xref: /openbmc/qemu/include/block/block.h (revision 2a53cff4)
1 #ifndef BLOCK_H
2 #define BLOCK_H
3 
4 #include "block/aio.h"
5 #include "qapi/qapi-types-block-core.h"
6 #include "block/aio-wait.h"
7 #include "qemu/iov.h"
8 #include "qemu/coroutine.h"
9 #include "block/accounting.h"
10 #include "block/dirty-bitmap.h"
11 #include "block/blockjob.h"
12 #include "qemu/hbitmap.h"
13 
14 /* block.c */
15 typedef struct BlockDriver BlockDriver;
16 typedef struct BdrvChild BdrvChild;
17 typedef struct BdrvChildRole BdrvChildRole;
18 
19 typedef struct BlockDriverInfo {
20     /* in bytes, 0 if irrelevant */
21     int cluster_size;
22     /* offset at which the VM state can be saved (0 if not possible) */
23     int64_t vm_state_offset;
24     bool is_dirty;
25     /*
26      * True if unallocated blocks read back as zeroes. This is equivalent
27      * to the LBPRZ flag in the SCSI logical block provisioning page.
28      */
29     bool unallocated_blocks_are_zero;
30     /*
31      * True if this block driver only supports compressed writes
32      */
33     bool needs_compressed_writes;
34 } BlockDriverInfo;
35 
36 typedef struct BlockFragInfo {
37     uint64_t allocated_clusters;
38     uint64_t total_clusters;
39     uint64_t fragmented_clusters;
40     uint64_t compressed_clusters;
41 } BlockFragInfo;
42 
43 typedef enum {
44     BDRV_REQ_COPY_ON_READ       = 0x1,
45     BDRV_REQ_ZERO_WRITE         = 0x2,
46 
47     /*
48      * The BDRV_REQ_MAY_UNMAP flag is used in write_zeroes requests to indicate
49      * that the block driver should unmap (discard) blocks if it is guaranteed
50      * that the result will read back as zeroes. The flag is only passed to the
51      * driver if the block device is opened with BDRV_O_UNMAP.
52      */
53     BDRV_REQ_MAY_UNMAP          = 0x4,
54 
55     /*
56      * The BDRV_REQ_NO_SERIALISING flag is only valid for reads and means that
57      * we don't want wait_serialising_requests() during the read operation.
58      *
59      * This flag is used for backup copy-on-write operations, when we need to
60      * read old data before write (write notifier triggered). It is okay since
61      * we already waited for other serializing requests in the initiating write
62      * (see bdrv_aligned_pwritev), and it is necessary if the initiating write
63      * is already serializing (without the flag, the read would deadlock
64      * waiting for the serialising write to complete).
65      */
66     BDRV_REQ_NO_SERIALISING     = 0x8,
67     BDRV_REQ_FUA                = 0x10,
68     BDRV_REQ_WRITE_COMPRESSED   = 0x20,
69 
70     /* Signifies that this write request will not change the visible disk
71      * content. */
72     BDRV_REQ_WRITE_UNCHANGED    = 0x40,
73 
74     /*
75      * BDRV_REQ_SERIALISING forces request serialisation for writes.
76      * It is used to ensure that writes to the backing file of a backup process
77      * target cannot race with a read of the backup target that defers to the
78      * backing file.
79      *
80      * Note, that BDRV_REQ_SERIALISING is _not_ opposite in meaning to
81      * BDRV_REQ_NO_SERIALISING. A more descriptive name for the latter might be
82      * _DO_NOT_WAIT_FOR_SERIALISING, except that is too long.
