xref: /openbmc/qemu/include/block/block.h (revision c63ca4ff)
1 #ifndef BLOCK_H
2 #define BLOCK_H
3 
4 #include "block/aio.h"
5 #include "block/aio-wait.h"
6 #include "qemu/iov.h"
7 #include "qemu/coroutine.h"
8 #include "block/accounting.h"
9 #include "block/dirty-bitmap.h"
10 #include "block/blockjob.h"
11 #include "qemu/hbitmap.h"
12 
13 /*
14  * generated_co_wrapper
15  *
16  * Function specifier, which does nothing but mark functions to be
17  * generated by scripts/block-coroutine-wrapper.py
18  *
19  * Read more in docs/devel/block-coroutine-wrapper.rst
20  */
21 #define generated_co_wrapper
22 
23 /* block.c */
24 typedef struct BlockDriver BlockDriver;
25 typedef struct BdrvChild BdrvChild;
26 typedef struct BdrvChildClass BdrvChildClass;
27 
28 typedef struct BlockDriverInfo {
29     /* in bytes, 0 if irrelevant */
30     int cluster_size;
31     /* offset at which the VM state can be saved (0 if not possible) */
32     int64_t vm_state_offset;
33     bool is_dirty;
34     /*
35      * True if this block driver only supports compressed writes
36      */
37     bool needs_compressed_writes;
38 } BlockDriverInfo;
39 
40 typedef struct BlockFragInfo {
41     uint64_t allocated_clusters;
42     uint64_t total_clusters;
43     uint64_t fragmented_clusters;
44     uint64_t compressed_clusters;
45 } BlockFragInfo;
46 
47 typedef enum {
48     BDRV_REQ_COPY_ON_READ       = 0x1,
49     BDRV_REQ_ZERO_WRITE         = 0x2,
50 
51     /*
52      * The BDRV_REQ_MAY_UNMAP flag is used in write_zeroes requests to indicate
53      * that the block driver should unmap (discard) blocks if it is guaranteed
54      * that the result will read back as zeroes. The flag is only passed to the
55      * driver if the block device is opened with BDRV_O_UNMAP.
56      */
57     BDRV_REQ_MAY_UNMAP          = 0x4,
58 
59     BDRV_REQ_FUA                = 0x10,
60     BDRV_REQ_WRITE_COMPRESSED   = 0x20,
61 
62     /* Signifies that this write request will not change the visible disk
63      * content. */
64     BDRV_REQ_WRITE_UNCHANGED    = 0x40,
65 
66     /*
67      * BDRV_REQ_SERIALISING forces request serialisation for writes.
68      * It is used to ensure that writes to the backing file of a backup process
69      * target cannot race with a read of the backup target that defers to the
70      * backing file.
71      *
72      * Note, that BDRV_REQ_SERIALISING is _not_ opposite in meaning to
73      * BDRV_REQ_NO_SERIALISING. A more descriptive name for the latter might be
74      * _DO_NOT_WAIT_FOR_SERIALISING, except that is too long.
75      */
76     BDRV_REQ_SERIALISING        = 0x80,
77 
78     /* Execute the request only if the operation can be offloaded or otherwise
79      * be executed efficiently, but return an error instead of using a slow
80      * fallback. */
81     BDRV_REQ_NO_FALLBACK        = 0x100,
82 
83     /*
84      * BDRV_REQ_PREFETCH may be used only together with BDRV_REQ_COPY_ON_READ
85      * on read request and means that caller doesn't really need data to be
86      * written to qiov parameter which may be NULL.
