xref: /openbmc/qemu/include/block/block_int.h (revision 72f463bc0803b74cabf0655df1ef4b749ef8dbbd)
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/main-loop.h"
37 #include "qemu/throttle.h"
38 
39 #define BLOCK_FLAG_LAZY_REFCOUNTS   8
40 
41 #define BLOCK_OPT_SIZE              "size"
42 #define BLOCK_OPT_ENCRYPT           "encryption"
43 #define BLOCK_OPT_ENCRYPT_FORMAT    "encrypt.format"
44 #define BLOCK_OPT_COMPAT6           "compat6"
45 #define BLOCK_OPT_HWVERSION         "hwversion"
46 #define BLOCK_OPT_BACKING_FILE      "backing_file"
47 #define BLOCK_OPT_BACKING_FMT       "backing_fmt"
48 #define BLOCK_OPT_CLUSTER_SIZE      "cluster_size"
49 #define BLOCK_OPT_TABLE_SIZE        "table_size"
50 #define BLOCK_OPT_PREALLOC          "preallocation"
51 #define BLOCK_OPT_SUBFMT            "subformat"
52 #define BLOCK_OPT_COMPAT_LEVEL      "compat"
53 #define BLOCK_OPT_LAZY_REFCOUNTS    "lazy_refcounts"
54 #define BLOCK_OPT_ADAPTER_TYPE      "adapter_type"
55 #define BLOCK_OPT_REDUNDANCY        "redundancy"
56 #define BLOCK_OPT_NOCOW             "nocow"
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 
62 #define BLOCK_PROBE_BUF_SIZE        512
63 
64 enum BdrvTrackedRequestType {
65     BDRV_TRACKED_READ,
66     BDRV_TRACKED_WRITE,
67     BDRV_TRACKED_DISCARD,
68     BDRV_TRACKED_TRUNCATE,
69 };
70 
71 typedef struct BdrvTrackedRequest {
72     BlockDriverState *bs;
73     int64_t offset;
74     uint64_t bytes;
75     enum BdrvTrackedRequestType type;
76 
77     bool serialising;
78     int64_t overlap_offset;
79     uint64_t overlap_bytes;
80 
81     QLIST_ENTRY(BdrvTrackedRequest) list;
82     Coroutine *co; /* owner, used for deadlock detection */
83     CoQueue wait_queue; /* coroutines blocked on this request */
84 
85     struct BdrvTrackedRequest *waiting_for;
86 } BdrvTrackedRequest;
87 
88 struct BlockDriver {
89     const char *format_name;
90     int instance_size;
91 
92     /* set to true if the BlockDriver is a block filter. Block filters pass
93      * certain callbacks that refer to data (see block.c) to their bs->file if
94      * the driver doesn't implement them. Drivers that do not wish to forward
95      * must implement them and return -ENOTSUP.
96      */
97     bool is_filter;
98     /* for snapshots block filter like Quorum can implement the
99      * following recursive callback.
100      * It's purpose is to recurse on the filter children while calling
101      * bdrv_recurse_is_first_non_filter on them.
102      * For a sample implementation look in the future Quorum block filter.
103      */
104     bool (*bdrv_recurse_is_first_non_filter)(BlockDriverState *bs,
105                                              BlockDriverState *candidate);
106 
107     int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename);
108     int (*bdrv_probe_device)(const char *filename);
109 
110     /* Any driver implementing this callback is expected to be able to handle
111      * NULL file names in its .bdrv_open() implementation */
112     void (*bdrv_parse_filename)(const char *filename, QDict *options, Error **errp);
113     /* Drivers not implementing bdrv_parse_filename nor bdrv_open should have
114      * this field set to true, except ones that are defined only by their
115      * child's bs.
116      * An example of the last type will be the quorum block driver.
117      */
118     bool bdrv_needs_filename;
119 
120     /* Set if a driver can support backing files */
121     bool supports_backing;
122 
123     /* For handling image reopen for split or non-split files */
124     int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state,
125                                BlockReopenQueue *queue, Error **errp);
126     void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state);
127     void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state);
128     void (*bdrv_join_options)(QDict *options, QDict *old_options);
129 
130     int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags,
131                      Error **errp);
132 
133     /* Protocol drivers should implement this instead of bdrv_open */
134     int (*bdrv_file_open)(BlockDriverState *bs, QDict *options, int flags,
135                           Error **errp);
136     void (*bdrv_close)(BlockDriverState *bs);
137     int coroutine_fn (*bdrv_co_create)(BlockdevCreateOptions *opts,
138                                        Error **errp);
139     int coroutine_fn (*bdrv_co_create_opts)(const char *filename,
140                                             QemuOpts *opts,
141                                             Error **errp);
142     int (*bdrv_make_empty)(BlockDriverState *bs);
143 
144     /*
145      * Refreshes the bs->exact_filename field. If that is impossible,
146      * bs->exact_filename has to be left empty.
147      */
148     void (*bdrv_refresh_filename)(BlockDriverState *bs);
149 
150     /*
151      * Gathers the open options for all children into @target.
152      * A simple format driver (without backing file support) might
153      * implement this function like this:
154      *
155      *     QINCREF(bs->file->bs->full_open_options);
156      *     qdict_put(target, "file", bs->file->bs->full_open_options);
157      *
158      * If not specified, the generic implementation will simply put
159      * all children's options under their respective name.
160      *
161      * @backing_overridden is true when bs->backing seems not to be
162      * the child that would result from opening bs->backing_file.
163      * Therefore, if it is true, the backing child's options should be
164      * gathered; otherwise, there is no need since the backing child
165      * is the one implied by the image header.
166      *
167      * Note that ideally this function would not be needed.  Every
168      * block driver which implements it is probably doing something
169      * shady regarding its runtime option structure.
170      */
171     void (*bdrv_gather_child_options)(BlockDriverState *bs, QDict *target,
172                                       bool backing_overridden);
173 
174     /*
175      * Returns an allocated string which is the directory name of this BDS: It
176      * will be used to make relative filenames absolute by prepending this
177      * function's return value to them.
