xref: /openbmc/qemu/include/block/block_int.h (revision 2df9f571)
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 #define BLOCK_OPT_COMPRESSION_TYPE  "compression_type"
61 
62 #define BLOCK_PROBE_BUF_SIZE        512
63 
64 enum BdrvTrackedRequestType {
65     BDRV_TRACKED_READ,
66     BDRV_TRACKED_WRITE,
67     BDRV_TRACKED_DISCARD,
68     BDRV_TRACKED_TRUNCATE,
69 };
70 
71 typedef struct BdrvTrackedRequest {
72     BlockDriverState *bs;
73     int64_t offset;
74     uint64_t bytes;
75     enum BdrvTrackedRequestType type;
76 
77     bool serialising;
78     int64_t overlap_offset;
79     uint64_t overlap_bytes;
80 
81     QLIST_ENTRY(BdrvTrackedRequest) list;
82     Coroutine *co; /* owner, used for deadlock detection */
83     CoQueue wait_queue; /* coroutines blocked on this request */
84 
85     struct BdrvTrackedRequest *waiting_for;
86 } BdrvTrackedRequest;
87 
88 struct BlockDriver {
89     const char *format_name;
90     int instance_size;
91 
92     /* set to true if the BlockDriver is a block filter. Block filters pass
93      * certain callbacks that refer to data (see block.c) to their bs->file if
94      * the driver doesn't implement them. Drivers that do not wish to forward
95      * must implement them and return -ENOTSUP.
96      */
97     bool is_filter;
98     /*
99      * Return true if @to_replace can be replaced by a BDS with the
100      * same data as @bs without it affecting @bs's behavior (that is,
101      * without it being visible to @bs's parents).
102      */
103     bool (*bdrv_recurse_can_replace)(BlockDriverState *bs,
104                                      BlockDriverState *to_replace);
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_commit_post)(BDRVReopenState *reopen_state);
127     void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state);
128     void (*bdrv_join_options)(QDict *options, QDict *old_options);
129 
130     int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags,
131                      Error **errp);
132 
133     /* Protocol drivers should implement this instead of bdrv_open */
134     int (*bdrv_file_open)(BlockDriverState *bs, QDict *options, int flags,
135                           Error **errp);
136     void (*bdrv_close)(BlockDriverState *bs);
137     int coroutine_fn (*bdrv_co_create)(BlockdevCreateOptions *opts,
138                                        Error **errp);
139     int coroutine_fn (*bdrv_co_create_opts)(BlockDriver *drv,
140                                             const char *filename,
141                                             QemuOpts *opts,
142                                             Error **errp);
143     int (*bdrv_make_empty)(BlockDriverState *bs);
144 
145     /*
146      * Refreshes the bs->exact_filename field. If that is impossible,
147      * bs->exact_filename has to be left empty.
148      */
149     void (*bdrv_refresh_filename)(BlockDriverState *bs);
150 
151     /*
152      * Gathers the open options for all children into @target.
153      * A simple format driver (without backing file support) might
154      * implement this function like this:
155      *
156      *     QINCREF(bs->file->bs->full_open_options);
157      *     qdict_put(target, "file", bs->file->bs->full_open_options);
158      *
159      * If not specified, the generic implementation will simply put
160      * all children's options under their respective name.
161      *
162      * @backing_overridden is true when bs->backing seems not to be
163      * the child that would result from opening bs->backing_file.
164      * Therefore, if it is true, the backing child's options should be
165      * gathered; otherwise, there is no need since the backing child
166      * is the one implied by the image header.
167      *
168      * Note that ideally this function would not be needed.  Every
169      * block driver which implements it is probably doing something
170      * shady regarding its runtime option structure.
171      */
172     void (*bdrv_gather_child_options)(BlockDriverState *bs, QDict *target,
173                                       bool backing_overridden);
174 
175     /*
176      * Returns an allocated string which is the directory name of this BDS: It
177      * will be used to make relative filenames absolute by prepending this
178      * function's return value to them.
179      */
180     char *(*bdrv_dirname)(BlockDriverState *bs, Error **errp);
181 
182     /* aio */
183     BlockAIOCB *(*bdrv_aio_preadv)(BlockDriverState *bs,
184         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags,
185         BlockCompletionFunc *cb, void *opaque);
186     BlockAIOCB *(*bdrv_aio_pwritev)(BlockDriverState *bs,
187         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags,
188         BlockCompletionFunc *cb, void *opaque);
189     BlockAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs,
190         BlockCompletionFunc *cb, void *opaque);
191     BlockAIOCB *(*bdrv_aio_pdiscard)(BlockDriverState *bs,
192         int64_t offset, int bytes,
193         BlockCompletionFunc *cb, void *opaque);
194 
195     int coroutine_fn (*bdrv_co_readv)(BlockDriverState *bs,
196         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov);
197 
198     /**
199      * @offset: position in bytes to read at
200      * @bytes: number of bytes to read
201      * @qiov: the buffers to fill with read data
202      * @flags: currently unused, always 0
203      *
204      * @offset and @bytes will be a multiple of 'request_alignment',
205      * but the length of individual @qiov elements does not have to
206      * be a multiple.
207      *
208      * @bytes will always equal the total size of @qiov, and will be
209      * no larger than 'max_transfer'.
210      *
211      * The buffer in @qiov may point directly to guest memory.
212      */
213     int coroutine_fn (*bdrv_co_preadv)(BlockDriverState *bs,
214         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags);
215     int coroutine_fn (*bdrv_co_preadv_part)(BlockDriverState *bs,
216         uint64_t offset, uint64_t bytes,
217         QEMUIOVector *qiov, size_t qiov_offset, int flags);
218     int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs,
219         int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, int flags);
220     /**
221      * @offset: position in bytes to write at
222      * @bytes: number of bytes to write
223      * @qiov: the buffers containing data to write
224      * @flags: zero or more bits allowed by 'supported_write_flags'
225      *
226      * @offset and @bytes will be a multiple of 'request_alignment',
227      * but the length of individual @qiov elements does not have to
228      * be a multiple.
229      *
230      * @bytes will always equal the total size of @qiov, and will be
231      * no larger than 'max_transfer'.
232      *
233      * The buffer in @qiov may point directly to guest memory.
234      */
235     int coroutine_fn (*bdrv_co_pwritev)(BlockDriverState *bs,
236         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, int flags);
237     int coroutine_fn (*bdrv_co_pwritev_part)(BlockDriverState *bs,
238         uint64_t offset, uint64_t bytes,
239         QEMUIOVector *qiov, size_t qiov_offset, int flags);
240 
241     /*
242      * Efficiently zero a region of the disk image.  Typically an image format
243      * would use a compact metadata representation to implement this.  This
244      * function pointer may be NULL or return -ENOSUP and .bdrv_co_writev()
245      * will be called instead.
