xref: /openbmc/qemu/include/block/block_int.h (revision dd205025)
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_EXTENT_SIZE_HINT  "extent_size_hint"
57 #define BLOCK_OPT_OBJECT_SIZE       "object_size"
58 #define BLOCK_OPT_REFCOUNT_BITS     "refcount_bits"
59 #define BLOCK_OPT_DATA_FILE         "data_file"
60 #define BLOCK_OPT_DATA_FILE_RAW     "data_file_raw"
61 #define BLOCK_OPT_COMPRESSION_TYPE  "compression_type"
62 #define BLOCK_OPT_EXTL2             "extended_l2"
63 
64 #define BLOCK_PROBE_BUF_SIZE        512
65 
66 enum BdrvTrackedRequestType {
67     BDRV_TRACKED_READ,
68     BDRV_TRACKED_WRITE,
69     BDRV_TRACKED_DISCARD,
70     BDRV_TRACKED_TRUNCATE,
71 };
72 
73 typedef struct BdrvTrackedRequest {
74     BlockDriverState *bs;
75     int64_t offset;
76     uint64_t bytes;
77     enum BdrvTrackedRequestType type;
78 
79     bool serialising;
80     int64_t overlap_offset;
81     uint64_t overlap_bytes;
82 
83     QLIST_ENTRY(BdrvTrackedRequest) list;
84     Coroutine *co; /* owner, used for deadlock detection */
85     CoQueue wait_queue; /* coroutines blocked on this request */
86 
87     struct BdrvTrackedRequest *waiting_for;
88 } BdrvTrackedRequest;
89 
90 struct BlockDriver {
91     const char *format_name;
92     int instance_size;
93 
94     /* set to true if the BlockDriver is a block filter. Block filters pass
95      * certain callbacks that refer to data (see block.c) to their bs->file if
96      * the driver doesn't implement them. Drivers that do not wish to forward
97      * must implement them and return -ENOTSUP.
98      */
99     bool is_filter;
100     /*
101      * Set to true if the BlockDriver is a format driver.  Format nodes
102      * generally do not expect their children to be other format nodes
103      * (except for backing files), and so format probing is disabled
104      * on those children.
105      */
106     bool is_format;
107     /*
108      * Return true if @to_replace can be replaced by a BDS with the
109      * same data as @bs without it affecting @bs's behavior (that is,
110      * without it being visible to @bs's parents).
111      */
112     bool (*bdrv_recurse_can_replace)(BlockDriverState *bs,
113                                      BlockDriverState *to_replace);
114 
115     int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename);
116     int (*bdrv_probe_device)(const char *filename);
117 
118     /* Any driver implementing this callback is expected to be able to handle
119      * NULL file names in its .bdrv_open() implementation */
120     void (*bdrv_parse_filename)(const char *filename, QDict *options, Error **errp);
121     /* Drivers not implementing bdrv_parse_filename nor bdrv_open should have
122      * this field set to true, except ones that are defined only by their
123      * child's bs.
124      * An example of the last type will be the quorum block driver.
125      */
126     bool bdrv_needs_filename;
127 
128     /*
129      * Set if a driver can support backing files. This also implies the
130      * following semantics:
131      *
132      *  - Return status 0 of .bdrv_co_block_status means that corresponding
133      *    blocks are not allocated in this layer of backing-chain
134      *  - For such (unallocated) blocks, read will:
135      *    - fill buffer with zeros if there is no backing file
136      *    - read from the backing file otherwise, where the block layer
137      *      takes care of reading zeros beyond EOF if backing file is short
138      */
139     bool supports_backing;
140 
141     /* For handling image reopen for split or non-split files */
142     int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state,
143                                BlockReopenQueue *queue, Error **errp);
144     void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state);
145     void (*bdrv_reopen_commit_post)(BDRVReopenState *reopen_state);
146     void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state);
147     void (*bdrv_join_options)(QDict *options, QDict *old_options);
148 
149     int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags,
150                      Error **errp);
151 
152     /* Protocol drivers should implement this instead of bdrv_open */
153     int (*bdrv_file_open)(BlockDriverState *bs, QDict *options, int flags,
154                           Error **errp);
155     void (*bdrv_close)(BlockDriverState *bs);
156 
157 
158     int coroutine_fn (*bdrv_co_create)(BlockdevCreateOptions *opts,
159                                        Error **errp);
160     int coroutine_fn (*bdrv_co_create_opts)(BlockDriver *drv,
161                                             const char *filename,
162                                             QemuOpts *opts,
163                                             Error **errp);
164 
165     int coroutine_fn (*bdrv_co_amend)(BlockDriverState *bs,
166                                       BlockdevAmendOptions *opts,
167                                       bool force,
168                                       Error **errp);
169 
170     int (*bdrv_amend_options)(BlockDriverState *bs,
171                               QemuOpts *opts,
172                               BlockDriverAmendStatusCB *status_cb,
173                               void *cb_opaque,
174                               bool force,
175                               Error **errp);
176 
177     int (*bdrv_make_empty)(BlockDriverState *bs);
178 
179     /*
180      * Refreshes the bs->exact_filename field. If that is impossible,
181      * bs->exact_filename has to be left empty.
182      */
183     void (*bdrv_refresh_filename)(BlockDriverState *bs);
184 
185     /*
186      * Gathers the open options for all children into @target.
187      * A simple format driver (without backing file support) might
188      * implement this function like this:
189      *
190      *     QINCREF(bs->file->bs->full_open_options);
191      *     qdict_put(target, "file", bs->file->bs->full_open_options);
192      *
193      * If not specified, the generic implementation will simply put
194      * all children's options under their respective name.
195      *
196      * @backing_overridden is true when bs->backing seems not to be
197      * the child that would result from opening bs->backing_file.
198      * Therefore, if it is true, the backing child's options should be
199      * gathered; otherwise, there is no need since the backing child
200      * is the one implied by the image header.
201      *
202      * Note that ideally this function would not be needed.  Every
203      * block driver which implements it is probably doing something
204      * shady regarding its runtime option structure.
205      */
206     void (*bdrv_gather_child_options)(BlockDriverState *bs, QDict *target,
207                                       bool backing_overridden);
208 
209     /*
210      * Returns an allocated string which is the directory name of this BDS: It
211      * will be used to make relative filenames absolute by prepending this
212      * function's return value to them.
213      */
214     char *(*bdrv_dirname)(BlockDriverState *bs, Error **errp);
215 
216     /* aio */
217     BlockAIOCB *(*bdrv_aio_preadv)(BlockDriverState *bs,
218         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags,
219         BlockCompletionFunc *cb, void *opaque);
220     BlockAIOCB *(*bdrv_aio_pwritev)(BlockDriverState *bs,
221         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags,
222         BlockCompletionFunc *cb, void *opaque);
223     BlockAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs,
224         BlockCompletionFunc *cb, void *opaque);
225     BlockAIOCB *(*bdrv_aio_pdiscard)(BlockDriverState *bs,
226         int64_t offset, int bytes,
227         BlockCompletionFunc *cb, void *opaque);
228 
229     int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs,
230         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
231 
232     /**
233      * @offset: position in bytes to read at
234      * @bytes: number of bytes to read
235      * @qiov: the buffers to fill with read data
236      * @flags: currently unused, always 0
237      *
238      * @offset and @bytes will be a multiple of 'request_alignment',
239      * but the length of individual @qiov elements does not have to
240      * be a multiple.
