xref: /openbmc/qemu/include/block/block_int.h (revision 135b03cb)
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_writev)(BlockDriverState *bs,
214         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, int flags);
215     /**
216      * @offset: position in bytes to write at
217      * @bytes: number of bytes to write
218      * @qiov: the buffers containing data to write
219      * @flags: zero or more bits allowed by 'supported_write_flags'
220      *
221      * @offset and @bytes will be a multiple of 'request_alignment',
222      * but the length of individual @qiov elements does not have to
223      * be a multiple.
224      *
225      * @bytes will always equal the total size of @qiov, and will be
226      * no larger than 'max_transfer'.
227      *
228      * The buffer in @qiov may point directly to guest memory.
229      */
230     int coroutine_fn (*bdrv_co_pwritev)(BlockDriverState *bs,
231         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags);
232 
233     /*
234      * Efficiently zero a region of the disk image.  Typically an image format
235      * would use a compact metadata representation to implement this.  This
236      * function pointer may be NULL or return -ENOSUP and .bdrv_co_writev()
237      * will be called instead.
238      */
239     int coroutine_fn (*bdrv_co_pwrite_zeroes)(BlockDriverState *bs,
240         int64_t offset, int bytes, BdrvRequestFlags flags);
241     int coroutine_fn (*bdrv_co_pdiscard)(BlockDriverState *bs,
242         int64_t offset, int bytes);
243 
244     /* Map [offset, offset + nbytes) range onto a child of @bs to copy from,
245      * and invoke bdrv_co_copy_range_from(child, ...), or invoke
246      * bdrv_co_copy_range_to() if @bs is the leaf child to copy data from.
247      *
248      * See the comment of bdrv_co_copy_range for the parameter and return value
249      * semantics.
250      */
251     int coroutine_fn (*bdrv_co_copy_range_from)(BlockDriverState *bs,
252                                                 BdrvChild *src,
253                                                 uint64_t offset,
254                                                 BdrvChild *dst,
255                                                 uint64_t dst_offset,
256                                                 uint64_t bytes,
257                                                 BdrvRequestFlags read_flags,
258                                                 BdrvRequestFlags write_flags);
259 
260     /* Map [offset, offset + nbytes) range onto a child of bs to copy data to,
261      * and invoke bdrv_co_copy_range_to(child, src, ...), or perform the copy
262      * operation if @bs is the leaf and @src has the same BlockDriver.  Return
263      * -ENOTSUP if @bs is the leaf but @src has a different BlockDriver.
264      *
265      * See the comment of bdrv_co_copy_range for the parameter and return value
266      * semantics.
267      */
268     int coroutine_fn (*bdrv_co_copy_range_to)(BlockDriverState *bs,
269                                               BdrvChild *src,
270                                               uint64_t src_offset,
271                                               BdrvChild *dst,
272                                               uint64_t dst_offset,
273                                               uint64_t bytes,
274                                               BdrvRequestFlags read_flags,
275                                               BdrvRequestFlags write_flags);
276 
277     /*
278      * Building block for bdrv_block_status[_above] and
279      * bdrv_is_allocated[_above].  The driver should answer only
280      * according to the current layer, and should only need to set
281      * BDRV_BLOCK_DATA, BDRV_BLOCK_ZERO, BDRV_BLOCK_OFFSET_VALID,
282      * and/or BDRV_BLOCK_RAW; if the current layer defers to a backing
283      * layer, the result should be 0 (and not BDRV_BLOCK_ZERO).  See
284      * block.h for the overall meaning of the bits.  As a hint, the
285      * flag want_zero is true if the caller cares more about precise
286      * mappings (favor accurate _OFFSET_VALID/_ZERO) or false for
287      * overall allocation (favor larger *pnum, perhaps by reporting
288      * _DATA instead of _ZERO).  The block layer guarantees input
289      * clamped to bdrv_getlength() and aligned to request_alignment,
290      * as well as non-NULL pnum, map, and file; in turn, the driver
291      * must return an error or set pnum to an aligned non-zero value.
292      */
293     int coroutine_fn (*bdrv_co_block_status)(BlockDriverState *bs,
294         bool want_zero, int64_t offset, int64_t bytes, int64_t *pnum,
295         int64_t *map, BlockDriverState **file);
296 
297     /*
298      * Invalidate any cached meta-data.
299      */
300     void coroutine_fn (*bdrv_co_invalidate_cache)(BlockDriverState *bs,
301                                                   Error **errp);
302     int (*bdrv_inactivate)(BlockDriverState *bs);
303 
304     /*
305      * Flushes all data for all layers by calling bdrv_co_flush for underlying
306      * layers, if needed. This function is needed for deterministic
307      * synchronization of the flush finishing callback.
308      */
309     int coroutine_fn (*bdrv_co_flush)(BlockDriverState *bs);
310 
311     /*
312      * Flushes all data that was already written to the OS all the way down to
313      * the disk (for example file-posix.c calls fsync()).
314      */
315     int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs);
316 
317     /*
318      * Flushes all internal caches to the OS. The data may still sit in a
319      * writeback cache of the host OS, but it will survive a crash of the qemu
320      * process.
321      */
322     int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs);
323 
324     /*
325      * Drivers setting this field must be able to work with just a plain
326      * filename with '<protocol_name>:' as a prefix, and no other options.
327      * Options may be extracted from the filename by implementing
328      * bdrv_parse_filename.