87      */
88     BDRV_REQ_PREFETCH  = 0x200,
89     /* Mask of valid flags */
90     BDRV_REQ_MASK               = 0x3ff,
91 } BdrvRequestFlags;
92 
93 typedef struct BlockSizes {
94     uint32_t phys;
95     uint32_t log;
96 } BlockSizes;
97 
98 typedef struct HDGeometry {
99     uint32_t heads;
100     uint32_t sectors;
101     uint32_t cylinders;
102 } HDGeometry;
103 
104 #define BDRV_O_RDWR        0x0002
105 #define BDRV_O_RESIZE      0x0004 /* request permission for resizing the node */
106 #define BDRV_O_SNAPSHOT    0x0008 /* open the file read only and save writes in a snapshot */
107 #define BDRV_O_TEMPORARY   0x0010 /* delete the file after use */
108 #define BDRV_O_NOCACHE     0x0020 /* do not use the host page cache */
109 #define BDRV_O_NATIVE_AIO  0x0080 /* use native AIO instead of the thread pool */
110 #define BDRV_O_NO_BACKING  0x0100 /* don't open the backing file */
111 #define BDRV_O_NO_FLUSH    0x0200 /* disable flushing on this disk */
112 #define BDRV_O_COPY_ON_READ 0x0400 /* copy read backing sectors into image */
113 #define BDRV_O_INACTIVE    0x0800  /* consistency hint for migration handoff */
114 #define BDRV_O_CHECK       0x1000  /* open solely for consistency check */
115 #define BDRV_O_ALLOW_RDWR  0x2000  /* allow reopen to change from r/o to r/w */
116 #define BDRV_O_UNMAP       0x4000  /* execute guest UNMAP/TRIM operations */
117 #define BDRV_O_PROTOCOL    0x8000  /* if no block driver is explicitly given:
118                                       select an appropriate protocol driver,
119                                       ignoring the format layer */
120 #define BDRV_O_NO_IO       0x10000 /* don't initialize for I/O */
121 #define BDRV_O_AUTO_RDONLY 0x20000 /* degrade to read-only if opening read-write fails */
122 #define BDRV_O_IO_URING    0x40000 /* use io_uring instead of the thread pool */
123 
124 #define BDRV_O_CACHE_MASK  (BDRV_O_NOCACHE | BDRV_O_NO_FLUSH)
125 
126 
127 /* Option names of options parsed by the block layer */
128 
129 #define BDRV_OPT_CACHE_WB       "cache.writeback"
130 #define BDRV_OPT_CACHE_DIRECT   "cache.direct"
131 #define BDRV_OPT_CACHE_NO_FLUSH "cache.no-flush"
132 #define BDRV_OPT_READ_ONLY      "read-only"
133 #define BDRV_OPT_AUTO_READ_ONLY "auto-read-only"
134 #define BDRV_OPT_DISCARD        "discard"
135 #define BDRV_OPT_FORCE_SHARE    "force-share"
136 
137 
138 #define BDRV_SECTOR_BITS   9
139 #define BDRV_SECTOR_SIZE   (1ULL << BDRV_SECTOR_BITS)
140 
141 #define BDRV_REQUEST_MAX_SECTORS MIN_CONST(SIZE_MAX >> BDRV_SECTOR_BITS, \
142                                            INT_MAX >> BDRV_SECTOR_BITS)
143 #define BDRV_REQUEST_MAX_BYTES (BDRV_REQUEST_MAX_SECTORS << BDRV_SECTOR_BITS)
144 
145 /*
146  * We want allow aligning requests and disk length up to any 32bit alignment
147  * and don't afraid of overflow.
148  * To achieve it, and in the same time use some pretty number as maximum disk
149  * size, let's define maximum "length" (a limit for any offset/bytes request and
150  * for disk size) to be the greatest power of 2 less than INT64_MAX.
151  */
152 #define BDRV_MAX_ALIGNMENT (1L << 30)
153 #define BDRV_MAX_LENGTH (QEMU_ALIGN_DOWN(INT64_MAX, BDRV_MAX_ALIGNMENT))
154 
155 /*
156  * Allocation status flags for bdrv_block_status() and friends.
157  *
158  * Public flags:
159  * BDRV_BLOCK_DATA: allocation for data at offset is tied to this layer
160  * BDRV_BLOCK_ZERO: offset reads as zero
161  * BDRV_BLOCK_OFFSET_VALID: an associated offset exists for accessing raw data
162  * BDRV_BLOCK_ALLOCATED: the content of the block is determined by this
163  *                       layer rather than any backing, set by block layer
164  * BDRV_BLOCK_EOF: the returned pnum covers through end of file for this
165  *                 layer, set by block layer
166  *
167  * Internal flags:
168  * BDRV_BLOCK_RAW: for use by passthrough drivers, such as raw, to request
169  *                 that the block layer recompute the answer from the returned
170  *                 BDS; must be accompanied by just BDRV_BLOCK_OFFSET_VALID.
171  * BDRV_BLOCK_RECURSE: request that the block layer will recursively search for
172  *                     zeroes in file child of current block node inside
173  *                     returned region. Only valid together with both
174  *                     BDRV_BLOCK_DATA and BDRV_BLOCK_OFFSET_VALID. Should not
175  *                     appear with BDRV_BLOCK_ZERO.
