xref: /openbmc/qemu/include/block/block_int.h (revision 19f70347)
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_preadv_part)(BlockDriverState *bs,
214         uint64_t offset, uint64_t bytes,
215         QEMUIOVector *qiov, size_t qiov_offset, int flags);
216     int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs,
217         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, int flags);
218     /**
219      * @offset: position in bytes to write at
220      * @bytes: number of bytes to write
221      * @qiov: the buffers containing data to write
222      * @flags: zero or more bits allowed by 'supported_write_flags'
223      *
224      * @offset and @bytes will be a multiple of 'request_alignment',
225      * but the length of individual @qiov elements does not have to
226      * be a multiple.
227      *
228      * @bytes will always equal the total size of @qiov, and will be
229      * no larger than 'max_transfer'.
230      *
231      * The buffer in @qiov may point directly to guest memory.
232      */
233     int coroutine_fn (*bdrv_co_pwritev)(BlockDriverState *bs,
234         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags);
235     int coroutine_fn (*bdrv_co_pwritev_part)(BlockDriverState *bs,
236         uint64_t offset, uint64_t bytes,
237         QEMUIOVector *qiov, size_t qiov_offset, int flags);
238 
239     /*
240      * Efficiently zero a region of the disk image.  Typically an image format
241      * would use a compact metadata representation to implement this.  This
242      * function pointer may be NULL or return -ENOSUP and .bdrv_co_writev()
243      * will be called instead.
244      */
245     int coroutine_fn (*bdrv_co_pwrite_zeroes)(BlockDriverState *bs,
246         int64_t offset, int bytes, BdrvRequestFlags flags);
247     int coroutine_fn (*bdrv_co_pdiscard)(BlockDriverState *bs,
248         int64_t offset, int bytes);
249 
250     /* Map [offset, offset + nbytes) range onto a child of @bs to copy from,
251      * and invoke bdrv_co_copy_range_from(child, ...), or invoke
252      * bdrv_co_copy_range_to() if @bs is the leaf child to copy data from.
253      *
254      * See the comment of bdrv_co_copy_range for the parameter and return value
255      * semantics.
256      */
257     int coroutine_fn (*bdrv_co_copy_range_from)(BlockDriverState *bs,
258                                                 BdrvChild *src,
259                                                 uint64_t offset,
260                                                 BdrvChild *dst,
261                                                 uint64_t dst_offset,
262                                                 uint64_t bytes,
263                                                 BdrvRequestFlags read_flags,
264                                                 BdrvRequestFlags write_flags);
265 
266     /* Map [offset, offset + nbytes) range onto a child of bs to copy data to,
267      * and invoke bdrv_co_copy_range_to(child, src, ...), or perform the copy
268      * operation if @bs is the leaf and @src has the same BlockDriver.  Return
269      * -ENOTSUP if @bs is the leaf but @src has a different BlockDriver.
270      *
271      * See the comment of bdrv_co_copy_range for the parameter and return value
272      * semantics.
273      */
274     int coroutine_fn (*bdrv_co_copy_range_to)(BlockDriverState *bs,
275                                               BdrvChild *src,
276                                               uint64_t src_offset,
277                                               BdrvChild *dst,
278                                               uint64_t dst_offset,
279                                               uint64_t bytes,
280                                               BdrvRequestFlags read_flags,
281                                               BdrvRequestFlags write_flags);
282 
283     /*
284      * Building block for bdrv_block_status[_above] and
285      * bdrv_is_allocated[_above].  The driver should answer only
286      * according to the current layer, and should only need to set
287      * BDRV_BLOCK_DATA, BDRV_BLOCK_ZERO, BDRV_BLOCK_OFFSET_VALID,
288      * and/or BDRV_BLOCK_RAW; if the current layer defers to a backing
289      * layer, the result should be 0 (and not BDRV_BLOCK_ZERO).  See
290      * block.h for the overall meaning of the bits.  As a hint, the
291      * flag want_zero is true if the caller cares more about precise
292      * mappings (favor accurate _OFFSET_VALID/_ZERO) or false for
293      * overall allocation (favor larger *pnum, perhaps by reporting
294      * _DATA instead of _ZERO).  The block layer guarantees input
295      * clamped to bdrv_getlength() and aligned to request_alignment,
296      * as well as non-NULL pnum, map, and file; in turn, the driver
297      * must return an error or set pnum to an aligned non-zero value.
298      */
299     int coroutine_fn (*bdrv_co_block_status)(BlockDriverState *bs,
300         bool want_zero, int64_t offset, int64_t bytes, int64_t *pnum,
301         int64_t *map, BlockDriverState **file);
302 
303     /*
304      * Invalidate any cached meta-data.
305      */
306     void coroutine_fn (*bdrv_co_invalidate_cache)(BlockDriverState *bs,
307                                                   Error **errp);
308     int (*bdrv_inactivate)(BlockDriverState *bs);
309 
310     /*
311      * Flushes all data for all layers by calling bdrv_co_flush for underlying
312      * layers, if needed. This function is needed for deterministic
313      * synchronization of the flush finishing callback.
314      */
315     int coroutine_fn (*bdrv_co_flush)(BlockDriverState *bs);
316 
317     /*
318      * Flushes all data that was already written to the OS all the way down to
319      * the disk (for example file-posix.c calls fsync()).
320      */
321     int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs);
322 
323     /*
324      * Flushes all internal caches to the OS. The data may still sit in a
325      * writeback cache of the host OS, but it will survive a crash of the qemu
326      * process.
327      */
328     int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs);
329 
330     /*
331      * Drivers setting this field must be able to work with just a plain
332      * filename with '<protocol_name>:' as a prefix, and no other options.
333      * Options may be extracted from the filename by implementing
334      * bdrv_parse_filename.
335      */
336     const char *protocol_name;
337 
338     /*
339      * Truncate @bs to @offset bytes using the given @prealloc mode
340      * when growing.  Modes other than PREALLOC_MODE_OFF should be
341      * rejected when shrinking @bs.
342      *
343      * If @exact is true, @bs must be resized to exactly @offset.