329      */
330     const char *protocol_name;
331     int coroutine_fn (*bdrv_co_truncate)(BlockDriverState *bs, int64_t offset,
332                                          PreallocMode prealloc, Error **errp);
333 
334     int64_t (*bdrv_getlength)(BlockDriverState *bs);
335     bool has_variable_length;
336     int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
337     BlockMeasureInfo *(*bdrv_measure)(QemuOpts *opts, BlockDriverState *in_bs,
338                                       Error **errp);
339 
340     int coroutine_fn (*bdrv_co_pwritev_compressed)(BlockDriverState *bs,
341         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov);
342 
343     int (*bdrv_snapshot_create)(BlockDriverState *bs,
344                                 QEMUSnapshotInfo *sn_info);
345     int (*bdrv_snapshot_goto)(BlockDriverState *bs,
346                               const char *snapshot_id);
347     int (*bdrv_snapshot_delete)(BlockDriverState *bs,
348                                 const char *snapshot_id,
349                                 const char *name,
350                                 Error **errp);
351     int (*bdrv_snapshot_list)(BlockDriverState *bs,
352                               QEMUSnapshotInfo **psn_info);
353     int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
354                                   const char *snapshot_id,
355                                   const char *name,
356                                   Error **errp);
357     int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
358     ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs,
359                                                  Error **errp);
360 
361     int coroutine_fn (*bdrv_save_vmstate)(BlockDriverState *bs,
362                                           QEMUIOVector *qiov,
363                                           int64_t pos);
364     int coroutine_fn (*bdrv_load_vmstate)(BlockDriverState *bs,
365                                           QEMUIOVector *qiov,
366                                           int64_t pos);
367 
368     int (*bdrv_change_backing_file)(BlockDriverState *bs,
369         const char *backing_file, const char *backing_fmt);
370 
371     /* removable device specific */
372     bool (*bdrv_is_inserted)(BlockDriverState *bs);
373     void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
374     void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
375 
376     /* to control generic scsi devices */
377     BlockAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
378         unsigned long int req, void *buf,
379         BlockCompletionFunc *cb, void *opaque);
380     int coroutine_fn (*bdrv_co_ioctl)(BlockDriverState *bs,
381                                       unsigned long int req, void *buf);
382 
383     /* List of options for creating images, terminated by name == NULL */
384     QemuOptsList *create_opts;
385     /*
386      * If this driver supports reopening images this contains a
387      * NULL-terminated list of the runtime options that can be
388      * modified. If an option in this list is unspecified during
389      * reopen then it _must_ be reset to its default value or return
390      * an error.
391      */
392     const char *const *mutable_opts;
393 
394     /*
395      * Returns 0 for completed check, -errno for internal errors.
396      * The check results are stored in result.
397      */
398     int coroutine_fn (*bdrv_co_check)(BlockDriverState *bs,
399                                       BdrvCheckResult *result,
400                                       BdrvCheckMode fix);
401 
402     int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts,
403                               BlockDriverAmendStatusCB *status_cb,
404                               void *cb_opaque,
405                               Error **errp);
406 
407     void (*bdrv_debug_event)(BlockDriverState *bs, BlkdebugEvent event);
408 
409     /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
410     int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
411         const char *tag);
412     int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs,
413         const char *tag);
414     int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
415     bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
416 
417     void (*bdrv_refresh_limits)(BlockDriverState *bs, Error **errp);
418 
419     /*
420      * Returns 1 if newly created images are guaranteed to contain only
421      * zeros, 0 otherwise.
422      */
423     int (*bdrv_has_zero_init)(BlockDriverState *bs);
424 
425     /* Remove fd handlers, timers, and other event loop callbacks so the event
426      * loop is no longer in use.  Called with no in-flight requests and in
427      * depth-first traversal order with parents before child nodes.
428      */
429     void (*bdrv_detach_aio_context)(BlockDriverState *bs);
430 
431     /* Add fd handlers, timers, and other event loop callbacks so I/O requests
432      * can be processed again.  Called with no in-flight requests and in
433      * depth-first traversal order with child nodes before parent nodes.
434      */
435     void (*bdrv_attach_aio_context)(BlockDriverState *bs,
436                                     AioContext *new_context);
437 
438     /* io queue for linux-aio */
439     void (*bdrv_io_plug)(BlockDriverState *bs);
440     void (*bdrv_io_unplug)(BlockDriverState *bs);
441 
442     /**
443      * Try to get @bs's logical and physical block size.
444      * On success, store them in @bsz and return zero.
445      * On failure, return negative errno.
446      */
447     int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz);
448     /**
449      * Try to get @bs's geometry (cyls, heads, sectors)
450      * On success, store them in @geo and return 0.
451      * On failure return -errno.
452      * Only drivers that want to override guest geometry implement this
453      * callback; see hd_geometry_guess().
454      */
455     int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo);
456 
457     /**
458      * bdrv_co_drain_begin is called if implemented in the beginning of a
459      * drain operation to drain and stop any internal sources of requests in
460      * the driver.
461      * bdrv_co_drain_end is called if implemented at the end of the drain.
462      *
463      * They should be used by the driver to e.g. manage scheduled I/O
464      * requests, or toggle an internal state. After the end of the drain new
465      * requests will continue normally.
466      */
467     void coroutine_fn (*bdrv_co_drain_begin)(BlockDriverState *bs);
468     void coroutine_fn (*bdrv_co_drain_end)(BlockDriverState *bs);
469 
470     void (*bdrv_add_child)(BlockDriverState *parent, BlockDriverState *child,
471                            Error **errp);
472     void (*bdrv_del_child)(BlockDriverState *parent, BdrvChild *child,
473                            Error **errp);
474 
475     /**
476      * Informs the block driver that a permission change is intended. The
477      * driver checks whether the change is permissible and may take other
478      * preparations for the change (e.g. get file system locks). This operation
479      * is always followed either by a call to either .bdrv_set_perm or
480      * .bdrv_abort_perm_update.
