xref: /openbmc/qemu/include/block/block_int.h (revision 358f6348)
1 /*
2  * QEMU System Emulator block driver
3  *
4  * Copyright (c) 2003 Fabrice Bellard
5  *
6  * Permission is hereby granted, free of charge, to any person obtaining a copy
7  * of this software and associated documentation files (the "Software"), to deal
8  * in the Software without restriction, including without limitation the rights
9  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10  * copies of the Software, and to permit persons to whom the Software is
11  * furnished to do so, subject to the following conditions:
12  *
13  * The above copyright notice and this permission notice shall be included in
14  * all copies or substantial portions of the Software.
15  *
16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22  * THE SOFTWARE.
23  */
24 #ifndef BLOCK_INT_H
25 #define BLOCK_INT_H
26 
27 #include "block/accounting.h"
28 #include "block/block.h"
29 #include "block/aio-wait.h"
30 #include "qemu/queue.h"
31 #include "qemu/coroutine.h"
32 #include "qemu/stats64.h"
33 #include "qemu/timer.h"
34 #include "qemu/hbitmap.h"
35 #include "block/snapshot.h"
36 #include "qemu/throttle.h"
37 
38 #define BLOCK_FLAG_LAZY_REFCOUNTS   8
39 
40 #define BLOCK_OPT_SIZE              "size"
41 #define BLOCK_OPT_ENCRYPT           "encryption"
42 #define BLOCK_OPT_ENCRYPT_FORMAT    "encrypt.format"
43 #define BLOCK_OPT_COMPAT6           "compat6"
44 #define BLOCK_OPT_HWVERSION         "hwversion"
45 #define BLOCK_OPT_BACKING_FILE      "backing_file"
46 #define BLOCK_OPT_BACKING_FMT       "backing_fmt"
47 #define BLOCK_OPT_CLUSTER_SIZE      "cluster_size"
48 #define BLOCK_OPT_TABLE_SIZE        "table_size"
49 #define BLOCK_OPT_PREALLOC          "preallocation"
50 #define BLOCK_OPT_SUBFMT            "subformat"
51 #define BLOCK_OPT_COMPAT_LEVEL      "compat"
52 #define BLOCK_OPT_LAZY_REFCOUNTS    "lazy_refcounts"
53 #define BLOCK_OPT_ADAPTER_TYPE      "adapter_type"
54 #define BLOCK_OPT_REDUNDANCY        "redundancy"
55 #define BLOCK_OPT_NOCOW             "nocow"
56 #define BLOCK_OPT_OBJECT_SIZE       "object_size"
57 #define BLOCK_OPT_REFCOUNT_BITS     "refcount_bits"
58 #define BLOCK_OPT_DATA_FILE         "data_file"
59 #define BLOCK_OPT_DATA_FILE_RAW     "data_file_raw"
60 
61 #define BLOCK_PROBE_BUF_SIZE        512
62 
63 enum BdrvTrackedRequestType {
64     BDRV_TRACKED_READ,
65     BDRV_TRACKED_WRITE,
66     BDRV_TRACKED_DISCARD,
67     BDRV_TRACKED_TRUNCATE,
68 };
69 
70 typedef struct BdrvTrackedRequest {
71     BlockDriverState *bs;
72     int64_t offset;
73     uint64_t bytes;
74     enum BdrvTrackedRequestType type;
75 
76     bool serialising;
77     int64_t overlap_offset;
78     uint64_t overlap_bytes;
79 
80     QLIST_ENTRY(BdrvTrackedRequest) list;
81     Coroutine *co; /* owner, used for deadlock detection */
82     CoQueue wait_queue; /* coroutines blocked on this request */
83 
84     struct BdrvTrackedRequest *waiting_for;
85 } BdrvTrackedRequest;
86 
87 struct BlockDriver {
88     const char *format_name;
89     int instance_size;
90 
91     /* set to true if the BlockDriver is a block filter. Block filters pass
92      * certain callbacks that refer to data (see block.c) to their bs->file if
93      * the driver doesn't implement them. Drivers that do not wish to forward
94      * must implement them and return -ENOTSUP.
95      */
96     bool is_filter;
97     /* for snapshots block filter like Quorum can implement the
98      * following recursive callback.
99      * It's purpose is to recurse on the filter children while calling
100      * bdrv_recurse_is_first_non_filter on them.
101      * For a sample implementation look in the future Quorum block filter.
102      */
103     bool (*bdrv_recurse_is_first_non_filter)(BlockDriverState *bs,
104                                              BlockDriverState *candidate);
105 
106     int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename);
107     int (*bdrv_probe_device)(const char *filename);
108 
109     /* Any driver implementing this callback is expected to be able to handle
110      * NULL file names in its .bdrv_open() implementation */
111     void (*bdrv_parse_filename)(const char *filename, QDict *options, Error **errp);
112     /* Drivers not implementing bdrv_parse_filename nor bdrv_open should have
113      * this field set to true, except ones that are defined only by their
114      * child's bs.
115      * An example of the last type will be the quorum block driver.
116      */
117     bool bdrv_needs_filename;
118 
119     /* Set if a driver can support backing files */
120     bool supports_backing;
121 
122     /* For handling image reopen for split or non-split files */
123     int (*bdrv_reopen_prepare)(BDRVReopenState *reopen_state,
124                                BlockReopenQueue *queue, Error **errp);
125     void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state);
126     void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state);
127     void (*bdrv_join_options)(QDict *options, QDict *old_options);
128 
129     int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags,
130                      Error **errp);
131 
132     /* Protocol drivers should implement this instead of bdrv_open */
133     int (*bdrv_file_open)(BlockDriverState *bs, QDict *options, int flags,
134                           Error **errp);
135     void (*bdrv_close)(BlockDriverState *bs);
136     int coroutine_fn (*bdrv_co_create)(BlockdevCreateOptions *opts,
137                                        Error **errp);
138     int coroutine_fn (*bdrv_co_create_opts)(const char *filename,
139                                             QemuOpts *opts,
140                                             Error **errp);
141     int (*bdrv_make_empty)(BlockDriverState *bs);
142 
143     /*
144      * Refreshes the bs->exact_filename field. If that is impossible,
145      * bs->exact_filename has to be left empty.
146      */
147     void (*bdrv_refresh_filename)(BlockDriverState *bs);
148 
149     /*
150      * Gathers the open options for all children into @target.
151      * A simple format driver (without backing file support) might
152      * implement this function like this:
153      *
154      *     QINCREF(bs->file->bs->full_open_options);
155      *     qdict_put(target, "file", bs->file->bs->full_open_options);
156      *
157      * If not specified, the generic implementation will simply put
158      * all children's options under their respective name.
159      *
160      * @backing_overridden is true when bs->backing seems not to be
161      * the child that would result from opening bs->backing_file.
162      * Therefore, if it is true, the backing child's options should be
163      * gathered; otherwise, there is no need since the backing child
164      * is the one implied by the image header.
165      *
166      * Note that ideally this function would not be needed.  Every
167      * block driver which implements it is probably doing something
168      * shady regarding its runtime option structure.