344      * Otherwise, it is sufficient for @bs (if it is a host block
345      * device and thus there is no way to resize it) to be at least
346      * @offset bytes in length.
347      *
348      * If @exact is true and this function fails but would succeed
349      * with @exact = false, it should return -ENOTSUP.
350      */
351     int coroutine_fn (*bdrv_co_truncate)(BlockDriverState *bs, int64_t offset,
352                                          bool exact, PreallocMode prealloc,
353                                          Error **errp);
354 
355     int64_t (*bdrv_getlength)(BlockDriverState *bs);
356     bool has_variable_length;
357     int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
358     BlockMeasureInfo *(*bdrv_measure)(QemuOpts *opts, BlockDriverState *in_bs,
359                                       Error **errp);
360 
361     int coroutine_fn (*bdrv_co_pwritev_compressed)(BlockDriverState *bs,
362         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov);
363     int coroutine_fn (*bdrv_co_pwritev_compressed_part)(BlockDriverState *bs,
364         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov,
365         size_t qiov_offset);
366 
367     int (*bdrv_snapshot_create)(BlockDriverState *bs,
368                                 QEMUSnapshotInfo *sn_info);
369     int (*bdrv_snapshot_goto)(BlockDriverState *bs,
370                               const char *snapshot_id);
371     int (*bdrv_snapshot_delete)(BlockDriverState *bs,
372                                 const char *snapshot_id,
373                                 const char *name,
374                                 Error **errp);
375     int (*bdrv_snapshot_list)(BlockDriverState *bs,
376                               QEMUSnapshotInfo **psn_info);
377     int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
378                                   const char *snapshot_id,
379                                   const char *name,
380                                   Error **errp);
381     int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
382     ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs,
383                                                  Error **errp);
384     BlockStatsSpecific *(*bdrv_get_specific_stats)(BlockDriverState *bs);
385 
386     int coroutine_fn (*bdrv_save_vmstate)(BlockDriverState *bs,
387                                           QEMUIOVector *qiov,
388                                           int64_t pos);
389     int coroutine_fn (*bdrv_load_vmstate)(BlockDriverState *bs,
390                                           QEMUIOVector *qiov,
391                                           int64_t pos);
392 
393     int (*bdrv_change_backing_file)(BlockDriverState *bs,
394         const char *backing_file, const char *backing_fmt);
395 
396     /* removable device specific */
397     bool (*bdrv_is_inserted)(BlockDriverState *bs);
398     void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
399     void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
400 
401     /* to control generic scsi devices */
402     BlockAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
403         unsigned long int req, void *buf,
404         BlockCompletionFunc *cb, void *opaque);
405     int coroutine_fn (*bdrv_co_ioctl)(BlockDriverState *bs,
406                                       unsigned long int req, void *buf);
407 
408     /* List of options for creating images, terminated by name == NULL */
409     QemuOptsList *create_opts;
410     /*
411      * If this driver supports reopening images this contains a
412      * NULL-terminated list of the runtime options that can be
413      * modified. If an option in this list is unspecified during
414      * reopen then it _must_ be reset to its default value or return
415      * an error.
416      */
417     const char *const *mutable_opts;
418 
419     /*
420      * Returns 0 for completed check, -errno for internal errors.
421      * The check results are stored in result.
422      */
423     int coroutine_fn (*bdrv_co_check)(BlockDriverState *bs,
424                                       BdrvCheckResult *result,
425                                       BdrvCheckMode fix);
426 
427     int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts,
428                               BlockDriverAmendStatusCB *status_cb,
429                               void *cb_opaque,
430                               Error **errp);
431 
432     void (*bdrv_debug_event)(BlockDriverState *bs, BlkdebugEvent event);
433 
434     /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
435     int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
436         const char *tag);
437     int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs,
438         const char *tag);
439     int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
440     bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
441 
442     void (*bdrv_refresh_limits)(BlockDriverState *bs, Error **errp);
443 
444     /*
445      * Returns 1 if newly created images are guaranteed to contain only
446      * zeros, 0 otherwise.
447      * Must return 0 if .bdrv_has_zero_init_truncate() returns 0.
448      */
449     int (*bdrv_has_zero_init)(BlockDriverState *bs);
450 
451     /*
452      * Returns 1 if new areas added by growing the image with
453      * PREALLOC_MODE_OFF contain only zeros, 0 otherwise.
454      */
455     int (*bdrv_has_zero_init_truncate)(BlockDriverState *bs);
456 
457     /* Remove fd handlers, timers, and other event loop callbacks so the event
458      * loop is no longer in use.  Called with no in-flight requests and in
459      * depth-first traversal order with parents before child nodes.
460      */
461     void (*bdrv_detach_aio_context)(BlockDriverState *bs);
462 
463     /* Add fd handlers, timers, and other event loop callbacks so I/O requests
464      * can be processed again.  Called with no in-flight requests and in
465      * depth-first traversal order with child nodes before parent nodes.
466      */
467     void (*bdrv_attach_aio_context)(BlockDriverState *bs,
468                                     AioContext *new_context);
469 
470     /* io queue for linux-aio */
471     void (*bdrv_io_plug)(BlockDriverState *bs);
472     void (*bdrv_io_unplug)(BlockDriverState *bs);
473 
474     /**
475      * Try to get @bs's logical and physical block size.
476      * On success, store them in @bsz and return zero.
477      * On failure, return negative errno.
478      */
479     int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz);
480     /**
481      * Try to get @bs's geometry (cyls, heads, sectors)
482      * On success, store them in @geo and return 0.
483      * On failure return -errno.
484      * Only drivers that want to override guest geometry implement this
485      * callback; see hd_geometry_guess().
486      */
487     int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo);
488 
489     /**
490      * bdrv_co_drain_begin is called if implemented in the beginning of a
491      * drain operation to drain and stop any internal sources of requests in
492      * the driver.
493      * bdrv_co_drain_end is called if implemented at the end of the drain.
494      *
495      * They should be used by the driver to e.g. manage scheduled I/O
496      * requests, or toggle an internal state. After the end of the drain new
497      * requests will continue normally.