481      *
482      * Checks whether the requested set of cumulative permissions in @perm
483      * can be granted for accessing @bs and whether no other users are using
484      * permissions other than those given in @shared (both arguments take
485      * BLK_PERM_* bitmasks).
486      *
487      * If both conditions are met, 0 is returned. Otherwise, -errno is returned
488      * and errp is set to an error describing the conflict.
489      */
490     int (*bdrv_check_perm)(BlockDriverState *bs, uint64_t perm,
491                            uint64_t shared, Error **errp);
492 
493     /**
494      * Called to inform the driver that the set of cumulative set of used
495      * permissions for @bs has changed to @perm, and the set of sharable
496      * permission to @shared. The driver can use this to propagate changes to
497      * its children (i.e. request permissions only if a parent actually needs
498      * them).
499      *
500      * This function is only invoked after bdrv_check_perm(), so block drivers
501      * may rely on preparations made in their .bdrv_check_perm implementation.
502      */
503     void (*bdrv_set_perm)(BlockDriverState *bs, uint64_t perm, uint64_t shared);
504 
505     /*
506      * Called to inform the driver that after a previous bdrv_check_perm()
507      * call, the permission update is not performed and any preparations made
508      * for it (e.g. taken file locks) need to be undone.
509      *
510      * This function can be called even for nodes that never saw a
511      * bdrv_check_perm() call. It is a no-op then.
512      */
513     void (*bdrv_abort_perm_update)(BlockDriverState *bs);
514 
515     /**
516      * Returns in @nperm and @nshared the permissions that the driver for @bs
517      * needs on its child @c, based on the cumulative permissions requested by
518      * the parents in @parent_perm and @parent_shared.
519      *
520      * If @c is NULL, return the permissions for attaching a new child for the
521      * given @role.
522      *
523      * If @reopen_queue is non-NULL, don't return the currently needed
524      * permissions, but those that will be needed after applying the
525      * @reopen_queue.
526      */
527      void (*bdrv_child_perm)(BlockDriverState *bs, BdrvChild *c,
528                              const BdrvChildRole *role,
529                              BlockReopenQueue *reopen_queue,
530                              uint64_t parent_perm, uint64_t parent_shared,
531                              uint64_t *nperm, uint64_t *nshared);
532 
533     /**
534      * Bitmaps should be marked as 'IN_USE' in the image on reopening image
535      * as rw. This handler should realize it. It also should unset readonly
536      * field of BlockDirtyBitmap's in case of success.
537      */
538     int (*bdrv_reopen_bitmaps_rw)(BlockDriverState *bs, Error **errp);
539     bool (*bdrv_can_store_new_dirty_bitmap)(BlockDriverState *bs,
540                                             const char *name,
541                                             uint32_t granularity,
542                                             Error **errp);
543     void (*bdrv_remove_persistent_dirty_bitmap)(BlockDriverState *bs,
544                                                 const char *name,
545                                                 Error **errp);
546 
547     /**
548      * Register/unregister a buffer for I/O. For example, when the driver is
549      * interested to know the memory areas that will later be used in iovs, so
550      * that it can do IOMMU mapping with VFIO etc., in order to get better
551      * performance. In the case of VFIO drivers, this callback is used to do
552      * DMA mapping for hot buffers.
553      */
554     void (*bdrv_register_buf)(BlockDriverState *bs, void *host, size_t size);
555     void (*bdrv_unregister_buf)(BlockDriverState *bs, void *host);
556     QLIST_ENTRY(BlockDriver) list;
557 
558     /* Pointer to a NULL-terminated array of names of strong options
559      * that can be specified for bdrv_open(). A strong option is one
560      * that changes the data of a BDS.
561      * If this pointer is NULL, the array is considered empty.
562      * "filename" and "driver" are always considered strong. */
563     const char *const *strong_runtime_opts;
564 };
565 
566 typedef struct BlockLimits {
567     /* Alignment requirement, in bytes, for offset/length of I/O
568      * requests. Must be a power of 2 less than INT_MAX; defaults to
569      * 1 for drivers with modern byte interfaces, and to 512
570      * otherwise. */
571     uint32_t request_alignment;
572 
573     /* Maximum number of bytes that can be discarded at once (since it
574      * is signed, it must be < 2G, if set). Must be multiple of
575      * pdiscard_alignment, but need not be power of 2. May be 0 if no
576      * inherent 32-bit limit */
577     int32_t max_pdiscard;
578 
579     /* Optimal alignment for discard requests in bytes. A power of 2
580      * is best but not mandatory.  Must be a multiple of
581      * bl.request_alignment, and must be less than max_pdiscard if
582      * that is set. May be 0 if bl.request_alignment is good enough */
583     uint32_t pdiscard_alignment;
584 
585     /* Maximum number of bytes that can zeroized at once (since it is
586      * signed, it must be < 2G, if set). Must be multiple of
587      * pwrite_zeroes_alignment. May be 0 if no inherent 32-bit limit */
588     int32_t max_pwrite_zeroes;
589 
590     /* Optimal alignment for write zeroes requests in bytes. A power
591      * of 2 is best but not mandatory.  Must be a multiple of
592      * bl.request_alignment, and must be less than max_pwrite_zeroes
593      * if that is set. May be 0 if bl.request_alignment is good
594      * enough */
595     uint32_t pwrite_zeroes_alignment;
596 
597     /* Optimal transfer length in bytes.  A power of 2 is best but not
598      * mandatory.  Must be a multiple of bl.request_alignment, or 0 if
599      * no preferred size */
600     uint32_t opt_transfer;
601 
602     /* Maximal transfer length in bytes.  Need not be power of 2, but