169      */
170     void (*bdrv_gather_child_options)(BlockDriverState *bs, QDict *target,
171                                       bool backing_overridden);
172 
173     /*
174      * Returns an allocated string which is the directory name of this BDS: It
175      * will be used to make relative filenames absolute by prepending this
176      * function's return value to them.
177      */
178     char *(*bdrv_dirname)(BlockDriverState *bs, Error **errp);
179 
180     /* aio */
181     BlockAIOCB *(*bdrv_aio_preadv)(BlockDriverState *bs,
182         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags,
183         BlockCompletionFunc *cb, void *opaque);
184     BlockAIOCB *(*bdrv_aio_pwritev)(BlockDriverState *bs,
185         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags,
186         BlockCompletionFunc *cb, void *opaque);
187     BlockAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs,
188         BlockCompletionFunc *cb, void *opaque);
189     BlockAIOCB *(*bdrv_aio_pdiscard)(BlockDriverState *bs,
190         int64_t offset, int bytes,
191         BlockCompletionFunc *cb, void *opaque);
192 
193     int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs,
194         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
195 
196     /**
197      * @offset: position in bytes to read at
198      * @bytes: number of bytes to read
199      * @qiov: the buffers to fill with read data
200      * @flags: currently unused, always 0
201      *
202      * @offset and @bytes will be a multiple of 'request_alignment',
203      * but the length of individual @qiov elements does not have to
204      * be a multiple.
205      *
206      * @bytes will always equal the total size of @qiov, and will be
207      * no larger than 'max_transfer'.
208      *
209      * The buffer in @qiov may point directly to guest memory.
210      */
211     int coroutine_fn (*bdrv_co_preadv)(BlockDriverState *bs,
212         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags);
213     int coroutine_fn (*bdrv_co_preadv_part)(BlockDriverState *bs,
214         uint64_t offset, uint64_t bytes,
215         QEMUIOVector *qiov, size_t qiov_offset, int flags);
216     int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs,
217         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, int flags);
218     /**
219      * @offset: position in bytes to write at
220      * @bytes: number of bytes to write
221      * @qiov: the buffers containing data to write
222      * @flags: zero or more bits allowed by 'supported_write_flags'
223      *
224      * @offset and @bytes will be a multiple of 'request_alignment',
225      * but the length of individual @qiov elements does not have to
226      * be a multiple.
227      *
228      * @bytes will always equal the total size of @qiov, and will be
229      * no larger than 'max_transfer'.
230      *
231      * The buffer in @qiov may point directly to guest memory.
232      */
233     int coroutine_fn (*bdrv_co_pwritev)(BlockDriverState *bs,
234         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags);
235     int coroutine_fn (*bdrv_co_pwritev_part)(BlockDriverState *bs,
236         uint64_t offset, uint64_t bytes,
237         QEMUIOVector *qiov, size_t qiov_offset, int flags);
238 
239     /*
240      * Efficiently zero a region of the disk image.  Typically an image format
241      * would use a compact metadata representation to implement this.  This
242      * function pointer may be NULL or return -ENOSUP and .bdrv_co_writev()
243      * will be called instead.
244      */
245     int coroutine_fn (*bdrv_co_pwrite_zeroes)(BlockDriverState *bs,
246         int64_t offset, int bytes, BdrvRequestFlags flags);
247     int coroutine_fn (*bdrv_co_pdiscard)(BlockDriverState *bs,
248         int64_t offset, int bytes);
249 
250     /* Map [offset, offset + nbytes) range onto a child of @bs to copy from,
251      * and invoke bdrv_co_copy_range_from(child, ...), or invoke
252      * bdrv_co_copy_range_to() if @bs is the leaf child to copy data from.
253      *
254      * See the comment of bdrv_co_copy_range for the parameter and return value
255      * semantics.
256      */
257     int coroutine_fn (*bdrv_co_copy_range_from)(BlockDriverState *bs,
258                                                 BdrvChild *src,
259                                                 uint64_t offset,
260                                                 BdrvChild *dst,
261                                                 uint64_t dst_offset,
262                                                 uint64_t bytes,
263                                                 BdrvRequestFlags read_flags,
264                                                 BdrvRequestFlags write_flags);
265 
266     /* Map [offset, offset + nbytes) range onto a child of bs to copy data to,
267      * and invoke bdrv_co_copy_range_to(child, src, ...), or perform the copy
268      * operation if @bs is the leaf and @src has the same BlockDriver.  Return
269      * -ENOTSUP if @bs is the leaf but @src has a different BlockDriver.
270      *
271      * See the comment of bdrv_co_copy_range for the parameter and return value
272      * semantics.
273      */
274     int coroutine_fn (*bdrv_co_copy_range_to)(BlockDriverState *bs,
275                                               BdrvChild *src,
276                                               uint64_t src_offset,
277                                               BdrvChild *dst,
278                                               uint64_t dst_offset,
279                                               uint64_t bytes,
280                                               BdrvRequestFlags read_flags,
281                                               BdrvRequestFlags write_flags);
282 
283     /*
284      * Building block for bdrv_block_status[_above] and
285      * bdrv_is_allocated[_above].  The driver should answer only
286      * according to the current layer, and should only need to set
287      * BDRV_BLOCK_DATA, BDRV_BLOCK_ZERO, BDRV_BLOCK_OFFSET_VALID,
288      * and/or BDRV_BLOCK_RAW; if the current layer defers to a backing
289      * layer, the result should be 0 (and not BDRV_BLOCK_ZERO).  See
290      * block.h for the overall meaning of the bits.  As a hint, the
291      * flag want_zero is true if the caller cares more about precise
292      * mappings (favor accurate _OFFSET_VALID/_ZERO) or false for
293      * overall allocation (favor larger *pnum, perhaps by reporting
294      * _DATA instead of _ZERO).  The block layer guarantees input
295      * clamped to bdrv_getlength() and aligned to request_alignment,
296      * as well as non-NULL pnum, map, and file; in turn, the driver
297      * must return an error or set pnum to an aligned non-zero value.
298      */
299     int coroutine_fn (*bdrv_co_block_status)(BlockDriverState *bs,
300         bool want_zero, int64_t offset, int64_t bytes, int64_t *pnum,
301         int64_t *map, BlockDriverState **file);
302 
303     /*
304      * Invalidate any cached meta-data.
305      */
306     void coroutine_fn (*bdrv_co_invalidate_cache)(BlockDriverState *bs,
307                                                   Error **errp);
308     int (*bdrv_inactivate)(BlockDriverState *bs);
309 
310     /*
311      * Flushes all data for all layers by calling bdrv_co_flush for underlying
312      * layers, if needed. This function is needed for deterministic
313      * synchronization of the flush finishing callback.
314      */
315     int coroutine_fn (*bdrv_co_flush)(BlockDriverState *bs);
316 
317     /*
318      * Flushes all data that was already written to the OS all the way down to
319      * the disk (for example file-posix.c calls fsync()).
