permissions in @perm
490      * can be granted for accessing @bs and whether no other users are using
491      * permissions other than those given in @shared (both arguments take
492      * BLK_PERM_* bitmasks).
493      *
494      * If both conditions are met, 0 is returned. Otherwise, -errno is returned
495      * and errp is set to an error describing the conflict.
496      */
497     int (*bdrv_check_perm)(BlockDriverState *bs, uint64_t perm,
498                            uint64_t shared, Error **errp);
499 
500     /**
501      * Called to inform the driver that the set of cumulative set of used
502      * permissions for @bs has changed to @perm, and the set of sharable
503      * permission to @shared. The driver can use this to propagate changes to
504      * its children (i.e. request permissions only if a parent actually needs
505      * them).
506      *
507      * This function is only invoked after bdrv_check_perm(), so block drivers
508      * may rely on preparations made in their .bdrv_check_perm implementation.
509      */
510     void (*bdrv_set_perm)(BlockDriverState *bs, uint64_t perm, uint64_t shared);
511 
512     /*
513      * Called to inform the driver that after a previous bdrv_check_perm()
514      * call, the permission update is not performed and any preparations made
515      * for it (e.g. taken file locks) need to be undone.
516      *
517      * This function can be called even for nodes that never saw a
518      * bdrv_check_perm() call. It is a no-op then.
519      */
520     void (*bdrv_abort_perm_update)(BlockDriverState *bs);
521 
522     /**
523      * Returns in @nperm and @nshared the permissions that the driver for @bs
524      * needs on its child @c, based on the cumulative permissions requested by
525      * the parents in @parent_perm and @parent_shared.
526      *
527      * If @c is NULL, return the permissions for attaching a new child for the
528      * given @role.
529      *
530      * If @reopen_queue is non-NULL, don't return the currently needed
531      * permissions, but those that will be needed after applying the
532      * @reopen_queue.
533      */
534      void (*bdrv_child_perm)(BlockDriverState *bs, BdrvChild *c,
535                              const BdrvChildRole *role,
536                              BlockReopenQueue *reopen_queue,
537                              uint64_t parent_perm, uint64_t parent_shared,
538                              uint64_t *nperm, uint64_t *nshared);
539 
540     /**
541      * Bitmaps should be marked as 'IN_USE' in the image on reopening image
542      * as rw. This handler should realize it. It also should unset readonly
543      * field of BlockDirtyBitmap's in case of success.
544      */
545     int (*bdrv_reopen_bitmaps_rw)(BlockDriverState *bs, Error **errp);
546     bool (*bdrv_can_store_new_dirty_bitmap)(BlockDriverState *bs,
547                                             const char *name,
548                                             uint32_t granularity,
549                                             Error **errp);
550     void (*bdrv_remove_persistent_dirty_bitmap)(BlockDriverState *bs,
551                                                 const char *name,
552                                                 Error **errp);
553 
554     /**
555      * Register/unregister a buffer for I/O. For example, when the driver is
556      * interested to know the memory areas that will later be used in iovs, so
557      * that it can do IOMMU mapping with VFIO etc., in order to get better
558      * performance. In the case of VFIO drivers, this callback is used to do
559      * DMA mapping for hot buffers.
560      */
561     void (*bdrv_register_buf)(BlockDriverState *bs, void *host, size_t size);
562     void (*bdrv_unregister_buf)(BlockDriverState *bs, void *host);
563     QLIST_ENTRY(BlockDriver) list;
564 
565     /* Pointer to a NULL-terminated array of names of strong options
566      * that can be specified for bdrv_open(). A strong option is one
567      * that changes the data of a BDS.
568      * If this pointer is NULL, the array is considered empty.
