xref: /openbmc/qemu/include/block/block_int.h (revision ed5abf46)
1 /*
2  * QEMU System Emulator block driver
3  *
4  * Copyright (c) 2003 Fabrice Bellard
5  *
6  * Permission is hereby granted, free of charge, to any person obtaining a copy
7  * of this software and associated documentation files (the "Software"), to deal
8  * in the Software without restriction, including without limitation the rights
9  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10  * copies of the Software, and to permit persons to whom the Software is
11  * furnished to do so, subject to the following conditions:
12  *
13  * The above copyright notice and this permission notice shall be included in
14  * all copies or substantial portions of the Software.
15  *
16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22  * THE SOFTWARE.
23  */
24 #ifndef BLOCK_INT_H
25 #define BLOCK_INT_H
26 
27 #include "block/accounting.h"
28 #include "block/block.h"
29 #include "block/aio-wait.h"
30 #include "qemu/queue.h"
31 #include "qemu/coroutine.h"
32 #include "qemu/stats64.h"
33 #include "qemu/timer.h"
34 #include "qemu/hbitmap.h"
35 #include "block/snapshot.h"
36 #include "qemu/throttle.h"
37 
38 #define BLOCK_FLAG_LAZY_REFCOUNTS   8
39 
40 #define BLOCK_OPT_SIZE              "size"
41 #define BLOCK_OPT_ENCRYPT           "encryption"
42 #define BLOCK_OPT_ENCRYPT_FORMAT    "encrypt.format"
43 #define BLOCK_OPT_COMPAT6           "compat6"
44 #define BLOCK_OPT_HWVERSION         "hwversion"
45 #define BLOCK_OPT_BACKING_FILE      "backing_file"
46 #define BLOCK_OPT_BACKING_FMT       "backing_fmt"
47 #define BLOCK_OPT_CLUSTER_SIZE      "cluster_size"
48 #define BLOCK_OPT_TABLE_SIZE        "table_size"
49 #define BLOCK_OPT_PREALLOC          "preallocation"
50 #define BLOCK_OPT_SUBFMT            "subformat"
51 #define BLOCK_OPT_COMPAT_LEVEL      "compat"
52 #define BLOCK_OPT_LAZY_REFCOUNTS    "lazy_refcounts"
53 #define BLOCK_OPT_ADAPTER_TYPE      "adapter_type"
54 #define BLOCK_OPT_REDUNDANCY        "redundancy"
55 #define BLOCK_OPT_NOCOW             "nocow"
56 #define BLOCK_OPT_OBJECT_SIZE       "object_size"
57 #define BLOCK_OPT_REFCOUNT_BITS     "refcount_bits"
58 #define BLOCK_OPT_DATA_FILE         "data_file"
59 #define BLOCK_OPT_DATA_FILE_RAW     "data_file_raw"
60 
61 #define BLOCK_PROBE_BUF_SIZE        512
62 
63 enum BdrvTrackedRequestType {
64     BDRV_TRACKED_READ,
65     BDRV_TRACKED_WRITE,
66     BDRV_TRACKED_DISCARD,
67     BDRV_TRACKED_TRUNCATE,
68 };
69 
70 typedef struct BdrvTrackedRequest {
71     BlockDriverState *bs;
72     int64_t offset;
73     uint64_t bytes;
74     enum BdrvTrackedRequestType type;
75 
76     bool serialising;
77     int64_t overlap_offset;
78     uint64_t overlap_bytes;
79 
80     QLIST_ENTRY(BdrvTrackedRequest) list;
81     Coroutine *co; /* owner, used for deadlock detection */
82     CoQueue wait_queue; /* coroutines blocked on this request */
83 
84     struct BdrvTrackedRequest *waiting_for;
85 } BdrvTrackedRequest;
86 
87 struct BlockDriver {
88     const char *format_name;
89     int instance_size;
90 
91     /* set to true if the BlockDriver is a block filter. Block filters pass
92      * certain callbacks that refer to data (see block.c) to their bs->file if
93      * the driver doesn't implement them. Drivers that do not wish to forward
94      * must implement them and return -ENOTSUP.
95      */
96     bool is_filter;
97     /*
98      * Return true if @to_replace can be replaced by a BDS with the
99      * same data as @bs without it affecting @bs's behavior (that is,
100      * without it being visible to @bs's parents).
101      */
102     bool (*bdrv_recurse_can_replace)(BlockDriverState *bs,
103                                      BlockDriverState *to_replace);
104 
105     int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename);
106     int (*bdrv_probe_device)(const char *filename);
107 
108     /* Any driver implementing this callback is expected to be able to handle
109      * NULL file names in its .bdrv_open() implementation */
110     void (*bdrv_parse_filename)(const char *filename, QDict *options, Error **errp);
111     /* Drivers not implementing bdrv_parse_filename nor bdrv_open should have
112      * this field set to true, except ones that are defined only by their
113      * child's bs.
114      * An example of the last type will be the quorum block driver.
115      */
116     bool bdrv_needs_filename;
117 
118     /* Set if a driver can support backing files */
119     bool supports_backing;
120 
121     /* For handling image reopen for split or non-split files */
122     int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state,
123                                BlockReopenQueue *queue, Error **errp);
124     void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state);
125     void (*bdrv_reopen_commit_post)(BDRVReopenState *reopen_state);
126     void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state);
127     void (*bdrv_join_options)(QDict *options, QDict *old_options);
128 
129     int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags,
130                      Error **errp);
131 
132     /* Protocol drivers should implement this instead of bdrv_open */
133     int (*bdrv_file_open)(BlockDriverState *bs, QDict *options, int flags,
134                           Error **errp);
135     void (*bdrv_close)(BlockDriverState *bs);
136     int coroutine_fn (*bdrv_co_create)(BlockdevCreateOptions *opts,
137                                        Error **errp);
138     int coroutine_fn (*bdrv_co_create_opts)(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     /* Delete a created file. */
318     int coroutine_fn (*bdrv_co_delete_file)(BlockDriverState *bs,
319                                             Error **errp);
320 
321     /*
322      * Flushes all data that was already written to the OS all the way down to
323      * the disk (for example file-posix.c calls fsync()).