320      */
321     int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs);
322 
323     /*
324      * Flushes all internal caches to the OS. The data may still sit in a
325      * writeback cache of the host OS, but it will survive a crash of the qemu
326      * process.
327      */
328     int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs);
329 
330     /*
331      * Drivers setting this field must be able to work with just a plain
332      * filename with '<protocol_name>:' as a prefix, and no other options.
333      * Options may be extracted from the filename by implementing
334      * bdrv_parse_filename.
335      */
336     const char *protocol_name;
337     int coroutine_fn (*bdrv_co_truncate)(BlockDriverState *bs, int64_t offset,
338                                          PreallocMode prealloc, Error **errp);
339 
340     int64_t (*bdrv_getlength)(BlockDriverState *bs);
341     bool has_variable_length;
342     int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
343     BlockMeasureInfo *(*bdrv_measure)(QemuOpts *opts, BlockDriverState *in_bs,
344                                       Error **errp);
345 
346     int coroutine_fn (*bdrv_co_pwritev_compressed)(BlockDriverState *bs,
347         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov);
348     int coroutine_fn (*bdrv_co_pwritev_compressed_part)(BlockDriverState *bs,
349         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov,
350         size_t qiov_offset);
351 
352     int (*bdrv_snapshot_create)(BlockDriverState *bs,
353                                 QEMUSnapshotInfo *sn_info);
354     int (*bdrv_snapshot_goto)(BlockDriverState *bs,
355                               const char *snapshot_id);
356     int (*bdrv_snapshot_delete)(BlockDriverState *bs,
357                                 const char *snapshot_id,
358                                 const char *name,
359                                 Error **errp);
360     int (*bdrv_snapshot_list)(BlockDriverState *bs,
361                               QEMUSnapshotInfo **psn_info);
362     int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
363                                   const char *snapshot_id,
364                                   const char *name,
365                                   Error **errp);
366     int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
367     ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs,
368                                                  Error **errp);
369 
370     int coroutine_fn (*bdrv_save_vmstate)(BlockDriverState *bs,
371                                           QEMUIOVector *qiov,
372                                           int64_t pos);
373     int coroutine_fn (*bdrv_load_vmstate)(BlockDriverState *bs,
374                                           QEMUIOVector *qiov,
375                                           int64_t pos);
376 
377     int (*bdrv_change_backing_file)(BlockDriverState *bs,
378         const char *backing_file, const char *backing_fmt);
379 
380     /* removable device specific */
381     bool (*bdrv_is_inserted)(BlockDriverState *bs);
382     void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
383     void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
384 
385     /* to control generic scsi devices */
386     BlockAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
387         unsigned long int req, void *buf,
388         BlockCompletionFunc *cb, void *opaque);
389     int coroutine_fn (*bdrv_co_ioctl)(BlockDriverState *bs,
390                                       unsigned long int req, void *buf);
391 
392     /* List of options for creating images, terminated by name == NULL */
393     QemuOptsList *create_opts;
394     /*
395      * If this driver supports reopening images this contains a
396      * NULL-terminated list of the runtime options that can be
397      * modified. If an option in this list is unspecified during
398      * reopen then it _must_ be reset to its default value or return
399      * an error.
400      */
401     const char *const *mutable_opts;
402 
403     /*
404      * Returns 0 for completed check, -errno for internal errors.
405      * The check results are stored in result.
406      */
407     int coroutine_fn (*bdrv_co_check)(BlockDriverState *bs,
408                                       BdrvCheckResult *result,
409                                       BdrvCheckMode fix);
410 
411     int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts,
412                               BlockDriverAmendStatusCB *status_cb,
413                               void *cb_opaque,
414                               Error **errp);
415 
416     void (*bdrv_debug_event)(BlockDriverState *bs, BlkdebugEvent event);
417 
418     /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
419     int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
420         const char *tag);
421     int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs,
422         const char *tag);
423     int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
424     bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
425 
426     void (*bdrv_refresh_limits)(BlockDriverState *bs, Error **errp);
427 
428     /*
429      * Returns 1 if newly created images are guaranteed to contain only
430      * zeros, 0 otherwise.
431      * Must return 0 if .bdrv_has_zero_init_truncate() returns 0.
432      */
433     int (*bdrv_has_zero_init)(BlockDriverState *bs);
434 
435     /*
436      * Returns 1 if new areas added by growing the image with
437      * PREALLOC_MODE_OFF contain only zeros, 0 otherwise.
438      */
439     int (*bdrv_has_zero_init_truncate)(BlockDriverState *bs);
440 
441     /* Remove fd handlers, timers, and other event loop callbacks so the event
442      * loop is no longer in use.  Called with no in-flight requests and in
443      * depth-first traversal order with parents before child nodes.
444      */
445     void (*bdrv_detach_aio_context)(BlockDriverState *bs);
446 
447     /* Add fd handlers, timers, and other event loop callbacks so I/O requests
448      * can be processed again.  Called with no in-flight requests and in
449      * depth-first traversal order with child nodes before parent nodes.
450      */
451     void (*bdrv_attach_aio_context)(BlockDriverState *bs,
452                                     AioContext *new_context);
453 
454     /* io queue for linux-aio */
455     void (*bdrv_io_plug)(BlockDriverState *bs);
456     void (*bdrv_io_unplug)(BlockDriverState *bs);
457 
458     /**
459      * Try to get @bs's logical and physical block size.
460      * On success, store them in @bsz and return zero.
461      * On failure, return negative errno.
462      */
463     int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz);
464     /**
465      * Try to get @bs's geometry (cyls, heads, sectors)
466      * On success, store them in @geo and return 0.
467      * On failure return -errno.
468      * Only drivers that want to override guest geometry implement this
469      * callback; see hd_geometry_guess().
470      */
471     int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo);
472 
473     /**
474      * bdrv_co_drain_begin is called if implemented in the beginning of a
475      * drain operation to drain and stop any internal sources of requests in
476      * the driver.
477      * bdrv_co_drain_end is called if implemented at the end of the drain.
478      *
479      * They should be used by the driver to e.g. manage scheduled I/O
480      * requests, or toggle an internal state. After the end of the drain new
481      * requests will continue normally.
482      */
483     void coroutine_fn (*bdrv_co_drain_begin)(BlockDriverState *bs);
484     void coroutine_fn (*bdrv_co_drain_end)(BlockDriverState *bs);
485 
486     void (*bdrv_add_child)(BlockDriverState *parent, BlockDriverState *child,
487                            Error **errp);
488     void (*bdrv_del_child)(BlockDriverState *parent, BdrvChild *child,
489                            Error **errp);
490 
491     /**
492      * Informs the block driver that a permission change is intended. The
493      * driver checks whether the change is permissible and may take other
494      * preparations for the change (e.g. get file system locks). This operation
495      * is always followed either by a call to either .bdrv_set_perm or
496      * .bdrv_abort_perm_update.
497      *
498      * Checks whether the requested set of cumulative permissions in @perm
499      * can be granted for accessing @bs and whether no other users are using
500      * permissions other than those given in @shared (both arguments take
501      * BLK_PERM_* bitmasks).