498      */
499     void coroutine_fn (*bdrv_co_drain_begin)(BlockDriverState *bs);
500     void coroutine_fn (*bdrv_co_drain_end)(BlockDriverState *bs);
501 
502     void (*bdrv_add_child)(BlockDriverState *parent, BlockDriverState *child,
503                            Error **errp);
504     void (*bdrv_del_child)(BlockDriverState *parent, BdrvChild *child,
505                            Error **errp);
506 
507     /**
508      * Informs the block driver that a permission change is intended. The
509      * driver checks whether the change is permissible and may take other
510      * preparations for the change (e.g. get file system locks). This operation
511      * is always followed either by a call to either .bdrv_set_perm or
512      * .bdrv_abort_perm_update.
513      *
514      * Checks whether the requested set of cumulative permissions in @perm
515      * can be granted for accessing @bs and whether no other users are using
516      * permissions other than those given in @shared (both arguments take
517      * BLK_PERM_* bitmasks).
518      *
519      * If both conditions are met, 0 is returned. Otherwise, -errno is returned
520      * and errp is set to an error describing the conflict.
521      */
522     int (*bdrv_check_perm)(BlockDriverState *bs, uint64_t perm,
523                            uint64_t shared, Error **errp);
524 
525     /**
526      * Called to inform the driver that the set of cumulative set of used
527      * permissions for @bs has changed to @perm, and the set of sharable
528      * permission to @shared. The driver can use this to propagate changes to
529      * its children (i.e. request permissions only if a parent actually needs
530      * them).
531      *
532      * This function is only invoked after bdrv_check_perm(), so block drivers
533      * may rely on preparations made in their .bdrv_check_perm implementation.
534      */
535     void (*bdrv_set_perm)(BlockDriverState *bs, uint64_t perm, uint64_t shared);
536 
537     /*
538      * Called to inform the driver that after a previous bdrv_check_perm()
539      * call, the permission update is not performed and any preparations made
540      * for it (e.g. taken file locks) need to be undone.
541      *
542      * This function can be called even for nodes that never saw a
543      * bdrv_check_perm() call. It is a no-op then.
544      */
545     void (*bdrv_abort_perm_update)(BlockDriverState *bs);
546 
547     /**
548      * Returns in @nperm and @nshared the permissions that the driver for @bs
549      * needs on its child @c, based on the cumulative permissions requested by
550      * the parents in @parent_perm and @parent_shared.
551      *
552      * If @c is NULL, return the permissions for attaching a new child for the
553      * given @role.
554      *
555      * If @reopen_queue is non-NULL, don't return the currently needed
556      * permissions, but those that will be needed after applying the
557      * @reopen_queue.
558      */
559      void (*bdrv_child_perm)(BlockDriverState *bs, BdrvChild *c,
560                              const BdrvChildRole *role,
561                              BlockReopenQueue *reopen_queue,
562                              uint64_t parent_perm, uint64_t parent_shared,
563                              uint64_t *nperm, uint64_t *nshared);
564 
565     bool (*bdrv_co_can_store_new_dirty_bitmap)(BlockDriverState *bs,
566                                                const char *name,
567                                                uint32_t granularity,
568                                                Error **errp);
569     int (*bdrv_co_remove_persistent_dirty_bitmap)(BlockDriverState *bs,
570                                                   const char *name,
571                                                   Error **errp);
572 
573     /**
574      * Register/unregister a buffer for I/O. For example, when the driver is
575      * interested to know the memory areas that will later be used in iovs, so
576      * that it can do IOMMU mapping with VFIO etc., in order to get better
577      * performance. In the case of VFIO drivers, this callback is used to do
578      * DMA mapping for hot buffers.
579      */
580     void (*bdrv_register_buf)(BlockDriverState *bs, void *host, size_t size);
581     void (*bdrv_unregister_buf)(BlockDriverState *bs, void *host);
582     QLIST_ENTRY(BlockDriver) list;
583 
584     /* Pointer to a NULL-terminated array of names of strong options
585      * that can be specified for bdrv_open(). A strong option is one
586      * that changes the data of a BDS.
587      * If this pointer is NULL, the array is considered empty.
588      * "filename" and "driver" are always considered strong. */
589     const char *const *strong_runtime_opts;
590 };
591 
592 static inline bool block_driver_can_compress(BlockDriver *drv)
593 {
594     return drv->bdrv_co_pwritev_compressed ||
595            drv->bdrv_co_pwritev_compressed_part;
596 }
597 
598 typedef struct BlockLimits {
599     /* Alignment requirement, in bytes, for offset/length of I/O
600      * requests. Must be a power of 2 less than INT_MAX; defaults to
601      * 1 for drivers with modern byte interfaces, and to 512
602      * otherwise. */
603     uint32_t request_alignment;
604 
605     /* Maximum number of bytes that can be discarded at once (since it
606      * is signed, it must be < 2G, if set). Must be multiple of
607      * pdiscard_alignment, but need not be power of 2. May be 0 if no
608      * inherent 32-bit limit */
609     int32_t max_pdiscard;
610 
611     /* Optimal alignment for discard requests in bytes. A power of 2
612      * is best but not mandatory.  Must be a multiple of
613      * bl.request_alignment, and must be less than max_pdiscard if
614      * that is set. May be 0 if bl.request_alignment is good enough */
615     uint32_t pdiscard_alignment;
616 
617     /* Maximum number of bytes that can zeroized at once (since it is
618      * signed, it must be < 2G, if set). Must be multiple of
619      * pwrite_zeroes_alignment. May be 0 if no inherent 32-bit limit */
620     int32_t max_pwrite_zeroes;
621 
622     /* Optimal alignment for write zeroes requests in bytes. A power
623      * of 2 is best but not mandatory.  Must be a multiple of
624      * bl.request_alignment, and must be less than max_pwrite_zeroes
625      * if that is set. May be 0 if bl.request_alignment is good
626      * enough */
627     uint32_t pwrite_zeroes_alignment;