324      */
325     int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs);
326 
327     /*
328      * Flushes all internal caches to the OS. The data may still sit in a
329      * writeback cache of the host OS, but it will survive a crash of the qemu
330      * process.
331      */
332     int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs);
333 
334     /*
335      * Drivers setting this field must be able to work with just a plain
336      * filename with '<protocol_name>:' as a prefix, and no other options.
337      * Options may be extracted from the filename by implementing
338      * bdrv_parse_filename.
339      */
340     const char *protocol_name;
341 
342     /*
343      * Truncate @bs to @offset bytes using the given @prealloc mode
344      * when growing.  Modes other than PREALLOC_MODE_OFF should be
345      * rejected when shrinking @bs.
346      *
347      * If @exact is true, @bs must be resized to exactly @offset.
348      * Otherwise, it is sufficient for @bs (if it is a host block
349      * device and thus there is no way to resize it) to be at least
350      * @offset bytes in length.
351      *
352      * If @exact is true and this function fails but would succeed
353      * with @exact = false, it should return -ENOTSUP.
354      */
355     int coroutine_fn (*bdrv_co_truncate)(BlockDriverState *bs, int64_t offset,
356                                          bool exact, PreallocMode prealloc,
357                                          Error **errp);
358 
359     int64_t (*bdrv_getlength)(BlockDriverState *bs);
360     bool has_variable_length;
361     int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
362     BlockMeasureInfo *(*bdrv_measure)(QemuOpts *opts, BlockDriverState *in_bs,
363                                       Error **errp);
364 
365     int coroutine_fn (*bdrv_co_pwritev_compressed)(BlockDriverState *bs,
366         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov);
367     int coroutine_fn (*bdrv_co_pwritev_compressed_part)(BlockDriverState *bs,
368         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov,
369         size_t qiov_offset);
370 
371     int (*bdrv_snapshot_create)(BlockDriverState *bs,
372                                 QEMUSnapshotInfo *sn_info);
373     int (*bdrv_snapshot_goto)(BlockDriverState *bs,
374                               const char *snapshot_id);
375     int (*bdrv_snapshot_delete)(BlockDriverState *bs,
376                                 const char *snapshot_id,
377                                 const char *name,
378                                 Error **errp);
379     int (*bdrv_snapshot_list)(BlockDriverState *bs,
380                               QEMUSnapshotInfo **psn_info);
381     int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
382                                   const char *snapshot_id,
383                                   const char *name,
384                                   Error **errp);
385     int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
386     ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs,
387                                                  Error **errp);
388     BlockStatsSpecific *(*bdrv_get_specific_stats)(BlockDriverState *bs);
389 
390     int coroutine_fn (*bdrv_save_vmstate)(BlockDriverState *bs,
391                                           QEMUIOVector *qiov,
392                                           int64_t pos);
393     int coroutine_fn (*bdrv_load_vmstate)(BlockDriverState *bs,
394                                           QEMUIOVector *qiov,
395                                           int64_t pos);
396 
397     int (*bdrv_change_backing_file)(BlockDriverState *bs,
398         const char *backing_file, const char *backing_fmt);
399 
400     /* removable device specific */
401     bool (*bdrv_is_inserted)(BlockDriverState *bs);
402     void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
403     void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
404 
405     /* to control generic scsi devices */
406     BlockAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
407         unsigned long int req, void *buf,
408         BlockCompletionFunc *cb, void *opaque);
409     int coroutine_fn (*bdrv_co_ioctl)(BlockDriverState *bs,
410                                       unsigned long int req, void *buf);
411 
412     /* List of options for creating images, terminated by name == NULL */
413     QemuOptsList *create_opts;
414     /*
415      * If this driver supports reopening images this contains a
416      * NULL-terminated list of the runtime options that can be
417      * modified. If an option in this list is unspecified during
418      * reopen then it _must_ be reset to its default value or return
419      * an error.
420      */
421     const char *const *mutable_opts;
422 
423     /*
424      * Returns 0 for completed check, -errno for internal errors.
425      * The check results are stored in result.
426      */
427     int coroutine_fn (*bdrv_co_check)(BlockDriverState *bs,
428                                       BdrvCheckResult *result,
429                                       BdrvCheckMode fix);
430 
431     int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts,
432                               BlockDriverAmendStatusCB *status_cb,
433                               void *cb_opaque,
434                               Error **errp);
435 
436     void (*bdrv_debug_event)(BlockDriverState *bs, BlkdebugEvent event);
437 
438     /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
439     int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
440         const char *tag);
441     int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs,
442         const char *tag);
443     int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
444     bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
445 
446     void (*bdrv_refresh_limits)(BlockDriverState *bs, Error **errp);
447 
448     /*
449      * Returns 1 if newly created images are guaranteed to contain only
450      * zeros, 0 otherwise.
451      * Must return 0 if .bdrv_has_zero_init_truncate() returns 0.
452      */
453     int (*bdrv_has_zero_init)(BlockDriverState *bs);
454 
455     /*
456      * Returns 1 if new areas added by growing the image with
457      * PREALLOC_MODE_OFF contain only zeros, 0 otherwise.
458      */
459     int (*bdrv_has_zero_init_truncate)(BlockDriverState *bs);
460 
461     /* Remove fd handlers, timers, and other event loop callbacks so the event
462      * loop is no longer in use.  Called with no in-flight requests and in
463      * depth-first traversal order with parents before child nodes.
464      */
465     void (*bdrv_detach_aio_context)(BlockDriverState *bs);
466 
467     /* Add fd handlers, timers, and other event loop callbacks so I/O requests
468      * can be processed again.  Called with no in-flight requests and in
469      * depth-first traversal order with child nodes before parent nodes.
470      */
471     void (*bdrv_attach_aio_context)(BlockDriverState *bs,
472                                     AioContext *new_context);
473 
474     /* io queue for linux-aio */
475     void (*bdrv_io_plug)(BlockDriverState *bs);
476     void (*bdrv_io_unplug)(BlockDriverState *bs);
477 
478     /**
479      * Try to get @bs's logical and physical block size.
480      * On success, store them in @bsz and return zero.
481      * On failure, return negative errno.
482      */
483     int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz);
484     /**
485      * Try to get @bs's geometry (cyls, heads, sectors)
486      * On success, store them in @geo and return 0.