569      * "filename" and "driver" are always considered strong. */
570     const char *const *strong_runtime_opts;
571 };
572 
573 typedef struct BlockLimits {
574     /* Alignment requirement, in bytes, for offset/length of I/O
575      * requests. Must be a power of 2 less than INT_MAX; defaults to
576      * 1 for drivers with modern byte interfaces, and to 512
577      * otherwise. */
578     uint32_t request_alignment;
579 
580     /* Maximum number of bytes that can be discarded at once (since it
581      * is signed, it must be < 2G, if set). Must be multiple of
582      * pdiscard_alignment, but need not be power of 2. May be 0 if no
583      * inherent 32-bit limit */
584     int32_t max_pdiscard;
585 
586     /* Optimal alignment for discard requests in bytes. A power of 2
587      * is best but not mandatory.  Must be a multiple of
588      * bl.request_alignment, and must be less than max_pdiscard if
589      * that is set. May be 0 if bl.request_alignment is good enough */
590     uint32_t pdiscard_alignment;
591 
592     /* Maximum number of bytes that can zeroized at once (since it is
593      * signed, it must be < 2G, if set). Must be multiple of
594      * pwrite_zeroes_alignment. May be 0 if no inherent 32-bit limit */
595     int32_t max_pwrite_zeroes;
596 
597     /* Optimal alignment for write zeroes requests in bytes. A power
598      * of 2 is best but not mandatory.  Must be a multiple of
599      * bl.request_alignment, and must be less than max_pwrite_zeroes
600      * if that is set. May be 0 if bl.request_alignment is good
601      * enough */
602     uint32_t pwrite_zeroes_alignment;
603 
604     /* Optimal transfer length in bytes.  A power of 2 is best but not
605      * mandatory.  Must be a multiple of bl.request_alignment, or 0 if
606      * no preferred size */
607     uint32_t opt_transfer;
608 
609     /* Maximal transfer length in bytes.  Need not be power of 2, but
610      * must be multiple of opt_transfer and bl.request_alignment, or 0
611      * for no 32-bit limit.  For now, anything larger than INT_MAX is
612      * clamped down. */
613     uint32_t max_transfer;
614 
615     /* memory alignment, in bytes so that no bounce buffer is needed */
616     size_t min_mem_alignment;
617 
618     /* memory alignment, in bytes, for bounce buffer */
619     size_t opt_mem_alignment;
620 
621     /* maximum number of iovec elements */
622     int max_iov;
623 } BlockLimits;
624 
625 typedef struct BdrvOpBlocker BdrvOpBlocker;
626 
627 typedef struct BdrvAioNotifier {
628     void (*attached_aio_context)(AioContext *new_context, void *opaque);
629     void (*detach_aio_context)(void *opaque);
630 
631     void *opaque;
632     bool deleted;
633 
634     QLIST_ENTRY(BdrvAioNotifier) list;
635 } BdrvAioNotifier;
636 
637 struct BdrvChildRole {
638     /* If true, bdrv_replace_node() doesn't change the node this BdrvChild
639      * points to. */
640     bool stay_at_node;
641 
642     /* If true, the parent is a BlockDriverState and bdrv_next_all_states()
643      * will return it. This information is used for drain_all, where every node
644      * will be drained separately, so the drain only needs to be propagated to
645      * non-BDS parents. */
646     bool parent_is_bds;
647 
648     void (*inherit_options)(int *child_flags, QDict *child_options,
649                             int parent_flags, QDict *parent_options);
650 
651     void (*change_media)(BdrvChild *child, bool load);
652     void (*resize)(BdrvChild *child);
653 
654     /* Returns a name that is supposedly more useful for human users than the
655      * node name for identifying the node in question (in particular, a BB
656      * name), or NULL if the parent can't provide a better name. */
657     const char *(*get_name)(BdrvChild *child);
658 
659     /* Returns a malloced string that describes the parent of the child for a
660      * human reader. This could be a node-name, BlockBackend name, qdev ID or
661      * QOM path of the device owning the BlockBackend, job type and ID etc. The
662      * caller is responsible for freeing the memory. */
663     char *(*get_parent_desc)(BdrvChild *child);
664 
665     /*
666      * If this pair of functions is implemented, the parent doesn't issue new
667      * requests after returning from .drained_begin() until .drained_end() is
668      * called.
669      *
670      * These functions must not change the graph (and therefore also must not
671      * call aio_poll(), which could change the graph indirectly).
672      *
673      * If drained_end() schedules background operations, it must atomically
674      * increment *drained_end_counter for each such operation and atomically
675      * decrement it once the operation has settled.
676      *
677      * Note that this can be nested. If drained_begin() was called twice, new
678      * I/O is allowed only after drained_end() was called twice, too.
679      */
680     void (*drained_begin)(BdrvChild *child);
681     void (*drained_end)(BdrvChild *child, int *drained_end_counter);
682 
683     /*
684      * Returns whether the parent has pending requests for the child. This
685      * callback is polled after .drained_begin() has been called until all
686      * activity on the child has stopped.