241      *
242      * @bytes will always equal the total size of @qiov, and will be
243      * no larger than 'max_transfer'.
244      *
245      * The buffer in @qiov may point directly to guest memory.
246      */
247     int coroutine_fn (*bdrv_co_preadv)(BlockDriverState *bs,
248         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags);
249     int coroutine_fn (*bdrv_co_preadv_part)(BlockDriverState *bs,
250         uint64_t offset, uint64_t bytes,
251         QEMUIOVector *qiov, size_t qiov_offset, int flags);
252     int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs,
253         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, int flags);
254     /**
255      * @offset: position in bytes to write at
256      * @bytes: number of bytes to write
257      * @qiov: the buffers containing data to write
258      * @flags: zero or more bits allowed by 'supported_write_flags'
259      *
260      * @offset and @bytes will be a multiple of 'request_alignment',
261      * but the length of individual @qiov elements does not have to
262      * be a multiple.
263      *
264      * @bytes will always equal the total size of @qiov, and will be
265      * no larger than 'max_transfer'.
266      *
267      * The buffer in @qiov may point directly to guest memory.
268      */
269     int coroutine_fn (*bdrv_co_pwritev)(BlockDriverState *bs,
270         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags);
271     int coroutine_fn (*bdrv_co_pwritev_part)(BlockDriverState *bs,
272         uint64_t offset, uint64_t bytes,
273         QEMUIOVector *qiov, size_t qiov_offset, int flags);
274 
275     /*
276      * Efficiently zero a region of the disk image.  Typically an image format
277      * would use a compact metadata representation to implement this.  This
278      * function pointer may be NULL or return -ENOSUP and .bdrv_co_writev()
279      * will be called instead.
280      */
281     int coroutine_fn (*bdrv_co_pwrite_zeroes)(BlockDriverState *bs,
282         int64_t offset, int bytes, BdrvRequestFlags flags);
283     int coroutine_fn (*bdrv_co_pdiscard)(BlockDriverState *bs,
284         int64_t offset, int bytes);
285 
286     /* Map [offset, offset + nbytes) range onto a child of @bs to copy from,
287      * and invoke bdrv_co_copy_range_from(child, ...), or invoke
288      * bdrv_co_copy_range_to() if @bs is the leaf child to copy data from.
289      *
290      * See the comment of bdrv_co_copy_range for the parameter and return value
291      * semantics.
292      */
293     int coroutine_fn (*bdrv_co_copy_range_from)(BlockDriverState *bs,
294                                                 BdrvChild *src,
295                                                 uint64_t offset,
296                                                 BdrvChild *dst,
297                                                 uint64_t dst_offset,
298                                                 uint64_t bytes,
299                                                 BdrvRequestFlags read_flags,
300                                                 BdrvRequestFlags write_flags);
301 
302     /* Map [offset, offset + nbytes) range onto a child of bs to copy data to,
303      * and invoke bdrv_co_copy_range_to(child, src, ...), or perform the copy
304      * operation if @bs is the leaf and @src has the same BlockDriver.  Return
305      * -ENOTSUP if @bs is the leaf but @src has a different BlockDriver.
306      *
307      * See the comment of bdrv_co_copy_range for the parameter and return value
308      * semantics.
309      */
310     int coroutine_fn (*bdrv_co_copy_range_to)(BlockDriverState *bs,
311                                               BdrvChild *src,
312                                               uint64_t src_offset,
313                                               BdrvChild *dst,
314                                               uint64_t dst_offset,
315                                               uint64_t bytes,
316                                               BdrvRequestFlags read_flags,
317                                               BdrvRequestFlags write_flags);
318 
319     /*
320      * Building block for bdrv_block_status[_above] and
321      * bdrv_is_allocated[_above].  The driver should answer only
322      * according to the current layer, and should only need to set
323      * BDRV_BLOCK_DATA, BDRV_BLOCK_ZERO, BDRV_BLOCK_OFFSET_VALID,
324      * and/or BDRV_BLOCK_RAW; if the current layer defers to a backing
325      * layer, the result should be 0 (and not BDRV_BLOCK_ZERO).  See
326      * block.h for the overall meaning of the bits.  As a hint, the
327      * flag want_zero is true if the caller cares more about precise
328      * mappings (favor accurate _OFFSET_VALID/_ZERO) or false for
329      * overall allocation (favor larger *pnum, perhaps by reporting
330      * _DATA instead of _ZERO).  The block layer guarantees input
331      * clamped to bdrv_getlength() and aligned to request_alignment,
332      * as well as non-NULL pnum, map, and file; in turn, the driver
333      * must return an error or set pnum to an aligned non-zero value.
334      */
335     int coroutine_fn (*bdrv_co_block_status)(BlockDriverState *bs,
336         bool want_zero, int64_t offset, int64_t bytes, int64_t *pnum,
337         int64_t *map, BlockDriverState **file);
338 
339     /*
340      * Invalidate any cached meta-data.
341      */
342     void coroutine_fn (*bdrv_co_invalidate_cache)(BlockDriverState *bs,
343                                                   Error **errp);
344     int (*bdrv_inactivate)(BlockDriverState *bs);
345 
346     /*
347      * Flushes all data for all layers by calling bdrv_co_flush for underlying
348      * layers, if needed. This function is needed for deterministic
349      * synchronization of the flush finishing callback.
350      */
351     int coroutine_fn (*bdrv_co_flush)(BlockDriverState *bs);
352 
353     /* Delete a created file. */
354     int coroutine_fn (*bdrv_co_delete_file)(BlockDriverState *bs,
355                                             Error **errp);
356 
357     /*
358      * Flushes all data that was already written to the OS all the way down to
359      * the disk (for example file-posix.c calls fsync()).
360      */
361     int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs);
362 
363     /*
364      * Flushes all internal caches to the OS. The data may still sit in a
365      * writeback cache of the host OS, but it will survive a crash of the qemu
366      * process.
367      */
368     int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs);
369 
370     /*
371      * Drivers setting this field must be able to work with just a plain
372      * filename with '<protocol_name>:' as a prefix, and no other options.
373      * Options may be extracted from the filename by implementing
374      * bdrv_parse_filename.
375      */
376     const char *protocol_name;
377 
378     /*
379      * Truncate @bs to @offset bytes using the given @prealloc mode
380      * when growing.  Modes other than PREALLOC_MODE_OFF should be
381      * rejected when shrinking @bs.
382      *
383      * If @exact is true, @bs must be resized to exactly @offset.
384      * Otherwise, it is sufficient for @bs (if it is a host block
385      * device and thus there is no way to resize it) to be at least
386      * @offset bytes in length.
387      *
388      * If @exact is true and this function fails but would succeed
389      * with @exact = false, it should return -ENOTSUP.