487      * On failure return -errno.
488      * Only drivers that want to override guest geometry implement this
489      * callback; see hd_geometry_guess().
490      */
491     int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo);
492 
493     /**
494      * bdrv_co_drain_begin is called if implemented in the beginning of a
495      * drain operation to drain and stop any internal sources of requests in
496      * the driver.
497      * bdrv_co_drain_end is called if implemented at the end of the drain.
498      *
499      * They should be used by the driver to e.g. manage scheduled I/O
500      * requests, or toggle an internal state. After the end of the drain new
501      * requests will continue normally.
502      */
503     void coroutine_fn (*bdrv_co_drain_begin)(BlockDriverState *bs);
504     void coroutine_fn (*bdrv_co_drain_end)(BlockDriverState *bs);
505 
506     void (*bdrv_add_child)(BlockDriverState *parent, BlockDriverState *child,
507                            Error **errp);
508     void (*bdrv_del_child)(BlockDriverState *parent, BdrvChild *child,
509                            Error **errp);
510 
511     /**
512      * Informs the block driver that a permission change is intended. The
513      * driver checks whether the change is permissible and may take other
514      * preparations for the change (e.g. get file system locks). This operation
515      * is always followed either by a call to either .bdrv_set_perm or
516      * .bdrv_abort_perm_update.
517      *
518      * Checks whether the requested set of cumulative permissions in @perm
519      * can be granted for accessing @bs and whether no other users are using
520      * permissions other than those given in @shared (both arguments take
521      * BLK_PERM_* bitmasks).
522      *
523      * If both conditions are met, 0 is returned. Otherwise, -errno is returned
524      * and errp is set to an error describing the conflict.
525      */
526     int (*bdrv_check_perm)(BlockDriverState *bs, uint64_t perm,
527                            uint64_t shared, Error **errp);
528 
529     /**
530      * Called to inform the driver that the set of cumulative set of used
531      * permissions for @bs has changed to @perm, and the set of sharable
532      * permission to @shared. The driver can use this to propagate changes to
533      * its children (i.e. request permissions only if a parent actually needs
534      * them).
535      *
536      * This function is only invoked after bdrv_check_perm(), so block drivers
537      * may rely on preparations made in their .bdrv_check_perm implementation.
538      */
539     void (*bdrv_set_perm)(BlockDriverState *bs, uint64_t perm, uint64_t shared);
540 
541     /*
542      * Called to inform the driver that after a previous bdrv_check_perm()
543      * call, the permission update is not performed and any preparations made
544      * for it (e.g. taken file locks) need to be undone.
545      *
546      * This function can be called even for nodes that never saw a
547      * bdrv_check_perm() call. It is a no-op then.
548      */
549     void (*bdrv_abort_perm_update)(BlockDriverState *bs);
550 
551     /**
552      * Returns in @nperm and @nshared the permissions that the driver for @bs
553      * needs on its child @c, based on the cumulative permissions requested by
554      * the parents in @parent_perm and @parent_shared.
555      *
556      * If @c is NULL, return the permissions for attaching a new child for the
557      * given @role.
558      *
559      * If @reopen_queue is non-NULL, don't return the currently needed
560      * permissions, but those that will be needed after applying the
561      * @reopen_queue.
562      */
563      void (*bdrv_child_perm)(BlockDriverState *bs, BdrvChild *c,
564                              const BdrvChildRole *role,
565                              BlockReopenQueue *reopen_queue,
566                              uint64_t parent_perm, uint64_t parent_shared,
567                              uint64_t *nperm, uint64_t *nshared);
568 
569     bool (*bdrv_co_can_store_new_dirty_bitmap)(BlockDriverState *bs,
570                                                const char *name,
571                                                uint32_t granularity,
572                                                Error **errp);
573     int (*bdrv_co_remove_persistent_dirty_bitmap)(BlockDriverState *bs,
574                                                   const char *name,
575                                                   Error **errp);
576 
577     /**
578      * Register/unregister a buffer for I/O. For example, when the driver is
579      * interested to know the memory areas that will later be used in iovs, so
580      * that it can do IOMMU mapping with VFIO etc., in order to get better
581      * performance. In the case of VFIO drivers, this callback is used to do
582      * DMA mapping for hot buffers.
583      */
584     void (*bdrv_register_buf)(BlockDriverState *bs, void *host, size_t size);
585     void (*bdrv_unregister_buf)(BlockDriverState *bs, void *host);
586     QLIST_ENTRY(BlockDriver) list;
587 
588     /* Pointer to a NULL-terminated array of names of strong options
589      * that can be specified for bdrv_open(). A strong option is one
590      * that changes the data of a BDS.
591      * If this pointer is NULL, the array is considered empty.