502      *
503      * If both conditions are met, 0 is returned. Otherwise, -errno is returned
504      * and errp is set to an error describing the conflict.
505      */
506     int (*bdrv_check_perm)(BlockDriverState *bs, uint64_t perm,
507                            uint64_t shared, Error **errp);
508 
509     /**
510      * Called to inform the driver that the set of cumulative set of used
511      * permissions for @bs has changed to @perm, and the set of sharable
512      * permission to @shared. The driver can use this to propagate changes to
513      * its children (i.e. request permissions only if a parent actually needs
514      * them).
515      *
516      * This function is only invoked after bdrv_check_perm(), so block drivers
517      * may rely on preparations made in their .bdrv_check_perm implementation.
518      */
519     void (*bdrv_set_perm)(BlockDriverState *bs, uint64_t perm, uint64_t shared);
520 
521     /*
522      * Called to inform the driver that after a previous bdrv_check_perm()
523      * call, the permission update is not performed and any preparations made
524      * for it (e.g. taken file locks) need to be undone.
525      *
526      * This function can be called even for nodes that never saw a
527      * bdrv_check_perm() call. It is a no-op then.
528      */
529     void (*bdrv_abort_perm_update)(BlockDriverState *bs);
530 
531     /**
532      * Returns in @nperm and @nshared the permissions that the driver for @bs
533      * needs on its child @c, based on the cumulative permissions requested by
534      * the parents in @parent_perm and @parent_shared.
535      *
536      * If @c is NULL, return the permissions for attaching a new child for the
537      * given @role.
538      *
539      * If @reopen_queue is non-NULL, don't return the currently needed
540      * permissions, but those that will be needed after applying the
541      * @reopen_queue.
542      */
543      void (*bdrv_child_perm)(BlockDriverState *bs, BdrvChild *c,
544                              const BdrvChildRole *role,
545                              BlockReopenQueue *reopen_queue,
546                              uint64_t parent_perm, uint64_t parent_shared,
547                              uint64_t *nperm, uint64_t *nshared);
548 
549     /**
550      * Bitmaps should be marked as 'IN_USE' in the image on reopening image
551      * as rw. This handler should realize it. It also should unset readonly
552      * field of BlockDirtyBitmap's in case of success.
553      */
554     int (*bdrv_reopen_bitmaps_rw)(BlockDriverState *bs, Error **errp);
555     bool (*bdrv_can_store_new_dirty_bitmap)(BlockDriverState *bs,
556                                             const char *name,
557                                             uint32_t granularity,
558                                             Error **errp);
559     void (*bdrv_remove_persistent_dirty_bitmap)(BlockDriverState *bs,
560                                                 const char *name,
561                                                 Error **errp);
562 
563     /**
564      * Register/unregister a buffer for I/O. For example, when the driver is
565      * interested to know the memory areas that will later be used in iovs, so
566      * that it can do IOMMU mapping with VFIO etc., in order to get better
567      * performance. In the case of VFIO drivers, this callback is used to do
568      * DMA mapping for hot buffers.
569      */
570     void (*bdrv_register_buf)(BlockDriverState *bs, void *host, size_t size);
571     void (*bdrv_unregister_buf)(BlockDriverState *bs, void *host);
572     QLIST_ENTRY(BlockDriver) list;
573 
574     /* Pointer to a NULL-terminated array of names of strong options
575      * that can be specified for bdrv_open(). A strong option is one
576      * that changes the data of a BDS.
577      * If this pointer is NULL, the array is considered empty.
578      * "filename" and "driver" are always considered strong. */
579     const char *const *strong_runtime_opts;
580 };
581 
582 static inline bool block_driver_can_compress(BlockDriver *drv)
583 {
584     return drv->bdrv_co_pwritev_compressed ||
585            drv->bdrv_co_pwritev_compressed_part;
586 }
587 
588 typedef struct BlockLimits {
589     /* Alignment requirement, in bytes, for offset/length of I/O
590      * requests. Must be a power of 2 less than INT_MAX; defaults to
591      * 1 for drivers with modern byte interfaces, and to 512
592      * otherwise. */
593     uint32_t request_alignment;
594 
595     /* Maximum number of bytes that can be discarded at once (since it
596      * is signed, it must be < 2G, if set). Must be multiple of
597      * pdiscard_alignment, but need not be power of 2. May be 0 if no
598      * inherent 32-bit limit */
599     int32_t max_pdiscard;
600 
601     /* Optimal alignment for discard requests in bytes. A power of 2
602      * is best but not mandatory.  Must be a multiple of
603      * bl.request_alignment, and must be less than max_pdiscard if
604      * that is set. May be 0 if bl.request_alignment is good enough */
605     uint32_t pdiscard_alignment;
606 
607     /* Maximum number of bytes that can zeroized at once (since it is
608      * signed, it must be < 2G, if set). Must be multiple of
609      * pwrite_zeroes_alignment. May be 0 if no inherent 32-bit limit */
610     int32_t max_pwrite_zeroes;
611 
612     /* Optimal alignment for write zeroes requests in bytes. A power
613      * of 2 is best but not mandatory.  Must be a multiple of
614      * bl.request_alignment, and must be less than max_pwrite_zeroes
615      * if that is set. May be 0 if bl.request_alignment is good
616      * enough */
617     uint32_t pwrite_zeroes_alignment;
618 
619     /* Optimal transfer length in bytes.  A power of 2 is best but not
620      * mandatory.  Must be a multiple of bl.request_alignment, or 0 if
621      * no preferred size */
622     uint32_t opt_transfer;
623 
624     /* Maximal transfer length in bytes.  Need not be power of 2, but
625      * must be multiple of opt_transfer and bl.request_alignment, or 0
626      * for no 32-bit limit.  For now, anything larger than INT_MAX is
627      * clamped down. */
628     uint32_t max_transfer;
629 
630     /* memory alignment, in bytes so that no bounce buffer is needed */
631     size_t min_mem_alignment;
632 
633     /* memory alignment, in bytes, for bounce buffer */
634     size_t opt_mem_alignment;
635 
636     /* maximum number of iovec elements */
637     int max_iov;
638 } BlockLimits;
639 
640 typedef struct BdrvOpBlocker BdrvOpBlocker;
641 
642 typedef struct BdrvAioNotifier {
643     void (*attached_aio_context)(AioContext *new_context, void *opaque);
644     void (*detach_aio_context)(void *opaque);
645 
646     void *opaque;
647     bool deleted;
648 
649     QLIST_ENTRY(BdrvAioNotifier) list;
650 } BdrvAioNotifier;
651 
652 struct BdrvChildRole {
653     /* If true, bdrv_replace_node() doesn't change the node this BdrvChild
654      * points to. */
655     bool stay_at_node;
656 
657     /* If true, the parent is a BlockDriverState and bdrv_next_all_states()
658      * will return it. This information is used for drain_all, where every node
659      * will be drained separately, so the drain only needs to be propagated to
660      * non-BDS parents. */
661     bool parent_is_bds;
662 
663     void (*inherit_options)(int *child_flags, QDict *child_options,
664                             int parent_flags, QDict *parent_options);
665 
666     void (*change_media)(BdrvChild *child, bool load);
667     void (*resize)(BdrvChild *child);
668 
669     /* Returns a name that is supposedly more useful for human users than the
670      * node name for identifying the node in question (in particular, a BB
671      * name), or NULL if the parent can't provide a better name. */
672     const char *(*get_name)(BdrvChild *child);
673 
674     /* Returns a malloced string that describes the parent of the child for a
675      * human reader. This could be a node-name, BlockBackend name, qdev ID or
676      * QOM path of the device owning the BlockBackend, job type and ID etc. The
677      * caller is responsible for freeing the memory. */
678     char *(*get_parent_desc)(BdrvChild *child);
679 
680     /*
681      * If this pair of functions is implemented, the parent doesn't issue new
682      * requests after returning from .drained_begin() until .drained_end() is
683      * called.