176  *
177  * If BDRV_BLOCK_OFFSET_VALID is set, the map parameter represents the
178  * host offset within the returned BDS that is allocated for the
179  * corresponding raw guest data.  However, whether that offset
180  * actually contains data also depends on BDRV_BLOCK_DATA, as follows:
181  *
182  * DATA ZERO OFFSET_VALID
183  *  t    t        t       sectors read as zero, returned file is zero at offset
184  *  t    f        t       sectors read as valid from file at offset
185  *  f    t        t       sectors preallocated, read as zero, returned file not
186  *                        necessarily zero at offset
187  *  f    f        t       sectors preallocated but read from backing_hd,
188  *                        returned file contains garbage at offset
189  *  t    t        f       sectors preallocated, read as zero, unknown offset
190  *  t    f        f       sectors read from unknown file or offset
191  *  f    t        f       not allocated or unknown offset, read as zero
192  *  f    f        f       not allocated or unknown offset, read from backing_hd
193  */
194 #define BDRV_BLOCK_DATA         0x01
195 #define BDRV_BLOCK_ZERO         0x02
196 #define BDRV_BLOCK_OFFSET_VALID 0x04
197 #define BDRV_BLOCK_RAW          0x08
198 #define BDRV_BLOCK_ALLOCATED    0x10
199 #define BDRV_BLOCK_EOF          0x20
200 #define BDRV_BLOCK_RECURSE      0x40
201 
202 typedef QTAILQ_HEAD(BlockReopenQueue, BlockReopenQueueEntry) BlockReopenQueue;
203 
204 typedef struct BDRVReopenState {
205     BlockDriverState *bs;
206     int flags;
207     BlockdevDetectZeroesOptions detect_zeroes;
208     bool backing_missing;
209     bool replace_backing_bs;  /* new_backing_bs is ignored if this is false */
210     BlockDriverState *new_backing_bs; /* If NULL then detach the current bs */
211     uint64_t perm, shared_perm;
212     QDict *options;
213     QDict *explicit_options;
214     void *opaque;
215 } BDRVReopenState;
216 
217 /*
218  * Block operation types
219  */
220 typedef enum BlockOpType {
221     BLOCK_OP_TYPE_BACKUP_SOURCE,
222     BLOCK_OP_TYPE_BACKUP_TARGET,
223     BLOCK_OP_TYPE_CHANGE,
224     BLOCK_OP_TYPE_COMMIT_SOURCE,
225     BLOCK_OP_TYPE_COMMIT_TARGET,
226     BLOCK_OP_TYPE_DATAPLANE,
227     BLOCK_OP_TYPE_DRIVE_DEL,
228     BLOCK_OP_TYPE_EJECT,
229     BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
230     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT,
231     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE,
232     BLOCK_OP_TYPE_MIRROR_SOURCE,
233     BLOCK_OP_TYPE_MIRROR_TARGET,
234     BLOCK_OP_TYPE_RESIZE,
235     BLOCK_OP_TYPE_STREAM,
236     BLOCK_OP_TYPE_REPLACE,
237     BLOCK_OP_TYPE_MAX,
238 } BlockOpType;
239 
240 /* Block node permission constants */
241 enum {
242     /**
243      * A user that has the "permission" of consistent reads is guaranteed that
244      * their view of the contents of the block device is complete and
245      * self-consistent, representing the contents of a disk at a specific
246      * point.
247      *
248      * For most block devices (including their backing files) this is true, but
249      * the property cannot be maintained in a few situations like for
250      * intermediate nodes of a commit block job.
251      */
252     BLK_PERM_CONSISTENT_READ    = 0x01,
253 
254     /** This permission is required to change the visible disk contents. */
255     BLK_PERM_WRITE              = 0x02,
256 
257     /**
258      * This permission (which is weaker than BLK_PERM_WRITE) is both enough and
259      * required for writes to the block node when the caller promises that
260      * the visible disk content doesn't change.
261      *
262      * As the BLK_PERM_WRITE permission is strictly stronger, either is
263      * sufficient to perform an unchanging write.
264      */
265     BLK_PERM_WRITE_UNCHANGED    = 0x04,
266 
267     /** This permission is required to change the size of a block node. */
268     BLK_PERM_RESIZE             = 0x08,
269 
270     /**
271      * This permission is required to change the node that this BdrvChild
272      * points to.
273      */
274     BLK_PERM_GRAPH_MOD          = 0x10,
275 
276     BLK_PERM_ALL                = 0x1f,
277 
278     DEFAULT_PERM_PASSTHROUGH    = BLK_PERM_CONSISTENT_READ
279                                  | BLK_PERM_WRITE
280                                  | BLK_PERM_WRITE_UNCHANGED
281                                  | BLK_PERM_RESIZE,
282 
283     DEFAULT_PERM_UNCHANGED      = BLK_PERM_ALL & ~DEFAULT_PERM_PASSTHROUGH,
284 };
285 
286 /*
287  * Flags that parent nodes assign to child nodes to specify what kind of
288  * role(s) they take.