246      */
247     int coroutine_fn (*bdrv_co_pwrite_zeroes)(BlockDriverState *bs,
248         int64_t offset, int bytes, BdrvRequestFlags flags);
249     int coroutine_fn (*bdrv_co_pdiscard)(BlockDriverState *bs,
250         int64_t offset, int bytes);
251 
252     /* Map [offset, offset + nbytes) range onto a child of @bs to copy from,
253      * and invoke bdrv_co_copy_range_from(child, ...), or invoke
254      * bdrv_co_copy_range_to() if @bs is the leaf child to copy data from.
255      *
256      * See the comment of bdrv_co_copy_range for the parameter and return value
257      * semantics.
258      */
259     int coroutine_fn (*bdrv_co_copy_range_from)(BlockDriverState *bs,
260                                                 BdrvChild *src,
261                                                 uint64_t offset,
262                                                 BdrvChild *dst,
263                                                 uint64_t dst_offset,
264                                                 uint64_t bytes,
265                                                 BdrvRequestFlags read_flags,
266                                                 BdrvRequestFlags write_flags);
267 
268     /* Map [offset, offset + nbytes) range onto a child of bs to copy data to,
269      * and invoke bdrv_co_copy_range_to(child, src, ...), or perform the copy
270      * operation if @bs is the leaf and @src has the same BlockDriver.  Return
271      * -ENOTSUP if @bs is the leaf but @src has a different BlockDriver.
272      *
273      * See the comment of bdrv_co_copy_range for the parameter and return value
274      * semantics.
275      */
276     int coroutine_fn (*bdrv_co_copy_range_to)(BlockDriverState *bs,
277                                               BdrvChild *src,
278                                               uint64_t src_offset,
279                                               BdrvChild *dst,
280                                               uint64_t dst_offset,
281                                               uint64_t bytes,
282                                               BdrvRequestFlags read_flags,
283                                               BdrvRequestFlags write_flags);
284 
285     /*
286      * Building block for bdrv_block_status[_above] and
287      * bdrv_is_allocated[_above].  The driver should answer only
288      * according to the current layer, and should only need to set
289      * BDRV_BLOCK_DATA, BDRV_BLOCK_ZERO, BDRV_BLOCK_OFFSET_VALID,
290      * and/or BDRV_BLOCK_RAW; if the current layer defers to a backing
291      * layer, the result should be 0 (and not BDRV_BLOCK_ZERO).  See
292      * block.h for the overall meaning of the bits.  As a hint, the
293      * flag want_zero is true if the caller cares more about precise
294      * mappings (favor accurate _OFFSET_VALID/_ZERO) or false for
295      * overall allocation (favor larger *pnum, perhaps by reporting
296      * _DATA instead of _ZERO).  The block layer guarantees input
297      * clamped to bdrv_getlength() and aligned to request_alignment,
298      * as well as non-NULL pnum, map, and file; in turn, the driver
299      * must return an error or set pnum to an aligned non-zero value.
300      */
301     int coroutine_fn (*bdrv_co_block_status)(BlockDriverState *bs,
302         bool want_zero, int64_t offset, int64_t bytes, int64_t *pnum,
303         int64_t *map, BlockDriverState **file);
304 
305     /*
306      * Invalidate any cached meta-data.
307      */
308     void coroutine_fn (*bdrv_co_invalidate_cache)(BlockDriverState *bs,
309                                                   Error **errp);
310     int (*bdrv_inactivate)(BlockDriverState *bs);
311 
312     /*
313      * Flushes all data for all layers by calling bdrv_co_flush for underlying
314      * layers, if needed. This function is needed for deterministic
315      * synchronization of the flush finishing callback.
316      */
317     int coroutine_fn (*bdrv_co_flush)(BlockDriverState *bs);
318 
319     /* Delete a created file. */
320     int coroutine_fn (*bdrv_co_delete_file)(BlockDriverState *bs,
321                                             Error **errp);
322 
323     /*
324      * Flushes all data that was already written to the OS all the way down to
325      * the disk (for example file-posix.c calls fsync()).
326      */
327     int coroutine_fn (*bdrv_co_flush_to_disk)(BlockDriverState *bs);
328 
329     /*
330      * Flushes all internal caches to the OS. The data may still sit in a
331      * writeback cache of the host OS, but it will survive a crash of the qemu
332      * process.
333      */
334     int coroutine_fn (*bdrv_co_flush_to_os)(BlockDriverState *bs);
335 
336     /*
337      * Drivers setting this field must be able to work with just a plain
338      * filename with '<protocol_name>:' as a prefix, and no other options.
339      * Options may be extracted from the filename by implementing
340      * bdrv_parse_filename.
341      */
342     const char *protocol_name;
343 
344     /*
345      * Truncate @bs to @offset bytes using the given @prealloc mode
346      * when growing.  Modes other than PREALLOC_MODE_OFF should be
347      * rejected when shrinking @bs.
348      *
349      * If @exact is true, @bs must be resized to exactly @offset.
350      * Otherwise, it is sufficient for @bs (if it is a host block
351      * device and thus there is no way to resize it) to be at least
352      * @offset bytes in length.
353      *
354      * If @exact is true and this function fails but would succeed
355      * with @exact = false, it should return -ENOTSUP.
356      */
357     int coroutine_fn (*bdrv_co_truncate)(BlockDriverState *bs, int64_t offset,
358                                          bool exact, PreallocMode prealloc,
359                                          BdrvRequestFlags flags, Error **errp);
360 
361     int64_t (*bdrv_getlength)(BlockDriverState *bs);
362     bool has_variable_length;
363     int64_t (*bdrv_get_allocated_file_size)(BlockDriverState *bs);
364     BlockMeasureInfo *(*bdrv_measure)(QemuOpts *opts, BlockDriverState *in_bs,
365                                       Error **errp);
366 
367     int coroutine_fn (*bdrv_co_pwritev_compressed)(BlockDriverState *bs,
368         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov);
369     int coroutine_fn (*bdrv_co_pwritev_compressed_part)(BlockDriverState *bs,
370         uint64_t offset, uint64_t bytes, QEMUIOVector *qiov,
371         size_t qiov_offset);
372 
373     int (*bdrv_snapshot_create)(BlockDriverState *bs,
374                                 QEMUSnapshotInfo *sn_info);
375     int (*bdrv_snapshot_goto)(BlockDriverState *bs,
376                               const char *snapshot_id);
377     int (*bdrv_snapshot_delete)(BlockDriverState *bs,
378                                 const char *snapshot_id,
379                                 const char *name,
380                                 Error **errp);
381     int (*bdrv_snapshot_list)(BlockDriverState *bs,
382                               QEMUSnapshotInfo **psn_info);
383     int (*bdrv_snapshot_load_tmp)(BlockDriverState *bs,
384                                   const char *snapshot_id,
385                                   const char *name,
386                                   Error **errp);
387     int (*bdrv_get_info)(BlockDriverState *bs, BlockDriverInfo *bdi);
388     ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs,
389                                                  Error **errp);
390     BlockStatsSpecific *(*bdrv_get_specific_stats)(BlockDriverState *bs);
391 
392     int coroutine_fn (*bdrv_save_vmstate)(BlockDriverState *bs,
393                                           QEMUIOVector *qiov,
394                                           int64_t pos);
395     int coroutine_fn (*bdrv_load_vmstate)(BlockDriverState *bs,
396                                           QEMUIOVector *qiov,
397                                           int64_t pos);
398 
399     int (*bdrv_change_backing_file)(BlockDriverState *bs,
400         const char *backing_file, const char *backing_fmt);
401 
402     /* removable device specific */
403     bool (*bdrv_is_inserted)(BlockDriverState *bs);
404     void (*bdrv_eject)(BlockDriverState *bs, bool eject_flag);
405     void (*bdrv_lock_medium)(BlockDriverState *bs, bool locked);
406 
407     /* to control generic scsi devices */
408     BlockAIOCB *(*bdrv_aio_ioctl)(BlockDriverState *bs,
409         unsigned long int req, void *buf,
410         BlockCompletionFunc *cb, void *opaque);
411     int coroutine_fn (*bdrv_co_ioctl)(BlockDriverState *bs,
412                                       unsigned long int req, void *buf);
413 
414     /* List of options for creating images, terminated by name == NULL */
415     QemuOptsList *create_opts;
416     /*
417      * If this driver supports reopening images this contains a
418      * NULL-terminated list of the runtime options that can be
419      * modified. If an option in this list is unspecified during
420      * reopen then it _must_ be reset to its default value or return
421      * an error.