592      * "filename" and "driver" are always considered strong. */
593     const char *const *strong_runtime_opts;
594 };
595 
596 static inline bool block_driver_can_compress(BlockDriver *drv)
597 {
598     return drv->bdrv_co_pwritev_compressed ||
599            drv->bdrv_co_pwritev_compressed_part;
600 }
601 
602 typedef struct BlockLimits {
603     /* Alignment requirement, in bytes, for offset/length of I/O
604      * requests. Must be a power of 2 less than INT_MAX; defaults to
605      * 1 for drivers with modern byte interfaces, and to 512
606      * otherwise. */
607     uint32_t request_alignment;
608 
609     /* Maximum number of bytes that can be discarded at once (since it
610      * is signed, it must be < 2G, if set). Must be multiple of
611      * pdiscard_alignment, but need not be power of 2. May be 0 if no
612      * inherent 32-bit limit */
613     int32_t max_pdiscard;
614 
615     /* Optimal alignment for discard requests in bytes. A power of 2
616      * is best but not mandatory.  Must be a multiple of
617      * bl.request_alignment, and must be less than max_pdiscard if
618      * that is set. May be 0 if bl.request_alignment is good enough */
619     uint32_t pdiscard_alignment;
620 
621     /* Maximum number of bytes that can zeroized at once (since it is
622      * signed, it must be < 2G, if set). Must be multiple of
623      * pwrite_zeroes_alignment. May be 0 if no inherent 32-bit limit */
624     int32_t max_pwrite_zeroes;
625 
626     /* Optimal alignment for write zeroes requests in bytes. A power
627      * of 2 is best but not mandatory.  Must be a multiple of
628      * bl.request_alignment, and must be less than max_pwrite_zeroes
629      * if that is set. May be 0 if bl.request_alignment is good
630      * enough */
631     uint32_t pwrite_zeroes_alignment;
632 
633     /* Optimal transfer length in bytes.  A power of 2 is best but not
634      * mandatory.  Must be a multiple of bl.request_alignment, or 0 if
635      * no preferred size */
636     uint32_t opt_transfer;
637 
638     /* Maximal transfer length in bytes.  Need not be power of 2, but
639      * must be multiple of opt_transfer and bl.request_alignment, or 0
640      * for no 32-bit limit.  For now, anything larger than INT_MAX is
641      * clamped down. */
642     uint32_t max_transfer;
643 
644     /* memory alignment, in bytes so that no bounce buffer is needed */
645     size_t min_mem_alignment;
646 
647     /* memory alignment, in bytes, for bounce buffer */
648     size_t opt_mem_alignment;
649 
650     /* maximum number of iovec elements */
651     int max_iov;
652 } BlockLimits;
653 
654 typedef struct BdrvOpBlocker BdrvOpBlocker;
655 
656 typedef struct BdrvAioNotifier {
657     void (*attached_aio_context)(AioContext *new_context, void *opaque);
658     void (*detach_aio_context)(void *opaque);
659 
660     void *opaque;
661     bool deleted;
662 
663     QLIST_ENTRY(BdrvAioNotifier) list;
664 } BdrvAioNotifier;
665 
666 struct BdrvChildRole {
667     /* If true, bdrv_replace_node() doesn't change the node this BdrvChild
668      * points to. */
669     bool stay_at_node;
670 
671     /* If true, the parent is a BlockDriverState and bdrv_next_all_states()
672      * will return it. This information is used for drain_all, where every node
673      * will be drained separately, so the drain only needs to be propagated to
674      * non-BDS parents. */
675     bool parent_is_bds;
676 
677     void (*inherit_options)(int *child_flags, QDict *child_options,
678                             int parent_flags, QDict *parent_options);
679 
680     void (*change_media)(BdrvChild *child, bool load);
681     void (*resize)(BdrvChild *child);
682 
683     /* Returns a name that is supposedly more useful for human users than the
684      * node name for identifying the node in question (in particular, a BB
685      * name), or NULL if the parent can't provide a better name. */
686     const char *(*get_name)(BdrvChild *child);
687 
688     /* Returns a malloced string that describes the parent of the child for a
689      * human reader. This could be a node-name, BlockBackend name, qdev ID or
690      * QOM path of the device owning the BlockBackend, job type and ID etc. The
691      * caller is responsible for freeing the memory. */
692     char *(*get_parent_desc)(BdrvChild *child);
693 
694     /*
695      * If this pair of functions is implemented, the parent doesn't issue new
696      * requests after returning from .drained_begin() until .drained_end() is
697      * called.
698      *
699      * These functions must not change the graph (and therefore also must not
700      * call aio_poll(), which could change the graph indirectly).
701      *
702      * If drained_end() schedules background operations, it must atomically
703      * increment *drained_end_counter for each such operation and atomically
704      * decrement it once the operation has settled.
705      *
706      * Note that this can be nested. If drained_begin() was called twice, new
707      * I/O is allowed only after drained_end() was called twice, too.
708      */
709     void (*drained_begin)(BdrvChild *child);
710     void (*drained_end)(BdrvChild *child, int *drained_end_counter);
711 
712     /*
713      * Returns whether the parent has pending requests for the child. This
714      * callback is polled after .drained_begin() has been called until all
715      * activity on the child has stopped.
716      */
717     bool (*drained_poll)(BdrvChild *child);
718 
719     /* Notifies the parent that the child has been activated/inactivated (e.g.
720      * when migration is completing) and it can start/stop requesting
721      * permissions and doing I/O on it. */
722     void (*activate)(BdrvChild *child, Error **errp);
723     int (*inactivate)(BdrvChild *child);
724 
725     void (*attach)(BdrvChild *child);
726     void (*detach)(BdrvChild *child);
727 
728     /* Notifies the parent that the filename of its child has changed (e.g.
729      * because the direct child was removed from the backing chain), so that it
730      * can update its reference. */
731     int (*update_filename)(BdrvChild *child, BlockDriverState *new_base,
732                            const char *filename, Error **errp);
733 
734     bool (*can_set_aio_ctx)(BdrvChild *child, AioContext *ctx,
735                             GSList **ignore, Error **errp);
736     void (*set_aio_ctx)(BdrvChild *child, AioContext *ctx, GSList **ignore);
737 };
738 
739 extern const BdrvChildRole child_file;
740 extern const BdrvChildRole child_format;
741 extern const BdrvChildRole child_backing;
742 
743 struct BdrvChild {
744     BlockDriverState *bs;
745     char *name;
746     const BdrvChildRole *role;
747     void *opaque;
748 
749     /**
750      * Granted permissions for operating on this BdrvChild (BLK_PERM_* bitmask)
751      */
752     uint64_t perm;
753 
754     /**
755      * Permissions that can still be granted to other users of @bs while this
756      * BdrvChild is still attached to it. (BLK_PERM_* bitmask)
757      */
758     uint64_t shared_perm;
759 
760     /* backup of permissions during permission update procedure */
761     bool has_backup_perm;
762     uint64_t backup_perm;
763     uint64_t backup_shared_perm;
764 
765     /*
766      * This link is frozen: the child can neither be replaced nor
767      * detached from the parent.
768      */
769     bool frozen;
770 
771     /*
772      * How many times the parent of this child has been drained
773      * (through role->drained_*).
774      * Usually, this is equal to bs->quiesce_counter (potentially
775      * reduced by bdrv_drain_all_count).  It may differ while the
776      * child is entering or leaving a drained section.
777      */
778     int parent_quiesce_counter;
779 
780     QLIST_ENTRY(BdrvChild) next;
781     QLIST_ENTRY(BdrvChild) next_parent;
782 };
783 
784 /*
785  * Note: the function bdrv_append() copies and swaps contents of
786  * BlockDriverStates, so if you add new fields to this struct, please
787  * inspect bdrv_append() to determine if the new fields need to be
788  * copied as well.