178      */
179     char *(*bdrv_dirname)(BlockDriverState *bs, Error **errp);
180 
181     /* aio */
182     BlockAIOCB *(*bdrv_aio_preadv)(BlockDriverState *bs,
183         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags,
184         BlockCompletionFunc *cb, void *opaque);
185     BlockAIOCB *(*bdrv_aio_pwritev)(BlockDriverState *bs,
186         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags,
187         BlockCompletionFunc *cb, void *opaque);
188     BlockAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs,
189         BlockCompletionFunc *cb, void *opaque);
190     BlockAIOCB *(*bdrv_aio_pdiscard)(BlockDriverState *bs,
191         int64_t offset, int bytes,
192         BlockCompletionFunc *cb, void *opaque);
193 
194     int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs,
195         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
196 
197     /**
198      * @offset: position in bytes to read at
199      * @bytes: number of bytes to read
200      * @qiov: the buffers to fill with read data
201      * @flags: currently unused, always 0
202      *
203      * @offset and @bytes will be a multiple of 'request_alignment',
204      * but the length of individual @qiov elements does not have to
205      * be a multiple.
206      *
207      * @bytes will always equal the total size of @qiov, and will be
208      * no larger than 'max_transfer'.
209      *
210      * The buffer in @qiov may point directly to guest memory.
211      */
212     int coroutine_fn (*bdrv_co_preadv)(BlockDriverState *bs,
213         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags);
214     int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs,
215         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, int flags);
216     /**
217      * @offset: position in bytes to write at
218      * @bytes: number of bytes to write
219      * @qiov: the buffers containing data to write
220      * @flags: zero or more bits allowed by 'supported_write_flags'
221      *
222      * @offset and @bytes will be a multiple of 'request_alignment',
223      * but the length of individual @qiov elements does not have to
224      * be a multiple.
225      *
226      * @bytes will always equal the total size of @qiov, and will be
227      * no larger than 'max_transfer'.
228      *
229      * The buffer in @qiov may point directly to guest memory.
230      */
231     int coroutine_fn (*bdrv_co_pwritev)(BlockDriverState *bs,
232         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags);
233 
234     /*
235      * Efficiently zero a region of the disk image.  Typically an image format
236      * would use a compact metadata representation to implement this.  This
237      * function pointer may be NULL or return -ENOSUP and .bdrv_co_writev()
238      * will be called instead.
239      */
240     int coroutine_fn (*bdrv_co_pwrite_zeroes)(BlockDriverState *bs,
241         int64_t offset, int bytes, BdrvRequestFlags flags);
242     int coroutine_fn (*bdrv_co_pdiscard)(BlockDriverState *bs,
243         int64_t offset, int bytes);
244 
245     /* Map [offset, offset + nbytes) range onto a child of @bs to copy from,
246      * and invoke bdrv_co_copy_range_from(child, ...), or invoke
247      * bdrv_co_copy_range_to() if @bs is the leaf child to copy data from.
248      *
249      * See the comment of bdrv_co_copy_range for the parameter and return value
250      * semantics.
251      */
252     int coroutine_fn (*bdrv_co_copy_range_from)(BlockDriverState *bs,
253                                                 BdrvChild *src,
254                                                 uint64_t offset,
255                                                 BdrvChild *dst,
256                                                 uint64_t dst_offset,
257                                                 uint64_t bytes,
258                                                 BdrvRequestFlags read_flags,
259                                                 BdrvRequestFlags write_flags);
260 
261     /* Map [offset, offset + nbytes) range onto a child of bs to copy data to,
262      * and invoke bdrv_co_copy_range_to(child, src, ...), or perform the copy
263      * operation if @bs is the leaf and @src has the same BlockDriver.  Return
264      * -ENOTSUP if @bs is the leaf but @src has a different BlockDriver.
265      *
266      * See the comment of bdrv_co_copy_range for the parameter and return value
267      * semantics.
268      */
269     int coroutine_fn (*bdrv_co_copy_range_to)(BlockDriverState *bs,
270                                               BdrvChild *src,
271                                               uint64_t src_offset,
272                                               BdrvChild *dst,
273                                               uint64_t dst_offset,
274                                               uint64_t bytes,
275                                               BdrvRequestFlags read_flags,
276                                               BdrvRequestFlags write_flags);
277 
278     /*
279      * Building block for bdrv_block_status[_above] and
280      * bdrv_is_allocated[_above].  The driver should answer only
281      * according to the current layer, and should only need to set
282      * BDRV_BLOCK_DATA, BDRV_BLOCK_ZERO, BDRV_BLOCK_OFFSET_VALID,
283      * and/or BDRV_BLOCK_RAW; if the current layer defers to a backing
284      * layer, the result should be 0 (and not BDRV_BLOCK_ZERO).  See
285      * block.h for the overall meaning of the bits.  As a hint, the
286      * flag want_zero is true if the caller cares more about precise
287      * mappings (favor accurate _OFFSET_VALID/_ZERO) or false for
288      * overall allocation (favor larger *pnum, perhaps by reporting
289      * _DATA instead of _ZERO).  The block layer guarantees input
290      * clamped to bdrv_getlength() and aligned to request_alignment,
291      * as well as non-NULL pnum, map, and file; in turn, the driver
292      * must return an error or set pnum to an aligned non-zero value.
293      */
294     int coroutine_fn (*bdrv_co_block_status)(BlockDriverState *bs,
295         bool want_zero, int64_t offset, int64_t bytes, int64_t *pnum,
296         int64_t *map, BlockDriverState **file);
297 
298     /*
299      * Invalidate any cached meta-data.
300      */
301     void coroutine_fn (*bdrv_co_invalidate_cache)(BlockDriverState *bs,
302                                                   Error **errp);
303     int (*bdrv_inactivate)(BlockDriverState *bs);
304 
305     /*
306      * Flushes all data for all layers by calling bdrv_co_flush for underlying
307      * layers, if needed. This function is needed for deterministic
308      * synchronization of the flush finishing callback.
309      */
310     int coroutine_fn (*bdrv_co_flush)(BlockDriverState *bs);
311 
312     /*
313      * Flushes all data that was already written to the OS all the way down to
314      * the disk (for example file-posix.c calls fsync()).
315      */
316     int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs);
317 
318     /*
319      * Flushes all internal caches to the OS. The data may still sit in a
320      * writeback cache of the host OS, but it will survive a crash of the qemu
321      * process.
322      */
323     int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs);
324 
325     /*
326      * Drivers setting this field must be able to work with just a plain
327      * filename with '<protocol_name>:' as a prefix, and no other options.
328      * Options may be extracted from the filename by implementing
329      * bdrv_parse_filename.