687      */
688     bool (*drained_poll)(BdrvChild *child);
689 
690     /* Notifies the parent that the child has been activated/inactivated (e.g.
691      * when migration is completing) and it can start/stop requesting
692      * permissions and doing I/O on it. */
693     void (*activate)(BdrvChild *child, Error **errp);
694     int (*inactivate)(BdrvChild *child);
695 
696     void (*attach)(BdrvChild *child);
697     void (*detach)(BdrvChild *child);
698 
699     /* Notifies the parent that the filename of its child has changed (e.g.
700      * because the direct child was removed from the backing chain), so that it
701      * can update its reference. */
702     int (*update_filename)(BdrvChild *child, BlockDriverState *new_base,
703                            const char *filename, Error **errp);
704 
705     bool (*can_set_aio_ctx)(BdrvChild *child, AioContext *ctx,
706                             GSList **ignore, Error **errp);
707     void (*set_aio_ctx)(BdrvChild *child, AioContext *ctx, GSList **ignore);
708 };
709 
710 extern const BdrvChildRole child_file;
711 extern const BdrvChildRole child_format;
712 extern const BdrvChildRole child_backing;
713 
714 struct BdrvChild {
715     BlockDriverState *bs;
716     char *name;
717     const BdrvChildRole *role;
718     void *opaque;
719 
720     /**
721      * Granted permissions for operating on this BdrvChild (BLK_PERM_* bitmask)
722      */
723     uint64_t perm;
724 
725     /**
726      * Permissions that can still be granted to other users of @bs while this
727      * BdrvChild is still attached to it. (BLK_PERM_* bitmask)
728      */
729     uint64_t shared_perm;
730 
731     /* backup of permissions during permission update procedure */
732     bool has_backup_perm;
733     uint64_t backup_perm;
734     uint64_t backup_shared_perm;
735 
736     /*
737      * This link is frozen: the child can neither be replaced nor
738      * detached from the parent.
739      */
740     bool frozen;
741 
742     /*
743      * How many times the parent of this child has been drained
744      * (through role->drained_*).
745      * Usually, this is equal to bs->quiesce_counter (potentially
746      * reduced by bdrv_drain_all_count).  It may differ while the
747      * child is entering or leaving a drained section.
748      */
749     int parent_quiesce_counter;
750 
751     QLIST_ENTRY(BdrvChild) next;
752     QLIST_ENTRY(BdrvChild) next_parent;
753 };
754 
755 /*
756  * Note: the function bdrv_append() copies and swaps contents of
757  * BlockDriverStates, so if you add new fields to this struct, please
758  * inspect bdrv_append() to determine if the new fields need to be
759  * copied as well.
760  */
761 struct BlockDriverState {
762     /* Protected by big QEMU lock or read-only after opening.  No special
763      * locking needed during I/O...
764      */
765     int open_flags; /* flags used to open the file, re-used for re-open */
766     bool read_only; /* if true, the media is read only */
767     bool encrypted; /* if true, the media is encrypted */
768     bool sg;        /* if true, the device is a /dev/sg* */
769     bool probed;    /* if true, format was probed rather than specified */
770     bool force_share; /* if true, always allow all shared permissions */
771     bool implicit;  /* if true, this filter node was automatically inserted */
772 
773     BlockDriver *drv; /* NULL means no media */
774     void *opaque;
775 
776     AioContext *aio_context; /* event loop used for fd handlers, timers, etc */
777     /* long-running tasks intended to always use the same AioContext as this
778      * BDS may register themselves in this list to be notified of changes
779      * regarding this BDS's context */
780     QLIST_HEAD(, BdrvAioNotifier) aio_notifiers;
781     bool walking_aio_notifiers; /* to make removal during iteration safe */
782 
783     char filename[PATH_MAX];
784     char backing_file[PATH_MAX]; /* if non zero, the image is a diff of
785                                     this file image */
786     /* The backing filename indicated by the image header; if we ever
787      * open this file, then this is replaced by the resulting BDS's
788      * filename (i.e. after a bdrv_refresh_filename() run). */
789     char auto_backing_file[PATH_MAX];
790     char backing_format[16]; /* if non-zero and backing_file exists */
791 
792     QDict *full_open_options;
793     char exact_filename[PATH_MAX];
794 
795     BdrvChild *backing;
796     BdrvChild *file;
797 
798     /* I/O Limits */
799     BlockLimits bl;
800 
801     /* Flags honored during pwrite (so far: BDRV_REQ_FUA,
802      * BDRV_REQ_WRITE_UNCHANGED).