603      * must be multiple of opt_transfer and bl.request_alignment, or 0
604      * for no 32-bit limit.  For now, anything larger than INT_MAX is
605      * clamped down. */
606     uint32_t max_transfer;
607 
608     /* memory alignment, in bytes so that no bounce buffer is needed */
609     size_t min_mem_alignment;
610 
611     /* memory alignment, in bytes, for bounce buffer */
612     size_t opt_mem_alignment;
613 
614     /* maximum number of iovec elements */
615     int max_iov;
616 } BlockLimits;
617 
618 typedef struct BdrvOpBlocker BdrvOpBlocker;
619 
620 typedef struct BdrvAioNotifier {
621     void (*attached_aio_context)(AioContext *new_context, void *opaque);
622     void (*detach_aio_context)(void *opaque);
623 
624     void *opaque;
625     bool deleted;
626 
627     QLIST_ENTRY(BdrvAioNotifier) list;
628 } BdrvAioNotifier;
629 
630 struct BdrvChildRole {
631     /* If true, bdrv_replace_node() doesn't change the node this BdrvChild
632      * points to. */
633     bool stay_at_node;
634 
635     /* If true, the parent is a BlockDriverState and bdrv_next_all_states()
636      * will return it. This information is used for drain_all, where every node
637      * will be drained separately, so the drain only needs to be propagated to
638      * non-BDS parents. */
639     bool parent_is_bds;
640 
641     void (*inherit_options)(int *child_flags, QDict *child_options,
642                             int parent_flags, QDict *parent_options);
643 
644     void (*change_media)(BdrvChild *child, bool load);
645     void (*resize)(BdrvChild *child);
646 
647     /* Returns a name that is supposedly more useful for human users than the
648      * node name for identifying the node in question (in particular, a BB
649      * name), or NULL if the parent can't provide a better name. */
650     const char *(*get_name)(BdrvChild *child);
651 
652     /* Returns a malloced string that describes the parent of the child for a
653      * human reader. This could be a node-name, BlockBackend name, qdev ID or
654      * QOM path of the device owning the BlockBackend, job type and ID etc. The
655      * caller is responsible for freeing the memory. */
656     char *(*get_parent_desc)(BdrvChild *child);
657 
658     /*
659      * If this pair of functions is implemented, the parent doesn't issue new
660      * requests after returning from .drained_begin() until .drained_end() is
661      * called.
662      *
663      * These functions must not change the graph (and therefore also must not
664      * call aio_poll(), which could change the graph indirectly).
665      *
666      * If drained_end() schedules background operations, it must atomically
667      * increment *drained_end_counter for each such operation and atomically
668      * decrement it once the operation has settled.
669      *
670      * Note that this can be nested. If drained_begin() was called twice, new
671      * I/O is allowed only after drained_end() was called twice, too.
672      */
673     void (*drained_begin)(BdrvChild *child);
674     void (*drained_end)(BdrvChild *child, int *drained_end_counter);
675 
676     /*
677      * Returns whether the parent has pending requests for the child. This
678      * callback is polled after .drained_begin() has been called until all
679      * activity on the child has stopped.
680      */
681     bool (*drained_poll)(BdrvChild *child);
682 
683     /* Notifies the parent that the child has been activated/inactivated (e.g.
684      * when migration is completing) and it can start/stop requesting
685      * permissions and doing I/O on it. */
686     void (*activate)(BdrvChild *child, Error **errp);
687     int (*inactivate)(BdrvChild *child);
688 
689     void (*attach)(BdrvChild *child);
690     void (*detach)(BdrvChild *child);
691 
692     /* Notifies the parent that the filename of its child has changed (e.g.
693      * because the direct child was removed from the backing chain), so that it
694      * can update its reference. */
695     int (*update_filename)(BdrvChild *child, BlockDriverState *new_base,
696                            const char *filename, Error **errp);
697 
698     bool (*can_set_aio_ctx)(BdrvChild *child, AioContext *ctx,
699                             GSList **ignore, Error **errp);
700     void (*set_aio_ctx)(BdrvChild *child, AioContext *ctx, GSList **ignore);
701 };
702 
703 extern const BdrvChildRole child_file;
704 extern const BdrvChildRole child_format;
705 extern const BdrvChildRole child_backing;
706 
707 struct BdrvChild {
708     BlockDriverState *bs;
709     char *name;
710     const BdrvChildRole *role;
711     void *opaque;
712 
713     /**
714      * Granted permissions for operating on this BdrvChild (BLK_PERM_* bitmask)
715      */
716     uint64_t perm;
717 
718     /**
719      * Permissions that can still be granted to other users of @bs while this
720      * BdrvChild is still attached to it. (BLK_PERM_* bitmask)
721      */
722     uint64_t shared_perm;
723 
724     /* backup of permissions during permission update procedure */
725     bool has_backup_perm;
726     uint64_t backup_perm;
727     uint64_t backup_shared_perm;
728 
729     /*
730      * This link is frozen: the child can neither be replaced nor
731      * detached from the parent.
732      */
733     bool frozen;
734 
735     /*
736      * How many times the parent of this child has been drained
737      * (through role->drained_*).
738      * Usually, this is equal to bs->quiesce_counter (potentially
739      * reduced by bdrv_drain_all_count).  It may differ while the
740      * child is entering or leaving a drained section.
741      */
742     int parent_quiesce_counter;
743 
744     QLIST_ENTRY(BdrvChild) next;
745     QLIST_ENTRY(BdrvChild) next_parent;
746 };
747 
748 /*
749  * Note: the function bdrv_append() copies and swaps contents of
750  * BlockDriverStates, so if you add new fields to this struct, please
751  * inspect bdrv_append() to determine if the new fields need to be
752  * copied as well.
753  */
754 struct BlockDriverState {
755     /* Protected by big QEMU lock or read-only after opening.  No special
756      * locking needed during I/O...