684      *
685      * These functions must not change the graph (and therefore also must not
686      * call aio_poll(), which could change the graph indirectly).
687      *
688      * If drained_end() schedules background operations, it must atomically
689      * increment *drained_end_counter for each such operation and atomically
690      * decrement it once the operation has settled.
691      *
692      * Note that this can be nested. If drained_begin() was called twice, new
693      * I/O is allowed only after drained_end() was called twice, too.
694      */
695     void (*drained_begin)(BdrvChild *child);
696     void (*drained_end)(BdrvChild *child, int *drained_end_counter);
697 
698     /*
699      * Returns whether the parent has pending requests for the child. This
700      * callback is polled after .drained_begin() has been called until all
701      * activity on the child has stopped.
702      */
703     bool (*drained_poll)(BdrvChild *child);
704 
705     /* Notifies the parent that the child has been activated/inactivated (e.g.
706      * when migration is completing) and it can start/stop requesting
707      * permissions and doing I/O on it. */
708     void (*activate)(BdrvChild *child, Error **errp);
709     int (*inactivate)(BdrvChild *child);
710 
711     void (*attach)(BdrvChild *child);
712     void (*detach)(BdrvChild *child);
713 
714     /* Notifies the parent that the filename of its child has changed (e.g.
715      * because the direct child was removed from the backing chain), so that it
716      * can update its reference. */
717     int (*update_filename)(BdrvChild *child, BlockDriverState *new_base,
718                            const char *filename, Error **errp);
719 
720     bool (*can_set_aio_ctx)(BdrvChild *child, AioContext *ctx,
721                             GSList **ignore, Error **errp);
722     void (*set_aio_ctx)(BdrvChild *child, AioContext *ctx, GSList **ignore);
723 };
724 
725 extern const BdrvChildRole child_file;
726 extern const BdrvChildRole child_format;
727 extern const BdrvChildRole child_backing;
728 
729 struct BdrvChild {
730     BlockDriverState *bs;
731     char *name;
732     const BdrvChildRole *role;
733     void *opaque;
734 
735     /**
736      * Granted permissions for operating on this BdrvChild (BLK_PERM_* bitmask)
737      */
738     uint64_t perm;
739 
740     /**
741      * Permissions that can still be granted to other users of @bs while this
742      * BdrvChild is still attached to it. (BLK_PERM_* bitmask)
743      */
744     uint64_t shared_perm;
745 
746     /* backup of permissions during permission update procedure */
747     bool has_backup_perm;
748     uint64_t backup_perm;
749     uint64_t backup_shared_perm;
750 
751     /*
752      * This link is frozen: the child can neither be replaced nor
753      * detached from the parent.
754      */
755     bool frozen;
756 
757     /*
758      * How many times the parent of this child has been drained
759      * (through role->drained_*).
760      * Usually, this is equal to bs->quiesce_counter (potentially
761      * reduced by bdrv_drain_all_count).  It may differ while the
762      * child is entering or leaving a drained section.
763      */
764     int parent_quiesce_counter;
765 
766     QLIST_ENTRY(BdrvChild) next;
767     QLIST_ENTRY(BdrvChild) next_parent;
768 };
769 
770 /*
771  * Note: the function bdrv_append() copies and swaps contents of
772  * BlockDriverStates, so if you add new fields to this struct, please
773  * inspect bdrv_append() to determine if the new fields need to be
774  * copied as well.
775  */
776 struct BlockDriverState {
777     /* Protected by big QEMU lock or read-only after opening.  No special
778      * locking needed during I/O...
779      */
780     int open_flags; /* flags used to open the file, re-used for re-open */
781     bool read_only; /* if true, the media is read only */
782     bool encrypted; /* if true, the media is encrypted */
783     bool sg;        /* if true, the device is a /dev/sg* */
784     bool probed;    /* if true, format was probed rather than specified */
785     bool force_share; /* if true, always allow all shared permissions */
786     bool implicit;  /* if true, this filter node was automatically inserted */
787 
788     BlockDriver *drv; /* NULL means no media */
789     void *opaque;
790 
791     AioContext *aio_context; /* event loop used for fd handlers, timers, etc */
792     /* long-running tasks intended to always use the same AioContext as this
793      * BDS may register themselves in this list to be notified of changes
794      * regarding this BDS's context */
795     QLIST_HEAD(, BdrvAioNotifier) aio_notifiers;
796     bool walking_aio_notifiers; /* to make removal during iteration safe */
797 
798     char filename[PATH_MAX];
799     char backing_file[PATH_MAX]; /* if non zero, the image is a diff of
800                                     this file image */
801     /* The backing filename indicated by the image header; if we ever
802      * open this file, then this is replaced by the resulting BDS's
803      * filename (i.e. after a bdrv_refresh_filename() run). */
804     char auto_backing_file[PATH_MAX];
805     char backing_format[16]; /* if non-zero and backing_file exists */
806 
807     QDict *full_open_options;
808     char exact_filename[PATH_MAX];
809 
810     BdrvChild *backing;
811     BdrvChild *file;
812 
813     /* I/O Limits */
814     BlockLimits bl;
815 
816     /* Flags honored during pwrite (so far: BDRV_REQ_FUA,
817      * BDRV_REQ_WRITE_UNCHANGED).
818      * If a driver does not support BDRV_REQ_WRITE_UNCHANGED, those
819      * writes will be issued as normal writes without the flag set.