803      * If a driver does not support BDRV_REQ_WRITE_UNCHANGED, those
804      * writes will be issued as normal writes without the flag set.
805      * This is important to note for drivers that do not explicitly
806      * request a WRITE permission for their children and instead take
807      * the same permissions as their parent did (this is commonly what
808      * block filters do).  Such drivers have to be aware that the
809      * parent may have taken a WRITE_UNCHANGED permission only and is
810      * issuing such requests.  Drivers either must make sure that
811      * these requests do not result in plain WRITE accesses (usually
812      * by supporting BDRV_REQ_WRITE_UNCHANGED, and then forwarding
813      * every incoming write request as-is, including potentially that
814      * flag), or they have to explicitly take the WRITE permission for
815      * their children. */
816     unsigned int supported_write_flags;
817     /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA,
818      * BDRV_REQ_MAY_UNMAP, BDRV_REQ_WRITE_UNCHANGED) */
819     unsigned int supported_zero_flags;
820 
821     /* the following member gives a name to every node on the bs graph. */
822     char node_name[32];
823     /* element of the list of named nodes building the graph */
824     QTAILQ_ENTRY(BlockDriverState) node_list;
825     /* element of the list of all BlockDriverStates (all_bdrv_states) */
826     QTAILQ_ENTRY(BlockDriverState) bs_list;
827     /* element of the list of monitor-owned BDS */
828     QTAILQ_ENTRY(BlockDriverState) monitor_list;
829     int refcnt;
830 
831     /* operation blockers */
832     QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX];
833 
834     /* The node that this node inherited default options from (and a reopen on
835      * which can affect this node by changing these defaults). This is always a
836      * parent node of this node. */
837     BlockDriverState *inherits_from;
838     QLIST_HEAD(, BdrvChild) children;
839     QLIST_HEAD(, BdrvChild) parents;
840 
841     QDict *options;
842     QDict *explicit_options;
843     BlockdevDetectZeroesOptions detect_zeroes;
844 
845     /* The error object in use for blocking operations on backing_hd */
846     Error *backing_blocker;
847 
848     /* Protected by AioContext lock */
849 
850     /* If we are reading a disk image, give its size in sectors.
851      * Generally read-only; it is written to by load_snapshot and
852      * save_snaphost, but the block layer is quiescent during those.
853      */
854     int64_t total_sectors;
855 
856     /* Callback before write request is processed */
857     NotifierWithReturnList before_write_notifiers;
858 
859     /* threshold limit for writes, in bytes. "High water mark". */
860     uint64_t write_threshold_offset;
861     NotifierWithReturn write_threshold_notifier;
862 
863     /* Writing to the list requires the BQL _and_ the dirty_bitmap_mutex.
864      * Reading from the list can be done with either the BQL or the
865      * dirty_bitmap_mutex.  Modifying a bitmap only requires
866      * dirty_bitmap_mutex.  */
867     QemuMutex dirty_bitmap_mutex;
868     QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps;
869 
870     /* Offset after the highest byte written to */
871     Stat64 wr_highest_offset;
872 
873     /* If true, copy read backing sectors into image.  Can be >1 if more
874      * than one client has requested copy-on-read.  Accessed with atomic
875      * ops.
876      */
877     int copy_on_read;
878 
879     /* number of in-flight requests; overall and serialising.
880      * Accessed with atomic ops.
881      */
882     unsigned int in_flight;
883     unsigned int serialising_in_flight;
884 
885     /* counter for nested bdrv_io_plug.
886      * Accessed with atomic ops.
887     */
888     unsigned io_plugged;
889 
890     /* do we need to tell the quest if we have a volatile write cache? */
891     int enable_write_cache;
892 
893     /* Accessed with atomic ops.  */
894     int quiesce_counter;
895     int recursive_quiesce_counter;
896 
897     unsigned int write_gen;               /* Current data generation */
898 
899     /* Protected by reqs_lock.  */
900     CoMutex reqs_lock;
901     QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
902     CoQueue flush_queue;                  /* Serializing flush queue */
903     bool active_flush_req;                /* Flush request in flight? */
904 
905     /* Only read/written by whoever has set active_flush_req to true.  */
906     unsigned int flushed_gen;             /* Flushed write generation */
907 
908     /* BdrvChild links to this node may never be frozen */
909     bool never_freeze;
910 };
911 
912 struct BlockBackendRootState {
913     int open_flags;
914     bool read_only;
915     BlockdevDetectZeroesOptions detect_zeroes;
916 };
917 
918 typedef enum BlockMirrorBackingMode {
919     /* Reuse the existing backing chain from the source for the target.