757      */
758     int open_flags; /* flags used to open the file, re-used for re-open */
759     bool read_only; /* if true, the media is read only */
760     bool encrypted; /* if true, the media is encrypted */
761     bool sg;        /* if true, the device is a /dev/sg* */
762     bool probed;    /* if true, format was probed rather than specified */
763     bool force_share; /* if true, always allow all shared permissions */
764     bool implicit;  /* if true, this filter node was automatically inserted */
765 
766     BlockDriver *drv; /* NULL means no media */
767     void *opaque;
768 
769     AioContext *aio_context; /* event loop used for fd handlers, timers, etc */
770     /* long-running tasks intended to always use the same AioContext as this
771      * BDS may register themselves in this list to be notified of changes
772      * regarding this BDS's context */
773     QLIST_HEAD(, BdrvAioNotifier) aio_notifiers;
774     bool walking_aio_notifiers; /* to make removal during iteration safe */
775 
776     char filename[PATH_MAX];
777     char backing_file[PATH_MAX]; /* if non zero, the image is a diff of
778                                     this file image */
779     /* The backing filename indicated by the image header; if we ever
780      * open this file, then this is replaced by the resulting BDS's
781      * filename (i.e. after a bdrv_refresh_filename() run). */
782     char auto_backing_file[PATH_MAX];
783     char backing_format[16]; /* if non-zero and backing_file exists */
784 
785     QDict *full_open_options;
786     char exact_filename[PATH_MAX];
787 
788     BdrvChild *backing;
789     BdrvChild *file;
790 
791     /* I/O Limits */
792     BlockLimits bl;
793 
794     /* Flags honored during pwrite (so far: BDRV_REQ_FUA,
795      * BDRV_REQ_WRITE_UNCHANGED).
796      * If a driver does not support BDRV_REQ_WRITE_UNCHANGED, those
797      * writes will be issued as normal writes without the flag set.
798      * This is important to note for drivers that do not explicitly
799      * request a WRITE permission for their children and instead take
800      * the same permissions as their parent did (this is commonly what
801      * block filters do).  Such drivers have to be aware that the
802      * parent may have taken a WRITE_UNCHANGED permission only and is
803      * issuing such requests.  Drivers either must make sure that
804      * these requests do not result in plain WRITE accesses (usually
805      * by supporting BDRV_REQ_WRITE_UNCHANGED, and then forwarding
806      * every incoming write request as-is, including potentially that
807      * flag), or they have to explicitly take the WRITE permission for
808      * their children. */
809     unsigned int supported_write_flags;
810     /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA,
811      * BDRV_REQ_MAY_UNMAP, BDRV_REQ_WRITE_UNCHANGED) */
812     unsigned int supported_zero_flags;
813 
814     /* the following member gives a name to every node on the bs graph. */
815     char node_name[32];
816     /* element of the list of named nodes building the graph */
817     QTAILQ_ENTRY(BlockDriverState) node_list;
818     /* element of the list of all BlockDriverStates (all_bdrv_states) */
819     QTAILQ_ENTRY(BlockDriverState) bs_list;
820     /* element of the list of monitor-owned BDS */
821     QTAILQ_ENTRY(BlockDriverState) monitor_list;
822     int refcnt;
823 
824     /* operation blockers */
825     QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX];
826 
827     /* The node that this node inherited default options from (and a reopen on
828      * which can affect this node by changing these defaults). This is always a
829      * parent node of this node. */
830     BlockDriverState *inherits_from;
831     QLIST_HEAD(, BdrvChild) children;
832     QLIST_HEAD(, BdrvChild) parents;
833 
834     QDict *options;
835     QDict *explicit_options;
836     BlockdevDetectZeroesOptions detect_zeroes;
837 
838     /* The error object in use for blocking operations on backing_hd */
839     Error *backing_blocker;
840 
841     /* Protected by AioContext lock */
842 
843     /* If we are reading a disk image, give its size in sectors.
844      * Generally read-only; it is written to by load_snapshot and
845      * save_snaphost, but the block layer is quiescent during those.
846      */
847     int64_t total_sectors;
848 
849     /* Callback before write request is processed */
850     NotifierWithReturnList before_write_notifiers;
851 
852     /* threshold limit for writes, in bytes. "High water mark". */
853     uint64_t write_threshold_offset;
854     NotifierWithReturn write_threshold_notifier;
855 
856     /* Writing to the list requires the BQL _and_ the dirty_bitmap_mutex.
857      * Reading from the list can be done with either the BQL or the
858      * dirty_bitmap_mutex.  Modifying a bitmap only requires
859      * dirty_bitmap_mutex.  */
860     QemuMutex dirty_bitmap_mutex;
861     QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps;
862 
863     /* Offset after the highest byte written to */
864     Stat64 wr_highest_offset;
865 
866     /* If true, copy read backing sectors into image.  Can be >1 if more
867      * than one client has requested copy-on-read.  Accessed with atomic
868      * ops.
869      */
870     int copy_on_read;
871 
872     /* number of in-flight requests; overall and serialising.
873      * Accessed with atomic ops.
874      */
875     unsigned int in_flight;
876     unsigned int serialising_in_flight;
877 
878     /* counter for nested bdrv_io_plug.
879      * Accessed with atomic ops.