289  *
290  * At least one of DATA, METADATA, FILTERED, or COW must be set for
291  * every child.
292  */
293 enum BdrvChildRoleBits {
294     /*
295      * This child stores data.
296      * Any node may have an arbitrary number of such children.
297      */
298     BDRV_CHILD_DATA         = (1 << 0),
299 
300     /*
301      * This child stores metadata.
302      * Any node may have an arbitrary number of metadata-storing
303      * children.
304      */
305     BDRV_CHILD_METADATA     = (1 << 1),
306 
307     /*
308      * A child that always presents exactly the same visible data as
309      * the parent, e.g. by virtue of the parent forwarding all reads
310      * and writes.
311      * This flag is mutually exclusive with DATA, METADATA, and COW.
312      * Any node may have at most one filtered child at a time.
313      */
314     BDRV_CHILD_FILTERED     = (1 << 2),
315 
316     /*
317      * Child from which to read all data that isn't allocated in the
318      * parent (i.e., the backing child); such data is copied to the
319      * parent through COW (and optionally COR).
320      * This field is mutually exclusive with DATA, METADATA, and
321      * FILTERED.
322      * Any node may have at most one such backing child at a time.
323      */
324     BDRV_CHILD_COW          = (1 << 3),
325 
326     /*
327      * The primary child.  For most drivers, this is the child whose
328      * filename applies best to the parent node.
329      * Any node may have at most one primary child at a time.
330      */
331     BDRV_CHILD_PRIMARY      = (1 << 4),
332 
333     /* Useful combination of flags */
334     BDRV_CHILD_IMAGE        = BDRV_CHILD_DATA
335                               | BDRV_CHILD_METADATA
336                               | BDRV_CHILD_PRIMARY,
337 };
338 
339 /* Mask of BdrvChildRoleBits values */
340 typedef unsigned int BdrvChildRole;
341 
342 char *bdrv_perm_names(uint64_t perm);
343 uint64_t bdrv_qapi_perm_to_blk_perm(BlockPermission qapi_perm);
344 
345 /* disk I/O throttling */
346 void bdrv_init(void);
347 void bdrv_init_with_whitelist(void);
348 bool bdrv_uses_whitelist(void);
349 int bdrv_is_whitelisted(BlockDriver *drv, bool read_only);
350 BlockDriver *bdrv_find_protocol(const char *filename,
351                                 bool allow_protocol_prefix,
352                                 Error **errp);
353 BlockDriver *bdrv_find_format(const char *format_name);
354 int bdrv_create(BlockDriver *drv, const char* filename,
355                 QemuOpts *opts, Error **errp);
356 int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp);
357 
358 BlockDriverState *bdrv_new(void);
359 void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top,
360                  Error **errp);
361 void bdrv_replace_node(BlockDriverState *from, BlockDriverState *to,
362                        Error **errp);
363 
364 int bdrv_parse_aio(const char *mode, int *flags);
365 int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough);
366 int bdrv_parse_discard_flags(const char *mode, int *flags);
367 BdrvChild *bdrv_open_child(const char *filename,
368                            QDict *options, const char *bdref_key,
369                            BlockDriverState* parent,
370                            const BdrvChildClass *child_class,
371                            BdrvChildRole child_role,
372                            bool allow_none, Error **errp);
373 BlockDriverState *bdrv_open_blockdev_ref(BlockdevRef *ref, Error **errp);
374 void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd,
375                          Error **errp);
376 int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options,
377                            const char *bdref_key, Error **errp);
378 BlockDriverState *bdrv_open(const char *filename, const char *reference,
379                             QDict *options, int flags, Error **errp);
380 BlockDriverState *bdrv_new_open_driver(BlockDriver *drv, const char *node_name,
381                                        int flags, Error **errp);
382 BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue,
383                                     BlockDriverState *bs, QDict *options,
384                                     bool keep_old_opts);
385 int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp);
386 int bdrv_reopen_set_read_only(BlockDriverState *bs, bool read_only,
387                               Error **errp);
388 int bdrv_reopen_prepare(BDRVReopenState *reopen_state,
389                         BlockReopenQueue *queue, Error **errp);
390 void bdrv_reopen_commit(BDRVReopenState *reopen_state);
391 void bdrv_reopen_abort(BDRVReopenState *reopen_state);
392 int bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset,
393                        int bytes, BdrvRequestFlags flags);
394 int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags);
395 int bdrv_pread(BdrvChild *child, int64_t offset, void *buf, int bytes);
396 int bdrv_pwrite(BdrvChild *child, int64_t offset, const void *buf, int bytes);
397 int bdrv_pwrite_sync(BdrvChild *child, int64_t offset,
398                      const void *buf, int count);
399 /*
400  * Efficiently zero a region of the disk image.  Note that this is a regular
401  * I/O request like read or write and should have a reasonable size.  This
402  * function is not suitable for zeroing the entire image in a single request
403  * because it may allocate memory for the entire region.