628 
629     /* Optimal transfer length in bytes.  A power of 2 is best but not
630      * mandatory.  Must be a multiple of bl.request_alignment, or 0 if
631      * no preferred size */
632     uint32_t opt_transfer;
633 
634     /* Maximal transfer length in bytes.  Need not be power of 2, but
635      * must be multiple of opt_transfer and bl.request_alignment, or 0
636      * for no 32-bit limit.  For now, anything larger than INT_MAX is
637      * clamped down. */
638     uint32_t max_transfer;
639 
640     /* memory alignment, in bytes so that no bounce buffer is needed */
641     size_t min_mem_alignment;
642 
643     /* memory alignment, in bytes, for bounce buffer */
644     size_t opt_mem_alignment;
645 
646     /* maximum number of iovec elements */
647     int max_iov;
648 } BlockLimits;
649 
650 typedef struct BdrvOpBlocker BdrvOpBlocker;
651 
652 typedef struct BdrvAioNotifier {
653     void (*attached_aio_context)(AioContext *new_context, void *opaque);
654     void (*detach_aio_context)(void *opaque);
655 
656     void *opaque;
657     bool deleted;
658 
659     QLIST_ENTRY(BdrvAioNotifier) list;
660 } BdrvAioNotifier;
661 
662 struct BdrvChildRole {
663     /* If true, bdrv_replace_node() doesn't change the node this BdrvChild
664      * points to. */
665     bool stay_at_node;
666 
667     /* If true, the parent is a BlockDriverState and bdrv_next_all_states()
668      * will return it. This information is used for drain_all, where every node
669      * will be drained separately, so the drain only needs to be propagated to
670      * non-BDS parents. */
671     bool parent_is_bds;
672 
673     void (*inherit_options)(int *child_flags, QDict *child_options,
674                             int parent_flags, QDict *parent_options);
675 
676     void (*change_media)(BdrvChild *child, bool load);
677     void (*resize)(BdrvChild *child);
678 
679     /* Returns a name that is supposedly more useful for human users than the
680      * node name for identifying the node in question (in particular, a BB
681      * name), or NULL if the parent can't provide a better name. */
682     const char *(*get_name)(BdrvChild *child);
683 
684     /* Returns a malloced string that describes the parent of the child for a
685      * human reader. This could be a node-name, BlockBackend name, qdev ID or
686      * QOM path of the device owning the BlockBackend, job type and ID etc. The
687      * caller is responsible for freeing the memory. */
688     char *(*get_parent_desc)(BdrvChild *child);
689 
690     /*
691      * If this pair of functions is implemented, the parent doesn't issue new
692      * requests after returning from .drained_begin() until .drained_end() is
693      * called.
694      *
695      * These functions must not change the graph (and therefore also must not
696      * call aio_poll(), which could change the graph indirectly).
697      *
698      * If drained_end() schedules background operations, it must atomically
699      * increment *drained_end_counter for each such operation and atomically
700      * decrement it once the operation has settled.
701      *
702      * Note that this can be nested. If drained_begin() was called twice, new
703      * I/O is allowed only after drained_end() was called twice, too.
704      */
705     void (*drained_begin)(BdrvChild *child);
706     void (*drained_end)(BdrvChild *child, int *drained_end_counter);
707 
708     /*
709      * Returns whether the parent has pending requests for the child. This
710      * callback is polled after .drained_begin() has been called until all
711      * activity on the child has stopped.
712      */
713     bool (*drained_poll)(BdrvChild *child);
714 
715     /* Notifies the parent that the child has been activated/inactivated (e.g.
716      * when migration is completing) and it can start/stop requesting
717      * permissions and doing I/O on it. */
718     void (*activate)(BdrvChild *child, Error **errp);
719     int (*inactivate)(BdrvChild *child);
720 
721     void (*attach)(BdrvChild *child);
722     void (*detach)(BdrvChild *child);
723 
724     /* Notifies the parent that the filename of its child has changed (e.g.
725      * because the direct child was removed from the backing chain), so that it
726      * can update its reference. */
727     int (*update_filename)(BdrvChild *child, BlockDriverState *new_base,
728                            const char *filename, Error **errp);
729 
730     bool (*can_set_aio_ctx)(BdrvChild *child, AioContext *ctx,
731                             GSList **ignore, Error **errp);
732     void (*set_aio_ctx)(BdrvChild *child, AioContext *ctx, GSList **ignore);
733 };
734 
735 extern const BdrvChildRole child_file;
736 extern const BdrvChildRole child_format;
737 extern const BdrvChildRole child_backing;
738 
739 struct BdrvChild {
740     BlockDriverState *bs;
741     char *name;
742     const BdrvChildRole *role;
743     void *opaque;
744 
745     /**
746      * Granted permissions for operating on this BdrvChild (BLK_PERM_* bitmask)
747      */
748     uint64_t perm;
749 
750     /**
751      * Permissions that can still be granted to other users of @bs while this
752      * BdrvChild is still attached to it. (BLK_PERM_* bitmask)
753      */
754     uint64_t shared_perm;
755 
756     /* backup of permissions during permission update procedure */
757     bool has_backup_perm;
758     uint64_t backup_perm;
759     uint64_t backup_shared_perm;
760 
761     /*
762      * This link is frozen: the child can neither be replaced nor
763      * detached from the parent.
764      */
765     bool frozen;
766 
767     /*
768      * How many times the parent of this child has been drained
769      * (through role->drained_*).
770      * Usually, this is equal to bs->quiesce_counter (potentially
771      * reduced by bdrv_drain_all_count).  It may differ while the
772      * child is entering or leaving a drained section.
773      */
774     int parent_quiesce_counter;
775 
776     QLIST_ENTRY(BdrvChild) next;
777     QLIST_ENTRY(BdrvChild) next_parent;
778 };
779 
780 /*
781  * Note: the function bdrv_append() copies and swaps contents of
782  * BlockDriverStates, so if you add new fields to this struct, please
783  * inspect bdrv_append() to determine if the new fields need to be
784  * copied as well.
785  */
786 struct BlockDriverState {
787     /* Protected by big QEMU lock or read-only after opening.  No special
788      * locking needed during I/O...