880     */
881     unsigned io_plugged;
882 
883     /* do we need to tell the quest if we have a volatile write cache? */
884     int enable_write_cache;
885 
886     /* Accessed with atomic ops.  */
887     int quiesce_counter;
888     int recursive_quiesce_counter;
889 
890     unsigned int write_gen;               /* Current data generation */
891 
892     /* Protected by reqs_lock.  */
893     CoMutex reqs_lock;
894     QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
895     CoQueue flush_queue;                  /* Serializing flush queue */
896     bool active_flush_req;                /* Flush request in flight? */
897 
898     /* Only read/written by whoever has set active_flush_req to true.  */
899     unsigned int flushed_gen;             /* Flushed write generation */
900 
901     /* BdrvChild links to this node may never be frozen */
902     bool never_freeze;
903 };
904 
905 struct BlockBackendRootState {
906     int open_flags;
907     bool read_only;
908     BlockdevDetectZeroesOptions detect_zeroes;
909 };
910 
911 typedef enum BlockMirrorBackingMode {
912     /* Reuse the existing backing chain from the source for the target.
913      * - sync=full: Set backing BDS to NULL.
914      * - sync=top:  Use source's backing BDS.
915      * - sync=none: Use source as the backing BDS. */
916     MIRROR_SOURCE_BACKING_CHAIN,
917 
918     /* Open the target's backing chain completely anew */
919     MIRROR_OPEN_BACKING_CHAIN,
920 
921     /* Do not change the target's backing BDS after job completion */
922     MIRROR_LEAVE_BACKING_CHAIN,
923 } BlockMirrorBackingMode;
924 
925 static inline BlockDriverState *backing_bs(BlockDriverState *bs)
926 {
927     return bs->backing ? bs->backing->bs : NULL;
928 }
929 
930 
931 /* Essential block drivers which must always be statically linked into qemu, and
932  * which therefore can be accessed without using bdrv_find_format() */
933 extern BlockDriver bdrv_file;
934 extern BlockDriver bdrv_raw;
935 extern BlockDriver bdrv_qcow2;
936 
937 int coroutine_fn bdrv_co_preadv(BdrvChild *child,
938     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
939     BdrvRequestFlags flags);
940 int coroutine_fn bdrv_co_pwritev(BdrvChild *child,
941     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
942     BdrvRequestFlags flags);
943 
944 static inline int coroutine_fn bdrv_co_pread(BdrvChild *child,
945     int64_t offset, unsigned int bytes, void *buf, BdrvRequestFlags flags)
946 {
947     QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes);
948 
949     return bdrv_co_preadv(child, offset, bytes, &qiov, flags);
950 }
951 
952 static inline int coroutine_fn bdrv_co_pwrite(BdrvChild *child,
953     int64_t offset, unsigned int bytes, void *buf, BdrvRequestFlags flags)
954 {
955     QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes);
956 
957     return bdrv_co_pwritev(child, offset, bytes, &qiov, flags);
958 }
959 
960 extern unsigned int bdrv_drain_all_count;
961 void bdrv_apply_subtree_drain(BdrvChild *child, BlockDriverState *new_parent);
962 void bdrv_unapply_subtree_drain(BdrvChild *child, BlockDriverState *old_parent);
963 
964 int get_tmp_filename(char *filename, int size);
965 BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size,
966                             const char *filename);
967 
968 void bdrv_parse_filename_strip_prefix(const char *filename, const char *prefix,
969                                       QDict *options);
970 
971 
972 /**
973  * bdrv_add_before_write_notifier:
974  *
975  * Register a callback that is invoked before write requests are processed but
976  * after any throttling or waiting for overlapping requests.
977  */
978 void bdrv_add_before_write_notifier(BlockDriverState *bs,
979                                     NotifierWithReturn *notifier);
980 
981 /**
982  * bdrv_add_aio_context_notifier:
983  *
984  * If a long-running job intends to be always run in the same AioContext as a
985  * certain BDS, it may use this function to be notified of changes regarding the
986  * association of the BDS to an AioContext.
987  *
988  * attached_aio_context() is called after the target BDS has been attached to a
989  * new AioContext; detach_aio_context() is called before the target BDS is being
990  * detached from its old AioContext.
991  */
992 void bdrv_add_aio_context_notifier(BlockDriverState *bs,
993         void (*attached_aio_context)(AioContext *new_context, void *opaque),
994         void (*detach_aio_context)(void *opaque), void *opaque);
995 
996 /**
997  * bdrv_remove_aio_context_notifier:
998  *
999  * Unsubscribe of change notifications regarding the BDS's AioContext. The
1000  * parameters given here have to be the same as those given to
1001  * bdrv_add_aio_context_notifier().
1002  */
1003 void bdrv_remove_aio_context_notifier(BlockDriverState *bs,
1004                                       void (*aio_context_attached)(AioContext *,
1005                                                                    void *),
1006                                       void (*aio_context_detached)(void *),
1007                                       void *opaque);
1008 
1009 /**
1010  * bdrv_wakeup:
1011  * @bs: The BlockDriverState for which an I/O operation has been completed.
1012  *
1013  * Wake up the main thread if it is waiting on BDRV_POLL_WHILE.  During
1014  * synchronous I/O on a BlockDriverState that is attached to another
1015  * I/O thread, the main thread lets the I/O thread's event loop run,
1016  * waiting for the I/O operation to complete.  A bdrv_wakeup will wake
1017  * up the main thread if necessary.
1018  *
1019  * Manual calls to bdrv_wakeup are rarely necessary, because
1020  * bdrv_dec_in_flight already calls it.
1021  */
1022 void bdrv_wakeup(BlockDriverState *bs);
1023 
1024 #ifdef _WIN32
1025 int is_windows_drive(const char *filename);
1026 #endif
1027 
1028 /**
1029  * stream_start:
1030  * @job_id: The id of the newly-created job, or %NULL to use the
1031  * device name of @bs.
1032  * @bs: Block device to operate on.
1033  * @base: Block device that will become the new base, or %NULL to
1034  * flatten the whole backing file chain onto @bs.
1035  * @backing_file_str: The file name that will be written to @bs as the
1036  * the new backing file if the job completes. Ignored if @base is %NULL.