920      * - sync=full: Set backing BDS to NULL.
921      * - sync=top:  Use source's backing BDS.
922      * - sync=none: Use source as the backing BDS. */
923     MIRROR_SOURCE_BACKING_CHAIN,
924 
925     /* Open the target's backing chain completely anew */
926     MIRROR_OPEN_BACKING_CHAIN,
927 
928     /* Do not change the target's backing BDS after job completion */
929     MIRROR_LEAVE_BACKING_CHAIN,
930 } BlockMirrorBackingMode;
931 
932 static inline BlockDriverState *backing_bs(BlockDriverState *bs)
933 {
934     return bs->backing ? bs->backing->bs : NULL;
935 }
936 
937 
938 /* Essential block drivers which must always be statically linked into qemu, and
939  * which therefore can be accessed without using bdrv_find_format() */
940 extern BlockDriver bdrv_file;
941 extern BlockDriver bdrv_raw;
942 extern BlockDriver bdrv_qcow2;
943 
944 int coroutine_fn bdrv_co_preadv(BdrvChild *child,
945     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
946     BdrvRequestFlags flags);
947 int coroutine_fn bdrv_co_pwritev(BdrvChild *child,
948     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
949     BdrvRequestFlags flags);
950 
951 static inline int coroutine_fn bdrv_co_pread(BdrvChild *child,
952     int64_t offset, unsigned int bytes, void *buf, BdrvRequestFlags flags)
953 {
954     QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes);
955 
956     return bdrv_co_preadv(child, offset, bytes, &qiov, flags);
957 }
958 
959 static inline int coroutine_fn bdrv_co_pwrite(BdrvChild *child,
960     int64_t offset, unsigned int bytes, void *buf, BdrvRequestFlags flags)
961 {
962     QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes);
963 
964     return bdrv_co_pwritev(child, offset, bytes, &qiov, flags);
965 }
966 
967 extern unsigned int bdrv_drain_all_count;
968 void bdrv_apply_subtree_drain(BdrvChild *child, BlockDriverState *new_parent);
969 void bdrv_unapply_subtree_drain(BdrvChild *child, BlockDriverState *old_parent);
970 
971 int get_tmp_filename(char *filename, int size);
972 BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size,
973                             const char *filename);
974 
975 void bdrv_parse_filename_strip_prefix(const char *filename, const char *prefix,
976                                       QDict *options);
977 
978 
979 /**
980  * bdrv_add_before_write_notifier:
981  *
982  * Register a callback that is invoked before write requests are processed but
983  * after any throttling or waiting for overlapping requests.
984  */
985 void bdrv_add_before_write_notifier(BlockDriverState *bs,
986                                     NotifierWithReturn *notifier);
987 
988 /**
989  * bdrv_add_aio_context_notifier:
990  *
991  * If a long-running job intends to be always run in the same AioContext as a
992  * certain BDS, it may use this function to be notified of changes regarding the
993  * association of the BDS to an AioContext.
994  *
995  * attached_aio_context() is called after the target BDS has been attached to a
996  * new AioContext; detach_aio_context() is called before the target BDS is being
997  * detached from its old AioContext.
998  */
999 void bdrv_add_aio_context_notifier(BlockDriverState *bs,
1000         void (*attached_aio_context)(AioContext *new_context, void *opaque),
1001         void (*detach_aio_context)(void *opaque), void *opaque);
1002 
1003 /**
1004  * bdrv_remove_aio_context_notifier:
1005  *
1006  * Unsubscribe of change notifications regarding the BDS's AioContext. The
1007  * parameters given here have to be the same as those given to
1008  * bdrv_add_aio_context_notifier().
1009  */
1010 void bdrv_remove_aio_context_notifier(BlockDriverState *bs,
1011                                       void (*aio_context_attached)(AioContext *,
1012                                                                    void *),
1013                                       void (*aio_context_detached)(void *),
1014                                       void *opaque);
1015 
1016 /**
1017  * bdrv_wakeup:
1018  * @bs: The BlockDriverState for which an I/O operation has been completed.