789      */
790     int open_flags; /* flags used to open the file, re-used for re-open */
791     bool read_only; /* if true, the media is read only */
792     bool encrypted; /* if true, the media is encrypted */
793     bool sg;        /* if true, the device is a /dev/sg* */
794     bool probed;    /* if true, format was probed rather than specified */
795     bool force_share; /* if true, always allow all shared permissions */
796     bool implicit;  /* if true, this filter node was automatically inserted */
797 
798     BlockDriver *drv; /* NULL means no media */
799     void *opaque;
800 
801     AioContext *aio_context; /* event loop used for fd handlers, timers, etc */
802     /* long-running tasks intended to always use the same AioContext as this
803      * BDS may register themselves in this list to be notified of changes
804      * regarding this BDS's context */
805     QLIST_HEAD(, BdrvAioNotifier) aio_notifiers;
806     bool walking_aio_notifiers; /* to make removal during iteration safe */
807 
808     char filename[PATH_MAX];
809     char backing_file[PATH_MAX]; /* if non zero, the image is a diff of
810                                     this file image */
811     /* The backing filename indicated by the image header; if we ever
812      * open this file, then this is replaced by the resulting BDS's
813      * filename (i.e. after a bdrv_refresh_filename() run). */
814     char auto_backing_file[PATH_MAX];
815     char backing_format[16]; /* if non-zero and backing_file exists */
816 
817     QDict *full_open_options;
818     char exact_filename[PATH_MAX];
819 
820     BdrvChild *backing;
821     BdrvChild *file;
822 
823     /* I/O Limits */
824     BlockLimits bl;
825 
826     /* Flags honored during pwrite (so far: BDRV_REQ_FUA,
827      * BDRV_REQ_WRITE_UNCHANGED).
828      * If a driver does not support BDRV_REQ_WRITE_UNCHANGED, those
829      * writes will be issued as normal writes without the flag set.
830      * This is important to note for drivers that do not explicitly
831      * request a WRITE permission for their children and instead take
832      * the same permissions as their parent did (this is commonly what
833      * block filters do).  Such drivers have to be aware that the
834      * parent may have taken a WRITE_UNCHANGED permission only and is
835      * issuing such requests.  Drivers either must make sure that
836      * these requests do not result in plain WRITE accesses (usually
837      * by supporting BDRV_REQ_WRITE_UNCHANGED, and then forwarding
838      * every incoming write request as-is, including potentially that
839      * flag), or they have to explicitly take the WRITE permission for
840      * their children. */
841     unsigned int supported_write_flags;
842     /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA,
843      * BDRV_REQ_MAY_UNMAP, BDRV_REQ_WRITE_UNCHANGED) */
844     unsigned int supported_zero_flags;
845 
846     /* the following member gives a name to every node on the bs graph. */
847     char node_name[32];
848     /* element of the list of named nodes building the graph */
849     QTAILQ_ENTRY(BlockDriverState) node_list;
850     /* element of the list of all BlockDriverStates (all_bdrv_states) */
851     QTAILQ_ENTRY(BlockDriverState) bs_list;
852     /* element of the list of monitor-owned BDS */
853     QTAILQ_ENTRY(BlockDriverState) monitor_list;
854     int refcnt;
855 
856     /* operation blockers */
857     QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX];
858 
859     /* The node that this node inherited default options from (and a reopen on
860      * which can affect this node by changing these defaults). This is always a
861      * parent node of this node. */
862     BlockDriverState *inherits_from;
863     QLIST_HEAD(, BdrvChild) children;
864     QLIST_HEAD(, BdrvChild) parents;
865 
866     QDict *options;
867     QDict *explicit_options;
868     BlockdevDetectZeroesOptions detect_zeroes;
869 
870     /* The error object in use for blocking operations on backing_hd */
871     Error *backing_blocker;
872 
873     /* Protected by AioContext lock */
874 
875     /* If we are reading a disk image, give its size in sectors.
876      * Generally read-only; it is written to by load_snapshot and
877      * save_snaphost, but the block layer is quiescent during those.
878      */
879     int64_t total_sectors;
880 
881     /* Callback before write request is processed */
882     NotifierWithReturnList before_write_notifiers;
883 
884     /* threshold limit for writes, in bytes. "High water mark". */
885     uint64_t write_threshold_offset;
886     NotifierWithReturn write_threshold_notifier;
887 
888     /* Writing to the list requires the BQL _and_ the dirty_bitmap_mutex.
889      * Reading from the list can be done with either the BQL or the
890      * dirty_bitmap_mutex.  Modifying a bitmap only requires
891      * dirty_bitmap_mutex.  */
892     QemuMutex dirty_bitmap_mutex;
893     QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps;
894 
895     /* Offset after the highest byte written to */
896     Stat64 wr_highest_offset;
897 
898     /* If true, copy read backing sectors into image.  Can be >1 if more
899      * than one client has requested copy-on-read.  Accessed with atomic
900      * ops.
901      */
902     int copy_on_read;
903 
904     /* number of in-flight requests; overall and serialising.
905      * Accessed with atomic ops.
906      */
907     unsigned int in_flight;
908     unsigned int serialising_in_flight;
909 
910     /* counter for nested bdrv_io_plug.
911      * Accessed with atomic ops.
912     */
913     unsigned io_plugged;
914 
915     /* do we need to tell the quest if we have a volatile write cache? */
916     int enable_write_cache;
917 
918     /* Accessed with atomic ops.  */
919     int quiesce_counter;
920     int recursive_quiesce_counter;
921 
922     unsigned int write_gen;               /* Current data generation */
923 
924     /* Protected by reqs_lock.  */
925     CoMutex reqs_lock;
926     QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
927     CoQueue flush_queue;                  /* Serializing flush queue */
928     bool active_flush_req;                /* Flush request in flight? */
929 
930     /* Only read/written by whoever has set active_flush_req to true.  */
931     unsigned int flushed_gen;             /* Flushed write generation */
932 
933     /* BdrvChild links to this node may never be frozen */
934     bool never_freeze;
935 };
936 
937 struct BlockBackendRootState {
938     int open_flags;
939     bool read_only;
940     BlockdevDetectZeroesOptions detect_zeroes;
941 };
942 
943 typedef enum BlockMirrorBackingMode {
944     /* Reuse the existing backing chain from the source for the target.