330      */
331     const char *protocol_name;
332     int coroutine_fn (*bdrv_co_truncate)(BlockDriverState *bs, int64_t offset,
333                                          PreallocMode prealloc, Error **errp);
334 
335     int64_t (*bdrv_getlength)(BlockDriverState *bs);
336     bool has_variable_length;
337     int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
338     BlockMeasureInfo *(*bdrv_measure)(QemuOpts *opts, BlockDriverState *in_bs,
339                                       Error **errp);
340 
341     int coroutine_fn (*bdrv_co_pwritev_compressed)(BlockDriverState *bs,
342         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov);
343 
344     int (*bdrv_snapshot_create)(BlockDriverState *bs,
345                                 QEMUSnapshotInfo *sn_info);
346     int (*bdrv_snapshot_goto)(BlockDriverState *bs,
347                               const char *snapshot_id);
348     int (*bdrv_snapshot_delete)(BlockDriverState *bs,
349                                 const char *snapshot_id,
350                                 const char *name,
351                                 Error **errp);
352     int (*bdrv_snapshot_list)(BlockDriverState *bs,
353                               QEMUSnapshotInfo **psn_info);
354     int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
355                                   const char *snapshot_id,
356                                   const char *name,
357                                   Error **errp);
358     int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
359     ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs,
360                                                  Error **errp);
361 
362     int coroutine_fn (*bdrv_save_vmstate)(BlockDriverState *bs,
363                                           QEMUIOVector *qiov,
364                                           int64_t pos);
365     int coroutine_fn (*bdrv_load_vmstate)(BlockDriverState *bs,
366                                           QEMUIOVector *qiov,
367                                           int64_t pos);
368 
369     int (*bdrv_change_backing_file)(BlockDriverState *bs,
370         const char *backing_file, const char *backing_fmt);
371 
372     /* removable device specific */
373     bool (*bdrv_is_inserted)(BlockDriverState *bs);
374     void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
375     void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
376 
377     /* to control generic scsi devices */
378     BlockAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
379         unsigned long int req, void *buf,
380         BlockCompletionFunc *cb, void *opaque);
381     int coroutine_fn (*bdrv_co_ioctl)(BlockDriverState *bs,
382                                       unsigned long int req, void *buf);
383 
384     /* List of options for creating images, terminated by name == NULL */
385     QemuOptsList *create_opts;
386 
387     /*
388      * Returns 0 for completed check, -errno for internal errors.
389      * The check results are stored in result.
390      */
391     int coroutine_fn (*bdrv_co_check)(BlockDriverState *bs,
392                                       BdrvCheckResult *result,
393                                       BdrvCheckMode fix);
394 
395     int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts,
396                               BlockDriverAmendStatusCB *status_cb,
397                               void *cb_opaque,
398                               Error **errp);
399 
400     void (*bdrv_debug_event)(BlockDriverState *bs, BlkdebugEvent event);
401 
402     /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
403     int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
404         const char *tag);
405     int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs,
406         const char *tag);
407     int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
408     bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
409 
410     void (*bdrv_refresh_limits)(BlockDriverState *bs, Error **errp);
411 
412     /*
413      * Returns 1 if newly created images are guaranteed to contain only
414      * zeros, 0 otherwise.
415      */
416     int (*bdrv_has_zero_init)(BlockDriverState *bs);
417 
418     /* Remove fd handlers, timers, and other event loop callbacks so the event
419      * loop is no longer in use.  Called with no in-flight requests and in
420      * depth-first traversal order with parents before child nodes.
421      */
422     void (*bdrv_detach_aio_context)(BlockDriverState *bs);
423 
424     /* Add fd handlers, timers, and other event loop callbacks so I/O requests
425      * can be processed again.  Called with no in-flight requests and in
426      * depth-first traversal order with child nodes before parent nodes.
427      */
428     void (*bdrv_attach_aio_context)(BlockDriverState *bs,
429                                     AioContext *new_context);
430 
431     /* io queue for linux-aio */
432     void (*bdrv_io_plug)(BlockDriverState *bs);
433     void (*bdrv_io_unplug)(BlockDriverState *bs);
434 
435     /**
436      * Try to get @bs's logical and physical block size.
437      * On success, store them in @bsz and return zero.
438      * On failure, return negative errno.
439      */
440     int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz);
441     /**
442      * Try to get @bs's geometry (cyls, heads, sectors)
443      * On success, store them in @geo and return 0.
444      * On failure return -errno.
445      * Only drivers that want to override guest geometry implement this
446      * callback; see hd_geometry_guess().
447      */
448     int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo);
449 
450     /**
451      * bdrv_co_drain_begin is called if implemented in the beginning of a
452      * drain operation to drain and stop any internal sources of requests in
453      * the driver.
454      * bdrv_co_drain_end is called if implemented at the end of the drain.
455      *
456      * They should be used by the driver to e.g. manage scheduled I/O
457      * requests, or toggle an internal state. After the end of the drain new
458      * requests will continue normally.
459      */
460     void coroutine_fn (*bdrv_co_drain_begin)(BlockDriverState *bs);
461     void coroutine_fn (*bdrv_co_drain_end)(BlockDriverState *bs);
462 
463     void (*bdrv_add_child)(BlockDriverState *parent, BlockDriverState *child,
464                            Error **errp);
465     void (*bdrv_del_child)(BlockDriverState *parent, BdrvChild *child,
466                            Error **errp);
467 
468     /**
469      * Informs the block driver that a permission change is intended. The
470      * driver checks whether the change is permissible and may take other
471      * preparations for the change (e.g. get file system locks). This operation
472      * is always followed either by a call to either .bdrv_set_perm or
473      * .bdrv_abort_perm_update.
474      *
475      * Checks whether the requested set of cumulative permissions in @perm
476      * can be granted for accessing @bs and whether no other users are using
477      * permissions other than those given in @shared (both arguments take
478      * BLK_PERM_* bitmasks).
479      *
480      * If both conditions are met, 0 is returned. Otherwise, -errno is returned
481      * and errp is set to an error describing the conflict.
482      */
483     int (*bdrv_check_perm)(BlockDriverState *bs, uint64_t perm,
484                            uint64_t shared, Error **errp);
485 
486     /**
487      * Called to inform the driver that the set of cumulative set of used
488      * permissions for @bs has changed to @perm, and the set of sharable
489      * permission to @shared. The driver can use this to propagate changes to
490      * its children (i.e. request permissions only if a parent actually needs
491      * them).