422      */
423     const char *const *mutable_opts;
424 
425     /*
426      * Returns 0 for completed check, -errno for internal errors.
427      * The check results are stored in result.
428      */
429     int coroutine_fn (*bdrv_co_check)(BlockDriverState *bs,
430                                       BdrvCheckResult *result,
431                                       BdrvCheckMode fix);
432 
433     int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts,
434                               BlockDriverAmendStatusCB *status_cb,
435                               void *cb_opaque,
436                               Error **errp);
437 
438     void (*bdrv_debug_event)(BlockDriverState *bs, BlkdebugEvent event);
439 
440     /* TODO Better pass a option string/QDict/QemuOpts to add any rule? */
441     int (*bdrv_debug_breakpoint)(BlockDriverState *bs, const char *event,
442         const char *tag);
443     int (*bdrv_debug_remove_breakpoint)(BlockDriverState *bs,
444         const char *tag);
445     int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag);
446     bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag);
447 
448     void (*bdrv_refresh_limits)(BlockDriverState *bs, Error **errp);
449 
450     /*
451      * Returns 1 if newly created images are guaranteed to contain only
452      * zeros, 0 otherwise.
453      */
454     int (*bdrv_has_zero_init)(BlockDriverState *bs);
455 
456     /* Remove fd handlers, timers, and other event loop callbacks so the event
457      * loop is no longer in use.  Called with no in-flight requests and in
458      * depth-first traversal order with parents before child nodes.
459      */
460     void (*bdrv_detach_aio_context)(BlockDriverState *bs);
461 
462     /* Add fd handlers, timers, and other event loop callbacks so I/O requests
463      * can be processed again.  Called with no in-flight requests and in
464      * depth-first traversal order with child nodes before parent nodes.
465      */
466     void (*bdrv_attach_aio_context)(BlockDriverState *bs,
467                                     AioContext *new_context);
468 
469     /* io queue for linux-aio */
470     void (*bdrv_io_plug)(BlockDriverState *bs);
471     void (*bdrv_io_unplug)(BlockDriverState *bs);
472 
473     /**
474      * Try to get @bs's logical and physical block size.
475      * On success, store them in @bsz and return zero.
476      * On failure, return negative errno.
477      */
478     int (*bdrv_probe_blocksizes)(BlockDriverState *bs, BlockSizes *bsz);
479     /**
480      * Try to get @bs's geometry (cyls, heads, sectors)
481      * On success, store them in @geo and return 0.
482      * On failure return -errno.
483      * Only drivers that want to override guest geometry implement this
484      * callback; see hd_geometry_guess().
485      */
486     int (*bdrv_probe_geometry)(BlockDriverState *bs, HDGeometry *geo);
487 
488     /**
489      * bdrv_co_drain_begin is called if implemented in the beginning of a
490      * drain operation to drain and stop any internal sources of requests in
491      * the driver.
492      * bdrv_co_drain_end is called if implemented at the end of the drain.
493      *
494      * They should be used by the driver to e.g. manage scheduled I/O
495      * requests, or toggle an internal state. After the end of the drain new
496      * requests will continue normally.
497      */
498     void coroutine_fn (*bdrv_co_drain_begin)(BlockDriverState *bs);
499     void coroutine_fn (*bdrv_co_drain_end)(BlockDriverState *bs);
500 
501     void (*bdrv_add_child)(BlockDriverState *parent, BlockDriverState *child,
502                            Error **errp);
503     void (*bdrv_del_child)(BlockDriverState *parent, BdrvChild *child,
504                            Error **errp);
505 
506     /**
507      * Informs the block driver that a permission change is intended. The
508      * driver checks whether the change is permissible and may take other
509      * preparations for the change (e.g. get file system locks). This operation
510      * is always followed either by a call to either .bdrv_set_perm or
511      * .bdrv_abort_perm_update.
512      *
513      * Checks whether the requested set of cumulative permissions in @perm
514      * can be granted for accessing @bs and whether no other users are using
515      * permissions other than those given in @shared (both arguments take
516      * BLK_PERM_* bitmasks).
517      *
518      * If both conditions are met, 0 is returned. Otherwise, -errno is returned
519      * and errp is set to an error describing the conflict.
520      */
521     int (*bdrv_check_perm)(BlockDriverState *bs, uint64_t perm,
522                            uint64_t shared, Error **errp);
523 
524     /**
525      * Called to inform the driver that the set of cumulative set of used
526      * permissions for @bs has changed to @perm, and the set of sharable
527      * permission to @shared. The driver can use this to propagate changes to
528      * its children (i.e. request permissions only if a parent actually needs
529      * them).
530      *
531      * This function is only invoked after bdrv_check_perm(), so block drivers
532      * may rely on preparations made in their .bdrv_check_perm implementation.
533      */
534     void (*bdrv_set_perm)(BlockDriverState *bs, uint64_t perm, uint64_t shared);
535 
536     /*
537      * Called to inform the driver that after a previous bdrv_check_perm()
538      * call, the permission update is not performed and any preparations made
539      * for it (e.g. taken file locks) need to be undone.
540      *
541      * This function can be called even for nodes that never saw a
542      * bdrv_check_perm() call. It is a no-op then.
543      */
544     void (*bdrv_abort_perm_update)(BlockDriverState *bs);
545 
546     /**
547      * Returns in @nperm and @nshared the permissions that the driver for @bs
548      * needs on its child @c, based on the cumulative permissions requested by
549      * the parents in @parent_perm and @parent_shared.
550      *
551      * If @c is NULL, return the permissions for attaching a new child for the
552      * given @role.
553      *
554      * If @reopen_queue is non-NULL, don't return the currently needed
555      * permissions, but those that will be needed after applying the
556      * @reopen_queue.