789  */
790 struct BlockDriverState {
791     /* Protected by big QEMU lock or read-only after opening.  No special
792      * locking needed during I/O...
793      */
794     int open_flags; /* flags used to open the file, re-used for re-open */
795     bool read_only; /* if true, the media is read only */
796     bool encrypted; /* if true, the media is encrypted */
797     bool sg;        /* if true, the device is a /dev/sg* */
798     bool probed;    /* if true, format was probed rather than specified */
799     bool force_share; /* if true, always allow all shared permissions */
800     bool implicit;  /* if true, this filter node was automatically inserted */
801 
802     BlockDriver *drv; /* NULL means no media */
803     void *opaque;
804 
805     AioContext *aio_context; /* event loop used for fd handlers, timers, etc */
806     /* long-running tasks intended to always use the same AioContext as this
807      * BDS may register themselves in this list to be notified of changes
808      * regarding this BDS's context */
809     QLIST_HEAD(, BdrvAioNotifier) aio_notifiers;
810     bool walking_aio_notifiers; /* to make removal during iteration safe */
811 
812     char filename[PATH_MAX];
813     char backing_file[PATH_MAX]; /* if non zero, the image is a diff of
814                                     this file image */
815     /* The backing filename indicated by the image header; if we ever
816      * open this file, then this is replaced by the resulting BDS's
817      * filename (i.e. after a bdrv_refresh_filename() run). */
818     char auto_backing_file[PATH_MAX];
819     char backing_format[16]; /* if non-zero and backing_file exists */
820 
821     QDict *full_open_options;
822     char exact_filename[PATH_MAX];
823 
824     BdrvChild *backing;
825     BdrvChild *file;
826 
827     /* I/O Limits */
828     BlockLimits bl;
829 
830     /* Flags honored during pwrite (so far: BDRV_REQ_FUA,
831      * BDRV_REQ_WRITE_UNCHANGED).
832      * If a driver does not support BDRV_REQ_WRITE_UNCHANGED, those
833      * writes will be issued as normal writes without the flag set.
834      * This is important to note for drivers that do not explicitly
835      * request a WRITE permission for their children and instead take
836      * the same permissions as their parent did (this is commonly what
837      * block filters do).  Such drivers have to be aware that the
838      * parent may have taken a WRITE_UNCHANGED permission only and is
839      * issuing such requests.  Drivers either must make sure that
840      * these requests do not result in plain WRITE accesses (usually
841      * by supporting BDRV_REQ_WRITE_UNCHANGED, and then forwarding
842      * every incoming write request as-is, including potentially that
843      * flag), or they have to explicitly take the WRITE permission for
844      * their children. */
845     unsigned int supported_write_flags;
846     /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA,
847      * BDRV_REQ_MAY_UNMAP, BDRV_REQ_WRITE_UNCHANGED) */
848     unsigned int supported_zero_flags;
849 
850     /* the following member gives a name to every node on the bs graph. */
851     char node_name[32];
852     /* element of the list of named nodes building the graph */
853     QTAILQ_ENTRY(BlockDriverState) node_list;
854     /* element of the list of all BlockDriverStates (all_bdrv_states) */
855     QTAILQ_ENTRY(BlockDriverState) bs_list;
856     /* element of the list of monitor-owned BDS */
857     QTAILQ_ENTRY(BlockDriverState) monitor_list;
858     int refcnt;
859 
860     /* operation blockers */
861     QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX];
862 
863     /* The node that this node inherited default options from (and a reopen on
864      * which can affect this node by changing these defaults). This is always a
865      * parent node of this node. */
866     BlockDriverState *inherits_from;
867     QLIST_HEAD(, BdrvChild) children;
868     QLIST_HEAD(, BdrvChild) parents;
869 
870     QDict *options;
871     QDict *explicit_options;
872     BlockdevDetectZeroesOptions detect_zeroes;
873 
874     /* The error object in use for blocking operations on backing_hd */
875     Error *backing_blocker;
876 
877     /* Protected by AioContext lock */
878 
879     /* If we are reading a disk image, give its size in sectors.
880      * Generally read-only; it is written to by load_snapshot and
881      * save_snaphost, but the block layer is quiescent during those.
882      */
883     int64_t total_sectors;
884 
885     /* Callback before write request is processed */
886     NotifierWithReturnList before_write_notifiers;
887 
888     /* threshold limit for writes, in bytes. "High water mark". */
889     uint64_t write_threshold_offset;
890     NotifierWithReturn write_threshold_notifier;
891 
892     /* Writing to the list requires the BQL _and_ the dirty_bitmap_mutex.
893      * Reading from the list can be done with either the BQL or the
894      * dirty_bitmap_mutex.  Modifying a bitmap only requires
895      * dirty_bitmap_mutex.  */
896     QemuMutex dirty_bitmap_mutex;
897     QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps;
898 
899     /* Offset after the highest byte written to */
900     Stat64 wr_highest_offset;
901 
902     /* If true, copy read backing sectors into image.  Can be >1 if more
903      * than one client has requested copy-on-read.  Accessed with atomic
904      * ops.
905      */
906     int copy_on_read;
907 
908     /* number of in-flight requests; overall and serialising.
909      * Accessed with atomic ops.
910      */
911     unsigned int in_flight;
912     unsigned int serialising_in_flight;
913 
914     /* counter for nested bdrv_io_plug.
915      * Accessed with atomic ops.
916     */
917     unsigned io_plugged;
918 
919     /* do we need to tell the quest if we have a volatile write cache? */
920     int enable_write_cache;
921 
922     /* Accessed with atomic ops.  */
923     int quiesce_counter;
924     int recursive_quiesce_counter;
925 
926     unsigned int write_gen;               /* Current data generation */
927 
928     /* Protected by reqs_lock.  */
929     CoMutex reqs_lock;
930     QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
931     CoQueue flush_queue;                  /* Serializing flush queue */
932     bool active_flush_req;                /* Flush request in flight? */
933 
934     /* Only read/written by whoever has set active_flush_req to true.  */
935     unsigned int flushed_gen;             /* Flushed write generation */
936 
937     /* BdrvChild links to this node may never be frozen */
938     bool never_freeze;
939 };
940 
941 struct BlockBackendRootState {
942     int open_flags;
943     bool read_only;
944     BlockdevDetectZeroesOptions detect_zeroes;
945 };
946 
947 typedef enum BlockMirrorBackingMode {
948     /* Reuse the existing backing chain from the source for the target.