404  */
405 int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset,
406                                        int bytes, BdrvRequestFlags flags);
407 BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,
408     const char *backing_file);
409 void bdrv_refresh_filename(BlockDriverState *bs);
410 
411 int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset, bool exact,
412                                   PreallocMode prealloc, BdrvRequestFlags flags,
413                                   Error **errp);
414 int generated_co_wrapper
415 bdrv_truncate(BdrvChild *child, int64_t offset, bool exact,
416               PreallocMode prealloc, BdrvRequestFlags flags, Error **errp);
417 
418 int64_t bdrv_nb_sectors(BlockDriverState *bs);
419 int64_t bdrv_getlength(BlockDriverState *bs);
420 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs);
421 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
422                                BlockDriverState *in_bs, Error **errp);
423 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr);
424 void bdrv_refresh_limits(BlockDriverState *bs, Error **errp);
425 int bdrv_commit(BlockDriverState *bs);
426 int bdrv_make_empty(BdrvChild *c, Error **errp);
427 int bdrv_change_backing_file(BlockDriverState *bs, const char *backing_file,
428                              const char *backing_fmt, bool warn);
429 void bdrv_register(BlockDriver *bdrv);
430 int bdrv_drop_intermediate(BlockDriverState *top, BlockDriverState *base,
431                            const char *backing_file_str);
432 BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
433                                     BlockDriverState *bs);
434 BlockDriverState *bdrv_find_base(BlockDriverState *bs);
435 bool bdrv_is_backing_chain_frozen(BlockDriverState *bs, BlockDriverState *base,
436                                   Error **errp);
437 int bdrv_freeze_backing_chain(BlockDriverState *bs, BlockDriverState *base,
438                               Error **errp);
439 void bdrv_unfreeze_backing_chain(BlockDriverState *bs, BlockDriverState *base);
440 int coroutine_fn bdrv_co_delete_file(BlockDriverState *bs, Error **errp);
441 
442 
443 typedef struct BdrvCheckResult {
444     int corruptions;
445     int leaks;
446     int check_errors;
447     int corruptions_fixed;
448     int leaks_fixed;
449     int64_t image_end_offset;
450     BlockFragInfo bfi;
451 } BdrvCheckResult;
452 
453 typedef enum {
454     BDRV_FIX_LEAKS    = 1,
455     BDRV_FIX_ERRORS   = 2,
456 } BdrvCheckMode;
457 
458 int generated_co_wrapper bdrv_check(BlockDriverState *bs, BdrvCheckResult *res,
459                                     BdrvCheckMode fix);
460 
461 /* The units of offset and total_work_size may be chosen arbitrarily by the
462  * block driver; total_work_size may change during the course of the amendment
463  * operation */
464 typedef void BlockDriverAmendStatusCB(BlockDriverState *bs, int64_t offset,
465                                       int64_t total_work_size, void *opaque);
466 int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts,
467                        BlockDriverAmendStatusCB *status_cb, void *cb_opaque,
468                        bool force,
469                        Error **errp);
470 
471 /* check if a named node can be replaced when doing drive-mirror */
472 BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs,
473                                         const char *node_name, Error **errp);
474 
475 /* async block I/O */
476 void bdrv_aio_cancel(BlockAIOCB *acb);
477 void bdrv_aio_cancel_async(BlockAIOCB *acb);
478 
479 /* sg packet commands */
480 int bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf);
481 
482 /* Invalidate any cached metadata used by image formats */
483 int generated_co_wrapper bdrv_invalidate_cache(BlockDriverState *bs,
484                                                Error **errp);
485 void bdrv_invalidate_cache_all(Error **errp);
486 int bdrv_inactivate_all(void);
487 
488 /* Ensure contents are flushed to disk.  */
489 int generated_co_wrapper bdrv_flush(BlockDriverState *bs);
490 int coroutine_fn bdrv_co_flush(BlockDriverState *bs);
491 int bdrv_flush_all(void);
492 void bdrv_close_all(void);
493 void bdrv_drain(BlockDriverState *bs);
494 void coroutine_fn bdrv_co_drain(BlockDriverState *bs);
495 void bdrv_drain_all_begin(void);
496 void bdrv_drain_all_end(void);
497 void bdrv_drain_all(void);
498 
499 #define BDRV_POLL_WHILE(bs, cond) ({                       \
500     BlockDriverState *bs_ = (bs);                          \
501     AIO_WAIT_WHILE(bdrv_get_aio_context(bs_),              \
502                    cond); })
503 
504 int generated_co_wrapper bdrv_pdiscard(BdrvChild *child, int64_t offset,
505                                        int64_t bytes);
506 int bdrv_co_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes);
507 int bdrv_has_zero_init_1(BlockDriverState *bs);
508 int bdrv_has_zero_init(BlockDriverState *bs);
509 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs);
510 int bdrv_block_status(BlockDriverState *bs, int64_t offset,