492      *
493      * This function is only invoked after bdrv_check_perm(), so block drivers
494      * may rely on preparations made in their .bdrv_check_perm implementation.
495      */
496     void (*bdrv_set_perm)(BlockDriverState *bs, uint64_t perm, uint64_t shared);
497 
498     /*
499      * Called to inform the driver that after a previous bdrv_check_perm()
500      * call, the permission update is not performed and any preparations made
501      * for it (e.g. taken file locks) need to be undone.
502      *
503      * This function can be called even for nodes that never saw a
504      * bdrv_check_perm() call. It is a no-op then.
505      */
506     void (*bdrv_abort_perm_update)(BlockDriverState *bs);
507 
508     /**
509      * Returns in @nperm and @nshared the permissions that the driver for @bs
510      * needs on its child @c, based on the cumulative permissions requested by
511      * the parents in @parent_perm and @parent_shared.
512      *
513      * If @c is NULL, return the permissions for attaching a new child for the
514      * given @role.
515      *
516      * If @reopen_queue is non-NULL, don't return the currently needed
517      * permissions, but those that will be needed after applying the
518      * @reopen_queue.
519      */
520      void (*bdrv_child_perm)(BlockDriverState *bs, BdrvChild *c,
521                              const BdrvChildRole *role,
522                              BlockReopenQueue *reopen_queue,
523                              uint64_t parent_perm, uint64_t parent_shared,
524                              uint64_t *nperm, uint64_t *nshared);
525 
526     /**
527      * Bitmaps should be marked as 'IN_USE' in the image on reopening image
528      * as rw. This handler should realize it. It also should unset readonly
529      * field of BlockDirtyBitmap's in case of success.
530      */
531     int (*bdrv_reopen_bitmaps_rw)(BlockDriverState *bs, Error **errp);
532     bool (*bdrv_can_store_new_dirty_bitmap)(BlockDriverState *bs,
533                                             const char *name,
534                                             uint32_t granularity,
535                                             Error **errp);
536     void (*bdrv_remove_persistent_dirty_bitmap)(BlockDriverState *bs,
537                                                 const char *name,
538                                                 Error **errp);
539 
540     /**
541      * Register/unregister a buffer for I/O. For example, when the driver is
542      * interested to know the memory areas that will later be used in iovs, so
543      * that it can do IOMMU mapping with VFIO etc., in order to get better
544      * performance. In the case of VFIO drivers, this callback is used to do
545      * DMA mapping for hot buffers.
546      */
547     void (*bdrv_register_buf)(BlockDriverState *bs, void *host, size_t size);
548     void (*bdrv_unregister_buf)(BlockDriverState *bs, void *host);
549     QLIST_ENTRY(BlockDriver) list;
550 
551     /* Pointer to a NULL-terminated array of names of strong options
552      * that can be specified for bdrv_open(). A strong option is one
553      * that changes the data of a BDS.
554      * If this pointer is NULL, the array is considered empty.
555      * "filename" and "driver" are always considered strong. */
556     const char *const *strong_runtime_opts;
557 };
558 
559 typedef struct BlockLimits {
560     /* Alignment requirement, in bytes, for offset/length of I/O
561      * requests. Must be a power of 2 less than INT_MAX; defaults to
562      * 1 for drivers with modern byte interfaces, and to 512
563      * otherwise. */
564     uint32_t request_alignment;
565 
566     /* Maximum number of bytes that can be discarded at once (since it
567      * is signed, it must be < 2G, if set). Must be multiple of
568      * pdiscard_alignment, but need not be power of 2. May be 0 if no
569      * inherent 32-bit limit */
570     int32_t max_pdiscard;
571 
572     /* Optimal alignment for discard requests in bytes. A power of 2
573      * is best but not mandatory.  Must be a multiple of
574      * bl.request_alignment, and must be less than max_pdiscard if
575      * that is set. May be 0 if bl.request_alignment is good enough */
576     uint32_t pdiscard_alignment;
577 
578     /* Maximum number of bytes that can zeroized at once (since it is
579      * signed, it must be < 2G, if set). Must be multiple of
580      * pwrite_zeroes_alignment. May be 0 if no inherent 32-bit limit */
581     int32_t max_pwrite_zeroes;
582 
583     /* Optimal alignment for write zeroes requests in bytes. A power
584      * of 2 is best but not mandatory.  Must be a multiple of
585      * bl.request_alignment, and must be less than max_pwrite_zeroes
586      * if that is set. May be 0 if bl.request_alignment is good
587      * enough */
588     uint32_t pwrite_zeroes_alignment;
589 
590     /* Optimal transfer length in bytes.  A power of 2 is best but not
591      * mandatory.  Must be a multiple of bl.request_alignment, or 0 if
592      * no preferred size */
593     uint32_t opt_transfer;
594 
595     /* Maximal transfer length in bytes.  Need not be power of 2, but
596      * must be multiple of opt_transfer and bl.request_alignment, or 0
597      * for no 32-bit limit.  For now, anything larger than INT_MAX is
598      * clamped down. */
599     uint32_t max_transfer;
600 
601     /* memory alignment, in bytes so that no bounce buffer is needed */
602     size_t min_mem_alignment;
603 
604     /* memory alignment, in bytes, for bounce buffer */
605     size_t opt_mem_alignment;
606 
607     /* maximum number of iovec elements */
608     int max_iov;
609 } BlockLimits;
610 
611 typedef struct BdrvOpBlocker BdrvOpBlocker;
612 
613 typedef struct BdrvAioNotifier {
614     void (*attached_aio_context)(AioContext *new_context, void *opaque);
615     void (*detach_aio_context)(void *opaque);
616 
617     void *opaque;
618     bool deleted;
619 
620     QLIST_ENTRY(BdrvAioNotifier) list;
621 } BdrvAioNotifier;
622 
623 struct BdrvChildRole {
624     /* If true, bdrv_replace_node() doesn't change the node this BdrvChild
625      * points to. */
626     bool stay_at_node;
627 
628     /* If true, the parent is a BlockDriverState and bdrv_next_all_states()
629      * will return it. This information is used for drain_all, where every node
630      * will be drained separately, so the drain only needs to be propagated to
631      * non-BDS parents. */
632     bool parent_is_bds;
633 
634     void (*inherit_options)(int *child_flags, QDict *child_options,
635                             int parent_flags, QDict *parent_options);
636 
637     void (*change_media)(BdrvChild *child, bool load);
638     void (*resize)(BdrvChild *child);
639 
640     /* Returns a name that is supposedly more useful for human users than the
641      * node name for identifying the node in question (in particular, a BB
642      * name), or NULL if the parent can't provide a better name. */
643     const char *(*get_name)(BdrvChild *child);
644 
645     /* Returns a malloced string that describes the parent of the child for a
646      * human reader. This could be a node-name, BlockBackend name, qdev ID or
647      * QOM path of the device owning the BlockBackend, job type and ID etc. The
648      * caller is responsible for freeing the memory. */
649     char *(*get_parent_desc)(BdrvChild *child);
650 
651     /*
652      * If this pair of functions is implemented, the parent doesn't issue new
653      * requests after returning from .drained_begin() until .drained_end() is
654      * called.