557      */
558      void (*bdrv_child_perm)(BlockDriverState *bs, BdrvChild *c,
559                              const BdrvChildRole *role,
560                              BlockReopenQueue *reopen_queue,
561                              uint64_t parent_perm, uint64_t parent_shared,
562                              uint64_t *nperm, uint64_t *nshared);
563 
564     bool (*bdrv_co_can_store_new_dirty_bitmap)(BlockDriverState *bs,
565                                                const char *name,
566                                                uint32_t granularity,
567                                                Error **errp);
568     int (*bdrv_co_remove_persistent_dirty_bitmap)(BlockDriverState *bs,
569                                                   const char *name,
570                                                   Error **errp);
571 
572     /**
573      * Register/unregister a buffer for I/O. For example, when the driver is
574      * interested to know the memory areas that will later be used in iovs, so
575      * that it can do IOMMU mapping with VFIO etc., in order to get better
576      * performance. In the case of VFIO drivers, this callback is used to do
577      * DMA mapping for hot buffers.
578      */
579     void (*bdrv_register_buf)(BlockDriverState *bs, void *host, size_t size);
580     void (*bdrv_unregister_buf)(BlockDriverState *bs, void *host);
581     QLIST_ENTRY(BlockDriver) list;
582 
583     /* Pointer to a NULL-terminated array of names of strong options
584      * that can be specified for bdrv_open(). A strong option is one
585      * that changes the data of a BDS.
586      * If this pointer is NULL, the array is considered empty.
587      * "filename" and "driver" are always considered strong. */
588     const char *const *strong_runtime_opts;
589 };
590 
591 static inline bool block_driver_can_compress(BlockDriver *drv)
592 {
593     return drv->bdrv_co_pwritev_compressed ||
594            drv->bdrv_co_pwritev_compressed_part;
595 }
596 
597 typedef struct BlockLimits {
598     /* Alignment requirement, in bytes, for offset/length of I/O
599      * requests. Must be a power of 2 less than INT_MAX; defaults to
600      * 1 for drivers with modern byte interfaces, and to 512
601      * otherwise. */
602     uint32_t request_alignment;
603 
604     /* Maximum number of bytes that can be discarded at once (since it
605      * is signed, it must be < 2G, if set). Must be multiple of
606      * pdiscard_alignment, but need not be power of 2. May be 0 if no
607      * inherent 32-bit limit */
608     int32_t max_pdiscard;
609 
610     /* Optimal alignment for discard requests in bytes. A power of 2
611      * is best but not mandatory.  Must be a multiple of
612      * bl.request_alignment, and must be less than max_pdiscard if
613      * that is set. May be 0 if bl.request_alignment is good enough */
614     uint32_t pdiscard_alignment;
615 
616     /* Maximum number of bytes that can zeroized at once (since it is
617      * signed, it must be < 2G, if set). Must be multiple of
618      * pwrite_zeroes_alignment. May be 0 if no inherent 32-bit limit */
619     int32_t max_pwrite_zeroes;
620 
621     /* Optimal alignment for write zeroes requests in bytes. A power
622      * of 2 is best but not mandatory.  Must be a multiple of
623      * bl.request_alignment, and must be less than max_pwrite_zeroes
624      * if that is set. May be 0 if bl.request_alignment is good
625      * enough */
626     uint32_t pwrite_zeroes_alignment;
627 
628     /* Optimal transfer length in bytes.  A power of 2 is best but not
629      * mandatory.  Must be a multiple of bl.request_alignment, or 0 if
630      * no preferred size */
631     uint32_t opt_transfer;
632 
633     /* Maximal transfer length in bytes.  Need not be power of 2, but
634      * must be multiple of opt_transfer and bl.request_alignment, or 0
635      * for no 32-bit limit.  For now, anything larger than INT_MAX is
636      * clamped down. */
637     uint32_t max_transfer;
638 
639     /* memory alignment, in bytes so that no bounce buffer is needed */
640     size_t min_mem_alignment;
641 
642     /* memory alignment, in bytes, for bounce buffer */
643     size_t opt_mem_alignment;
644 
645     /* maximum number of iovec elements */
646     int max_iov;
647 } BlockLimits;
648 
649 typedef struct BdrvOpBlocker BdrvOpBlocker;
650 
651 typedef struct BdrvAioNotifier {
652     void (*attached_aio_context)(AioContext *new_context, void *opaque);
653     void (*detach_aio_context)(void *opaque);
654 
655     void *opaque;
656     bool deleted;
657 
658     QLIST_ENTRY(BdrvAioNotifier) list;
659 } BdrvAioNotifier;
660 
661 struct BdrvChildRole {
662     /* If true, bdrv_replace_node() doesn't change the node this BdrvChild
663      * points to. */
664     bool stay_at_node;
665 
666     /* If true, the parent is a BlockDriverState and bdrv_next_all_states()
667      * will return it. This information is used for drain_all, where every node
668      * will be drained separately, so the drain only needs to be propagated to
669      * non-BDS parents. */
670     bool parent_is_bds;
671 
672     void (*inherit_options)(int *child_flags, QDict *child_options,
673                             int parent_flags, QDict *parent_options);
674 
675     void (*change_media)(BdrvChild *child, bool load);
676     void (*resize)(BdrvChild *child);
677 
678     /* Returns a name that is supposedly more useful for human users than the
679      * node name for identifying the node in question (in particular, a BB
680      * name), or NULL if the parent can't provide a better name. */
681     const char *(*get_name)(BdrvChild *child);
682 
683     /* Returns a malloced string that describes the parent of the child for a
684      * human reader. This could be a node-name, BlockBackend name, qdev ID or
685      * QOM path of the device owning the BlockBackend, job type and ID etc. The
686      * caller is responsible for freeing the memory. */
687     char *(*get_parent_desc)(BdrvChild *child);
688 
689     /*
690      * If this pair of functions is implemented, the parent doesn't issue new
691      * requests after returning from .drained_begin() until .drained_end() is
692      * called.
693      *
694      * These functions must not change the graph (and therefore also must not
695      * call aio_poll(), which could change the graph indirectly).
696      *
697      * If drained_end() schedules background operations, it must atomically
698      * increment *drained_end_counter for each such operation and atomically
699      * decrement it once the operation has settled.
700      *
701      * Note that this can be nested. If drained_begin() was called twice, new
702      * I/O is allowed only after drained_end() was called twice, too.
703      */
704     void (*drained_begin)(BdrvChild *child);
705     void (*drained_end)(BdrvChild *child, int *drained_end_counter);
706 
707     /*
708      * Returns whether the parent has pending requests for the child. This
709      * callback is polled after .drained_begin() has been called until all
710      * activity on the child has stopped.
711      */
712     bool (*drained_poll)(BdrvChild *child);
713 
714     /* Notifies the parent that the child has been activated/inactivated (e.g.