945      * - sync=full: Set backing BDS to NULL.
946      * - sync=top:  Use source's backing BDS.
947      * - sync=none: Use source as the backing BDS. */
948     MIRROR_SOURCE_BACKING_CHAIN,
949 
950     /* Open the target's backing chain completely anew */
951     MIRROR_OPEN_BACKING_CHAIN,
952 
953     /* Do not change the target's backing BDS after job completion */
954     MIRROR_LEAVE_BACKING_CHAIN,
955 } BlockMirrorBackingMode;
956 
957 static inline BlockDriverState *backing_bs(BlockDriverState *bs)
958 {
959     return bs->backing ? bs->backing->bs : NULL;
960 }
961 
962 
963 /* Essential block drivers which must always be statically linked into qemu, and
964  * which therefore can be accessed without using bdrv_find_format() */
965 extern BlockDriver bdrv_file;
966 extern BlockDriver bdrv_raw;
967 extern BlockDriver bdrv_qcow2;
968 
969 int coroutine_fn bdrv_co_preadv(BdrvChild *child,
970     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
971     BdrvRequestFlags flags);
972 int coroutine_fn bdrv_co_preadv_part(BdrvChild *child,
973     int64_t offset, unsigned int bytes,
974     QEMUIOVector *qiov, size_t qiov_offset, BdrvRequestFlags flags);
975 int coroutine_fn bdrv_co_pwritev(BdrvChild *child,
976     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
977     BdrvRequestFlags flags);
978 int coroutine_fn bdrv_co_pwritev_part(BdrvChild *child,
979     int64_t offset, unsigned int bytes,
980     QEMUIOVector *qiov, size_t qiov_offset, BdrvRequestFlags flags);
981 
982 static inline int coroutine_fn bdrv_co_pread(BdrvChild *child,
983     int64_t offset, unsigned int bytes, void *buf, BdrvRequestFlags flags)
984 {
985     QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes);
986 
987     return bdrv_co_preadv(child, offset, bytes, &qiov, flags);
988 }
989 
990 static inline int coroutine_fn bdrv_co_pwrite(BdrvChild *child,
991     int64_t offset, unsigned int bytes, void *buf, BdrvRequestFlags flags)
992 {
993     QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes);
994 
995     return bdrv_co_pwritev(child, offset, bytes, &qiov, flags);
996 }
997 
998 extern unsigned int bdrv_drain_all_count;
999 void bdrv_apply_subtree_drain(BdrvChild *child, BlockDriverState *new_parent);
1000 void bdrv_unapply_subtree_drain(BdrvChild *child, BlockDriverState *old_parent);
1001 
1002 bool coroutine_fn bdrv_mark_request_serialising(BdrvTrackedRequest *req, uint64_t align);
1003 BdrvTrackedRequest *coroutine_fn bdrv_co_get_self_request(BlockDriverState *bs);
1004 
1005 int get_tmp_filename(char *filename, int size);
1006 BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size,
1007                             const char *filename);
1008 
1009 void bdrv_parse_filename_strip_prefix(const char *filename, const char *prefix,
1010                                       QDict *options);
1011 
1012 
1013 /**
1014  * bdrv_add_before_write_notifier:
1015  *
1016  * Register a callback that is invoked before write requests are processed but
1017  * after any throttling or waiting for overlapping requests.
1018  */
1019 void bdrv_add_before_write_notifier(BlockDriverState *bs,
1020                                     NotifierWithReturn *notifier);
1021 
1022 /**
1023  * bdrv_add_aio_context_notifier:
1024  *
1025  * If a long-running job intends to be always run in the same AioContext as a
1026  * certain BDS, it may use this function to be notified of changes regarding the
1027  * association of the BDS to an AioContext.
1028  *
1029  * attached_aio_context() is called after the target BDS has been attached to a
1030  * new AioContext; detach_aio_context() is called before the target BDS is being
1031  * detached from its old AioContext.
1032  */
1033 void bdrv_add_aio_context_notifier(BlockDriverState *bs,
1034         void (*attached_aio_context)(AioContext *new_context, void *opaque),
1035         void (*detach_aio_context)(void *opaque), void *opaque);
1036 
1037 /**
1038  * bdrv_remove_aio_context_notifier:
1039  *
1040  * Unsubscribe of change notifications regarding the BDS's AioContext. The
1041  * parameters given here have to be the same as those given to
1042  * bdrv_add_aio_context_notifier().
1043  */
1044 void bdrv_remove_aio_context_notifier(BlockDriverState *bs,
1045                                       void (*aio_context_attached)(AioContext *,
1046                                                                    void *),
1047                                       void (*aio_context_detached)(void *),
1048                                       void *opaque);
1049 
1050 /**
1051  * bdrv_wakeup:
1052  * @bs: The BlockDriverState for which an I/O operation has been completed.
1053  *
1054  * Wake up the main thread if it is waiting on BDRV_POLL_WHILE.  During
1055  * synchronous I/O on a BlockDriverState that is attached to another
1056  * I/O thread, the main thread lets the I/O thread's event loop run,
1057  * waiting for the I/O operation to complete.  A bdrv_wakeup will wake
1058  * up the main thread if necessary.
1059  *
1060  * Manual calls to bdrv_wakeup are rarely necessary, because
1061  * bdrv_dec_in_flight already calls it.
1062  */
1063 void bdrv_wakeup(BlockDriverState *bs);
1064 
1065 #ifdef _WIN32
1066 int is_windows_drive(const char *filename);
1067 #endif
1068 
1069 /**
1070  * stream_start:
1071  * @job_id: The id of the newly-created job, or %NULL to use the
1072  * device name of @bs.
1073  * @bs: Block device to operate on.
1074  * @base: Block device that will become the new base, or %NULL to
1075  * flatten the whole backing file chain onto @bs.