655      *
656      * These functions must not change the graph (and therefore also must not
657      * call aio_poll(), which could change the graph indirectly).
658      *
659      * Note that this can be nested. If drained_begin() was called twice, new
660      * I/O is allowed only after drained_end() was called twice, too.
661      */
662     void (*drained_begin)(BdrvChild *child);
663     void (*drained_end)(BdrvChild *child);
664 
665     /*
666      * Returns whether the parent has pending requests for the child. This
667      * callback is polled after .drained_begin() has been called until all
668      * activity on the child has stopped.
669      */
670     bool (*drained_poll)(BdrvChild *child);
671 
672     /* Notifies the parent that the child has been activated/inactivated (e.g.
673      * when migration is completing) and it can start/stop requesting
674      * permissions and doing I/O on it. */
675     void (*activate)(BdrvChild *child, Error **errp);
676     int (*inactivate)(BdrvChild *child);
677 
678     void (*attach)(BdrvChild *child);
679     void (*detach)(BdrvChild *child);
680 
681     /* Notifies the parent that the filename of its child has changed (e.g.
682      * because the direct child was removed from the backing chain), so that it
683      * can update its reference. */
684     int (*update_filename)(BdrvChild *child, BlockDriverState *new_base,
685                            const char *filename, Error **errp);
686 };
687 
688 extern const BdrvChildRole child_file;
689 extern const BdrvChildRole child_format;
690 extern const BdrvChildRole child_backing;
691 
692 struct BdrvChild {
693     BlockDriverState *bs;
694     char *name;
695     const BdrvChildRole *role;
696     void *opaque;
697 
698     /**
699      * Granted permissions for operating on this BdrvChild (BLK_PERM_* bitmask)
700      */
701     uint64_t perm;
702 
703     /**
704      * Permissions that can still be granted to other users of @bs while this
705      * BdrvChild is still attached to it. (BLK_PERM_* bitmask)
706      */
707     uint64_t shared_perm;
708 
709     /* backup of permissions during permission update procedure */
710     bool has_backup_perm;
711     uint64_t backup_perm;
712     uint64_t backup_shared_perm;
713 
714     QLIST_ENTRY(BdrvChild) next;
715     QLIST_ENTRY(BdrvChild) next_parent;
716 };
717 
718 /*
719  * Note: the function bdrv_append() copies and swaps contents of
720  * BlockDriverStates, so if you add new fields to this struct, please
721  * inspect bdrv_append() to determine if the new fields need to be
722  * copied as well.
723  */
724 struct BlockDriverState {
725     /* Protected by big QEMU lock or read-only after opening.  No special
726      * locking needed during I/O...
727      */
728     int open_flags; /* flags used to open the file, re-used for re-open */
729     bool read_only; /* if true, the media is read only */
730     bool encrypted; /* if true, the media is encrypted */
731     bool sg;        /* if true, the device is a /dev/sg* */
732     bool probed;    /* if true, format was probed rather than specified */
733     bool force_share; /* if true, always allow all shared permissions */
734     bool implicit;  /* if true, this filter node was automatically inserted */
735 
736     BlockDriver *drv; /* NULL means no media */
737     void *opaque;
738 
739     AioContext *aio_context; /* event loop used for fd handlers, timers, etc */
740     /* long-running tasks intended to always use the same AioContext as this
741      * BDS may register themselves in this list to be notified of changes
742      * regarding this BDS's context */
743     QLIST_HEAD(, BdrvAioNotifier) aio_notifiers;
744     bool walking_aio_notifiers; /* to make removal during iteration safe */
745 
746     char filename[PATH_MAX];
747     char backing_file[PATH_MAX]; /* if non zero, the image is a diff of
748                                     this file image */
749     /* The backing filename indicated by the image header; if we ever
750      * open this file, then this is replaced by the resulting BDS's
751      * filename (i.e. after a bdrv_refresh_filename() run). */
752     char auto_backing_file[PATH_MAX];
753     char backing_format[16]; /* if non-zero and backing_file exists */
754 
755     QDict *full_open_options;
756     char exact_filename[PATH_MAX];
757 
758     BdrvChild *backing;
759     BdrvChild *file;
760 
761     /* I/O Limits */
762     BlockLimits bl;
763 
764     /* Flags honored during pwrite (so far: BDRV_REQ_FUA,
765      * BDRV_REQ_WRITE_UNCHANGED).
766      * If a driver does not support BDRV_REQ_WRITE_UNCHANGED, those
767      * writes will be issued as normal writes without the flag set.
768      * This is important to note for drivers that do not explicitly
769      * request a WRITE permission for their children and instead take
770      * the same permissions as their parent did (this is commonly what
771      * block filters do).  Such drivers have to be aware that the
772      * parent may have taken a WRITE_UNCHANGED permission only and is
773      * issuing such requests.  Drivers either must make sure that
774      * these requests do not result in plain WRITE accesses (usually
775      * by supporting BDRV_REQ_WRITE_UNCHANGED, and then forwarding
776      * every incoming write request as-is, including potentially that
777      * flag), or they have to explicitly take the WRITE permission for
778      * their children. */
779     unsigned int supported_write_flags;
780     /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA,
781      * BDRV_REQ_MAY_UNMAP, BDRV_REQ_WRITE_UNCHANGED) */
782     unsigned int supported_zero_flags;
783 
784     /* the following member gives a name to every node on the bs graph. */
785     char node_name[32];
786     /* element of the list of named nodes building the graph */
787     QTAILQ_ENTRY(BlockDriverState) node_list;
788     /* element of the list of all BlockDriverStates (all_bdrv_states) */
789     QTAILQ_ENTRY(BlockDriverState) bs_list;
790     /* element of the list of monitor-owned BDS */
791     QTAILQ_ENTRY(BlockDriverState) monitor_list;
792     int refcnt;
793 
794     /* operation blockers */
795     QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX];
796 
797     /* long-running background operation */
798     BlockJob *job;
799 
800     /* The node that this node inherited default options from (and a reopen on
801      * which can affect this node by changing these defaults). This is always a
802      * parent node of this node. */
803     BlockDriverState *inherits_from;
804     QLIST_HEAD(, BdrvChild) children;
805     QLIST_HEAD(, BdrvChild) parents;
806 
807     QDict *options;
808     QDict *explicit_options;
809     BlockdevDetectZeroesOptions detect_zeroes;
810 
811     /* The error object in use for blocking operations on backing_hd */
812     Error *backing_blocker;
813 
814     /* Protected by AioContext lock */
815 
816     /* If we are reading a disk image, give its size in sectors.