390      */
391     int coroutine_fn (*bdrv_co_truncate)(BlockDriverState *bs, int64_t offset,
392                                          bool exact, PreallocMode prealloc,
393                                          BdrvRequestFlags flags, Error **errp);
394 
395     int64_t (*bdrv_getlength)(BlockDriverState *bs);
396     bool has_variable_length;
397     int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
398     BlockMeasureInfo *(*bdrv_measure)(QemuOpts *opts, BlockDriverState *in_bs,
399                                       Error **errp);
400 
401     int coroutine_fn (*bdrv_co_pwritev_compressed)(BlockDriverState *bs,
402         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov);
403     int coroutine_fn (*bdrv_co_pwritev_compressed_part)(BlockDriverState *bs,
404         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov,
405         size_t qiov_offset);
406 
407     int (*bdrv_snapshot_create)(BlockDriverState *bs,
408                                 QEMUSnapshotInfo *sn_info);
409     int (*bdrv_snapshot_goto)(BlockDriverState *bs,
410                               const char *snapshot_id);
411     int (*bdrv_snapshot_delete)(BlockDriverState *bs,
412                                 const char *snapshot_id,
413                                 const char *name,
414                                 Error **errp);
415     int (*bdrv_snapshot_list)(BlockDriverState *bs,
416                               QEMUSnapshotInfo **psn_info);
417     int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
418                                   const char *snapshot_id,
419                                   const char *name,
420                                   Error **errp);
421     int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
422     ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs,
423                                                  Error **errp);
424     BlockStatsSpecific *(*bdrv_get_specific_stats)(BlockDriverState *bs);
425 
426     int coroutine_fn (*bdrv_save_vmstate)(BlockDriverState *bs,
427                                           QEMUIOVector *qiov,
428                                           int64_t pos);
429     int coroutine_fn (*bdrv_load_vmstate)(BlockDriverState *bs,
430                                           QEMUIOVector *qiov,
431                                           int64_t pos);
432 
433     int (*bdrv_change_backing_file)(BlockDriverState *bs,
434         const char *backing_file, const char *backing_fmt);
435 
436     /* removable device specific */
437     bool (*bdrv_is_inserted)(BlockDriverState *bs);
438     void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
439     void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
440 
441     /* to control generic scsi devices */
442     BlockAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
443         unsigned long int req, void *buf,
444         BlockCompletionFunc *cb, void *opaque);
445     int coroutine_fn (*bdrv_co_ioctl)(BlockDriverState *bs,
446                                       unsigned long int req, void *buf);
447 
448     /* List of options for creating images, terminated by name == NULL */
449     QemuOptsList *create_opts;
450 
451     /* List of options for image amend */
452     QemuOptsList *amend_opts;
453 
454     /*
455      * If this driver supports reopening images this contains a
456      * NULL-terminated list of the runtime options that can be
457      * modified. If an option in this list is unspecified during
458      * reopen then it _must_ be reset to its default value or return
459      * an error.
460      */
461     const char *const *mutable_opts;
462 
463     /*
464      * Returns 0 for completed check, -errno for internal errors.
465      * The check results are stored in result.
466      */
467     int coroutine_fn (*bdrv_co_check)(BlockDriverState *bs,
468                                       BdrvCheckResult *result,
469                                       BdrvCheckMode fix);
470 
471     void (*bdrv_debug_event)(BlockDriverState *bs, BlkdebugEvent event);
472 
473     /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
474     int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
475         const char *tag);
476     int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs,
477         const char *tag);
478     int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
479     bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
480 
481     void (*bdrv_refresh_limits)(BlockDriverState *bs, Error **errp);
482 
483     /*
484      * Returns 1 if newly created images are guaranteed to contain only
485      * zeros, 0 otherwise.
486      */
487     int (*bdrv_has_zero_init)(BlockDriverState *bs);
488 
489     /* Remove fd handlers, timers, and other event loop callbacks so the event
490      * loop is no longer in use.  Called with no in-flight requests and in
491      * depth-first traversal order with parents before child nodes.
492      */
493     void (*bdrv_detach_aio_context)(BlockDriverState *bs);
494 
495     /* Add fd handlers, timers, and other event loop callbacks so I/O requests
496      * can be processed again.  Called with no in-flight requests and in
497      * depth-first traversal order with child nodes before parent nodes.
498      */
499     void (*bdrv_attach_aio_context)(BlockDriverState *bs,
500                                     AioContext *new_context);
501 
502     /* io queue for linux-aio */
503     void (*bdrv_io_plug)(BlockDriverState *bs);
504     void (*bdrv_io_unplug)(BlockDriverState *bs);
505 
506     /**
507      * Try to get @bs's logical and physical block size.
508      * On success, store them in @bsz and return zero.
509      * On failure, return negative errno.
510      */
511     int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz);
512     /**
513      * Try to get @bs's geometry (cyls, heads, sectors)
514      * On success, store them in @geo and return 0.
515      * On failure return -errno.
516      * Only drivers that want to override guest geometry implement this
517      * callback; see hd_geometry_guess().
518      */
519     int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo);
520 
521     /**
522      * bdrv_co_drain_begin is called if implemented in the beginning of a
523      * drain operation to drain and stop any internal sources of requests in
524      * the driver.
525      * bdrv_co_drain_end is called if implemented at the end of the drain.
526      *
527      * They should be used by the driver to e.g. manage scheduled I/O
528      * requests, or toggle an internal state. After the end of the drain new
529      * requests will continue normally.
530      */
531     void coroutine_fn (*bdrv_co_drain_begin)(BlockDriverState *bs);
532     void coroutine_fn (*bdrv_co_drain_end)(BlockDriverState *bs);
533 
534     void (*bdrv_add_child)(BlockDriverState *parent, BlockDriverState *child,
535                            Error **errp);
536     void (*bdrv_del_child)(BlockDriverState *parent, BdrvChild *child,
537                            Error **errp);
538 
539     /**
540      * Informs the block driver that a permission change is intended. The
541      * driver checks whether the change is permissible and may take other
542      * preparations for the change (e.g. get file system locks). This operation
543      * is always followed either by a call to either .bdrv_set_perm or
544      * .bdrv_abort_perm_update.
545      *
546      * Checks whether the requested set of cumulative permissions in @perm
547      * can be granted for accessing @bs and whether no other users are using
548      * permissions other than those given in @shared (both arguments take
549      * BLK_PERM_* bitmasks).
550      *
551      * If both conditions are met, 0 is returned. Otherwise, -errno is returned
552      * and errp is set to an error describing the conflict.
553      */
554     int (*bdrv_check_perm)(BlockDriverState *bs, uint64_t perm,
555                            uint64_t shared, Error **errp);
556 
557     /**
558      * Called to inform the driver that the set of cumulative set of used
559      * permissions for @bs has changed to @perm, and the set of sharable
560      * permission to @shared. The driver can use this to propagate changes to
561      * its children (i.e. request permissions only if a parent actually needs
562      * them).
563      *
564      * This function is only invoked after bdrv_check_perm(), so block drivers
565      * may rely on preparations made in their .bdrv_check_perm implementation.
566      */
567     void (*bdrv_set_perm)(BlockDriverState *bs, uint64_t perm, uint64_t shared);
568 
569     /*
570      * Called to inform the driver that after a previous bdrv_check_perm()
571      * call, the permission update is not performed and any preparations made
572      * for it (e.g. taken file locks) need to be undone.