1019  *
1020  * Wake up the main thread if it is waiting on BDRV_POLL_WHILE.  During
1021  * synchronous I/O on a BlockDriverState that is attached to another
1022  * I/O thread, the main thread lets the I/O thread's event loop run,
1023  * waiting for the I/O operation to complete.  A bdrv_wakeup will wake
1024  * up the main thread if necessary.
1025  *
1026  * Manual calls to bdrv_wakeup are rarely necessary, because
1027  * bdrv_dec_in_flight already calls it.
1028  */
1029 void bdrv_wakeup(BlockDriverState *bs);
1030 
1031 #ifdef _WIN32
1032 int is_windows_drive(const char *filename);
1033 #endif
1034 
1035 /**
1036  * stream_start:
1037  * @job_id: The id of the newly-created job, or %NULL to use the
1038  * device name of @bs.
1039  * @bs: Block device to operate on.
1040  * @base: Block device that will become the new base, or %NULL to
1041  * flatten the whole backing file chain onto @bs.
1042  * @backing_file_str: The file name that will be written to @bs as the
1043  * the new backing file if the job completes. Ignored if @base is %NULL.
1044  * @creation_flags: Flags that control the behavior of the Job lifetime.
1045  *                  See @BlockJobCreateFlags
1046  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1047  * @on_error: The action to take upon error.
1048  * @errp: Error object.
1049  *
1050  * Start a streaming operation on @bs.  Clusters that are unallocated
1051  * in @bs, but allocated in any image between @base and @bs (both
1052  * exclusive) will be written to @bs.  At the end of a successful
1053  * streaming job, the backing file of @bs will be changed to
1054  * @backing_file_str in the written image and to @base in the live
1055  * BlockDriverState.
1056  */
1057 void stream_start(const char *job_id, BlockDriverState *bs,
1058                   BlockDriverState *base, const char *backing_file_str,
1059                   int creation_flags, int64_t speed,
1060                   BlockdevOnError on_error, Error **errp);
1061 
1062 /**
1063  * commit_start:
1064  * @job_id: The id of the newly-created job, or %NULL to use the
1065  * device name of @bs.
1066  * @bs: Active block device.
1067  * @top: Top block device to be committed.
1068  * @base: Block device that will be written into, and become the new top.
1069  * @creation_flags: Flags that control the behavior of the Job lifetime.
1070  *                  See @BlockJobCreateFlags
1071  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1072  * @on_error: The action to take upon error.
1073  * @backing_file_str: String to use as the backing file in @top's overlay
1074  * @filter_node_name: The node name that should be assigned to the filter
1075  * driver that the commit job inserts into the graph above @top. NULL means
1076  * that a node name should be autogenerated.
1077  * @errp: Error object.
1078  *
1079  */
1080 void commit_start(const char *job_id, BlockDriverState *bs,
1081                   BlockDriverState *base, BlockDriverState *top,
1082                   int creation_flags, int64_t speed,
1083                   BlockdevOnError on_error, const char *backing_file_str,
1084                   const char *filter_node_name, Error **errp);
1085 /**
1086  * commit_active_start:
1087  * @job_id: The id of the newly-created job, or %NULL to use the
1088  * device name of @bs.
1089  * @bs: Active block device to be committed.
1090  * @base: Block device that will be written into, and become the new top.
1091  * @creation_flags: Flags that control the behavior of the Job lifetime.
1092  *                  See @BlockJobCreateFlags
1093  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1094  * @on_error: The action to take upon error.
1095  * @filter_node_name: The node name that should be assigned to the filter
1096  * driver that the commit job inserts into the graph above @bs. NULL means that
1097  * a node name should be autogenerated.
1098  * @cb: Completion function for the job.
1099  * @opaque: Opaque pointer value passed to @cb.
1100  * @auto_complete: Auto complete the job.
1101  * @errp: Error object.
1102  *
1103  */
1104 BlockJob *commit_active_start(const char *job_id, BlockDriverState *bs,
1105                               BlockDriverState *base, int creation_flags,
1106                               int64_t speed, BlockdevOnError on_error,
1107                               const char *filter_node_name,
1108                               BlockCompletionFunc *cb, void *opaque,
1109                               bool auto_complete, Error **errp);
1110 /*
1111  * mirror_start:
1112  * @job_id: The id of the newly-created job, or %NULL to use the
1113  * device name of @bs.
1114  * @bs: Block device to operate on.