949      * - sync=full: Set backing BDS to NULL.
950      * - sync=top:  Use source's backing BDS.
951      * - sync=none: Use source as the backing BDS. */
952     MIRROR_SOURCE_BACKING_CHAIN,
953 
954     /* Open the target's backing chain completely anew */
955     MIRROR_OPEN_BACKING_CHAIN,
956 
957     /* Do not change the target's backing BDS after job completion */
958     MIRROR_LEAVE_BACKING_CHAIN,
959 } BlockMirrorBackingMode;
960 
961 static inline BlockDriverState *backing_bs(BlockDriverState *bs)
962 {
963     return bs->backing ? bs->backing->bs : NULL;
964 }
965 
966 
967 /* Essential block drivers which must always be statically linked into qemu, and
968  * which therefore can be accessed without using bdrv_find_format() */
969 extern BlockDriver bdrv_file;
970 extern BlockDriver bdrv_raw;
971 extern BlockDriver bdrv_qcow2;
972 
973 int coroutine_fn bdrv_co_preadv(BdrvChild *child,
974     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
975     BdrvRequestFlags flags);
976 int coroutine_fn bdrv_co_preadv_part(BdrvChild *child,
977     int64_t offset, unsigned int bytes,
978     QEMUIOVector *qiov, size_t qiov_offset, BdrvRequestFlags flags);
979 int coroutine_fn bdrv_co_pwritev(BdrvChild *child,
980     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
981     BdrvRequestFlags flags);
982 int coroutine_fn bdrv_co_pwritev_part(BdrvChild *child,
983     int64_t offset, unsigned int bytes,
984     QEMUIOVector *qiov, size_t qiov_offset, BdrvRequestFlags flags);
985 
986 static inline int coroutine_fn bdrv_co_pread(BdrvChild *child,
987     int64_t offset, unsigned int bytes, void *buf, BdrvRequestFlags flags)
988 {
989     QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes);
990 
991     return bdrv_co_preadv(child, offset, bytes, &qiov, flags);
992 }
993 
994 static inline int coroutine_fn bdrv_co_pwrite(BdrvChild *child,
995     int64_t offset, unsigned int bytes, void *buf, BdrvRequestFlags flags)
996 {
997     QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes);
998 
999     return bdrv_co_pwritev(child, offset, bytes, &qiov, flags);
1000 }
1001 
1002 extern unsigned int bdrv_drain_all_count;
1003 void bdrv_apply_subtree_drain(BdrvChild *child, BlockDriverState *new_parent);
1004 void bdrv_unapply_subtree_drain(BdrvChild *child, BlockDriverState *old_parent);
1005 
1006 bool coroutine_fn bdrv_mark_request_serialising(BdrvTrackedRequest *req, uint64_t align);
1007 BdrvTrackedRequest *coroutine_fn bdrv_co_get_self_request(BlockDriverState *bs);
1008 
1009 int get_tmp_filename(char *filename, int size);
1010 BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size,
1011                             const char *filename);
1012 
1013 void bdrv_parse_filename_strip_prefix(const char *filename, const char *prefix,
1014                                       QDict *options);
1015 
1016 
1017 /**
1018  * bdrv_add_before_write_notifier:
1019  *
1020  * Register a callback that is invoked before write requests are processed but
1021  * after any throttling or waiting for overlapping requests.
1022  */
1023 void bdrv_add_before_write_notifier(BlockDriverState *bs,
1024                                     NotifierWithReturn *notifier);
1025 
1026 /**
1027  * bdrv_add_aio_context_notifier:
1028  *
1029  * If a long-running job intends to be always run in the same AioContext as a
1030  * certain BDS, it may use this function to be notified of changes regarding the
1031  * association of the BDS to an AioContext.
1032  *
1033  * attached_aio_context() is called after the target BDS has been attached to a
1034  * new AioContext; detach_aio_context() is called before the target BDS is being
1035  * detached from its old AioContext.
1036  */
1037 void bdrv_add_aio_context_notifier(BlockDriverState *bs,
1038         void (*attached_aio_context)(AioContext *new_context, void *opaque),
1039         void (*detach_aio_context)(void *opaque), void *opaque);
1040 
1041 /**
1042  * bdrv_remove_aio_context_notifier:
1043  *
1044  * Unsubscribe of change notifications regarding the BDS's AioContext. The
1045  * parameters given here have to be the same as those given to
1046  * bdrv_add_aio_context_notifier().
1047  */
1048 void bdrv_remove_aio_context_notifier(BlockDriverState *bs,
1049                                       void (*aio_context_attached)(AioContext *,
1050                                                                    void *),
1051                                       void (*aio_context_detached)(void *),
1052                                       void *opaque);
1053 
1054 /**
1055  * bdrv_wakeup:
1056  * @bs: The BlockDriverState for which an I/O operation has been completed.
1057  *
1058  * Wake up the main thread if it is waiting on BDRV_POLL_WHILE.  During
1059  * synchronous I/O on a BlockDriverState that is attached to another
1060  * I/O thread, the main thread lets the I/O thread's event loop run,
1061  * waiting for the I/O operation to complete.  A bdrv_wakeup will wake
1062  * up the main thread if necessary.
1063  *
1064  * Manual calls to bdrv_wakeup are rarely necessary, because
1065  * bdrv_dec_in_flight already calls it.
1066  */
1067 void bdrv_wakeup(BlockDriverState *bs);
1068 
1069 #ifdef _WIN32
1070 int is_windows_drive(const char *filename);
1071 #endif
1072 
1073 /**
1074  * stream_start:
1075  * @job_id: The id of the newly-created job, or %NULL to use the
1076  * device name of @bs.
1077  * @bs: Block device to operate on.
1078  * @base: Block device that will become the new base, or %NULL to
1079  * flatten the whole backing file chain onto @bs.