573      *
574      * This function can be called even for nodes that never saw a
575      * bdrv_check_perm() call. It is a no-op then.
576      */
577     void (*bdrv_abort_perm_update)(BlockDriverState *bs);
578 
579     /**
580      * Returns in @nperm and @nshared the permissions that the driver for @bs
581      * needs on its child @c, based on the cumulative permissions requested by
582      * the parents in @parent_perm and @parent_shared.
583      *
584      * If @c is NULL, return the permissions for attaching a new child for the
585      * given @child_class and @role.
586      *
587      * If @reopen_queue is non-NULL, don't return the currently needed
588      * permissions, but those that will be needed after applying the
589      * @reopen_queue.
590      */
591      void (*bdrv_child_perm)(BlockDriverState *bs, BdrvChild *c,
592                              BdrvChildRole role,
593                              BlockReopenQueue *reopen_queue,
594                              uint64_t parent_perm, uint64_t parent_shared,
595                              uint64_t *nperm, uint64_t *nshared);
596 
597     bool (*bdrv_supports_persistent_dirty_bitmap)(BlockDriverState *bs);
598     bool (*bdrv_co_can_store_new_dirty_bitmap)(BlockDriverState *bs,
599                                                const char *name,
600                                                uint32_t granularity,
601                                                Error **errp);
602     int (*bdrv_co_remove_persistent_dirty_bitmap)(BlockDriverState *bs,
603                                                   const char *name,
604                                                   Error **errp);
605 
606     /**
607      * Register/unregister a buffer for I/O. For example, when the driver is
608      * interested to know the memory areas that will later be used in iovs, so
609      * that it can do IOMMU mapping with VFIO etc., in order to get better
610      * performance. In the case of VFIO drivers, this callback is used to do
611      * DMA mapping for hot buffers.
612      */
613     void (*bdrv_register_buf)(BlockDriverState *bs, void *host, size_t size);
614     void (*bdrv_unregister_buf)(BlockDriverState *bs, void *host);
615     QLIST_ENTRY(BlockDriver) list;
616 
617     /* Pointer to a NULL-terminated array of names of strong options
618      * that can be specified for bdrv_open(). A strong option is one
619      * that changes the data of a BDS.
620      * If this pointer is NULL, the array is considered empty.
621      * "filename" and "driver" are always considered strong. */
622     const char *const *strong_runtime_opts;
623 };
624 
625 static inline bool block_driver_can_compress(BlockDriver *drv)
626 {
627     return drv->bdrv_co_pwritev_compressed ||
628            drv->bdrv_co_pwritev_compressed_part;
629 }
630 
631 typedef struct BlockLimits {
632     /* Alignment requirement, in bytes, for offset/length of I/O
633      * requests. Must be a power of 2 less than INT_MAX; defaults to
634      * 1 for drivers with modern byte interfaces, and to 512
635      * otherwise. */
636     uint32_t request_alignment;
637 
638     /* Maximum number of bytes that can be discarded at once (since it
639      * is signed, it must be < 2G, if set). Must be multiple of
640      * pdiscard_alignment, but need not be power of 2. May be 0 if no
641      * inherent 32-bit limit */
642     int32_t max_pdiscard;
643 
644     /* Optimal alignment for discard requests in bytes. A power of 2
645      * is best but not mandatory.  Must be a multiple of
646      * bl.request_alignment, and must be less than max_pdiscard if
647      * that is set. May be 0 if bl.request_alignment is good enough */
648     uint32_t pdiscard_alignment;
649 
650     /* Maximum number of bytes that can zeroized at once (since it is
651      * signed, it must be < 2G, if set). Must be multiple of
652      * pwrite_zeroes_alignment. May be 0 if no inherent 32-bit limit */
653     int32_t max_pwrite_zeroes;
654 
655     /* Optimal alignment for write zeroes requests in bytes. A power
656      * of 2 is best but not mandatory.  Must be a multiple of
657      * bl.request_alignment, and must be less than max_pwrite_zeroes
658      * if that is set. May be 0 if bl.request_alignment is good
659      * enough */
660     uint32_t pwrite_zeroes_alignment;
661 
662     /* Optimal transfer length in bytes.  A power of 2 is best but not
663      * mandatory.  Must be a multiple of bl.request_alignment, or 0 if
664      * no preferred size */
665     uint32_t opt_transfer;
666 
667     /* Maximal transfer length in bytes.  Need not be power of 2, but
668      * must be multiple of opt_transfer and bl.request_alignment, or 0
669      * for no 32-bit limit.  For now, anything larger than INT_MAX is
670      * clamped down. */
671     uint32_t max_transfer;
672 
673     /* memory alignment, in bytes so that no bounce buffer is needed */
674     size_t min_mem_alignment;
675 
676     /* memory alignment, in bytes, for bounce buffer */
677     size_t opt_mem_alignment;
678 
679     /* maximum number of iovec elements */
680     int max_iov;
681 } BlockLimits;
682 
683 typedef struct BdrvOpBlocker BdrvOpBlocker;
684 
685 typedef struct BdrvAioNotifier {
686     void (*attached_aio_context)(AioContext *new_context, void *opaque);
687     void (*detach_aio_context)(void *opaque);
688 
689     void *opaque;
690     bool deleted;
691 
692     QLIST_ENTRY(BdrvAioNotifier) list;
693 } BdrvAioNotifier;
694 
695 struct BdrvChildClass {
696     /* If true, bdrv_replace_node() doesn't change the node this BdrvChild
697      * points to. */
698     bool stay_at_node;
699 
700     /* If true, the parent is a BlockDriverState and bdrv_next_all_states()
701      * will return it. This information is used for drain_all, where every node
702      * will be drained separately, so the drain only needs to be propagated to
703      * non-BDS parents. */
704     bool parent_is_bds;
705 
706     void (*inherit_options)(BdrvChildRole role, bool parent_is_format,
707                             int *child_flags, QDict *child_options,
708                             int parent_flags, QDict *parent_options);
709 
710     void (*change_media)(BdrvChild *child, bool load);
711     void (*resize)(BdrvChild *child);
712 
713     /* Returns a name that is supposedly more useful for human users than the
714      * node name for identifying the node in question (in particular, a BB
715      * name), or NULL if the parent can't provide a better name. */
716     const char *(*get_name)(BdrvChild *child);
717 
718     /* Returns a malloced string that describes the parent of the child for a
719      * human reader. This could be a node-name, BlockBackend name, qdev ID or
720      * QOM path of the device owning the BlockBackend, job type and ID etc. The
721      * caller is responsible for freeing the memory. */
722     char *(*get_parent_desc)(BdrvChild *child);
723 
724     /*
725      * If this pair of functions is implemented, the parent doesn't issue new
726      * requests after returning from .drained_begin() until .drained_end() is
727      * called.
728      *
729      * These functions must not change the graph (and therefore also must not
730      * call aio_poll(), which could change the graph indirectly).
731      *
732      * If drained_end() schedules background operations, it must atomically
733      * increment *drained_end_counter for each such operation and atomically
734      * decrement it once the operation has settled.