1115  * @target: Block device to write to.
1116  * @replaces: Block graph node name to replace once the mirror is done. Can
1117  *            only be used when full mirroring is selected.
1118  * @creation_flags: Flags that control the behavior of the Job lifetime.
1119  *                  See @BlockJobCreateFlags
1120  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1121  * @granularity: The chosen granularity for the dirty bitmap.
1122  * @buf_size: The amount of data that can be in flight at one time.
1123  * @mode: Whether to collapse all images in the chain to the target.
1124  * @backing_mode: How to establish the target's backing chain after completion.
1125  * @zero_target: Whether the target should be explicitly zero-initialized
1126  * @on_source_error: The action to take upon error reading from the source.
1127  * @on_target_error: The action to take upon error writing to the target.
1128  * @unmap: Whether to unmap target where source sectors only contain zeroes.
1129  * @filter_node_name: The node name that should be assigned to the filter
1130  * driver that the mirror job inserts into the graph above @bs. NULL means that
1131  * a node name should be autogenerated.
1132  * @copy_mode: When to trigger writes to the target.
1133  * @errp: Error object.
1134  *
1135  * Start a mirroring operation on @bs.  Clusters that are allocated
1136  * in @bs will be written to @target until the job is cancelled or
1137  * manually completed.  At the end of a successful mirroring job,
1138  * @bs will be switched to read from @target.
1139  */
1140 void mirror_start(const char *job_id, BlockDriverState *bs,
1141                   BlockDriverState *target, const char *replaces,
1142                   int creation_flags, int64_t speed,
1143                   uint32_t granularity, int64_t buf_size,
1144                   MirrorSyncMode mode, BlockMirrorBackingMode backing_mode,
1145                   bool zero_target,
1146                   BlockdevOnError on_source_error,
1147                   BlockdevOnError on_target_error,
1148                   bool unmap, const char *filter_node_name,
1149                   MirrorCopyMode copy_mode, Error **errp);
1150 
1151 /*
1152  * backup_job_create:
1153  * @job_id: The id of the newly-created job, or %NULL to use the
1154  * device name of @bs.
1155  * @bs: Block device to operate on.
1156  * @target: Block device to write to.
1157  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1158  * @sync_mode: What parts of the disk image should be copied to the destination.
1159  * @sync_bitmap: The dirty bitmap if sync_mode is 'bitmap' or 'incremental'
1160  * @bitmap_mode: The bitmap synchronization policy to use.
1161  * @on_source_error: The action to take upon error reading from the source.
1162  * @on_target_error: The action to take upon error writing to the target.
1163  * @creation_flags: Flags that control the behavior of the Job lifetime.
1164  *                  See @BlockJobCreateFlags
1165  * @cb: Completion function for the job.
1166  * @opaque: Opaque pointer value passed to @cb.
1167  * @txn: Transaction that this job is part of (may be NULL).
1168  *
1169  * Create a backup operation on @bs.  Clusters in @bs are written to @target
1170  * until the job is cancelled or manually completed.
1171  */
1172 BlockJob *backup_job_create(const char *job_id, BlockDriverState *bs,
1173                             BlockDriverState *target, int64_t speed,
1174                             MirrorSyncMode sync_mode,
1175                             BdrvDirtyBitmap *sync_bitmap,
1176                             BitmapSyncMode bitmap_mode,
1177                             bool compress,
1178                             BlockdevOnError on_source_error,
1179                             BlockdevOnError on_target_error,
1180                             int creation_flags,
1181                             BlockCompletionFunc *cb, void *opaque,
1182                             JobTxn *txn, Error **errp);
1183 
1184 void hmp_drive_add_node(Monitor *mon, const char *optstr);
1185 
1186 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
1187                                   const char *child_name,
1188                                   const BdrvChildRole *child_role,
1189                                   AioContext *ctx,
1190                                   uint64_t perm, uint64_t shared_perm,
1191                                   void *opaque, Error **errp);
1192 void bdrv_root_unref_child(BdrvChild *child);
1193 
1194 /**
1195  * Sets a BdrvChild's permissions.  Avoid if the parent is a BDS; use
1196  * bdrv_child_refresh_perms() instead and make the parent's
1197  * .bdrv_child_perm() implementation return the correct values.