1080  * @backing_file_str: The file name that will be written to @bs as the
1081  * the new backing file if the job completes. Ignored if @base is %NULL.
1082  * @creation_flags: Flags that control the behavior of the Job lifetime.
1083  *                  See @BlockJobCreateFlags
1084  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1085  * @on_error: The action to take upon error.
1086  * @errp: Error object.
1087  *
1088  * Start a streaming operation on @bs.  Clusters that are unallocated
1089  * in @bs, but allocated in any image between @base and @bs (both
1090  * exclusive) will be written to @bs.  At the end of a successful
1091  * streaming job, the backing file of @bs will be changed to
1092  * @backing_file_str in the written image and to @base in the live
1093  * BlockDriverState.
1094  */
1095 void stream_start(const char *job_id, BlockDriverState *bs,
1096                   BlockDriverState *base, const char *backing_file_str,
1097                   int creation_flags, int64_t speed,
1098                   BlockdevOnError on_error, Error **errp);
1099 
1100 /**
1101  * commit_start:
1102  * @job_id: The id of the newly-created job, or %NULL to use the
1103  * device name of @bs.
1104  * @bs: Active block device.
1105  * @top: Top block device to be committed.
1106  * @base: Block device that will be written into, and become the new top.
1107  * @creation_flags: Flags that control the behavior of the Job lifetime.
1108  *                  See @BlockJobCreateFlags
1109  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1110  * @on_error: The action to take upon error.
1111  * @backing_file_str: String to use as the backing file in @top's overlay
1112  * @filter_node_name: The node name that should be assigned to the filter
1113  * driver that the commit job inserts into the graph above @top. NULL means
1114  * that a node name should be autogenerated.
1115  * @errp: Error object.
1116  *
1117  */
1118 void commit_start(const char *job_id, BlockDriverState *bs,
1119                   BlockDriverState *base, BlockDriverState *top,
1120                   int creation_flags, int64_t speed,
1121                   BlockdevOnError on_error, const char *backing_file_str,
1122                   const char *filter_node_name, Error **errp);
1123 /**
1124  * commit_active_start:
1125  * @job_id: The id of the newly-created job, or %NULL to use the
1126  * device name of @bs.
1127  * @bs: Active block device to be committed.
1128  * @base: Block device that will be written into, and become the new top.
1129  * @creation_flags: Flags that control the behavior of the Job lifetime.
1130  *                  See @BlockJobCreateFlags
1131  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1132  * @on_error: The action to take upon error.
1133  * @filter_node_name: The node name that should be assigned to the filter
1134  * driver that the commit job inserts into the graph above @bs. NULL means that
1135  * a node name should be autogenerated.
1136  * @cb: Completion function for the job.
1137  * @opaque: Opaque pointer value passed to @cb.
1138  * @auto_complete: Auto complete the job.
1139  * @errp: Error object.
1140  *
1141  */
1142 BlockJob *commit_active_start(const char *job_id, BlockDriverState *bs,
1143                               BlockDriverState *base, int creation_flags,
1144                               int64_t speed, BlockdevOnError on_error,
1145                               const char *filter_node_name,
1146                               BlockCompletionFunc *cb, void *opaque,
1147                               bool auto_complete, Error **errp);
1148 /*
1149  * mirror_start:
1150  * @job_id: The id of the newly-created job, or %NULL to use the
1151  * device name of @bs.
1152  * @bs: Block device to operate on.
1153  * @target: Block device to write to.
1154  * @replaces: Block graph node name to replace once the mirror is done. Can
1155  *            only be used when full mirroring is selected.
1156  * @creation_flags: Flags that control the behavior of the Job lifetime.
1157  *                  See @BlockJobCreateFlags
1158  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1159  * @granularity: The chosen granularity for the dirty bitmap.
1160  * @buf_size: The amount of data that can be in flight at one time.
1161  * @mode: Whether to collapse all images in the chain to the target.
1162  * @backing_mode: How to establish the target's backing chain after completion.
1163  * @zero_target: Whether the target should be explicitly zero-initialized
1164  * @on_source_error: The action to take upon error reading from the source.
1165  * @on_target_error: The action to take upon error writing to the target.
1166  * @unmap: Whether to unmap target where source sectors only contain zeroes.
1167  * @filter_node_name: The node name that should be assigned to the filter
1168  * driver that the mirror job inserts into the graph above @bs. NULL means that
1169  * a node name should be autogenerated.
1170  * @copy_mode: When to trigger writes to the target.
1171  * @errp: Error object.
1172  *
1173  * Start a mirroring operation on @bs.  Clusters that are allocated
1174  * in @bs will be written to @target until the job is cancelled or
1175  * manually completed.  At the end of a successful mirroring job,
1176  * @bs will be switched to read from @target.
1177  */
1178 void mirror_start(const char *job_id, BlockDriverState *bs,
1179                   BlockDriverState *target, const char *replaces,
1180                   int creation_flags, int64_t speed,
1181                   uint32_t granularity, int64_t buf_size,
1182                   MirrorSyncMode mode, BlockMirrorBackingMode backing_mode,
1183                   bool zero_target,
1184                   BlockdevOnError on_source_error,
1185                   BlockdevOnError on_target_error,
1186                   bool unmap, const char *filter_node_name,
1187                   MirrorCopyMode copy_mode, Error **errp);
1188 
1189 /*
1190  * backup_job_create:
1191  * @job_id: The id of the newly-created job, or %NULL to use the
1192  * device name of @bs.
1193  * @bs: Block device to operate on.
1194  * @target: Block device to write to.
1195  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1196  * @sync_mode: What parts of the disk image should be copied to the destination.
1197  * @sync_bitmap: The dirty bitmap if sync_mode is 'bitmap' or 'incremental'
1198  * @bitmap_mode: The bitmap synchronization policy to use.
1199  * @on_source_error: The action to take upon error reading from the source.
1200  * @on_target_error: The action to take upon error writing to the target.
1201  * @creation_flags: Flags that control the behavior of the Job lifetime.
1202  *                  See @BlockJobCreateFlags
1203  * @cb: Completion function for the job.
1204  * @opaque: Opaque pointer value passed to @cb.
1205  * @txn: Transaction that this job is part of (may be NULL).