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  * @errp: Error object.
1042  *
1043  * Start a streaming operation on @bs.  Clusters that are unallocated
1044  * in @bs, but allocated in any image between @base and @bs (both
1045  * exclusive) will be written to @bs.  At the end of a successful
1046  * streaming job, the backing file of @bs will be changed to
1047  * @backing_file_str in the written image and to @base in the live
1048  * BlockDriverState.
1049  */
1050 void stream_start(const char *job_id, BlockDriverState *bs,
1051                   BlockDriverState *base, const char *backing_file_str,
1052                   int creation_flags, int64_t speed,
1053                   BlockdevOnError on_error, Error **errp);
1054 
1055 /**
1056  * commit_start:
1057  * @job_id: The id of the newly-created job, or %NULL to use the
1058  * device name of @bs.
1059  * @bs: Active block device.
1060  * @top: Top block device to be committed.
1061  * @base: Block device that will be written into, and become the new top.
1062  * @creation_flags: Flags that control the behavior of the Job lifetime.
1063  *                  See @BlockJobCreateFlags
1064  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1065  * @on_error: The action to take upon error.
1066  * @backing_file_str: String to use as the backing file in @top's overlay
1067  * @filter_node_name: The node name that should be assigned to the filter
1068  * driver that the commit job inserts into the graph above @top. NULL means
1069  * that a node name should be autogenerated.
1070  * @errp: Error object.
1071  *
1072  */
1073 void commit_start(const char *job_id, BlockDriverState *bs,
1074                   BlockDriverState *base, BlockDriverState *top,
1075                   int creation_flags, int64_t speed,
1076                   BlockdevOnError on_error, const char *backing_file_str,
1077                   const char *filter_node_name, Error **errp);
1078 /**
1079  * commit_active_start:
1080  * @job_id: The id of the newly-created job, or %NULL to use the
1081  * device name of @bs.
1082  * @bs: Active block device to be committed.
1083  * @base: Block device that will be written into, and become the new top.
1084  * @creation_flags: Flags that control the behavior of the Job lifetime.
1085  *                  See @BlockJobCreateFlags
1086  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1087  * @on_error: The action to take upon error.
1088  * @filter_node_name: The node name that should be assigned to the filter
1089  * driver that the commit job inserts into the graph above @bs. NULL means that
1090  * a node name should be autogenerated.
1091  * @cb: Completion function for the job.
1092  * @opaque: Opaque pointer value passed to @cb.
1093  * @auto_complete: Auto complete the job.
1094  * @errp: Error object.
1095  *
1096  */
1097 BlockJob *commit_active_start(const char *job_id, BlockDriverState *bs,
1098                               BlockDriverState *base, int creation_flags,
1099                               int64_t speed, BlockdevOnError on_error,
1100                               const char *filter_node_name,
1101                               BlockCompletionFunc *cb, void *opaque,
1102                               bool auto_complete, Error **errp);
1103 /*
1104  * mirror_start:
1105  * @job_id: The id of the newly-created job, or %NULL to use the
1106  * device name of @bs.
1107  * @bs: Block device to operate on.
1108  * @target: Block device to write to.
1109  * @replaces: Block graph node name to replace once the mirror is done. Can
1110  *            only be used when full mirroring is selected.
1111  * @creation_flags: Flags that control the behavior of the Job lifetime.
1112  *                  See @BlockJobCreateFlags
1113  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1114  * @granularity: The chosen granularity for the dirty bitmap.
1115  * @buf_size: The amount of data that can be in flight at one time.
1116  * @mode: Whether to collapse all images in the chain to the target.
1117  * @backing_mode: How to establish the target's backing chain after completion.
1118  * @on_source_error: The action to take upon error reading from the source.
1119  * @on_target_error: The action to take upon error writing to the target.
1120  * @unmap: Whether to unmap target where source sectors only contain zeroes.
1121  * @filter_node_name: The node name that should be assigned to the filter
1122  * driver that the mirror job inserts into the graph above @bs. NULL means that
1123  * a node name should be autogenerated.
1124  * @copy_mode: When to trigger writes to the target.
1125  * @errp: Error object.
1126  *
1127  * Start a mirroring operation on @bs.  Clusters that are allocated
1128  * in @bs will be written to @target until the job is cancelled or
1129  * manually completed.  At the end of a successful mirroring job,
1130  * @bs will be switched to read from @target.
1131  */
1132 void mirror_start(const char *job_id, BlockDriverState *bs,
1133                   BlockDriverState *target, const char *replaces,
1134                   int creation_flags, int64_t speed,
1135                   uint32_t granularity, int64_t buf_size,
1136                   MirrorSyncMode mode, BlockMirrorBackingMode backing_mode,
1137                   BlockdevOnError on_source_error,
1138                   BlockdevOnError on_target_error,
1139                   bool unmap, const char *filter_node_name,
1140                   MirrorCopyMode copy_mode, Error **errp);
1141 
1142 /*
1143  * backup_job_create:
1144  * @job_id: The id of the newly-created job, or %NULL to use the
1145  * device name of @bs.
1146  * @bs: Block device to operate on.
1147  * @target: Block device to write to.
1148  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1149  * @sync_mode: What parts of the disk image should be copied to the destination.
1150  * @sync_bitmap: The dirty bitmap if sync_mode is MIRROR_SYNC_MODE_INCREMENTAL.
1151  * @on_source_error: The action to take upon error reading from the source.
1152  * @on_target_error: The action to take upon error writing to the target.
1153  * @creation_flags: Flags that control the behavior of the Job lifetime.
1154  *                  See @BlockJobCreateFlags
1155  * @cb: Completion function for the job.