820      * This is important to note for drivers that do not explicitly
821      * request a WRITE permission for their children and instead take
822      * the same permissions as their parent did (this is commonly what
823      * block filters do).  Such drivers have to be aware that the
824      * parent may have taken a WRITE_UNCHANGED permission only and is
825      * issuing such requests.  Drivers either must make sure that
826      * these requests do not result in plain WRITE accesses (usually
827      * by supporting BDRV_REQ_WRITE_UNCHANGED, and then forwarding
828      * every incoming write request as-is, including potentially that
829      * flag), or they have to explicitly take the WRITE permission for
830      * their children. */
831     unsigned int supported_write_flags;
832     /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA,
833      * BDRV_REQ_MAY_UNMAP, BDRV_REQ_WRITE_UNCHANGED) */
834     unsigned int supported_zero_flags;
835 
836     /* the following member gives a name to every node on the bs graph. */
837     char node_name[32];
838     /* element of the list of named nodes building the graph */
839     QTAILQ_ENTRY(BlockDriverState) node_list;
840     /* element of the list of all BlockDriverStates (all_bdrv_states) */
841     QTAILQ_ENTRY(BlockDriverState) bs_list;
842     /* element of the list of monitor-owned BDS */
843     QTAILQ_ENTRY(BlockDriverState) monitor_list;
844     int refcnt;
845 
846     /* operation blockers */
847     QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX];
848 
849     /* The node that this node inherited default options from (and a reopen on
850      * which can affect this node by changing these defaults). This is always a
851      * parent node of this node. */
852     BlockDriverState *inherits_from;
853     QLIST_HEAD(, BdrvChild) children;
854     QLIST_HEAD(, BdrvChild) parents;
855 
856     QDict *options;
857     QDict *explicit_options;
858     BlockdevDetectZeroesOptions detect_zeroes;
859 
860     /* The error object in use for blocking operations on backing_hd */
861     Error *backing_blocker;
862 
863     /* Protected by AioContext lock */
864 
865     /* If we are reading a disk image, give its size in sectors.
866      * Generally read-only; it is written to by load_snapshot and
867      * save_snaphost, but the block layer is quiescent during those.
868      */
869     int64_t total_sectors;
870 
871     /* Callback before write request is processed */
872     NotifierWithReturnList before_write_notifiers;
873 
874     /* threshold limit for writes, in bytes. "High water mark". */
875     uint64_t write_threshold_offset;
876     NotifierWithReturn write_threshold_notifier;
877 
878     /* Writing to the list requires the BQL _and_ the dirty_bitmap_mutex.
879      * Reading from the list can be done with either the BQL or the
880      * dirty_bitmap_mutex.  Modifying a bitmap only requires
881      * dirty_bitmap_mutex.  */
882     QemuMutex dirty_bitmap_mutex;
883     QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps;
884 
885     /* Offset after the highest byte written to */
886     Stat64 wr_highest_offset;
887 
888     /* If true, copy read backing sectors into image.  Can be >1 if more
889      * than one client has requested copy-on-read.  Accessed with atomic
890      * ops.
891      */
892     int copy_on_read;
893 
894     /* number of in-flight requests; overall and serialising.
895      * Accessed with atomic ops.
896      */
897     unsigned int in_flight;
898     unsigned int serialising_in_flight;
899 
900     /* counter for nested bdrv_io_plug.
901      * Accessed with atomic ops.
902     */
903     unsigned io_plugged;
904 
905     /* do we need to tell the quest if we have a volatile write cache? */
906     int enable_write_cache;
907 
908     /* Accessed with atomic ops.  */
909     int quiesce_counter;
910     int recursive_quiesce_counter;
911 
912     unsigned int write_gen;               /* Current data generation */
913 
914     /* Protected by reqs_lock.  */
915     CoMutex reqs_lock;
916     QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
917     CoQueue flush_queue;                  /* Serializing flush queue */
918     bool active_flush_req;                /* Flush request in flight? */
919 
920     /* Only read/written by whoever has set active_flush_req to true.  */
921     unsigned int flushed_gen;             /* Flushed write generation */
922 
923     /* BdrvChild links to this node may never be frozen */
924     bool never_freeze;
925 };
926 
927 struct BlockBackendRootState {
928     int open_flags;
929     bool read_only;
930     BlockdevDetectZeroesOptions detect_zeroes;
931 };
932 
933 typedef enum BlockMirrorBackingMode {
934     /* Reuse the existing backing chain from the source for the target.
935      * - sync=full: Set backing BDS to NULL.
936      * - sync=top:  Use source's backing BDS.
937      * - sync=none: Use source as the backing BDS. */
938     MIRROR_SOURCE_BACKING_CHAIN,
939 
940     /* Open the target's backing chain completely anew */
941     MIRROR_OPEN_BACKING_CHAIN,
942 
943     /* Do not change the target's backing BDS after job completion */
944     MIRROR_LEAVE_BACKING_CHAIN,
945 } BlockMirrorBackingMode;
946 
947 static inline BlockDriverState *backing_bs(BlockDriverState *bs)
948 {
949     return bs->backing ? bs->backing->bs : NULL;
950 }
951 
952 
953 /* Essential block drivers which must always be statically linked into qemu, and
954  * which therefore can be accessed without using bdrv_find_format() */
955 extern BlockDriver bdrv_file;
956 extern BlockDriver bdrv_raw;
957 extern BlockDriver bdrv_qcow2;
958 
959 int coroutine_fn bdrv_co_preadv(BdrvChild *child,
960     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
961     BdrvRequestFlags flags);
962 int coroutine_fn bdrv_co_preadv_part(BdrvChild *child,
963     int64_t offset, unsigned int bytes,
964     QEMUIOVector *qiov, size_t qiov_offset, BdrvRequestFlags flags);
965 int coroutine_fn bdrv_co_pwritev(BdrvChild *child,
966     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
967     BdrvRequestFlags flags);
968 int coroutine_fn bdrv_co_pwritev_part(BdrvChild *child,
969     int64_t offset, unsigned int bytes,
970     QEMUIOVector *qiov, size_t qiov_offset, BdrvRequestFlags flags);
971 
972 static inline int coroutine_fn bdrv_co_pread(BdrvChild *child,
973     int64_t offset, unsigned int bytes, void *buf, BdrvRequestFlags flags)
974 {
975     QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes);
976 
977     return bdrv_co_preadv(child, offset, bytes, &qiov, flags);
978 }
979 
980 static inline int coroutine_fn bdrv_co_pwrite(BdrvChild *child,
981     int64_t offset, unsigned int bytes, void *buf, BdrvRequestFlags flags)
982 {
983     QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes);
984 
985     return bdrv_co_pwritev(child, offset, bytes, &qiov, flags);
986 }
987 
988 extern unsigned int bdrv_drain_all_count;
989 void bdrv_apply_subtree_drain(BdrvChild *child, BlockDriverState *new_parent);
990 void bdrv_unapply_subtree_drain(BdrvChild *child, BlockDriverState *old_parent);
991 
992 int get_tmp_filename(char *filename, int size);
993 BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size,
994                             const char *filename);
995 
996 void bdrv_parse_filename_strip_prefix(const char *filename, const char *prefix,
997                                       QDict *options);
998 
999 
1000 /**
1001  * bdrv_add_before_write_notifier:
1002  *
1003  * Register a callback that is invoked before write requests are processed but
1004  * after any throttling or waiting for overlapping requests.
1005  */
1006 void bdrv_add_before_write_notifier(BlockDriverState *bs,
1007                                     NotifierWithReturn *notifier);
1008 
1009 /**
1010  * bdrv_add_aio_context_notifier:
1011  *
1012  * If a long-running job intends to be always run in the same AioContext as a
1013  * certain BDS, it may use this function to be notified of changes regarding the
1014  * association of the BDS to an AioContext.