735      *
736      * Note that this can be nested. If drained_begin() was called twice, new
737      * I/O is allowed only after drained_end() was called twice, too.
738      */
739     void (*drained_begin)(BdrvChild *child);
740     void (*drained_end)(BdrvChild *child, int *drained_end_counter);
741 
742     /*
743      * Returns whether the parent has pending requests for the child. This
744      * callback is polled after .drained_begin() has been called until all
745      * activity on the child has stopped.
746      */
747     bool (*drained_poll)(BdrvChild *child);
748 
749     /* Notifies the parent that the child has been activated/inactivated (e.g.
750      * when migration is completing) and it can start/stop requesting
751      * permissions and doing I/O on it. */
752     void (*activate)(BdrvChild *child, Error **errp);
753     int (*inactivate)(BdrvChild *child);
754 
755     void (*attach)(BdrvChild *child);
756     void (*detach)(BdrvChild *child);
757 
758     /* Notifies the parent that the filename of its child has changed (e.g.
759      * because the direct child was removed from the backing chain), so that it
760      * can update its reference. */
761     int (*update_filename)(BdrvChild *child, BlockDriverState *new_base,
762                            const char *filename, Error **errp);
763 
764     bool (*can_set_aio_ctx)(BdrvChild *child, AioContext *ctx,
765                             GSList **ignore, Error **errp);
766     void (*set_aio_ctx)(BdrvChild *child, AioContext *ctx, GSList **ignore);
767 };
768 
769 extern const BdrvChildClass child_of_bds;
770 
771 struct BdrvChild {
772     BlockDriverState *bs;
773     char *name;
774     const BdrvChildClass *klass;
775     BdrvChildRole role;
776     void *opaque;
777 
778     /**
779      * Granted permissions for operating on this BdrvChild (BLK_PERM_* bitmask)
780      */
781     uint64_t perm;
782 
783     /**
784      * Permissions that can still be granted to other users of @bs while this
785      * BdrvChild is still attached to it. (BLK_PERM_* bitmask)
786      */
787     uint64_t shared_perm;
788 
789     /* backup of permissions during permission update procedure */
790     bool has_backup_perm;
791     uint64_t backup_perm;
792     uint64_t backup_shared_perm;
793 
794     /*
795      * This link is frozen: the child can neither be replaced nor
796      * detached from the parent.
797      */
798     bool frozen;
799 
800     /*
801      * How many times the parent of this child has been drained
802      * (through klass->drained_*).
803      * Usually, this is equal to bs->quiesce_counter (potentially
804      * reduced by bdrv_drain_all_count).  It may differ while the
805      * child is entering or leaving a drained section.
806      */
807     int parent_quiesce_counter;
808 
809     QLIST_ENTRY(BdrvChild) next;
810     QLIST_ENTRY(BdrvChild) next_parent;
811 };
812 
813 /*
814  * Note: the function bdrv_append() copies and swaps contents of
815  * BlockDriverStates, so if you add new fields to this struct, please
816  * inspect bdrv_append() to determine if the new fields need to be
817  * copied as well.
818  */
819 struct BlockDriverState {
820     /* Protected by big QEMU lock or read-only after opening.  No special
821      * locking needed during I/O...
822      */
823     int open_flags; /* flags used to open the file, re-used for re-open */
824     bool read_only; /* if true, the media is read only */
825     bool encrypted; /* if true, the media is encrypted */
826     bool sg;        /* if true, the device is a /dev/sg* */
827     bool probed;    /* if true, format was probed rather than specified */
828     bool force_share; /* if true, always allow all shared permissions */
829     bool implicit;  /* if true, this filter node was automatically inserted */
830 
831     BlockDriver *drv; /* NULL means no media */
832     void *opaque;
833 
834     AioContext *aio_context; /* event loop used for fd handlers, timers, etc */
835     /* long-running tasks intended to always use the same AioContext as this
836      * BDS may register themselves in this list to be notified of changes
837      * regarding this BDS's context */
838     QLIST_HEAD(, BdrvAioNotifier) aio_notifiers;
839     bool walking_aio_notifiers; /* to make removal during iteration safe */
840 
841     char filename[PATH_MAX];
842     char backing_file[PATH_MAX]; /* if non zero, the image is a diff of
843                                     this file image */
844     /* The backing filename indicated by the image header; if we ever
845      * open this file, then this is replaced by the resulting BDS's
846      * filename (i.e. after a bdrv_refresh_filename() run). */
847     char auto_backing_file[PATH_MAX];
848     char backing_format[16]; /* if non-zero and backing_file exists */
849 
850     QDict *full_open_options;
851     char exact_filename[PATH_MAX];
852 
853     BdrvChild *backing;
854     BdrvChild *file;
855 
856     /* I/O Limits */
857     BlockLimits bl;
858 
859     /* Flags honored during pwrite (so far: BDRV_REQ_FUA,
860      * BDRV_REQ_WRITE_UNCHANGED).
861      * If a driver does not support BDRV_REQ_WRITE_UNCHANGED, those
862      * writes will be issued as normal writes without the flag set.
863      * This is important to note for drivers that do not explicitly
864      * request a WRITE permission for their children and instead take
865      * the same permissions as their parent did (this is commonly what
866      * block filters do).  Such drivers have to be aware that the
867      * parent may have taken a WRITE_UNCHANGED permission only and is
868      * issuing such requests.  Drivers either must make sure that
869      * these requests do not result in plain WRITE accesses (usually
870      * by supporting BDRV_REQ_WRITE_UNCHANGED, and then forwarding
871      * every incoming write request as-is, including potentially that
872      * flag), or they have to explicitly take the WRITE permission for
873      * their children. */
874     unsigned int supported_write_flags;
875     /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA,
876      * BDRV_REQ_MAY_UNMAP, BDRV_REQ_WRITE_UNCHANGED) */
877     unsigned int supported_zero_flags;
878     /*
879      * Flags honoured during truncate (so far: BDRV_REQ_ZERO_WRITE).
880      *
881      * If BDRV_REQ_ZERO_WRITE is given, the truncate operation must make sure
882      * that any added space reads as all zeros. If this can't be guaranteed,
883      * the operation must fail.
884      */
885     unsigned int supported_truncate_flags;
886 
887     /* the following member gives a name to every node on the bs graph. */
888     char node_name[32];
889     /* element of the list of named nodes building the graph */
890     QTAILQ_ENTRY(BlockDriverState) node_list;
891     /* element of the list of all BlockDriverStates (all_bdrv_states) */
892     QTAILQ_ENTRY(BlockDriverState) bs_list;
893     /* element of the list of monitor-owned BDS */
894     QTAILQ_ENTRY(BlockDriverState) monitor_list;
895     int refcnt;
896 
897     /* operation blockers */
898     QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX];
899 
900     /* The node that this node inherited default options from (and a reopen on
901      * which can affect this node by changing these defaults). This is always a
902      * parent node of this node. */
903     BlockDriverState *inherits_from;
904     QLIST_HEAD(, BdrvChild) children;
905     QLIST_HEAD(, BdrvChild) parents;
906 
907     QDict *options;
908     QDict *explicit_options;
909     BlockdevDetectZeroesOptions detect_zeroes;
910 
911     /* The error object in use for blocking operations on backing_hd */
912     Error *backing_blocker;
913 
914     /* Protected by AioContext lock */
915 
916     /* If we are reading a disk image, give its size in sectors.