83      */
84     BDRV_REQ_SERIALISING        = 0x80,
85 
86     /* Mask of valid flags */
87     BDRV_REQ_MASK               = 0xff,
88 } BdrvRequestFlags;
89 
90 typedef struct BlockSizes {
91     uint32_t phys;
92     uint32_t log;
93 } BlockSizes;
94 
95 typedef struct HDGeometry {
96     uint32_t heads;
97     uint32_t sectors;
98     uint32_t cylinders;
99 } HDGeometry;
100 
101 #define BDRV_O_RDWR        0x0002
102 #define BDRV_O_RESIZE      0x0004 /* request permission for resizing the node */
103 #define BDRV_O_SNAPSHOT    0x0008 /* open the file read only and save writes in a snapshot */
104 #define BDRV_O_TEMPORARY   0x0010 /* delete the file after use */
105 #define BDRV_O_NOCACHE     0x0020 /* do not use the host page cache */
106 #define BDRV_O_NATIVE_AIO  0x0080 /* use native AIO instead of the thread pool */
107 #define BDRV_O_NO_BACKING  0x0100 /* don't open the backing file */
108 #define BDRV_O_NO_FLUSH    0x0200 /* disable flushing on this disk */
109 #define BDRV_O_COPY_ON_READ 0x0400 /* copy read backing sectors into image */
110 #define BDRV_O_INACTIVE    0x0800  /* consistency hint for migration handoff */
111 #define BDRV_O_CHECK       0x1000  /* open solely for consistency check */
112 #define BDRV_O_ALLOW_RDWR  0x2000  /* allow reopen to change from r/o to r/w */
113 #define BDRV_O_UNMAP       0x4000  /* execute guest UNMAP/TRIM operations */
114 #define BDRV_O_PROTOCOL    0x8000  /* if no block driver is explicitly given:
115                                       select an appropriate protocol driver,
116                                       ignoring the format layer */
117 #define BDRV_O_NO_IO       0x10000 /* don't initialize for I/O */
118 #define BDRV_O_AUTO_RDONLY 0x20000 /* degrade to read-only if opening read-write fails */
119 
120 #define BDRV_O_CACHE_MASK  (BDRV_O_NOCACHE | BDRV_O_NO_FLUSH)
121 
122 
123 /* Option names of options parsed by the block layer */
124 
125 #define BDRV_OPT_CACHE_WB       "cache.writeback"
126 #define BDRV_OPT_CACHE_DIRECT   "cache.direct"
127 #define BDRV_OPT_CACHE_NO_FLUSH "cache.no-flush"
128 #define BDRV_OPT_READ_ONLY      "read-only"
129 #define BDRV_OPT_AUTO_READ_ONLY "auto-read-only"
130 #define BDRV_OPT_DISCARD        "discard"
131 #define BDRV_OPT_FORCE_SHARE    "force-share"
132 
133 
134 #define BDRV_SECTOR_BITS   9
135 #define BDRV_SECTOR_SIZE   (1ULL << BDRV_SECTOR_BITS)
136 #define BDRV_SECTOR_MASK   ~(BDRV_SECTOR_SIZE - 1)
137 
138 #define BDRV_REQUEST_MAX_SECTORS MIN(SIZE_MAX >> BDRV_SECTOR_BITS, \
139                                      INT_MAX >> BDRV_SECTOR_BITS)
140 #define BDRV_REQUEST_MAX_BYTES (BDRV_REQUEST_MAX_SECTORS << BDRV_SECTOR_BITS)
141 
142 /*
143  * Allocation status flags for bdrv_block_status() and friends.
144  *
145  * Public flags:
146  * BDRV_BLOCK_DATA: allocation for data at offset is tied to this layer
147  * BDRV_BLOCK_ZERO: offset reads as zero
148  * BDRV_BLOCK_OFFSET_VALID: an associated offset exists for accessing raw data
149  * BDRV_BLOCK_ALLOCATED: the content of the block is determined by this
150  *                       layer rather than any backing, set by block layer
151  * BDRV_BLOCK_EOF: the returned pnum covers through end of file for this
152  *                 layer, set by block layer
153  *
154  * Internal flag:
155  * BDRV_BLOCK_RAW: for use by passthrough drivers, such as raw, to request
156  *                 that the block layer recompute the answer from the returned
157  *                 BDS; must be accompanied by just BDRV_BLOCK_OFFSET_VALID.