1076  * @backing_file_str: The file name that will be written to @bs as the
1077  * the new backing file if the job completes. Ignored if @base is %NULL.
1078  * @creation_flags: Flags that control the behavior of the Job lifetime.
1079  *                  See @BlockJobCreateFlags
1080  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1081  * @on_error: The action to take upon error.
1082  * @errp: Error object.
1083  *
1084  * Start a streaming operation on @bs.  Clusters that are unallocated
1085  * in @bs, but allocated in any image between @base and @bs (both
1086  * exclusive) will be written to @bs.  At the end of a successful
1087  * streaming job, the backing file of @bs will be changed to
1088  * @backing_file_str in the written image and to @base in the live
1089  * BlockDriverState.
1090  */
1091 void stream_start(const char *job_id, BlockDriverState *bs,
1092                   BlockDriverState *base, const char *backing_file_str,
1093                   int creation_flags, int64_t speed,
1094                   BlockdevOnError on_error, Error **errp);
1095 
1096 /**
1097  * commit_start:
1098  * @job_id: The id of the newly-created job, or %NULL to use the
1099  * device name of @bs.
1100  * @bs: Active block device.
1101  * @top: Top block device to be committed.
1102  * @base: Block device that will be written into, and become the new top.
1103  * @creation_flags: Flags that control the behavior of the Job lifetime.
1104  *                  See @BlockJobCreateFlags
1105  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1106  * @on_error: The action to take upon error.
1107  * @backing_file_str: String to use as the backing file in @top's overlay
1108  * @filter_node_name: The node name that should be assigned to the filter
1109  * driver that the commit job inserts into the graph above @top. NULL means
1110  * that a node name should be autogenerated.
1111  * @errp: Error object.
1112  *
1113  */
1114 void commit_start(const char *job_id, BlockDriverState *bs,
1115                   BlockDriverState *base, BlockDriverState *top,
1116                   int creation_flags, int64_t speed,
1117                   BlockdevOnError on_error, const char *backing_file_str,
1118                   const char *filter_node_name, Error **errp);
1119 /**
1120  * commit_active_start:
1121  * @job_id: The id of the newly-created job, or %NULL to use the
1122  * device name of @bs.
1123  * @bs: Active block device to be committed.
1124  * @base: Block device that will be written into, and become the new top.
1125  * @creation_flags: Flags that control the behavior of the Job lifetime.
1126  *                  See @BlockJobCreateFlags
1127  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1128  * @on_error: The action to take upon error.
1129  * @filter_node_name: The node name that should be assigned to the filter
1130  * driver that the commit job inserts into the graph above @bs. NULL means that
1131  * a node name should be autogenerated.
1132  * @cb: Completion function for the job.
1133  * @opaque: Opaque pointer value passed to @cb.
1134  * @auto_complete: Auto complete the job.
1135  * @errp: Error object.
1136  *
1137  */
1138 BlockJob *commit_active_start(const char *job_id, BlockDriverState *bs,
1139                               BlockDriverState *base, int creation_flags,
1140                               int64_t speed, BlockdevOnError on_error,
1141                               const char *filter_node_name,
1142                               BlockCompletionFunc *cb, void *opaque,
1143                               bool auto_complete, Error **errp);
1144 /*
1145  * mirror_start:
1146  * @job_id: The id of the newly-created job, or %NULL to use the
1147  * device name of @bs.
1148  * @bs: Block device to operate on.
1149  * @target: Block device to write to.
1150  * @replaces: Block graph node name to replace once the mirror is done. Can
1151  *            only be used when full mirroring is selected.
1152  * @creation_flags: Flags that control the behavior of the Job lifetime.
1153  *                  See @BlockJobCreateFlags
1154  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1155  * @granularity: The chosen granularity for the dirty bitmap.
1156  * @buf_size: The amount of data that can be in flight at one time.
1157  * @mode: Whether to collapse all images in the chain to the target.
1158  * @backing_mode: How to establish the target's backing chain after completion.
1159  * @zero_target: Whether the target should be explicitly zero-initialized
1160  * @on_source_error: The action to take upon error reading from the source.
1161  * @on_target_error: The action to take upon error writing to the target.
1162  * @unmap: Whether to unmap target where source sectors only contain zeroes.
1163  * @filter_node_name: The node name that should be assigned to the filter
1164  * driver that the mirror job inserts into the graph above @bs. NULL means that
1165  * a node name should be autogenerated.
1166  * @copy_mode: When to trigger writes to the target.
1167  * @errp: Error object.
1168  *
1169  * Start a mirroring operation on @bs.  Clusters that are allocated
1170  * in @bs will be written to @target until the job is cancelled or
1171  * manually completed.  At the end of a successful mirroring job,
1172  * @bs will be switched to read from @target.
1173  */
1174 void mirror_start(const char *job_id, BlockDriverState *bs,
1175                   BlockDriverState *target, const char *replaces,
1176                   int creation_flags, int64_t speed,
1177                   uint32_t granularity, int64_t buf_size,
1178                   MirrorSyncMode mode, BlockMirrorBackingMode backing_mode,
1179                   bool zero_target,
1180                   BlockdevOnError on_source_error,
1181                   BlockdevOnError on_target_error,
1182                   bool unmap, const char *filter_node_name,
1183                   MirrorCopyMode copy_mode, Error **errp);
1184 
1185 /*
1186  * backup_job_create:
1187  * @job_id: The id of the newly-created job, or %NULL to use the
1188  * device name of @bs.
1189  * @bs: Block device to operate on.
1190  * @target: Block device to write to.
1191  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1192  * @sync_mode: What parts of the disk image should be copied to the destination.
1193  * @sync_bitmap: The dirty bitmap if sync_mode is 'bitmap' or 'incremental'
1194  * @bitmap_mode: The bitmap synchronization policy to use.
1195  * @on_source_error: The action to take upon error reading from the source.
1196  * @on_target_error: The action to take upon error writing to the target.
1197  * @creation_flags: Flags that control the behavior of the Job lifetime.
1198  *                  See @BlockJobCreateFlags
1199  * @cb: Completion function for the job.