817      * Generally read-only; it is written to by load_snapshot and
818      * save_snaphost, but the block layer is quiescent during those.
819      */
820     int64_t total_sectors;
821 
822     /* Callback before write request is processed */
823     NotifierWithReturnList before_write_notifiers;
824 
825     /* threshold limit for writes, in bytes. "High water mark". */
826     uint64_t write_threshold_offset;
827     NotifierWithReturn write_threshold_notifier;
828 
829     /* Writing to the list requires the BQL _and_ the dirty_bitmap_mutex.
830      * Reading from the list can be done with either the BQL or the
831      * dirty_bitmap_mutex.  Modifying a bitmap only requires
832      * dirty_bitmap_mutex.  */
833     QemuMutex dirty_bitmap_mutex;
834     QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps;
835 
836     /* Offset after the highest byte written to */
837     Stat64 wr_highest_offset;
838 
839     /* If true, copy read backing sectors into image.  Can be >1 if more
840      * than one client has requested copy-on-read.  Accessed with atomic
841      * ops.
842      */
843     int copy_on_read;
844 
845     /* number of in-flight requests; overall and serialising.
846      * Accessed with atomic ops.
847      */
848     unsigned int in_flight;
849     unsigned int serialising_in_flight;
850 
851     /* counter for nested bdrv_io_plug.
852      * Accessed with atomic ops.
853     */
854     unsigned io_plugged;
855 
856     /* do we need to tell the quest if we have a volatile write cache? */
857     int enable_write_cache;
858 
859     /* Accessed with atomic ops.  */
860     int quiesce_counter;
861     int recursive_quiesce_counter;
862 
863     unsigned int write_gen;               /* Current data generation */
864 
865     /* Protected by reqs_lock.  */
866     CoMutex reqs_lock;
867     QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
868     CoQueue flush_queue;                  /* Serializing flush queue */
869     bool active_flush_req;                /* Flush request in flight? */
870 
871     /* Only read/written by whoever has set active_flush_req to true.  */
872     unsigned int flushed_gen;             /* Flushed write generation */
873 };
874 
875 struct BlockBackendRootState {
876     int open_flags;
877     bool read_only;
878     BlockdevDetectZeroesOptions detect_zeroes;
879 };
880 
881 typedef enum BlockMirrorBackingMode {
882     /* Reuse the existing backing chain from the source for the target.
883      * - sync=full: Set backing BDS to NULL.
884      * - sync=top:  Use source's backing BDS.
885      * - sync=none: Use source as the backing BDS. */
886     MIRROR_SOURCE_BACKING_CHAIN,
887 
888     /* Open the target's backing chain completely anew */
889     MIRROR_OPEN_BACKING_CHAIN,
890 
891     /* Do not change the target's backing BDS after job completion */
892     MIRROR_LEAVE_BACKING_CHAIN,
893 } BlockMirrorBackingMode;
894 
895 static inline BlockDriverState *backing_bs(BlockDriverState *bs)
896 {
897     return bs->backing ? bs->backing->bs : NULL;
898 }
899 
900 
901 /* Essential block drivers which must always be statically linked into qemu, and
902  * which therefore can be accessed without using bdrv_find_format() */
903 extern BlockDriver bdrv_file;
904 extern BlockDriver bdrv_raw;
905 extern BlockDriver bdrv_qcow2;
906 
907 int coroutine_fn bdrv_co_preadv(BdrvChild *child,
908     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
909     BdrvRequestFlags flags);
910 int coroutine_fn bdrv_co_pwritev(BdrvChild *child,
911     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
912     BdrvRequestFlags flags);
913 
914 extern unsigned int bdrv_drain_all_count;
915 void bdrv_apply_subtree_drain(BdrvChild *child, BlockDriverState *new_parent);
916 void bdrv_unapply_subtree_drain(BdrvChild *child, BlockDriverState *old_parent);
917 
918 int get_tmp_filename(char *filename, int size);
919 BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size,
920                             const char *filename);
921 
922 void bdrv_parse_filename_strip_prefix(const char *filename, const char *prefix,
923                                       QDict *options);
924 
925 
926 /**
927  * bdrv_add_before_write_notifier:
928  *
929  * Register a callback that is invoked before write requests are processed but
930  * after any throttling or waiting for overlapping requests.
931  */
932 void bdrv_add_before_write_notifier(BlockDriverState *bs,
933                                     NotifierWithReturn *notifier);
934 
935 /**
936  * bdrv_detach_aio_context:
937  *
938  * May be called from .bdrv_detach_aio_context() to detach children from the
939  * current #AioContext.  This is only needed by block drivers that manage their
940  * own children.  Both ->file and ->backing are automatically handled and
941  * block drivers should not call this function on them explicitly.
942  */
943 void bdrv_detach_aio_context(BlockDriverState *bs);
944 
945 /**
946  * bdrv_attach_aio_context:
947  *
948  * May be called from .bdrv_attach_aio_context() to attach children to the new
949  * #AioContext.  This is only needed by block drivers that manage their own
950  * children.  Both ->file and ->backing are automatically handled and block
951  * drivers should not call this function on them explicitly.
952  */
953 void bdrv_attach_aio_context(BlockDriverState *bs,
954                              AioContext *new_context);
955 
956 /**
957  * bdrv_add_aio_context_notifier:
958  *
959  * If a long-running job intends to be always run in the same AioContext as a
960  * certain BDS, it may use this function to be notified of changes regarding the
961  * association of the BDS to an AioContext.