715      * when migration is completing) and it can start/stop requesting
716      * permissions and doing I/O on it. */
717     void (*activate)(BdrvChild *child, Error **errp);
718     int (*inactivate)(BdrvChild *child);
719 
720     void (*attach)(BdrvChild *child);
721     void (*detach)(BdrvChild *child);
722 
723     /* Notifies the parent that the filename of its child has changed (e.g.
724      * because the direct child was removed from the backing chain), so that it
725      * can update its reference. */
726     int (*update_filename)(BdrvChild *child, BlockDriverState *new_base,
727                            const char *filename, Error **errp);
728 
729     bool (*can_set_aio_ctx)(BdrvChild *child, AioContext *ctx,
730                             GSList **ignore, Error **errp);
731     void (*set_aio_ctx)(BdrvChild *child, AioContext *ctx, GSList **ignore);
732 };
733 
734 extern const BdrvChildRole child_file;
735 extern const BdrvChildRole child_format;
736 extern const BdrvChildRole child_backing;
737 
738 struct BdrvChild {
739     BlockDriverState *bs;
740     char *name;
741     const BdrvChildRole *role;
742     void *opaque;
743 
744     /**
745      * Granted permissions for operating on this BdrvChild (BLK_PERM_* bitmask)
746      */
747     uint64_t perm;
748 
749     /**
750      * Permissions that can still be granted to other users of @bs while this
751      * BdrvChild is still attached to it. (BLK_PERM_* bitmask)
752      */
753     uint64_t shared_perm;
754 
755     /* backup of permissions during permission update procedure */
756     bool has_backup_perm;
757     uint64_t backup_perm;
758     uint64_t backup_shared_perm;
759 
760     /*
761      * This link is frozen: the child can neither be replaced nor
762      * detached from the parent.
763      */
764     bool frozen;
765 
766     /*
767      * How many times the parent of this child has been drained
768      * (through role->drained_*).
769      * Usually, this is equal to bs->quiesce_counter (potentially
770      * reduced by bdrv_drain_all_count).  It may differ while the
771      * child is entering or leaving a drained section.
772      */
773     int parent_quiesce_counter;
774 
775     QLIST_ENTRY(BdrvChild) next;
776     QLIST_ENTRY(BdrvChild) next_parent;
777 };
778 
779 /*
780  * Note: the function bdrv_append() copies and swaps contents of
781  * BlockDriverStates, so if you add new fields to this struct, please
782  * inspect bdrv_append() to determine if the new fields need to be
783  * copied as well.
784  */
785 struct BlockDriverState {
786     /* Protected by big QEMU lock or read-only after opening.  No special
787      * locking needed during I/O...
788      */
789     int open_flags; /* flags used to open the file, re-used for re-open */
790     bool read_only; /* if true, the media is read only */
791     bool encrypted; /* if true, the media is encrypted */
792     bool sg;        /* if true, the device is a /dev/sg* */
793     bool probed;    /* if true, format was probed rather than specified */
794     bool force_share; /* if true, always allow all shared permissions */
795     bool implicit;  /* if true, this filter node was automatically inserted */
796 
797     BlockDriver *drv; /* NULL means no media */
798     void *opaque;
799 
800     AioContext *aio_context; /* event loop used for fd handlers, timers, etc */
801     /* long-running tasks intended to always use the same AioContext as this
802      * BDS may register themselves in this list to be notified of changes
803      * regarding this BDS's context */
804     QLIST_HEAD(, BdrvAioNotifier) aio_notifiers;
805     bool walking_aio_notifiers; /* to make removal during iteration safe */
806 
807     char filename[PATH_MAX];
808     char backing_file[PATH_MAX]; /* if non zero, the image is a diff of
809                                     this file image */
810     /* The backing filename indicated by the image header; if we ever
811      * open this file, then this is replaced by the resulting BDS's
812      * filename (i.e. after a bdrv_refresh_filename() run). */
813     char auto_backing_file[PATH_MAX];
814     char backing_format[16]; /* if non-zero and backing_file exists */
815 
816     QDict *full_open_options;
817     char exact_filename[PATH_MAX];
818 
819     BdrvChild *backing;
820     BdrvChild *file;
821 
822     /* I/O Limits */
823     BlockLimits bl;
824 
825     /* Flags honored during pwrite (so far: BDRV_REQ_FUA,
826      * BDRV_REQ_WRITE_UNCHANGED).
827      * If a driver does not support BDRV_REQ_WRITE_UNCHANGED, those
828      * writes will be issued as normal writes without the flag set.
829      * This is important to note for drivers that do not explicitly
830      * request a WRITE permission for their children and instead take
831      * the same permissions as their parent did (this is commonly what
832      * block filters do).  Such drivers have to be aware that the
833      * parent may have taken a WRITE_UNCHANGED permission only and is
834      * issuing such requests.  Drivers either must make sure that
835      * these requests do not result in plain WRITE accesses (usually
836      * by supporting BDRV_REQ_WRITE_UNCHANGED, and then forwarding
837      * every incoming write request as-is, including potentially that
838      * flag), or they have to explicitly take the WRITE permission for
839      * their children. */
840     unsigned int supported_write_flags;
841     /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA,
842      * BDRV_REQ_MAY_UNMAP, BDRV_REQ_WRITE_UNCHANGED) */
843     unsigned int supported_zero_flags;
844     /*
845      * Flags honoured during truncate (so far: BDRV_REQ_ZERO_WRITE).
846      *
847      * If BDRV_REQ_ZERO_WRITE is given, the truncate operation must make sure
848      * that any added space reads as all zeros. If this can't be guaranteed,
849      * the operation must fail.
850      */
851     unsigned int supported_truncate_flags;
852 
853     /* the following member gives a name to every node on the bs graph. */
854     char node_name[32];
855     /* element of the list of named nodes building the graph */
856     QTAILQ_ENTRY(BlockDriverState) node_list;
857     /* element of the list of all BlockDriverStates (all_bdrv_states) */
858     QTAILQ_ENTRY(BlockDriverState) bs_list;
859     /* element of the list of monitor-owned BDS */
860     QTAILQ_ENTRY(BlockDriverState) monitor_list;
861     int refcnt;
862 
863     /* operation blockers */
864     QLIST_HEAD(, BdrvOpBlocker) op_blockers[BLOCK_OP_TYPE_MAX];
865 
866     /* The node that this node inherited default options from (and a reopen on
867      * which can affect this node by changing these defaults). This is always a
868      * parent node of this node. */
869     BlockDriverState *inherits_from;
870     QLIST_HEAD(, BdrvChild) children;
871     QLIST_HEAD(, BdrvChild) parents;
872 
873     QDict *options;
874     QDict *explicit_options;
875     BlockdevDetectZeroesOptions detect_zeroes;
876 
877     /* The error object in use for blocking operations on backing_hd */
878     Error *backing_blocker;
879 
880     /* Protected by AioContext lock */
881 
882     /* If we are reading a disk image, give its size in sectors.
883      * Generally read-only; it is written to by load_snapshot and
884      * save_snaphost, but the block layer is quiescent during those.