917      * Generally read-only; it is written to by load_snapshot and
918      * save_snaphost, but the block layer is quiescent during those.
919      */
920     int64_t total_sectors;
921 
922     /* Callback before write request is processed */
923     NotifierWithReturnList before_write_notifiers;
924 
925     /* threshold limit for writes, in bytes. "High water mark". */
926     uint64_t write_threshold_offset;
927     NotifierWithReturn write_threshold_notifier;
928 
929     /* Writing to the list requires the BQL _and_ the dirty_bitmap_mutex.
930      * Reading from the list can be done with either the BQL or the
931      * dirty_bitmap_mutex.  Modifying a bitmap only requires
932      * dirty_bitmap_mutex.  */
933     QemuMutex dirty_bitmap_mutex;
934     QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps;
935 
936     /* Offset after the highest byte written to */
937     Stat64 wr_highest_offset;
938 
939     /* If true, copy read backing sectors into image.  Can be >1 if more
940      * than one client has requested copy-on-read.  Accessed with atomic
941      * ops.
942      */
943     int copy_on_read;
944 
945     /* number of in-flight requests; overall and serialising.
946      * Accessed with atomic ops.
947      */
948     unsigned int in_flight;
949     unsigned int serialising_in_flight;
950 
951     /* counter for nested bdrv_io_plug.
952      * Accessed with atomic ops.
953     */
954     unsigned io_plugged;
955 
956     /* do we need to tell the quest if we have a volatile write cache? */
957     int enable_write_cache;
958 
959     /* Accessed with atomic ops.  */
960     int quiesce_counter;
961     int recursive_quiesce_counter;
962 
963     unsigned int write_gen;               /* Current data generation */
964 
965     /* Protected by reqs_lock.  */
966     CoMutex reqs_lock;
967     QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
968     CoQueue flush_queue;                  /* Serializing flush queue */
969     bool active_flush_req;                /* Flush request in flight? */
970 
971     /* Only read/written by whoever has set active_flush_req to true.  */
972     unsigned int flushed_gen;             /* Flushed write generation */
973 
974     /* BdrvChild links to this node may never be frozen */
975     bool never_freeze;
976 };
977 
978 struct BlockBackendRootState {
979     int open_flags;
980     bool read_only;
981     BlockdevDetectZeroesOptions detect_zeroes;
982 };
983 
984 typedef enum BlockMirrorBackingMode {
985     /* Reuse the existing backing chain from the source for the target.
986      * - sync=full: Set backing BDS to NULL.
987      * - sync=top:  Use source's backing BDS.
988      * - sync=none: Use source as the backing BDS. */
989     MIRROR_SOURCE_BACKING_CHAIN,
990 
991     /* Open the target's backing chain completely anew */
992     MIRROR_OPEN_BACKING_CHAIN,
993 
994     /* Do not change the target's backing BDS after job completion */
995     MIRROR_LEAVE_BACKING_CHAIN,
996 } BlockMirrorBackingMode;
997 
998 static inline BlockDriverState *backing_bs(BlockDriverState *bs)
999 {
1000     return bs->backing ? bs->backing->bs : NULL;
1001 }
1002 
1003 
1004 /* Essential block drivers which must always be statically linked into qemu, and
1005  * which therefore can be accessed without using bdrv_find_format() */
1006 extern BlockDriver bdrv_file;
1007 extern BlockDriver bdrv_raw;
1008 extern BlockDriver bdrv_qcow2;
1009 
1010 int coroutine_fn bdrv_co_preadv(BdrvChild *child,
1011     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
1012     BdrvRequestFlags flags);
1013 int coroutine_fn bdrv_co_preadv_part(BdrvChild *child,
1014     int64_t offset, unsigned int bytes,
1015     QEMUIOVector *qiov, size_t qiov_offset, BdrvRequestFlags flags);
1016 int coroutine_fn bdrv_co_pwritev(BdrvChild *child,
1017     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
1018     BdrvRequestFlags flags);
1019 int coroutine_fn bdrv_co_pwritev_part(BdrvChild *child,
1020     int64_t offset, unsigned int bytes,
1021     QEMUIOVector *qiov, size_t qiov_offset, BdrvRequestFlags flags);
1022 
1023 static inline int coroutine_fn bdrv_co_pread(BdrvChild *child,
1024     int64_t offset, unsigned int bytes, void *buf, BdrvRequestFlags flags)
1025 {
1026     QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes);
1027 
1028     return bdrv_co_preadv(child, offset, bytes, &qiov, flags);
1029 }
1030 
1031 static inline int coroutine_fn bdrv_co_pwrite(BdrvChild *child,
1032     int64_t offset, unsigned int bytes, void *buf, BdrvRequestFlags flags)
1033 {
1034     QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes);
1035 
1036     return bdrv_co_pwritev(child, offset, bytes, &qiov, flags);
1037 }
1038 
1039 extern unsigned int bdrv_drain_all_count;
1040 void bdrv_apply_subtree_drain(BdrvChild *child, BlockDriverState *new_parent);
1041 void bdrv_unapply_subtree_drain(BdrvChild *child, BlockDriverState *old_parent);
1042 
1043 bool coroutine_fn bdrv_mark_request_serialising(BdrvTrackedRequest *req, uint64_t align);
1044 BdrvTrackedRequest *coroutine_fn bdrv_co_get_self_request(BlockDriverState *bs);
1045 
1046 int get_tmp_filename(char *filename, int size);
1047 BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size,
1048                             const char *filename);
1049 
1050 void bdrv_parse_filename_strip_prefix(const char *filename, const char *prefix,
1051                                       QDict *options);
1052 
1053 
1054 /**
1055  * bdrv_add_before_write_notifier:
1056  *
1057  * Register a callback that is invoked before write requests are processed but
1058  * after any throttling or waiting for overlapping requests.
1059  */
1060 void bdrv_add_before_write_notifier(BlockDriverState *bs,
1061                                     NotifierWithReturn *notifier);
1062 
1063 /**
1064  * bdrv_add_aio_context_notifier:
1065  *
1066  * If a long-running job intends to be always run in the same AioContext as a
1067  * certain BDS, it may use this function to be notified of changes regarding the
1068  * association of the BDS to an AioContext.
1069  *
1070  * attached_aio_context() is called after the target BDS has been attached to a
1071  * new AioContext; detach_aio_context() is called before the target BDS is being
1072  * detached from its old AioContext.
1073  */
1074 void bdrv_add_aio_context_notifier(BlockDriverState *bs,
1075         void (*attached_aio_context)(AioContext *new_context, void *opaque),
1076         void (*detach_aio_context)(void *opaque), void *opaque);
1077 
1078 /**
1079  * bdrv_remove_aio_context_notifier:
1080  *
1081  * Unsubscribe of change notifications regarding the BDS's AioContext. The
1082  * parameters given here have to be the same as those given to
1083  * bdrv_add_aio_context_notifier().