962  *
963  * attached_aio_context() is called after the target BDS has been attached to a
964  * new AioContext; detach_aio_context() is called before the target BDS is being
965  * detached from its old AioContext.
966  */
967 void bdrv_add_aio_context_notifier(BlockDriverState *bs,
968         void (*attached_aio_context)(AioContext *new_context, void *opaque),
969         void (*detach_aio_context)(void *opaque), void *opaque);
970 
971 /**
972  * bdrv_remove_aio_context_notifier:
973  *
974  * Unsubscribe of change notifications regarding the BDS's AioContext. The
975  * parameters given here have to be the same as those given to
976  * bdrv_add_aio_context_notifier().
977  */
978 void bdrv_remove_aio_context_notifier(BlockDriverState *bs,
979                                       void (*aio_context_attached)(AioContext *,
980                                                                    void *),
981                                       void (*aio_context_detached)(void *),
982                                       void *opaque);
983 
984 /**
985  * bdrv_wakeup:
986  * @bs: The BlockDriverState for which an I/O operation has been completed.
987  *
988  * Wake up the main thread if it is waiting on BDRV_POLL_WHILE.  During
989  * synchronous I/O on a BlockDriverState that is attached to another
990  * I/O thread, the main thread lets the I/O thread's event loop run,
991  * waiting for the I/O operation to complete.  A bdrv_wakeup will wake
992  * up the main thread if necessary.
993  *
994  * Manual calls to bdrv_wakeup are rarely necessary, because
995  * bdrv_dec_in_flight already calls it.
996  */
997 void bdrv_wakeup(BlockDriverState *bs);
998 
999 #ifdef _WIN32
1000 int is_windows_drive(const char *filename);
1001 #endif
1002 
1003 /**
1004  * stream_start:
1005  * @job_id: The id of the newly-created job, or %NULL to use the
1006  * device name of @bs.
1007  * @bs: Block device to operate on.
1008  * @base: Block device that will become the new base, or %NULL to
1009  * flatten the whole backing file chain onto @bs.
1010  * @backing_file_str: The file name that will be written to @bs as the
1011  * the new backing file if the job completes. Ignored if @base is %NULL.
1012  * @creation_flags: Flags that control the behavior of the Job lifetime.
1013  *                  See @BlockJobCreateFlags
1014  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1015  * @on_error: The action to take upon error.
1016  * @errp: Error object.
1017  *
1018  * Start a streaming operation on @bs.  Clusters that are unallocated
1019  * in @bs, but allocated in any image between @base and @bs (both
1020  * exclusive) will be written to @bs.  At the end of a successful
1021  * streaming job, the backing file of @bs will be changed to
1022  * @backing_file_str in the written image and to @base in the live
1023  * BlockDriverState.
1024  */
1025 void stream_start(const char *job_id, BlockDriverState *bs,
1026                   BlockDriverState *base, const char *backing_file_str,
1027                   int creation_flags, int64_t speed,
1028                   BlockdevOnError on_error, Error **errp);
1029 
1030 /**
1031  * commit_start:
1032  * @job_id: The id of the newly-created job, or %NULL to use the
1033  * device name of @bs.
1034  * @bs: Active block device.
1035  * @top: Top block device to be committed.
1036  * @base: Block device that will be written into, and become the new top.
1037  * @creation_flags: Flags that control the behavior of the Job lifetime.
1038  *                  See @BlockJobCreateFlags
1039  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1040  * @on_error: The action to take upon error.
1041  * @backing_file_str: String to use as the backing file in @top's overlay
1042  * @filter_node_name: The node name that should be assigned to the filter
1043  * driver that the commit job inserts into the graph above @top. NULL means
1044  * that a node name should be autogenerated.
1045  * @errp: Error object.
1046  *
1047  */
1048 void commit_start(const char *job_id, BlockDriverState *bs,
1049                   BlockDriverState *base, BlockDriverState *top,
1050                   int creation_flags, int64_t speed,
1051                   BlockdevOnError on_error, const char *backing_file_str,
1052                   const char *filter_node_name, Error **errp);
1053 /**
1054  * commit_active_start:
1055  * @job_id: The id of the newly-created job, or %NULL to use the
1056  * device name of @bs.
1057  * @bs: Active block device to be committed.
1058  * @base: Block device that will be written into, and become the new top.
1059  * @creation_flags: Flags that control the behavior of the Job lifetime.
1060  *                  See @BlockJobCreateFlags
1061  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1062  * @on_error: The action to take upon error.
1063  * @filter_node_name: The node name that should be assigned to the filter
1064  * driver that the commit job inserts into the graph above @bs. NULL means that
1065  * a node name should be autogenerated.
1066  * @cb: Completion function for the job.
1067  * @opaque: Opaque pointer value passed to @cb.
1068  * @auto_complete: Auto complete the job.
1069  * @errp: Error object.
1070  *
1071  */
1072 void commit_active_start(const char *job_id, BlockDriverState *bs,
1073                          BlockDriverState *base, int creation_flags,
1074                          int64_t speed, BlockdevOnError on_error,
1075                          const char *filter_node_name,
1076                          BlockCompletionFunc *cb, void *opaque,
1077                          bool auto_complete, Error **errp);
1078 /*
1079  * mirror_start:
1080  * @job_id: The id of the newly-created job, or %NULL to use the
1081  * device name of @bs.
1082  * @bs: Block device to operate on.
1083  * @target: Block device to write to.
1084  * @replaces: Block graph node name to replace once the mirror is done. Can
1085  *            only be used when full mirroring is selected.
1086  * @creation_flags: Flags that control the behavior of the Job lifetime.
1087  *                  See @BlockJobCreateFlags
1088  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1089  * @granularity: The chosen granularity for the dirty bitmap.
1090  * @buf_size: The amount of data that can be in flight at one time.
1091  * @mode: Whether to collapse all images in the chain to the target.
1092  * @backing_mode: How to establish the target's backing chain after completion.
1093  * @on_source_error: The action to take upon error reading from the source.
1094  * @on_target_error: The action to take upon error writing to the target.
1095  * @unmap: Whether to unmap target where source sectors only contain zeroes.
1096  * @filter_node_name: The node name that should be assigned to the filter
1097  * driver that the mirror job inserts into the graph above @bs. NULL means that
1098  * a node name should be autogenerated.
1099  * @copy_mode: When to trigger writes to the target.
1100  * @errp: Error object.