885      */
886     int64_t total_sectors;
887 
888     /* Callback before write request is processed */
889     NotifierWithReturnList before_write_notifiers;
890 
891     /* threshold limit for writes, in bytes. "High water mark". */
892     uint64_t write_threshold_offset;
893     NotifierWithReturn write_threshold_notifier;
894 
895     /* Writing to the list requires the BQL _and_ the dirty_bitmap_mutex.
896      * Reading from the list can be done with either the BQL or the
897      * dirty_bitmap_mutex.  Modifying a bitmap only requires
898      * dirty_bitmap_mutex.  */
899     QemuMutex dirty_bitmap_mutex;
900     QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps;
901 
902     /* Offset after the highest byte written to */
903     Stat64 wr_highest_offset;
904 
905     /* If true, copy read backing sectors into image.  Can be >1 if more
906      * than one client has requested copy-on-read.  Accessed with atomic
907      * ops.
908      */
909     int copy_on_read;
910 
911     /* number of in-flight requests; overall and serialising.
912      * Accessed with atomic ops.
913      */
914     unsigned int in_flight;
915     unsigned int serialising_in_flight;
916 
917     /* counter for nested bdrv_io_plug.
918      * Accessed with atomic ops.
919     */
920     unsigned io_plugged;
921 
922     /* do we need to tell the quest if we have a volatile write cache? */
923     int enable_write_cache;
924 
925     /* Accessed with atomic ops.  */
926     int quiesce_counter;
927     int recursive_quiesce_counter;
928 
929     unsigned int write_gen;               /* Current data generation */
930 
931     /* Protected by reqs_lock.  */
932     CoMutex reqs_lock;
933     QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
934     CoQueue flush_queue;                  /* Serializing flush queue */
935     bool active_flush_req;                /* Flush request in flight? */
936 
937     /* Only read/written by whoever has set active_flush_req to true.  */
938     unsigned int flushed_gen;             /* Flushed write generation */
939 
940     /* BdrvChild links to this node may never be frozen */
941     bool never_freeze;
942 };
943 
944 struct BlockBackendRootState {
945     int open_flags;
946     bool read_only;
947     BlockdevDetectZeroesOptions detect_zeroes;
948 };
949 
950 typedef enum BlockMirrorBackingMode {
951     /* Reuse the existing backing chain from the source for the target.
952      * - sync=full: Set backing BDS to NULL.
953      * - sync=top:  Use source's backing BDS.
954      * - sync=none: Use source as the backing BDS. */
955     MIRROR_SOURCE_BACKING_CHAIN,
956 
957     /* Open the target's backing chain completely anew */
958     MIRROR_OPEN_BACKING_CHAIN,
959 
960     /* Do not change the target's backing BDS after job completion */
961     MIRROR_LEAVE_BACKING_CHAIN,
962 } BlockMirrorBackingMode;
963 
964 static inline BlockDriverState *backing_bs(BlockDriverState *bs)
965 {
966     return bs->backing ? bs->backing->bs : NULL;
967 }
968 
969 
970 /* Essential block drivers which must always be statically linked into qemu, and
971  * which therefore can be accessed without using bdrv_find_format() */
972 extern BlockDriver bdrv_file;
973 extern BlockDriver bdrv_raw;
974 extern BlockDriver bdrv_qcow2;
975 
976 int coroutine_fn bdrv_co_preadv(BdrvChild *child,
977     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
978     BdrvRequestFlags flags);
979 int coroutine_fn bdrv_co_preadv_part(BdrvChild *child,
980     int64_t offset, unsigned int bytes,
981     QEMUIOVector *qiov, size_t qiov_offset, BdrvRequestFlags flags);
982 int coroutine_fn bdrv_co_pwritev(BdrvChild *child,
983     int64_t offset, unsigned int bytes, QEMUIOVector *qiov,
984     BdrvRequestFlags flags);
985 int coroutine_fn bdrv_co_pwritev_part(BdrvChild *child,
986     int64_t offset, unsigned int bytes,
987     QEMUIOVector *qiov, size_t qiov_offset, BdrvRequestFlags flags);
988 
989 static inline int coroutine_fn bdrv_co_pread(BdrvChild *child,
990     int64_t offset, unsigned int bytes, void *buf, BdrvRequestFlags flags)
991 {
992     QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes);
993 
994     return bdrv_co_preadv(child, offset, bytes, &qiov, flags);
995 }
996 
997 static inline int coroutine_fn bdrv_co_pwrite(BdrvChild *child,
998     int64_t offset, unsigned int bytes, void *buf, BdrvRequestFlags flags)
999 {
1000     QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, buf, bytes);
1001 
1002     return bdrv_co_pwritev(child, offset, bytes, &qiov, flags);
1003 }
1004 
1005 extern unsigned int bdrv_drain_all_count;
1006 void bdrv_apply_subtree_drain(BdrvChild *child, BlockDriverState *new_parent);
1007 void bdrv_unapply_subtree_drain(BdrvChild *child, BlockDriverState *old_parent);
1008 
1009 bool coroutine_fn bdrv_mark_request_serialising(BdrvTrackedRequest *req, uint64_t align);
1010 BdrvTrackedRequest *coroutine_fn bdrv_co_get_self_request(BlockDriverState *bs);
1011 
1012 int get_tmp_filename(char *filename, int size);
1013 BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size,
1014                             const char *filename);
1015 
1016 void bdrv_parse_filename_strip_prefix(const char *filename, const char *prefix,
1017                                       QDict *options);
1018 
1019 
1020 /**
1021  * bdrv_add_before_write_notifier:
1022  *
1023  * Register a callback that is invoked before write requests are processed but
1024  * after any throttling or waiting for overlapping requests.
1025  */
1026 void bdrv_add_before_write_notifier(BlockDriverState *bs,
1027                                     NotifierWithReturn *notifier);
1028 
1029 /**
1030  * bdrv_add_aio_context_notifier:
1031  *
1032  * If a long-running job intends to be always run in the same AioContext as a
1033  * certain BDS, it may use this function to be notified of changes regarding the
1034  * association of the BDS to an AioContext.
1035  *
1036  * attached_aio_context() is called after the target BDS has been attached to a
1037  * new AioContext; detach_aio_context() is called before the target BDS is being
1038  * detached from its old AioContext.
1039  */
1040 void bdrv_add_aio_context_notifier(BlockDriverState *bs,
1041         void (*attached_aio_context)(AioContext *new_context, void *opaque),
1042         void (*detach_aio_context)(void *opaque), void *opaque);
1043 
1044 /**
1045  * bdrv_remove_aio_context_notifier:
1046  *
1047  * Unsubscribe of change notifications regarding the BDS's AioContext. The
1048  * parameters given here have to be the same as those given to
1049  * bdrv_add_aio_context_notifier().
1050  */
1051 void bdrv_remove_aio_context_notifier(BlockDriverState *bs,
1052                                       void (*aio_context_attached)(AioContext *,
1053                                                                    void *),
1054                                       void (*aio_context_detached)(void *),
1055                                       void *opaque);
1056 
1057 /**
1058  * bdrv_wakeup:
1059  * @bs: The BlockDriverState for which an I/O operation has been completed.