1084  */
1085 void bdrv_remove_aio_context_notifier(BlockDriverState *bs,
1086                                       void (*aio_context_attached)(AioContext *,
1087                                                                    void *),
1088                                       void (*aio_context_detached)(void *),
1089                                       void *opaque);
1090 
1091 /**
1092  * bdrv_wakeup:
1093  * @bs: The BlockDriverState for which an I/O operation has been completed.
1094  *
1095  * Wake up the main thread if it is waiting on BDRV_POLL_WHILE.  During
1096  * synchronous I/O on a BlockDriverState that is attached to another
1097  * I/O thread, the main thread lets the I/O thread's event loop run,
1098  * waiting for the I/O operation to complete.  A bdrv_wakeup will wake
1099  * up the main thread if necessary.
1100  *
1101  * Manual calls to bdrv_wakeup are rarely necessary, because
1102  * bdrv_dec_in_flight already calls it.
1103  */
1104 void bdrv_wakeup(BlockDriverState *bs);
1105 
1106 #ifdef _WIN32
1107 int is_windows_drive(const char *filename);
1108 #endif
1109 
1110 /**
1111  * stream_start:
1112  * @job_id: The id of the newly-created job, or %NULL to use the
1113  * device name of @bs.
1114  * @bs: Block device to operate on.
1115  * @base: Block device that will become the new base, or %NULL to
1116  * flatten the whole backing file chain onto @bs.
1117  * @backing_file_str: The file name that will be written to @bs as the
1118  * the new backing file if the job completes. Ignored if @base is %NULL.
1119  * @creation_flags: Flags that control the behavior of the Job lifetime.
1120  *                  See @BlockJobCreateFlags
1121  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1122  * @on_error: The action to take upon error.
1123  * @errp: Error object.
1124  *
1125  * Start a streaming operation on @bs.  Clusters that are unallocated
1126  * in @bs, but allocated in any image between @base and @bs (both
1127  * exclusive) will be written to @bs.  At the end of a successful
1128  * streaming job, the backing file of @bs will be changed to
1129  * @backing_file_str in the written image and to @base in the live
1130  * BlockDriverState.
1131  */
1132 void stream_start(const char *job_id, BlockDriverState *bs,
1133                   BlockDriverState *base, const char *backing_file_str,
1134                   int creation_flags, int64_t speed,
1135                   BlockdevOnError on_error, Error **errp);
1136 
1137 /**
1138  * commit_start:
1139  * @job_id: The id of the newly-created job, or %NULL to use the
1140  * device name of @bs.
1141  * @bs: Active block device.
1142  * @top: Top block device to be committed.
1143  * @base: Block device that will be written into, and become the new top.
1144  * @creation_flags: Flags that control the behavior of the Job lifetime.
1145  *                  See @BlockJobCreateFlags
1146  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1147  * @on_error: The action to take upon error.
1148  * @backing_file_str: String to use as the backing file in @top's overlay
1149  * @filter_node_name: The node name that should be assigned to the filter
1150  * driver that the commit job inserts into the graph above @top. NULL means
1151  * that a node name should be autogenerated.
1152  * @errp: Error object.
1153  *
1154  */
1155 void commit_start(const char *job_id, BlockDriverState *bs,
1156                   BlockDriverState *base, BlockDriverState *top,
1157                   int creation_flags, int64_t speed,
1158                   BlockdevOnError on_error, const char *backing_file_str,
1159                   const char *filter_node_name, Error **errp);
1160 /**
1161  * commit_active_start:
1162  * @job_id: The id of the newly-created job, or %NULL to use the
1163  * device name of @bs.
1164  * @bs: Active block device to be committed.
1165  * @base: Block device that will be written into, and become the new top.
1166  * @creation_flags: Flags that control the behavior of the Job lifetime.
1167  *                  See @BlockJobCreateFlags
1168  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1169  * @on_error: The action to take upon error.
1170  * @filter_node_name: The node name that should be assigned to the filter
1171  * driver that the commit job inserts into the graph above @bs. NULL means that
1172  * a node name should be autogenerated.
1173  * @cb: Completion function for the job.
1174  * @opaque: Opaque pointer value passed to @cb.
1175  * @auto_complete: Auto complete the job.
1176  * @errp: Error object.
1177  *
1178  */
1179 BlockJob *commit_active_start(const char *job_id, BlockDriverState *bs,
1180                               BlockDriverState *base, int creation_flags,
1181                               int64_t speed, BlockdevOnError on_error,
1182                               const char *filter_node_name,
1183                               BlockCompletionFunc *cb, void *opaque,
1184                               bool auto_complete, Error **errp);
1185 /*
1186  * mirror_start:
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  * @replaces: Block graph node name to replace once the mirror is done. Can
1192  *            only be used when full mirroring is selected.
1193  * @creation_flags: Flags that control the behavior of the Job lifetime.
1194  *                  See @BlockJobCreateFlags
1195  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1196  * @granularity: The chosen granularity for the dirty bitmap.
1197  * @buf_size: The amount of data that can be in flight at one time.
1198  * @mode: Whether to collapse all images in the chain to the target.
1199  * @backing_mode: How to establish the target's backing chain after completion.
1200  * @zero_target: Whether the target should be explicitly zero-initialized
1201  * @on_source_error: The action to take upon error reading from the source.
1202  * @on_target_error: The action to take upon error writing to the target.
1203  * @unmap: Whether to unmap target where source sectors only contain zeroes.
1204  * @filter_node_name: The node name that should be assigned to the filter
1205  * driver that the mirror job inserts into the graph above @bs. NULL means that
1206  * a node name should be autogenerated.
1207  * @copy_mode: When to trigger writes to the target.
1208  * @errp: Error object.
1209  *
1210  * Start a mirroring operation on @bs.  Clusters that are allocated
1211  * in @bs will be written to @target until the job is cancelled or
1212  * manually completed.  At the end of a successful mirroring job,
1213  * @bs will be switched to read from @target.
1214  */
1215 void mirror_start(const char *job_id, BlockDriverState *bs,
1216                   BlockDriverState *target, const char *replaces,
1217                   int creation_flags, int64_t speed,
1218                   uint32_t granularity, int64_t buf_size,
1219                   MirrorSyncMode mode, BlockMirrorBackingMode backing_mode,
1220                   bool zero_target,
1221                   BlockdevOnError on_source_error,
1222                   BlockdevOnError on_target_error,
1223                   bool unmap, const char *filter_node_name,
1224                   MirrorCopyMode copy_mode, Error **errp);
1225 
1226 /*
1227  * backup_job_create:
1228  * @job_id: The id of the newly-created job, or %NULL to use the
1229  * device name of @bs.
1230  * @bs: Block device to operate on.
1231  * @target: Block device to write to.
1232  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1233  * @sync_mode: What parts of the disk image should be copied to the destination.
1234  * @sync_bitmap: The dirty bitmap if sync_mode is 'bitmap' or 'incremental'
1235  * @bitmap_mode: The bitmap synchronization policy to use.
1236  * @on_source_error: The action to take upon error reading from the source.
1237  * @on_target_error: The action to take upon error writing to the target.
1238  * @creation_flags: Flags that control the behavior of the Job lifetime.