511                       int64_t bytes, int64_t *pnum, int64_t *map,
512                       BlockDriverState **file);
513 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base,
514                             int64_t offset, int64_t bytes, int64_t *pnum,
515                             int64_t *map, BlockDriverState **file);
516 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
517                       int64_t *pnum);
518 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
519                             bool include_base, int64_t offset, int64_t bytes,
520                             int64_t *pnum);
521 int coroutine_fn bdrv_co_is_zero_fast(BlockDriverState *bs, int64_t offset,
522                                       int64_t bytes);
523 
524 bool bdrv_is_read_only(BlockDriverState *bs);
525 int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only,
526                            bool ignore_allow_rdw, Error **errp);
527 int bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg,
528                               Error **errp);
529 bool bdrv_is_writable(BlockDriverState *bs);
530 bool bdrv_is_sg(BlockDriverState *bs);
531 bool bdrv_is_inserted(BlockDriverState *bs);
532 void bdrv_lock_medium(BlockDriverState *bs, bool locked);
533 void bdrv_eject(BlockDriverState *bs, bool eject_flag);
534 const char *bdrv_get_format_name(BlockDriverState *bs);
535 BlockDriverState *bdrv_find_node(const char *node_name);
536 BlockDeviceInfoList *bdrv_named_nodes_list(bool flat, Error **errp);
537 XDbgBlockGraph *bdrv_get_xdbg_block_graph(Error **errp);
538 BlockDriverState *bdrv_lookup_bs(const char *device,
539                                  const char *node_name,
540                                  Error **errp);
541 bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base);
542 BlockDriverState *bdrv_next_node(BlockDriverState *bs);
543 BlockDriverState *bdrv_next_all_states(BlockDriverState *bs);
544 
545 typedef struct BdrvNextIterator {
546     enum {
547         BDRV_NEXT_BACKEND_ROOTS,
548         BDRV_NEXT_MONITOR_OWNED,
549     } phase;
550     BlockBackend *blk;
551     BlockDriverState *bs;
552 } BdrvNextIterator;
553 
554 BlockDriverState *bdrv_first(BdrvNextIterator *it);
555 BlockDriverState *bdrv_next(BdrvNextIterator *it);
556 void bdrv_next_cleanup(BdrvNextIterator *it);
557 
558 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs);
559 bool bdrv_supports_compressed_writes(BlockDriverState *bs);
560 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
561                          void *opaque, bool read_only);
562 const char *bdrv_get_node_name(const BlockDriverState *bs);
563 const char *bdrv_get_device_name(const BlockDriverState *bs);
564 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
565 int bdrv_get_flags(BlockDriverState *bs);
566 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
567 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs,
568                                           Error **errp);
569 BlockStatsSpecific *bdrv_get_specific_stats(BlockDriverState *bs);
570 void bdrv_round_to_clusters(BlockDriverState *bs,
571                             int64_t offset, int64_t bytes,
572                             int64_t *cluster_offset,
573                             int64_t *cluster_bytes);
574 
575 void bdrv_get_backing_filename(BlockDriverState *bs,
576                                char *filename, int filename_size);
577 char *bdrv_get_full_backing_filename(BlockDriverState *bs, Error **errp);
578 char *bdrv_get_full_backing_filename_from_filename(const char *backed,
579                                                    const char *backing,
580                                                    Error **errp);
581 char *bdrv_dirname(BlockDriverState *bs, Error **errp);
582 
583 int path_has_protocol(const char *path);
584 int path_is_absolute(const char *path);
585 char *path_combine(const char *base_path, const char *filename);
586 
587 int generated_co_wrapper
588 bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
589 int generated_co_wrapper
590 bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
591 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
592                       int64_t pos, int size);
593 
594 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
595                       int64_t pos, int size);
596 
597 void bdrv_img_create(const char *filename, const char *fmt,
598                      const char *base_filename, const char *base_fmt,
599                      char *options, uint64_t img_size, int flags,
600                      bool quiet, Error **errp);
601 
602 /* Returns the alignment in bytes that is required so that no bounce buffer
603  * is required throughout the stack */
604 size_t bdrv_min_mem_align(BlockDriverState *bs);
605 /* Returns optimal alignment in bytes for bounce buffer */
606 size_t bdrv_opt_mem_align(BlockDriverState *bs);
607 void *qemu_blockalign(BlockDriverState *bs, size_t size);
608 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
609 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