1060  *
1061  * Wake up the main thread if it is waiting on BDRV_POLL_WHILE.  During
1062  * synchronous I/O on a BlockDriverState that is attached to another
1063  * I/O thread, the main thread lets the I/O thread's event loop run,
1064  * waiting for the I/O operation to complete.  A bdrv_wakeup will wake
1065  * up the main thread if necessary.
1066  *
1067  * Manual calls to bdrv_wakeup are rarely necessary, because
1068  * bdrv_dec_in_flight already calls it.
1069  */
1070 void bdrv_wakeup(BlockDriverState *bs);
1071 
1072 #ifdef _WIN32
1073 int is_windows_drive(const char *filename);
1074 #endif
1075 
1076 /**
1077  * stream_start:
1078  * @job_id: The id of the newly-created job, or %NULL to use the
1079  * device name of @bs.
1080  * @bs: Block device to operate on.
1081  * @base: Block device that will become the new base, or %NULL to
1082  * flatten the whole backing file chain onto @bs.
1083  * @backing_file_str: The file name that will be written to @bs as the
1084  * the new backing file if the job completes. Ignored if @base is %NULL.
1085  * @creation_flags: Flags that control the behavior of the Job lifetime.
1086  *                  See @BlockJobCreateFlags
1087  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1088  * @on_error: The action to take upon error.
1089  * @errp: Error object.
1090  *
1091  * Start a streaming operation on @bs.  Clusters that are unallocated
1092  * in @bs, but allocated in any image between @base and @bs (both
1093  * exclusive) will be written to @bs.  At the end of a successful
1094  * streaming job, the backing file of @bs will be changed to
1095  * @backing_file_str in the written image and to @base in the live
1096  * BlockDriverState.
1097  */
1098 void stream_start(const char *job_id, BlockDriverState *bs,
1099                   BlockDriverState *base, const char *backing_file_str,
1100                   int creation_flags, int64_t speed,
1101                   BlockdevOnError on_error, Error **errp);
1102 
1103 /**
1104  * commit_start:
1105  * @job_id: The id of the newly-created job, or %NULL to use the
1106  * device name of @bs.
1107  * @bs: Active block device.
1108  * @top: Top block device to be committed.
1109  * @base: Block device that will be written into, and become the new top.
1110  * @creation_flags: Flags that control the behavior of the Job lifetime.
1111  *                  See @BlockJobCreateFlags
1112  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1113  * @on_error: The action to take upon error.
1114  * @backing_file_str: String to use as the backing file in @top's overlay
1115  * @filter_node_name: The node name that should be assigned to the filter
1116  * driver that the commit job inserts into the graph above @top. NULL means
1117  * that a node name should be autogenerated.
1118  * @errp: Error object.
1119  *
1120  */
1121 void commit_start(const char *job_id, BlockDriverState *bs,
1122                   BlockDriverState *base, BlockDriverState *top,
1123                   int creation_flags, int64_t speed,
1124                   BlockdevOnError on_error, const char *backing_file_str,
1125                   const char *filter_node_name, Error **errp);
1126 /**
1127  * commit_active_start:
1128  * @job_id: The id of the newly-created job, or %NULL to use the
1129  * device name of @bs.
1130  * @bs: Active block device to be committed.
1131  * @base: Block device that will be written into, and become the new top.
1132  * @creation_flags: Flags that control the behavior of the Job lifetime.
1133  *                  See @BlockJobCreateFlags
1134  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1135  * @on_error: The action to take upon error.
1136  * @filter_node_name: The node name that should be assigned to the filter
1137  * driver that the commit job inserts into the graph above @bs. NULL means that
1138  * a node name should be autogenerated.
1139  * @cb: Completion function for the job.
1140  * @opaque: Opaque pointer value passed to @cb.
1141  * @auto_complete: Auto complete the job.
1142  * @errp: Error object.
1143  *
1144  */
1145 BlockJob *commit_active_start(const char *job_id, BlockDriverState *bs,
1146                               BlockDriverState *base, int creation_flags,
1147                               int64_t speed, BlockdevOnError on_error,
1148                               const char *filter_node_name,
1149                               BlockCompletionFunc *cb, void *opaque,
1150                               bool auto_complete, Error **errp);
1151 /*
1152  * mirror_start:
1153  * @job_id: The id of the newly-created job, or %NULL to use the
1154  * device name of @bs.
1155  * @bs: Block device to operate on.
1156  * @target: Block device to write to.
1157  * @replaces: Block graph node name to replace once the mirror is done. Can
1158  *            only be used when full mirroring is selected.
1159  * @creation_flags: Flags that control the behavior of the Job lifetime.
1160  *                  See @BlockJobCreateFlags
1161  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1162  * @granularity: The chosen granularity for the dirty bitmap.
1163  * @buf_size: The amount of data that can be in flight at one time.
1164  * @mode: Whether to collapse all images in the chain to the target.
1165  * @backing_mode: How to establish the target's backing chain after completion.
1166  * @zero_target: Whether the target should be explicitly zero-initialized
1167  * @on_source_error: The action to take upon error reading from the source.
1168  * @on_target_error: The action to take upon error writing to the target.
1169  * @unmap: Whether to unmap target where source sectors only contain zeroes.
1170  * @filter_node_name: The node name that should be assigned to the filter
1171  * driver that the mirror job inserts into the graph above @bs. NULL means that
1172  * a node name should be autogenerated.
1173  * @copy_mode: When to trigger writes to the target.
1174  * @errp: Error object.
1175  *
1176  * Start a mirroring operation on @bs.  Clusters that are allocated
1177  * in @bs will be written to @target until the job is cancelled or
1178  * manually completed.  At the end of a successful mirroring job,
1179  * @bs will be switched to read from @target.
1180  */
1181 void mirror_start(const char *job_id, BlockDriverState *bs,
1182                   BlockDriverState *target, const char *replaces,
1183                   int creation_flags, int64_t speed,
1184                   uint32_t granularity, int64_t buf_size,
1185                   MirrorSyncMode mode, BlockMirrorBackingMode backing_mode,
1186                   bool zero_target,
1187                   BlockdevOnError on_source_error,
1188                   BlockdevOnError on_target_error,
1189                   bool unmap, const char *filter_node_name,
1190                   MirrorCopyMode copy_mode, Error **errp);
1191 
1192 /*
1193  * backup_job_create:
1194  * @job_id: The id of the newly-created job, or %NULL to use the
1195  * device name of @bs.
1196  * @bs: Block device to operate on.
1197  * @target: Block device to write to.
1198  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
1199  * @sync_mode: What parts of the disk image should be copied to the destination.
1200  * @sync_bitmap: The dirty bitmap if sync_mode is 'bitmap' or 'incremental'
1201  * @bitmap_mode: The bitmap synchronization policy to use.
1202  * @on_source_error: The action to take upon error reading from the source.
1203  * @on_target_error: The action to take upon error writing to the target.
1204  * @creation_flags: Flags that control the behavior of the Job lifetime.