1015  *
1016  * attached_aio_context() is called after the target BDS has been attached to a
1017  * new AioContext; detach_aio_context() is called before the target BDS is being
1018  * detached from its old AioContext.
1019  */
1020 void bdrv_add_aio_context_notifier(BlockDriverState *bs,
1021         void (*attached_aio_context)(AioContext *new_context, void *opaque),
1022         void (*detach_aio_context)(void *opaque), void *opaque);
1023 
1024 /**
1025  * bdrv_remove_aio_context_notifier:
1026  *
1027  * Unsubscribe of change notifications regarding the BDS's AioContext. The
1028  * parameters given here have to be the same as those given to
1029  * bdrv_add_aio_context_notifier().
1030  */
1031 void bdrv_remove_aio_context_notifier(BlockDriverState *bs,
1032                                       void (*aio_context_attached)(AioContext *,
1033                                                                    void *),
1034                                       void (*aio_context_detached)(void *),
1035                                       void *opaque);
1036 
1037 /**
1038  * bdrv_wakeup:
1039  * @bs: The BlockDriverState for which an I/O operation has been completed.
1040  *
1041  * Wake up the main thread if it is waiting on BDRV_POLL_WHILE.  During
1042  * synchronous I/O on a BlockDriverState that is attached to another
1043  * I/O thread, the main thread lets the I/O thread's event loop run,
1044  * waiting for the I/O operation to complete.  A bdrv_wakeup will wake
1045  * up the main thread if necessary.
1046  *
1047  * Manual calls to bdrv_wakeup are rarely necessary, because
1048  * bdrv_dec_in_flight already calls it.
1049  */
1050 void bdrv_wakeup(BlockDriverState *bs);
1051 
1052 #ifdef _WIN32
1053 int is_windows_drive(const char *filename);
1054 #endif
1055 
1056 /**
1057  * stream_start:
1058  * @job_id: The id of the newly-created job, or %NULL to use the
1059  * device name of @bs.
1060  * @bs: Block device to operate on.
1061  * @base: Block device that will become the new base, or %NULL to
1062  * flatten the whole backing file chain onto @bs.
1063  * @backing_file_str: The file name that will be written to @bs as the
1064  * the new backing file if the job completes. Ignored if @base is %NULL.
1065  * @creation_flags: Flags that control the behavior of the Job lifetime.
1066  *                  See @BlockJobCreateFlags
1067  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1068  * @on_error: The action to take upon error.
1069  * @errp: Error object.
1070  *
1071  * Start a streaming operation on @bs.  Clusters that are unallocated
1072  * in @bs, but allocated in any image between @base and @bs (both
1073  * exclusive) will be written to @bs.  At the end of a successful
1074  * streaming job, the backing file of @bs will be changed to
1075  * @backing_file_str in the written image and to @base in the live
1076  * BlockDriverState.
1077  */
1078 void stream_start(const char *job_id, BlockDriverState *bs,
1079                   BlockDriverState *base, const char *backing_file_str,
1080                   int creation_flags, int64_t speed,
1081                   BlockdevOnError on_error, Error **errp);
1082 
1083 /**
1084  * commit_start:
1085  * @job_id: The id of the newly-created job, or %NULL to use the
1086  * device name of @bs.
1087  * @bs: Active block device.
1088  * @top: Top block device to be committed.
1089  * @base: Block device that will be written into, and become the new top.
1090  * @creation_flags: Flags that control the behavior of the Job lifetime.
1091  *                  See @BlockJobCreateFlags
1092  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1093  * @on_error: The action to take upon error.
1094  * @backing_file_str: String to use as the backing file in @top's overlay
1095  * @filter_node_name: The node name that should be assigned to the filter
1096  * driver that the commit job inserts into the graph above @top. NULL means
1097  * that a node name should be autogenerated.
1098  * @errp: Error object.
1099  *
1100  */
1101 void commit_start(const char *job_id, BlockDriverState *bs,
1102                   BlockDriverState *base, BlockDriverState *top,
1103                   int creation_flags, int64_t speed,
1104                   BlockdevOnError on_error, const char *backing_file_str,
1105                   const char *filter_node_name, Error **errp);
1106 /**
1107  * commit_active_start:
1108  * @job_id: The id of the newly-created job, or %NULL to use the
1109  * device name of @bs.
1110  * @bs: Active block device to be committed.
1111  * @base: Block device that will be written into, and become the new top.
1112  * @creation_flags: Flags that control the behavior of the Job lifetime.
1113  *                  See @BlockJobCreateFlags
1114  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1115  * @on_error: The action to take upon error.
1116  * @filter_node_name: The node name that should be assigned to the filter
1117  * driver that the commit job inserts into the graph above @bs. NULL means that
1118  * a node name should be autogenerated.
1119  * @cb: Completion function for the job.
1120  * @opaque: Opaque pointer value passed to @cb.
1121  * @auto_complete: Auto complete the job.
1122  * @errp: Error object.
1123  *
1124  */
1125 BlockJob *commit_active_start(const char *job_id, BlockDriverState *bs,
1126                               BlockDriverState *base, int creation_flags,
1127                               int64_t speed, BlockdevOnError on_error,
1128                               const char *filter_node_name,
1129                               BlockCompletionFunc *cb, void *opaque,
1130                               bool auto_complete, Error **errp);
1131 /*
1132  * mirror_start:
1133  * @job_id: The id of the newly-created job, or %NULL to use the
1134  * device name of @bs.
1135  * @bs: Block device to operate on.
1136  * @target: Block device to write to.
1137  * @replaces: Block graph node name to replace once the mirror is done. Can
1138  *            only be used when full mirroring is selected.
1139  * @creation_flags: Flags that control the behavior of the Job lifetime.
1140  *                  See @BlockJobCreateFlags
1141  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1142  * @granularity: The chosen granularity for the dirty bitmap.
1143  * @buf_size: The amount of data that can be in flight at one time.
1144  * @mode: Whether to collapse all images in the chain to the target.
1145  * @backing_mode: How to establish the target's backing chain after completion.
1146  * @zero_target: Whether the target should be explicitly zero-initialized
1147  * @on_source_error: The action to take upon error reading from the source.
1148  * @on_target_error: The action to take upon error writing to the target.
1149  * @unmap: Whether to unmap target where source sectors only contain zeroes.
1150  * @filter_node_name: The node name that should be assigned to the filter
1151  * driver that the mirror job inserts into the graph above @bs. NULL means that
1152  * a node name should be autogenerated.
1153  * @copy_mode: When to trigger writes to the target.
1154  * @errp: Error object.
1155  *
1156  * Start a mirroring operation on @bs.  Clusters that are allocated
1157  * in @bs will be written to @target until the job is cancelled or
1158  * manually completed.  At the end of a successful mirroring job,
1159  * @bs will be switched to read from @target.