1239  *                  See @BlockJobCreateFlags
1240  * @cb: Completion function for the job.
1241  * @opaque: Opaque pointer value passed to @cb.
1242  * @txn: Transaction that this job is part of (may be NULL).
1243  *
1244  * Create a backup operation on @bs.  Clusters in @bs are written to @target
1245  * until the job is cancelled or manually completed.
1246  */
1247 BlockJob *backup_job_create(const char *job_id, BlockDriverState *bs,
1248                             BlockDriverState *target, int64_t speed,
1249                             MirrorSyncMode sync_mode,
1250                             BdrvDirtyBitmap *sync_bitmap,
1251                             BitmapSyncMode bitmap_mode,
1252                             bool compress,
1253                             const char *filter_node_name,
1254                             BlockdevOnError on_source_error,
1255                             BlockdevOnError on_target_error,
1256                             int creation_flags,
1257                             BlockCompletionFunc *cb, void *opaque,
1258                             JobTxn *txn, Error **errp);
1259 
1260 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
1261                                   const char *child_name,
1262                                   const BdrvChildClass *child_class,
1263                                   BdrvChildRole child_role,
1264                                   AioContext *ctx,
1265                                   uint64_t perm, uint64_t shared_perm,
1266                                   void *opaque, Error **errp);
1267 void bdrv_root_unref_child(BdrvChild *child);
1268 
1269 void bdrv_get_cumulative_perm(BlockDriverState *bs, uint64_t *perm,
1270                               uint64_t *shared_perm);
1271 
1272 /**
1273  * Sets a BdrvChild's permissions.  Avoid if the parent is a BDS; use
1274  * bdrv_child_refresh_perms() instead and make the parent's
1275  * .bdrv_child_perm() implementation return the correct values.
1276  */
1277 int bdrv_child_try_set_perm(BdrvChild *c, uint64_t perm, uint64_t shared,
1278                             Error **errp);
1279 
1280 /**
1281  * Calls bs->drv->bdrv_child_perm() and updates the child's permission
1282  * masks with the result.
1283  * Drivers should invoke this function whenever an event occurs that
1284  * makes their .bdrv_child_perm() implementation return different
1285  * values than before, but which will not result in the block layer
1286  * automatically refreshing the permissions.
1287  */
1288 int bdrv_child_refresh_perms(BlockDriverState *bs, BdrvChild *c, Error **errp);
1289 
1290 bool bdrv_recurse_can_replace(BlockDriverState *bs,
1291                               BlockDriverState *to_replace);
1292 
1293 /*
1294  * Default implementation for BlockDriver.bdrv_child_perm() that can
1295  * be used by block filters and image formats, as long as they use the
1296  * child_of_bds child class and set an appropriate BdrvChildRole.
1297  */
1298 void bdrv_default_perms(BlockDriverState *bs, BdrvChild *c,
1299                         BdrvChildRole role, BlockReopenQueue *reopen_queue,
1300                         uint64_t perm, uint64_t shared,
1301                         uint64_t *nperm, uint64_t *nshared);
1302 
1303 /*
1304  * Default implementation for drivers to pass bdrv_co_block_status() to
1305  * their file.
1306  */
1307 int coroutine_fn bdrv_co_block_status_from_file(BlockDriverState *bs,
1308                                                 bool want_zero,
1309                                                 int64_t offset,
1310                                                 int64_t bytes,
1311                                                 int64_t *pnum,
1312                                                 int64_t *map,
1313                                                 BlockDriverState **file);
1314 /*
1315  * Default implementation for drivers to pass bdrv_co_block_status() to
1316  * their backing file.
1317  */
1318 int coroutine_fn bdrv_co_block_status_from_backing(BlockDriverState *bs,
1319                                                    bool want_zero,
1320                                                    int64_t offset,
1321                                                    int64_t bytes,
1322                                                    int64_t *pnum,
1323                                                    int64_t *map,
1324                                                    BlockDriverState **file);
1325 const char *bdrv_get_parent_name(const BlockDriverState *bs);
1326 void blk_dev_change_media_cb(BlockBackend *blk, bool load, Error **errp);
1327 bool blk_dev_has_removable_media(BlockBackend *blk);
1328 bool blk_dev_has_tray(BlockBackend *blk);
1329 void blk_dev_eject_request(BlockBackend *blk, bool force);
1330 bool blk_dev_is_tray_open(BlockBackend *blk);
1331 bool blk_dev_is_medium_locked(BlockBackend *blk);
1332 
1333 void bdrv_set_dirty(BlockDriverState *bs, int64_t offset, int64_t bytes);
1334 
1335 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap **out);
1336 void bdrv_restore_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap *backup);
1337 bool bdrv_dirty_bitmap_merge_internal(BdrvDirtyBitmap *dest,
1338                                       const BdrvDirtyBitmap *src,
1339                                       HBitmap **backup, bool lock);
1340 
1341 void bdrv_inc_in_flight(BlockDriverState *bs);
1342 void bdrv_dec_in_flight(BlockDriverState *bs);
1343 
1344 void blockdev_close_all_bdrv_states(void);
1345 
1346 int coroutine_fn bdrv_co_copy_range_from(BdrvChild *src, uint64_t src_offset,
1347                                          BdrvChild *dst, uint64_t dst_offset,
1348                                          uint64_t bytes,
1349                                          BdrvRequestFlags read_flags,
1350                                          BdrvRequestFlags write_flags);
1351 int coroutine_fn bdrv_co_copy_range_to(BdrvChild *src, uint64_t src_offset,
1352                                        BdrvChild *dst, uint64_t dst_offset,
1353                                        uint64_t bytes,
1354                                        BdrvRequestFlags read_flags,
1355                                        BdrvRequestFlags write_flags);
1356 
1357 int refresh_total_sectors(BlockDriverState *bs, int64_t hint);
1358 
1359 void bdrv_set_monitor_owned(BlockDriverState *bs);
1360 BlockDriverState *bds_tree_init(QDict *bs_opts, Error **errp);
1361 
1362 /**
1363  * Simple implementation of bdrv_co_create_opts for protocol drivers
1364  * which only support creation via opening a file
1365  * (usually existing raw storage device)
1366  */
1367 int coroutine_fn bdrv_co_create_opts_simple(BlockDriver *drv,
1368                                             const char *filename,
1369                                             QemuOpts *opts,
1370                                             Error **errp);
1371 extern QemuOptsList bdrv_create_opts_simple;
1372 
1373 BdrvDirtyBitmap *block_dirty_bitmap_lookup(const char *node,
1374                                            const char *name,
1375                                            BlockDriverState **pbs,
1376                                            Error **errp);
1377 BdrvDirtyBitmap *block_dirty_bitmap_merge(const char *node, const char *target,
1378                                           BlockDirtyBitmapMergeSourceList *bms,
1379                                           HBitmap **backup, Error **errp);
1380 BdrvDirtyBitmap *block_dirty_bitmap_remove(const char *node, const char *name,
1381                                            bool release,
1382                                            BlockDriverState **bitmap_bs,
1383                                            Error **errp);
1384 
1385 #endif /* BLOCK_INT_H */
1386