1205  *                  See @BlockJobCreateFlags
1206  * @cb: Completion function for the job.
1207  * @opaque: Opaque pointer value passed to @cb.
1208  * @txn: Transaction that this job is part of (may be NULL).
1209  *
1210  * Create a backup operation on @bs.  Clusters in @bs are written to @target
1211  * until the job is cancelled or manually completed.
1212  */
1213 BlockJob *backup_job_create(const char *job_id, BlockDriverState *bs,
1214                             BlockDriverState *target, int64_t speed,
1215                             MirrorSyncMode sync_mode,
1216                             BdrvDirtyBitmap *sync_bitmap,
1217                             BitmapSyncMode bitmap_mode,
1218                             bool compress,
1219                             const char *filter_node_name,
1220                             BlockdevOnError on_source_error,
1221                             BlockdevOnError on_target_error,
1222                             int creation_flags,
1223                             BlockCompletionFunc *cb, void *opaque,
1224                             JobTxn *txn, Error **errp);
1225 
1226 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
1227                                   const char *child_name,
1228                                   const BdrvChildRole *child_role,
1229                                   AioContext *ctx,
1230                                   uint64_t perm, uint64_t shared_perm,
1231                                   void *opaque, Error **errp);
1232 void bdrv_root_unref_child(BdrvChild *child);
1233 
1234 void bdrv_get_cumulative_perm(BlockDriverState *bs, uint64_t *perm,
1235                               uint64_t *shared_perm);
1236 
1237 /**
1238  * Sets a BdrvChild's permissions.  Avoid if the parent is a BDS; use
1239  * bdrv_child_refresh_perms() instead and make the parent's
1240  * .bdrv_child_perm() implementation return the correct values.
1241  */
1242 int bdrv_child_try_set_perm(BdrvChild *c, uint64_t perm, uint64_t shared,
1243                             Error **errp);
1244 
1245 /**
1246  * Calls bs->drv->bdrv_child_perm() and updates the child's permission
1247  * masks with the result.
1248  * Drivers should invoke this function whenever an event occurs that
1249  * makes their .bdrv_child_perm() implementation return different
1250  * values than before, but which will not result in the block layer
1251  * automatically refreshing the permissions.
1252  */
1253 int bdrv_child_refresh_perms(BlockDriverState *bs, BdrvChild *c, Error **errp);
1254 
1255 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1256  * block filters: Forward CONSISTENT_READ, WRITE, WRITE_UNCHANGED and RESIZE to
1257  * all children */
1258 void bdrv_filter_default_perms(BlockDriverState *bs, BdrvChild *c,
1259                                const BdrvChildRole *role,
1260                                BlockReopenQueue *reopen_queue,
1261                                uint64_t perm, uint64_t shared,
1262                                uint64_t *nperm, uint64_t *nshared);
1263 
1264 /* Default implementation for BlockDriver.bdrv_child_perm() that can be used by
1265  * (non-raw) image formats: Like above for bs->backing, but for bs->file it
1266  * requires WRITE | RESIZE for read-write images, always requires
1267  * CONSISTENT_READ and doesn't share WRITE. */
1268 void bdrv_format_default_perms(BlockDriverState *bs, BdrvChild *c,
1269                                const BdrvChildRole *role,
1270                                BlockReopenQueue *reopen_queue,
1271                                uint64_t perm, uint64_t shared,
1272                                uint64_t *nperm, uint64_t *nshared);
1273 
1274 bool bdrv_recurse_can_replace(BlockDriverState *bs,
1275                               BlockDriverState *to_replace);
1276 
1277 /*
1278  * Default implementation for drivers to pass bdrv_co_block_status() to
1279  * their file.
1280  */
1281 int coroutine_fn bdrv_co_block_status_from_file(BlockDriverState *bs,
1282                                                 bool want_zero,
1283                                                 int64_t offset,
1284                                                 int64_t bytes,
1285                                                 int64_t *pnum,
1286                                                 int64_t *map,
1287                                                 BlockDriverState **file);
1288 /*
1289  * Default implementation for drivers to pass bdrv_co_block_status() to
1290  * their backing file.
1291  */
1292 int coroutine_fn bdrv_co_block_status_from_backing(BlockDriverState *bs,
1293                                                    bool want_zero,
1294                                                    int64_t offset,
1295                                                    int64_t bytes,
1296                                                    int64_t *pnum,
1297                                                    int64_t *map,
1298                                                    BlockDriverState **file);
1299 const char *bdrv_get_parent_name(const BlockDriverState *bs);
1300 void blk_dev_change_media_cb(BlockBackend *blk, bool load, Error **errp);
1301 bool blk_dev_has_removable_media(BlockBackend *blk);
1302 bool blk_dev_has_tray(BlockBackend *blk);
1303 void blk_dev_eject_request(BlockBackend *blk, bool force);
1304 bool blk_dev_is_tray_open(BlockBackend *blk);
1305 bool blk_dev_is_medium_locked(BlockBackend *blk);
1306 
1307 void bdrv_set_dirty(BlockDriverState *bs, int64_t offset, int64_t bytes);
1308 
1309 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap **out);
1310 void bdrv_restore_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap *backup);
1311 bool bdrv_dirty_bitmap_merge_internal(BdrvDirtyBitmap *dest,
1312                                       const BdrvDirtyBitmap *src,
1313                                       HBitmap **backup, bool lock);
1314 
1315 void bdrv_inc_in_flight(BlockDriverState *bs);
1316 void bdrv_dec_in_flight(BlockDriverState *bs);
1317 
1318 void blockdev_close_all_bdrv_states(void);
1319 
1320 int coroutine_fn bdrv_co_copy_range_from(BdrvChild *src, uint64_t src_offset,
1321                                          BdrvChild *dst, uint64_t dst_offset,
1322                                          uint64_t bytes,
1323                                          BdrvRequestFlags read_flags,
1324                                          BdrvRequestFlags write_flags);
1325 int coroutine_fn bdrv_co_copy_range_to(BdrvChild *src, uint64_t src_offset,
1326                                        BdrvChild *dst, uint64_t dst_offset,
1327                                        uint64_t bytes,
1328                                        BdrvRequestFlags read_flags,
1329                                        BdrvRequestFlags write_flags);
1330 
1331 int refresh_total_sectors(BlockDriverState *bs, int64_t hint);
1332 
1333 void bdrv_set_monitor_owned(BlockDriverState *bs);
1334 BlockDriverState *bds_tree_init(QDict *bs_opts, Error **errp);
1335 
1336 /**
1337  * Simple implementation of bdrv_co_create_opts for protocol drivers
1338  * which only support creation via opening a file
1339  * (usually existing raw storage device)
1340  */
1341 int coroutine_fn bdrv_co_create_opts_simple(BlockDriver *drv,
1342                                             const char *filename,
1343                                             QemuOpts *opts,
1344                                             Error **errp);
1345 extern QemuOptsList bdrv_create_opts_simple;
1346 
1347 #endif /* BLOCK_INT_H */
1348