610 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
611 bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov);
612 
613 void bdrv_enable_copy_on_read(BlockDriverState *bs);
614 void bdrv_disable_copy_on_read(BlockDriverState *bs);
615 
616 void bdrv_ref(BlockDriverState *bs);
617 void bdrv_unref(BlockDriverState *bs);
618 void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child);
619 BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
620                              BlockDriverState *child_bs,
621                              const char *child_name,
622                              const BdrvChildClass *child_class,
623                              BdrvChildRole child_role,
624                              Error **errp);
625 
626 bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp);
627 void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason);
628 void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason);
629 void bdrv_op_block_all(BlockDriverState *bs, Error *reason);
630 void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason);
631 bool bdrv_op_blocker_is_empty(BlockDriverState *bs);
632 
633 #define BLKDBG_EVENT(child, evt) \
634     do { \
635         if (child) { \
636             bdrv_debug_event(child->bs, evt); \
637         } \
638     } while (0)
639 
640 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
641 
642 int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event,
643                            const char *tag);
644 int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag);
645 int bdrv_debug_resume(BlockDriverState *bs, const char *tag);
646 bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag);
647 
648 /**
649  * bdrv_get_aio_context:
650  *
651  * Returns: the currently bound #AioContext
652  */
653 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
654 
655 /**
656  * Move the current coroutine to the AioContext of @bs and return the old
657  * AioContext of the coroutine. Increase bs->in_flight so that draining @bs
658  * will wait for the operation to proceed until the corresponding
659  * bdrv_co_leave().
660  *
661  * Consequently, you can't call drain inside a bdrv_co_enter/leave() section as
662  * this will deadlock.
663  */
664 AioContext *coroutine_fn bdrv_co_enter(BlockDriverState *bs);
665 
666 /**
667  * Ends a section started by bdrv_co_enter(). Move the current coroutine back
668  * to old_ctx and decrease bs->in_flight again.
669  */
670 void coroutine_fn bdrv_co_leave(BlockDriverState *bs, AioContext *old_ctx);
671 
672 /**
673  * Locks the AioContext of @bs if it's not the current AioContext. This avoids
674  * double locking which could lead to deadlocks: This is a coroutine_fn, so we
675  * know we already own the lock of the current AioContext.
676  *
677  * May only be called in the main thread.
678  */
679 void coroutine_fn bdrv_co_lock(BlockDriverState *bs);
680 
681 /**
682  * Unlocks the AioContext of @bs if it's not the current AioContext.
683  */
684 void coroutine_fn bdrv_co_unlock(BlockDriverState *bs);
685 
686 /**
687  * Transfer control to @co in the aio context of @bs
688  */
689 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co);
690 
691 void bdrv_set_aio_context_ignore(BlockDriverState *bs,
692                                  AioContext *new_context, GSList **ignore);
693 int bdrv_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
694                              Error **errp);
695 int bdrv_child_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
696                                    BdrvChild *ignore_child, Error **errp);
697 bool bdrv_child_can_set_aio_context(BdrvChild *c, AioContext *ctx,
698                                     GSList **ignore, Error **errp);
699 bool bdrv_can_set_aio_context(BlockDriverState *bs, AioContext *ctx,
700                               GSList **ignore, Error **errp);
701 int bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz);
702 int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo);
703 
704 void bdrv_io_plug(BlockDriverState *bs);
705 void bdrv_io_unplug(BlockDriverState *bs);
706 
707 /**
708  * bdrv_parent_drained_begin_single:
709  *
710  * Begin a quiesced section for the parent of @c. If @poll is true, wait for
711  * any pending activity to cease.
712  */
713 void bdrv_parent_drained_begin_single(BdrvChild *c, bool poll);
714 
715 /**
716  * bdrv_parent_drained_end_single:
717  *
718  * End a quiesced section for the parent of @c.
719  *
720  * This polls @bs's AioContext until all scheduled sub-drained_ends
721  * have settled, which may result in graph changes.
722  */
723 void bdrv_parent_drained_end_single(BdrvChild *c);
724 
725 /**
726  * bdrv_drain_poll:
727  *
728  * Poll for pending requests in @bs, its parents (except for @ignore_parent),
729  * and if @recursive is true its children as well (used for subtree drain).
730  *
731  * If @ignore_bds_parents is true, parents that are BlockDriverStates must
732  * ignore the drain request because they will be drained separately (used for
733  * drain_all).