1156  * @opaque: Opaque pointer value passed to @cb.
1157  * @txn: Transaction that this job is part of (may be NULL).
1158  *
1159  * Create a backup operation on @bs.  Clusters in @bs are written to @target
1160  * until the job is cancelled or manually completed.
1161  */
1162 BlockJob *backup_job_create(const char *job_id, BlockDriverState *bs,
1163                             BlockDriverState *target, int64_t speed,
1164                             MirrorSyncMode sync_mode,
1165                             BdrvDirtyBitmap *sync_bitmap,
1166                             bool compress,
1167                             BlockdevOnError on_source_error,
1168                             BlockdevOnError on_target_error,
1169                             int creation_flags,
1170                             BlockCompletionFunc *cb, void *opaque,
1171                             JobTxn *txn, Error **errp);
1172 
1173 void hmp_drive_add_node(Monitor *mon, const char *optstr);
1174 
1175 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
1176                                   const char *child_name,
1177                                   const BdrvChildRole *child_role,
1178                                   AioContext *ctx,
1179                                   uint64_t perm, uint64_t shared_perm,
1180                                   void *opaque, Error **errp);
1181 void bdrv_root_unref_child(BdrvChild *child);
1182 
1183 /**
1184  * Sets a BdrvChild's permissions.  Avoid if the parent is a BDS; use
1185  * bdrv_child_refresh_perms() instead and make the parent's
1186  * .bdrv_child_perm() implementation return the correct values.
1187  */
1188 int bdrv_child_try_set_perm(BdrvChild *c, uint64_t perm, uint64_t shared,
1189                             Error **errp);
1190 
1191 /**
1192  * Calls bs->drv->bdrv_child_perm() and updates the child's permission
1193  * masks with the result.
1194  * Drivers should invoke this function whenever an event occurs that
1195  * makes their .bdrv_child_perm() implementation return different
1196  * values than before, but which will not result in the block layer
1197  * automatically refreshing the permissions.
1198  */
1199 int bdrv_child_refresh_perms(BlockDriverState *bs, BdrvChild *c, Error **errp);
1200 
1201 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1202  * block filters: Forward CONSISTENT_READ, WRITE, WRITE_UNCHANGED and RESIZE to
1203  * all children */
1204 void bdrv_filter_default_perms(BlockDriverState *bs, BdrvChild *c,
1205                                const BdrvChildRole *role,
1206                                BlockReopenQueue *reopen_queue,
1207                                uint64_t perm, uint64_t shared,
1208                                uint64_t *nperm, uint64_t *nshared);
1209 
1210 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1211  * (non-raw) image formats: Like above for bs->backing, but for bs->file it
1212  * requires WRITE | RESIZE for read-write images, always requires
1213  * CONSISTENT_READ and doesn't share WRITE. */
1214 void bdrv_format_default_perms(BlockDriverState *bs, BdrvChild *c,
1215                                const BdrvChildRole *role,
1216                                BlockReopenQueue *reopen_queue,
1217                                uint64_t perm, uint64_t shared,
1218                                uint64_t *nperm, uint64_t *nshared);
1219 
1220 /*
1221  * Default implementation for drivers to pass bdrv_co_block_status() to
1222  * their file.
1223  */
1224 int coroutine_fn bdrv_co_block_status_from_file(BlockDriverState *bs,
1225                                                 bool want_zero,
1226                                                 int64_t offset,
1227                                                 int64_t bytes,
1228                                                 int64_t *pnum,
1229                                                 int64_t *map,
1230                                                 BlockDriverState **file);
1231 /*
1232  * Default implementation for drivers to pass bdrv_co_block_status() to
1233  * their backing file.
1234  */
1235 int coroutine_fn bdrv_co_block_status_from_backing(BlockDriverState *bs,
1236                                                    bool want_zero,
1237                                                    int64_t offset,
1238                                                    int64_t bytes,
1239                                                    int64_t *pnum,
1240                                                    int64_t *map,
1241                                                    BlockDriverState **file);
1242 const char *bdrv_get_parent_name(const BlockDriverState *bs);
1243 void blk_dev_change_media_cb(BlockBackend *blk, bool load, Error **errp);
1244 bool blk_dev_has_removable_media(BlockBackend *blk);
1245 bool blk_dev_has_tray(BlockBackend *blk);
1246 void blk_dev_eject_request(BlockBackend *blk, bool force);
1247 bool blk_dev_is_tray_open(BlockBackend *blk);
1248 bool blk_dev_is_medium_locked(BlockBackend *blk);
1249 
1250 void bdrv_set_dirty(BlockDriverState *bs, int64_t offset, int64_t bytes);
1251 
1252 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap **out);
1253 void bdrv_restore_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap *backup);
1254 
1255 void bdrv_inc_in_flight(BlockDriverState *bs);
1256 void bdrv_dec_in_flight(BlockDriverState *bs);
1257 
1258 void blockdev_close_all_bdrv_states(void);
1259 
1260 int coroutine_fn bdrv_co_copy_range_from(BdrvChild *src, uint64_t src_offset,
1261                                          BdrvChild *dst, uint64_t dst_offset,
1262                                          uint64_t bytes,
1263                                          BdrvRequestFlags read_flags,
1264                                          BdrvRequestFlags write_flags);
1265 int coroutine_fn bdrv_co_copy_range_to(BdrvChild *src, uint64_t src_offset,
1266                                        BdrvChild *dst, uint64_t dst_offset,
1267                                        uint64_t bytes,
1268                                        BdrvRequestFlags read_flags,
1269                                        BdrvRequestFlags write_flags);
1270 
1271 int refresh_total_sectors(BlockDriverState *bs, int64_t hint);
1272 
1273 #endif /* BLOCK_INT_H */
1274