1160  */
1161 void mirror_start(const char *job_id, BlockDriverState *bs,
1162                   BlockDriverState *target, const char *replaces,
1163                   int creation_flags, int64_t speed,
1164                   uint32_t granularity, int64_t buf_size,
1165                   MirrorSyncMode mode, BlockMirrorBackingMode backing_mode,
1166                   bool zero_target,
1167                   BlockdevOnError on_source_error,
1168                   BlockdevOnError on_target_error,
1169                   bool unmap, const char *filter_node_name,
1170                   MirrorCopyMode copy_mode, Error **errp);
1171 
1172 /*
1173  * backup_job_create:
1174  * @job_id: The id of the newly-created job, or %NULL to use the
1175  * device name of @bs.
1176  * @bs: Block device to operate on.
1177  * @target: Block device to write to.
1178  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1179  * @sync_mode: What parts of the disk image should be copied to the destination.
1180  * @sync_bitmap: The dirty bitmap if sync_mode is 'bitmap' or 'incremental'
1181  * @bitmap_mode: The bitmap synchronization policy to use.
1182  * @on_source_error: The action to take upon error reading from the source.
1183  * @on_target_error: The action to take upon error writing to the target.
1184  * @creation_flags: Flags that control the behavior of the Job lifetime.
1185  *                  See @BlockJobCreateFlags
1186  * @cb: Completion function for the job.
1187  * @opaque: Opaque pointer value passed to @cb.
1188  * @txn: Transaction that this job is part of (may be NULL).
1189  *
1190  * Create a backup operation on @bs.  Clusters in @bs are written to @target
1191  * until the job is cancelled or manually completed.
1192  */
1193 BlockJob *backup_job_create(const char *job_id, BlockDriverState *bs,
1194                             BlockDriverState *target, int64_t speed,
1195                             MirrorSyncMode sync_mode,
1196                             BdrvDirtyBitmap *sync_bitmap,
1197                             BitmapSyncMode bitmap_mode,
1198                             bool compress,
1199                             BlockdevOnError on_source_error,
1200                             BlockdevOnError on_target_error,
1201                             int creation_flags,
1202                             BlockCompletionFunc *cb, void *opaque,
1203                             JobTxn *txn, Error **errp);
1204 
1205 void hmp_drive_add_node(Monitor *mon, const char *optstr);
1206 
1207 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
1208                                   const char *child_name,
1209                                   const BdrvChildRole *child_role,
1210                                   AioContext *ctx,
1211                                   uint64_t perm, uint64_t shared_perm,
1212                                   void *opaque, Error **errp);
1213 void bdrv_root_unref_child(BdrvChild *child);
1214 
1215 /**
1216  * Sets a BdrvChild's permissions.  Avoid if the parent is a BDS; use
1217  * bdrv_child_refresh_perms() instead and make the parent's
1218  * .bdrv_child_perm() implementation return the correct values.
1219  */
1220 int bdrv_child_try_set_perm(BdrvChild *c, uint64_t perm, uint64_t shared,
1221                             Error **errp);
1222 
1223 /**
1224  * Calls bs->drv->bdrv_child_perm() and updates the child's permission
1225  * masks with the result.
1226  * Drivers should invoke this function whenever an event occurs that
1227  * makes their .bdrv_child_perm() implementation return different
1228  * values than before, but which will not result in the block layer
1229  * automatically refreshing the permissions.
1230  */
1231 int bdrv_child_refresh_perms(BlockDriverState *bs, BdrvChild *c, Error **errp);
1232 
1233 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1234  * block filters: Forward CONSISTENT_READ, WRITE, WRITE_UNCHANGED and RESIZE to
1235  * all children */
1236 void bdrv_filter_default_perms(BlockDriverState *bs, BdrvChild *c,
1237                                const BdrvChildRole *role,
1238                                BlockReopenQueue *reopen_queue,
1239                                uint64_t perm, uint64_t shared,
1240                                uint64_t *nperm, uint64_t *nshared);
1241 
1242 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1243  * (non-raw) image formats: Like above for bs->backing, but for bs->file it
1244  * requires WRITE | RESIZE for read-write images, always requires
1245  * CONSISTENT_READ and doesn't share WRITE. */
1246 void bdrv_format_default_perms(BlockDriverState *bs, BdrvChild *c,
1247                                const BdrvChildRole *role,
1248                                BlockReopenQueue *reopen_queue,
1249                                uint64_t perm, uint64_t shared,
1250                                uint64_t *nperm, uint64_t *nshared);
1251 
1252 /*
1253  * Default implementation for drivers to pass bdrv_co_block_status() to
1254  * their file.
1255  */
1256 int coroutine_fn bdrv_co_block_status_from_file(BlockDriverState *bs,
1257                                                 bool want_zero,
1258                                                 int64_t offset,
1259                                                 int64_t bytes,
1260                                                 int64_t *pnum,
1261                                                 int64_t *map,
1262                                                 BlockDriverState **file);
1263 /*
1264  * Default implementation for drivers to pass bdrv_co_block_status() to
1265  * their backing file.
1266  */
1267 int coroutine_fn bdrv_co_block_status_from_backing(BlockDriverState *bs,
1268                                                    bool want_zero,
1269                                                    int64_t offset,
1270                                                    int64_t bytes,
1271                                                    int64_t *pnum,
1272                                                    int64_t *map,
1273                                                    BlockDriverState **file);
1274 const char *bdrv_get_parent_name(const BlockDriverState *bs);
1275 void blk_dev_change_media_cb(BlockBackend *blk, bool load, Error **errp);
1276 bool blk_dev_has_removable_media(BlockBackend *blk);
1277 bool blk_dev_has_tray(BlockBackend *blk);
1278 void blk_dev_eject_request(BlockBackend *blk, bool force);
1279 bool blk_dev_is_tray_open(BlockBackend *blk);
1280 bool blk_dev_is_medium_locked(BlockBackend *blk);
1281 
1282 void bdrv_set_dirty(BlockDriverState *bs, int64_t offset, int64_t bytes);
1283 
1284 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap **out);
1285 void bdrv_restore_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap *backup);
1286 bool bdrv_dirty_bitmap_merge_internal(BdrvDirtyBitmap *dest,
1287                                       const BdrvDirtyBitmap *src,
1288                                       HBitmap **backup, bool lock);
1289 
1290 void bdrv_inc_in_flight(BlockDriverState *bs);
1291 void bdrv_dec_in_flight(BlockDriverState *bs);
1292 
1293 void blockdev_close_all_bdrv_states(void);
1294 
1295 int coroutine_fn bdrv_co_copy_range_from(BdrvChild *src, uint64_t src_offset,
1296                                          BdrvChild *dst, uint64_t dst_offset,
1297                                          uint64_t bytes,
1298                                          BdrvRequestFlags read_flags,
1299                                          BdrvRequestFlags write_flags);
1300 int coroutine_fn bdrv_co_copy_range_to(BdrvChild *src, uint64_t src_offset,
1301                                        BdrvChild *dst, uint64_t dst_offset,
1302                                        uint64_t bytes,
1303                                        BdrvRequestFlags read_flags,
1304                                        BdrvRequestFlags write_flags);
1305 
1306 int refresh_total_sectors(BlockDriverState *bs, int64_t hint);
1307 
1308 #endif /* BLOCK_INT_H */
1309