734  *
735  * This is part of bdrv_drained_begin.
736  */
737 bool bdrv_drain_poll(BlockDriverState *bs, bool recursive,
738                      BdrvChild *ignore_parent, bool ignore_bds_parents);
739 
740 /**
741  * bdrv_drained_begin:
742  *
743  * Begin a quiesced section for exclusive access to the BDS, by disabling
744  * external request sources including NBD server and device model. Note that
745  * this doesn't block timers or coroutines from submitting more requests, which
746  * means block_job_pause is still necessary.
747  *
748  * This function can be recursive.
749  */
750 void bdrv_drained_begin(BlockDriverState *bs);
751 
752 /**
753  * bdrv_do_drained_begin_quiesce:
754  *
755  * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already
756  * running requests to complete.
757  */
758 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs,
759                                    BdrvChild *parent, bool ignore_bds_parents);
760 
761 /**
762  * Like bdrv_drained_begin, but recursively begins a quiesced section for
763  * exclusive access to all child nodes as well.
764  */
765 void bdrv_subtree_drained_begin(BlockDriverState *bs);
766 
767 /**
768  * bdrv_drained_end:
769  *
770  * End a quiescent section started by bdrv_drained_begin().
771  *
772  * This polls @bs's AioContext until all scheduled sub-drained_ends
773  * have settled.  On one hand, that may result in graph changes.  On
774  * the other, this requires that the caller either runs in the main
775  * loop; or that all involved nodes (@bs and all of its parents) are
776  * in the caller's AioContext.
777  */
778 void bdrv_drained_end(BlockDriverState *bs);
779 
780 /**
781  * bdrv_drained_end_no_poll:
782  *
783  * Same as bdrv_drained_end(), but do not poll for the subgraph to
784  * actually become unquiesced.  Therefore, no graph changes will occur
785  * with this function.
786  *
787  * *drained_end_counter is incremented for every background operation
788  * that is scheduled, and will be decremented for every operation once
789  * it settles.  The caller must poll until it reaches 0.  The counter
790  * should be accessed using atomic operations only.
791  */
792 void bdrv_drained_end_no_poll(BlockDriverState *bs, int *drained_end_counter);
793 
794 /**
795  * End a quiescent section started by bdrv_subtree_drained_begin().
796  */
797 void bdrv_subtree_drained_end(BlockDriverState *bs);
798 
799 void bdrv_add_child(BlockDriverState *parent, BlockDriverState *child,
800                     Error **errp);
801 void bdrv_del_child(BlockDriverState *parent, BdrvChild *child, Error **errp);
802 
803 bool bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
804                                      uint32_t granularity, Error **errp);
805 /**
806  *
807  * bdrv_register_buf/bdrv_unregister_buf:
808  *
809  * Register/unregister a buffer for I/O. For example, VFIO drivers are
810  * interested to know the memory areas that would later be used for I/O, so
811  * that they can prepare IOMMU mapping etc., to get better performance.
812  */
813 void bdrv_register_buf(BlockDriverState *bs, void *host, size_t size);
814 void bdrv_unregister_buf(BlockDriverState *bs, void *host);
815 
816 /**
817  *
818  * bdrv_co_copy_range:
819  *
820  * Do offloaded copy between two children. If the operation is not implemented
821  * by the driver, or if the backend storage doesn't support it, a negative
822  * error code will be returned.
823  *
824  * Note: block layer doesn't emulate or fallback to a bounce buffer approach
825  * because usually the caller shouldn't attempt offloaded copy any more (e.g.
826  * calling copy_file_range(2)) after the first error, thus it should fall back
827  * to a read+write path in the caller level.
828  *
829  * @src: Source child to copy data from
830  * @src_offset: offset in @src image to read data
831  * @dst: Destination child to copy data to
832  * @dst_offset: offset in @dst image to write data
833  * @bytes: number of bytes to copy
834  * @flags: request flags. Supported flags:
835  *         BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero
836  *                               write on @dst as if bdrv_co_pwrite_zeroes is
837  *                               called. Used to simplify caller code, or
838  *                               during BlockDriver.bdrv_co_copy_range_from()
839  *                               recursion.
840  *         BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping
841  *                                   requests currently in flight.
842  *
843  * Returns: 0 if succeeded; negative error code if failed.
844  **/
845 int coroutine_fn bdrv_co_copy_range(BdrvChild *src, uint64_t src_offset,
846                                     BdrvChild *dst, uint64_t dst_offset,
847                                     uint64_t bytes, BdrvRequestFlags read_flags,
848                                     BdrvRequestFlags write_flags);
849 #endif
850