1101  *
1102  * Start a mirroring operation on @bs.  Clusters that are allocated
1103  * in @bs will be written to @target until the job is cancelled or
1104  * manually completed.  At the end of a successful mirroring job,
1105  * @bs will be switched to read from @target.
1106  */
1107 void mirror_start(const char *job_id, BlockDriverState *bs,
1108                   BlockDriverState *target, const char *replaces,
1109                   int creation_flags, int64_t speed,
1110                   uint32_t granularity, int64_t buf_size,
1111                   MirrorSyncMode mode, BlockMirrorBackingMode backing_mode,
1112                   BlockdevOnError on_source_error,
1113                   BlockdevOnError on_target_error,
1114                   bool unmap, const char *filter_node_name,
1115                   MirrorCopyMode copy_mode, Error **errp);
1116 
1117 /*
1118  * backup_job_create:
1119  * @job_id: The id of the newly-created job, or %NULL to use the
1120  * device name of @bs.
1121  * @bs: Block device to operate on.
1122  * @target: Block device to write to.
1123  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1124  * @sync_mode: What parts of the disk image should be copied to the destination.
1125  * @sync_bitmap: The dirty bitmap if sync_mode is MIRROR_SYNC_MODE_INCREMENTAL.
1126  * @on_source_error: The action to take upon error reading from the source.
1127  * @on_target_error: The action to take upon error writing to the target.
1128  * @creation_flags: Flags that control the behavior of the Job lifetime.
1129  *                  See @BlockJobCreateFlags
1130  * @cb: Completion function for the job.
1131  * @opaque: Opaque pointer value passed to @cb.
1132  * @txn: Transaction that this job is part of (may be NULL).
1133  *
1134  * Create a backup operation on @bs.  Clusters in @bs are written to @target
1135  * until the job is cancelled or manually completed.
1136  */
1137 BlockJob *backup_job_create(const char *job_id, BlockDriverState *bs,
1138                             BlockDriverState *target, int64_t speed,
1139                             MirrorSyncMode sync_mode,
1140                             BdrvDirtyBitmap *sync_bitmap,
1141                             bool compress,
1142                             BlockdevOnError on_source_error,
1143                             BlockdevOnError on_target_error,
1144                             int creation_flags,
1145                             BlockCompletionFunc *cb, void *opaque,
1146                             JobTxn *txn, Error **errp);
1147 
1148 void hmp_drive_add_node(Monitor *mon, const char *optstr);
1149 
1150 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
1151                                   const char *child_name,
1152                                   const BdrvChildRole *child_role,
1153                                   uint64_t perm, uint64_t shared_perm,
1154                                   void *opaque, Error **errp);
1155 void bdrv_root_unref_child(BdrvChild *child);
1156 
1157 int bdrv_child_try_set_perm(BdrvChild *c, uint64_t perm, uint64_t shared,
1158                             Error **errp);
1159 
1160 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1161  * block filters: Forward CONSISTENT_READ, WRITE, WRITE_UNCHANGED and RESIZE to
1162  * all children */
1163 void bdrv_filter_default_perms(BlockDriverState *bs, BdrvChild *c,
1164                                const BdrvChildRole *role,
1165                                BlockReopenQueue *reopen_queue,
1166                                uint64_t perm, uint64_t shared,
1167                                uint64_t *nperm, uint64_t *nshared);
1168 
1169 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1170  * (non-raw) image formats: Like above for bs->backing, but for bs->file it
1171  * requires WRITE | RESIZE for read-write images, always requires
1172  * CONSISTENT_READ and doesn't share WRITE. */
1173 void bdrv_format_default_perms(BlockDriverState *bs, BdrvChild *c,
1174                                const BdrvChildRole *role,
1175                                BlockReopenQueue *reopen_queue,
1176                                uint64_t perm, uint64_t shared,
1177                                uint64_t *nperm, uint64_t *nshared);
1178 
1179 /*
1180  * Default implementation for drivers to pass bdrv_co_block_status() to
1181  * their file.
1182  */
1183 int coroutine_fn bdrv_co_block_status_from_file(BlockDriverState *bs,
1184                                                 bool want_zero,
1185                                                 int64_t offset,
1186                                                 int64_t bytes,
1187                                                 int64_t *pnum,
1188                                                 int64_t *map,
1189                                                 BlockDriverState **file);
1190 /*
1191  * Default implementation for drivers to pass bdrv_co_block_status() to
1192  * their backing file.
1193  */
1194 int coroutine_fn bdrv_co_block_status_from_backing(BlockDriverState *bs,
1195                                                    bool want_zero,
1196                                                    int64_t offset,
1197                                                    int64_t bytes,
1198                                                    int64_t *pnum,
1199                                                    int64_t *map,
1200                                                    BlockDriverState **file);
1201 const char *bdrv_get_parent_name(const BlockDriverState *bs);
1202 void blk_dev_change_media_cb(BlockBackend *blk, bool load, Error **errp);
1203 bool blk_dev_has_removable_media(BlockBackend *blk);
1204 bool blk_dev_has_tray(BlockBackend *blk);
1205 void blk_dev_eject_request(BlockBackend *blk, bool force);
1206 bool blk_dev_is_tray_open(BlockBackend *blk);
1207 bool blk_dev_is_medium_locked(BlockBackend *blk);
1208 
1209 void bdrv_set_dirty(BlockDriverState *bs, int64_t offset, int64_t bytes);
1210 
1211 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap **out);
1212 void bdrv_restore_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap *backup);
1213 
1214 void bdrv_inc_in_flight(BlockDriverState *bs);
1215 void bdrv_dec_in_flight(BlockDriverState *bs);
1216 
1217 void blockdev_close_all_bdrv_states(void);
1218 
1219 int coroutine_fn bdrv_co_copy_range_from(BdrvChild *src, uint64_t src_offset,
1220                                          BdrvChild *dst, uint64_t dst_offset,
1221                                          uint64_t bytes,
1222                                          BdrvRequestFlags read_flags,
1223                                          BdrvRequestFlags write_flags);
1224 int coroutine_fn bdrv_co_copy_range_to(BdrvChild *src, uint64_t src_offset,
1225                                        BdrvChild *dst, uint64_t dst_offset,
1226                                        uint64_t bytes,
1227                                        BdrvRequestFlags read_flags,
1228                                        BdrvRequestFlags write_flags);
1229 
1230 int refresh_total_sectors(BlockDriverState *bs, int64_t hint);
1231 
1232 #endif /* BLOCK_INT_H */
1233