1206  *
1207  * Create a backup operation on @bs.  Clusters in @bs are written to @target
1208  * until the job is cancelled or manually completed.
1209  */
1210 BlockJob *backup_job_create(const char *job_id, BlockDriverState *bs,
1211                             BlockDriverState *target, int64_t speed,
1212                             MirrorSyncMode sync_mode,
1213                             BdrvDirtyBitmap *sync_bitmap,
1214                             BitmapSyncMode bitmap_mode,
1215                             bool compress,
1216                             const char *filter_node_name,
1217                             BlockdevOnError on_source_error,
1218                             BlockdevOnError on_target_error,
1219                             int creation_flags,
1220                             BlockCompletionFunc *cb, void *opaque,
1221                             JobTxn *txn, Error **errp);
1222 
1223 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
1224                                   const char *child_name,
1225                                   const BdrvChildRole *child_role,
1226                                   AioContext *ctx,
1227                                   uint64_t perm, uint64_t shared_perm,
1228                                   void *opaque, Error **errp);
1229 void bdrv_root_unref_child(BdrvChild *child);
1230 
1231 void bdrv_get_cumulative_perm(BlockDriverState *bs, uint64_t *perm,
1232                               uint64_t *shared_perm);
1233 
1234 /**
1235  * Sets a BdrvChild's permissions.  Avoid if the parent is a BDS; use
1236  * bdrv_child_refresh_perms() instead and make the parent's
1237  * .bdrv_child_perm() implementation return the correct values.
1238  */
1239 int bdrv_child_try_set_perm(BdrvChild *c, uint64_t perm, uint64_t shared,
1240                             Error **errp);
1241 
1242 /**
1243  * Calls bs->drv->bdrv_child_perm() and updates the child's permission
1244  * masks with the result.
1245  * Drivers should invoke this function whenever an event occurs that
1246  * makes their .bdrv_child_perm() implementation return different
1247  * values than before, but which will not result in the block layer
1248  * automatically refreshing the permissions.
1249  */
1250 int bdrv_child_refresh_perms(BlockDriverState *bs, BdrvChild *c, Error **errp);
1251 
1252 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1253  * block filters: Forward CONSISTENT_READ, WRITE, WRITE_UNCHANGED and RESIZE to
1254  * all children */
1255 void bdrv_filter_default_perms(BlockDriverState *bs, BdrvChild *c,
1256                                const BdrvChildRole *role,
1257                                BlockReopenQueue *reopen_queue,
1258                                uint64_t perm, uint64_t shared,
1259                                uint64_t *nperm, uint64_t *nshared);
1260 
1261 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1262  * (non-raw) image formats: Like above for bs->backing, but for bs->file it
1263  * requires WRITE | RESIZE for read-write images, always requires
1264  * CONSISTENT_READ and doesn't share WRITE. */
1265 void bdrv_format_default_perms(BlockDriverState *bs, BdrvChild *c,
1266                                const BdrvChildRole *role,
1267                                BlockReopenQueue *reopen_queue,
1268                                uint64_t perm, uint64_t shared,
1269                                uint64_t *nperm, uint64_t *nshared);
1270 
1271 bool bdrv_recurse_can_replace(BlockDriverState *bs,
1272                               BlockDriverState *to_replace);
1273 
1274 /*
1275  * Default implementation for drivers to pass bdrv_co_block_status() to
1276  * their file.
1277  */
1278 int coroutine_fn bdrv_co_block_status_from_file(BlockDriverState *bs,
1279                                                 bool want_zero,
1280                                                 int64_t offset,
1281                                                 int64_t bytes,
1282                                                 int64_t *pnum,
1283                                                 int64_t *map,
1284                                                 BlockDriverState **file);
1285 /*
1286  * Default implementation for drivers to pass bdrv_co_block_status() to
1287  * their backing file.
1288  */
1289 int coroutine_fn bdrv_co_block_status_from_backing(BlockDriverState *bs,
1290                                                    bool want_zero,
1291                                                    int64_t offset,
1292                                                    int64_t bytes,
1293                                                    int64_t *pnum,
1294                                                    int64_t *map,
1295                                                    BlockDriverState **file);
1296 const char *bdrv_get_parent_name(const BlockDriverState *bs);
1297 void blk_dev_change_media_cb(BlockBackend *blk, bool load, Error **errp);
1298 bool blk_dev_has_removable_media(BlockBackend *blk);
1299 bool blk_dev_has_tray(BlockBackend *blk);
1300 void blk_dev_eject_request(BlockBackend *blk, bool force);
1301 bool blk_dev_is_tray_open(BlockBackend *blk);
1302 bool blk_dev_is_medium_locked(BlockBackend *blk);
1303 
1304 void bdrv_set_dirty(BlockDriverState *bs, int64_t offset, int64_t bytes);
1305 
1306 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap **out);
1307 void bdrv_restore_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap *backup);
1308 bool bdrv_dirty_bitmap_merge_internal(BdrvDirtyBitmap *dest,
1309                                       const BdrvDirtyBitmap *src,
1310                                       HBitmap **backup, bool lock);
1311 
1312 void bdrv_inc_in_flight(BlockDriverState *bs);
1313 void bdrv_dec_in_flight(BlockDriverState *bs);
1314 
1315 void blockdev_close_all_bdrv_states(void);
1316 
1317 int coroutine_fn bdrv_co_copy_range_from(BdrvChild *src, uint64_t src_offset,
1318                                          BdrvChild *dst, uint64_t dst_offset,
1319                                          uint64_t bytes,
1320                                          BdrvRequestFlags read_flags,
1321                                          BdrvRequestFlags write_flags);
1322 int coroutine_fn bdrv_co_copy_range_to(BdrvChild *src, uint64_t src_offset,
1323                                        BdrvChild *dst, uint64_t dst_offset,
1324                                        uint64_t bytes,
1325                                        BdrvRequestFlags read_flags,
1326                                        BdrvRequestFlags write_flags);
1327 
1328 int refresh_total_sectors(BlockDriverState *bs, int64_t hint);
1329 
1330 void bdrv_set_monitor_owned(BlockDriverState *bs);
1331 BlockDriverState *bds_tree_init(QDict *bs_opts, Error **errp);
1332 
1333 #endif /* BLOCK_INT_H */
1334