1200  * @opaque: Opaque pointer value passed to @cb.
1201  * @txn: Transaction that this job is part of (may be NULL).
1202  *
1203  * Create a backup operation on @bs.  Clusters in @bs are written to @target
1204  * until the job is cancelled or manually completed.
1205  */
1206 BlockJob *backup_job_create(const char *job_id, BlockDriverState *bs,
1207                             BlockDriverState *target, int64_t speed,
1208                             MirrorSyncMode sync_mode,
1209                             BdrvDirtyBitmap *sync_bitmap,
1210                             BitmapSyncMode bitmap_mode,
1211                             bool compress,
1212                             const char *filter_node_name,
1213                             BlockdevOnError on_source_error,
1214                             BlockdevOnError on_target_error,
1215                             int creation_flags,
1216                             BlockCompletionFunc *cb, void *opaque,
1217                             JobTxn *txn, Error **errp);
1218 
1219 void hmp_drive_add_node(Monitor *mon, const char *optstr);
1220 
1221 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
1222                                   const char *child_name,
1223                                   const BdrvChildRole *child_role,
1224                                   AioContext *ctx,
1225                                   uint64_t perm, uint64_t shared_perm,
1226                                   void *opaque, Error **errp);
1227 void bdrv_root_unref_child(BdrvChild *child);
1228 
1229 /**
1230  * Sets a BdrvChild's permissions.  Avoid if the parent is a BDS; use
1231  * bdrv_child_refresh_perms() instead and make the parent's
1232  * .bdrv_child_perm() implementation return the correct values.
1233  */
1234 int bdrv_child_try_set_perm(BdrvChild *c, uint64_t perm, uint64_t shared,
1235                             Error **errp);
1236 
1237 /**
1238  * Calls bs->drv->bdrv_child_perm() and updates the child's permission
1239  * masks with the result.
1240  * Drivers should invoke this function whenever an event occurs that
1241  * makes their .bdrv_child_perm() implementation return different
1242  * values than before, but which will not result in the block layer
1243  * automatically refreshing the permissions.
1244  */
1245 int bdrv_child_refresh_perms(BlockDriverState *bs, BdrvChild *c, Error **errp);
1246 
1247 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1248  * block filters: Forward CONSISTENT_READ, WRITE, WRITE_UNCHANGED and RESIZE to
1249  * all children */
1250 void bdrv_filter_default_perms(BlockDriverState *bs, BdrvChild *c,
1251                                const BdrvChildRole *role,
1252                                BlockReopenQueue *reopen_queue,
1253                                uint64_t perm, uint64_t shared,
1254                                uint64_t *nperm, uint64_t *nshared);
1255 
1256 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1257  * (non-raw) image formats: Like above for bs->backing, but for bs->file it
1258  * requires WRITE | RESIZE for read-write images, always requires
1259  * CONSISTENT_READ and doesn't share WRITE. */
1260 void bdrv_format_default_perms(BlockDriverState *bs, BdrvChild *c,
1261                                const BdrvChildRole *role,
1262                                BlockReopenQueue *reopen_queue,
1263                                uint64_t perm, uint64_t shared,
1264                                uint64_t *nperm, uint64_t *nshared);
1265 
1266 /*
1267  * Default implementation for drivers to pass bdrv_co_block_status() to
1268  * their file.
1269  */
1270 int coroutine_fn bdrv_co_block_status_from_file(BlockDriverState *bs,
1271                                                 bool want_zero,
1272                                                 int64_t offset,
1273                                                 int64_t bytes,
1274                                                 int64_t *pnum,
1275                                                 int64_t *map,
1276                                                 BlockDriverState **file);
1277 /*
1278  * Default implementation for drivers to pass bdrv_co_block_status() to
1279  * their backing file.
1280  */
1281 int coroutine_fn bdrv_co_block_status_from_backing(BlockDriverState *bs,
1282                                                    bool want_zero,
1283                                                    int64_t offset,
1284                                                    int64_t bytes,
1285                                                    int64_t *pnum,
1286                                                    int64_t *map,
1287                                                    BlockDriverState **file);
1288 const char *bdrv_get_parent_name(const BlockDriverState *bs);
1289 void blk_dev_change_media_cb(BlockBackend *blk, bool load, Error **errp);
1290 bool blk_dev_has_removable_media(BlockBackend *blk);
1291 bool blk_dev_has_tray(BlockBackend *blk);
1292 void blk_dev_eject_request(BlockBackend *blk, bool force);
1293 bool blk_dev_is_tray_open(BlockBackend *blk);
1294 bool blk_dev_is_medium_locked(BlockBackend *blk);
1295 
1296 void bdrv_set_dirty(BlockDriverState *bs, int64_t offset, int64_t bytes);
1297 
1298 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap **out);
1299 void bdrv_restore_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap *backup);
1300 bool bdrv_dirty_bitmap_merge_internal(BdrvDirtyBitmap *dest,
1301                                       const BdrvDirtyBitmap *src,
1302                                       HBitmap **backup, bool lock);
1303 
1304 void bdrv_inc_in_flight(BlockDriverState *bs);
1305 void bdrv_dec_in_flight(BlockDriverState *bs);
1306 
1307 void blockdev_close_all_bdrv_states(void);
1308 
1309 int coroutine_fn bdrv_co_copy_range_from(BdrvChild *src, uint64_t src_offset,
1310                                          BdrvChild *dst, uint64_t dst_offset,
1311                                          uint64_t bytes,
1312                                          BdrvRequestFlags read_flags,
1313                                          BdrvRequestFlags write_flags);
1314 int coroutine_fn bdrv_co_copy_range_to(BdrvChild *src, uint64_t src_offset,
1315                                        BdrvChild *dst, uint64_t dst_offset,
1316                                        uint64_t bytes,
1317                                        BdrvRequestFlags read_flags,
1318                                        BdrvRequestFlags write_flags);
1319 
1320 int refresh_total_sectors(BlockDriverState *bs, int64_t hint);
1321 
1322 #endif /* BLOCK_INT_H */
1323