1198  */
1199 int bdrv_child_try_set_perm(BdrvChild *c, uint64_t perm, uint64_t shared,
1200                             Error **errp);
1201 
1202 /**
1203  * Calls bs->drv->bdrv_child_perm() and updates the child's permission
1204  * masks with the result.
1205  * Drivers should invoke this function whenever an event occurs that
1206  * makes their .bdrv_child_perm() implementation return different
1207  * values than before, but which will not result in the block layer
1208  * automatically refreshing the permissions.
1209  */
1210 int bdrv_child_refresh_perms(BlockDriverState *bs, BdrvChild *c, Error **errp);
1211 
1212 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1213  * block filters: Forward CONSISTENT_READ, WRITE, WRITE_UNCHANGED and RESIZE to
1214  * all children */
1215 void bdrv_filter_default_perms(BlockDriverState *bs, BdrvChild *c,
1216                                const BdrvChildRole *role,
1217                                BlockReopenQueue *reopen_queue,
1218                                uint64_t perm, uint64_t shared,
1219                                uint64_t *nperm, uint64_t *nshared);
1220 
1221 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1222  * (non-raw) image formats: Like above for bs->backing, but for bs->file it
1223  * requires WRITE | RESIZE for read-write images, always requires
1224  * CONSISTENT_READ and doesn't share WRITE. */
1225 void bdrv_format_default_perms(BlockDriverState *bs, BdrvChild *c,
1226                                const BdrvChildRole *role,
1227                                BlockReopenQueue *reopen_queue,
1228                                uint64_t perm, uint64_t shared,
1229                                uint64_t *nperm, uint64_t *nshared);
1230 
1231 /*
1232  * Default implementation for drivers to pass bdrv_co_block_status() to
1233  * their file.
1234  */
1235 int coroutine_fn bdrv_co_block_status_from_file(BlockDriverState *bs,
1236                                                 bool want_zero,
1237                                                 int64_t offset,
1238                                                 int64_t bytes,
1239                                                 int64_t *pnum,
1240                                                 int64_t *map,
1241                                                 BlockDriverState **file);
1242 /*
1243  * Default implementation for drivers to pass bdrv_co_block_status() to
1244  * their backing file.
1245  */
1246 int coroutine_fn bdrv_co_block_status_from_backing(BlockDriverState *bs,
1247                                                    bool want_zero,
1248                                                    int64_t offset,
1249                                                    int64_t bytes,
1250                                                    int64_t *pnum,
1251                                                    int64_t *map,
1252                                                    BlockDriverState **file);
1253 const char *bdrv_get_parent_name(const BlockDriverState *bs);
1254 void blk_dev_change_media_cb(BlockBackend *blk, bool load, Error **errp);
1255 bool blk_dev_has_removable_media(BlockBackend *blk);
1256 bool blk_dev_has_tray(BlockBackend *blk);
1257 void blk_dev_eject_request(BlockBackend *blk, bool force);
1258 bool blk_dev_is_tray_open(BlockBackend *blk);
1259 bool blk_dev_is_medium_locked(BlockBackend *blk);
1260 
1261 void bdrv_set_dirty(BlockDriverState *bs, int64_t offset, int64_t bytes);
1262 
1263 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap **out);
1264 void bdrv_restore_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap *backup);
1265 bool bdrv_dirty_bitmap_merge_internal(BdrvDirtyBitmap *dest,
1266                                       const BdrvDirtyBitmap *src,
1267                                       HBitmap **backup, bool lock);
1268 
1269 void bdrv_inc_in_flight(BlockDriverState *bs);
1270 void bdrv_dec_in_flight(BlockDriverState *bs);
1271 
1272 void blockdev_close_all_bdrv_states(void);
1273 
1274 int coroutine_fn bdrv_co_copy_range_from(BdrvChild *src, uint64_t src_offset,
1275                                          BdrvChild *dst, uint64_t dst_offset,
1276                                          uint64_t bytes,
1277                                          BdrvRequestFlags read_flags,
1278                                          BdrvRequestFlags write_flags);
1279 int coroutine_fn bdrv_co_copy_range_to(BdrvChild *src, uint64_t src_offset,
1280                                        BdrvChild *dst, uint64_t dst_offset,
1281                                        uint64_t bytes,
1282                                        BdrvRequestFlags read_flags,
1283                                        BdrvRequestFlags write_flags);
1284 
1285 int refresh_total_sectors(BlockDriverState *bs, int64_t hint);
1286 
1287 #endif /* BLOCK_INT_H */
1288