158  *
159  * If BDRV_BLOCK_OFFSET_VALID is set, the map parameter represents the
160  * host offset within the returned BDS that is allocated for the
161  * corresponding raw guest data.  However, whether that offset
162  * actually contains data also depends on BDRV_BLOCK_DATA, as follows:
163  *
164  * DATA ZERO OFFSET_VALID
165  *  t    t        t       sectors read as zero, returned file is zero at offset
166  *  t    f        t       sectors read as valid from file at offset
167  *  f    t        t       sectors preallocated, read as zero, returned file not
168  *                        necessarily zero at offset
169  *  f    f        t       sectors preallocated but read from backing_hd,
170  *                        returned file contains garbage at offset
171  *  t    t        f       sectors preallocated, read as zero, unknown offset
172  *  t    f        f       sectors read from unknown file or offset
173  *  f    t        f       not allocated or unknown offset, read as zero
174  *  f    f        f       not allocated or unknown offset, read from backing_hd
175  */
176 #define BDRV_BLOCK_DATA         0x01
177 #define BDRV_BLOCK_ZERO         0x02
178 #define BDRV_BLOCK_OFFSET_VALID 0x04
179 #define BDRV_BLOCK_RAW          0x08
180 #define BDRV_BLOCK_ALLOCATED    0x10
181 #define BDRV_BLOCK_EOF          0x20
182 #define BDRV_BLOCK_OFFSET_MASK  BDRV_SECTOR_MASK
183 
184 typedef QSIMPLEQ_HEAD(BlockReopenQueue, BlockReopenQueueEntry) BlockReopenQueue;
185 
186 typedef struct BDRVReopenState {
187     BlockDriverState *bs;
188     int flags;
189     BlockdevDetectZeroesOptions detect_zeroes;
190     uint64_t perm, shared_perm;
191     QDict *options;
192     QDict *explicit_options;
193     void *opaque;
194 } BDRVReopenState;
195 
196 /*
197  * Block operation types
198  */
199 typedef enum BlockOpType {
200     BLOCK_OP_TYPE_BACKUP_SOURCE,
201     BLOCK_OP_TYPE_BACKUP_TARGET,
202     BLOCK_OP_TYPE_CHANGE,
203     BLOCK_OP_TYPE_COMMIT_SOURCE,
204     BLOCK_OP_TYPE_COMMIT_TARGET,
205     BLOCK_OP_TYPE_DATAPLANE,
206     BLOCK_OP_TYPE_DRIVE_DEL,
207     BLOCK_OP_TYPE_EJECT,
208     BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
209     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT,
210     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE,
211     BLOCK_OP_TYPE_MIRROR_SOURCE,
212     BLOCK_OP_TYPE_MIRROR_TARGET,
213     BLOCK_OP_TYPE_RESIZE,
214     BLOCK_OP_TYPE_STREAM,
215     BLOCK_OP_TYPE_REPLACE,
216     BLOCK_OP_TYPE_MAX,
217 } BlockOpType;
218 
219 /* Block node permission constants */
220 enum {
221     /**
222      * A user that has the "permission" of consistent reads is guaranteed that
223      * their view of the contents of the block device is complete and
224      * self-consistent, representing the contents of a disk at a specific
225      * point.
226      *
227      * For most block devices (including their backing files) this is true, but
228      * the property cannot be maintained in a few situations like for
229      * intermediate nodes of a commit block job.
230      */
231     BLK_PERM_CONSISTENT_READ    = 0x01,
232 
233     /** This permission is required to change the visible disk contents. */
234     BLK_PERM_WRITE              = 0x02,
235 
236     /**
237      * This permission (which is weaker than BLK_PERM_WRITE) is both enough and
238      * required for writes to the block node when the caller promises that
239      * the visible disk content doesn't change.
240      *
241      * As the BLK_PERM_WRITE permission is strictly stronger, either is
242      * sufficient to perform an unchanging write.
243      */
244     BLK_PERM_WRITE_UNCHANGED    = 0x04,
245 
246     /** This permission is required to change the size of a block node. */
247     BLK_PERM_RESIZE             = 0x08,
248 
249     /**
250      * This permission is required to change the node that this BdrvChild
251      * points to.
252      */
253     BLK_PERM_GRAPH_MOD          = 0x10,
254 
255     BLK_PERM_ALL                = 0x1f,
256 
257     DEFAULT_PERM_PASSTHROUGH    = BLK_PERM_CONSISTENT_READ
258                                  | BLK_PERM_WRITE
259                                  | BLK_PERM_WRITE_UNCHANGED
260                                  | BLK_PERM_RESIZE,
261 
262     DEFAULT_PERM_UNCHANGED      = BLK_PERM_ALL & ~DEFAULT_PERM_PASSTHROUGH,
263 };
264 
265 char *bdrv_perm_names(uint64_t perm);
266 
267 /* disk I/O throttling */
268 void bdrv_init(void);
269 void bdrv_init_with_whitelist(void);
270 bool bdrv_uses_whitelist(void);
271 int bdrv_is_whitelisted(BlockDriver *drv, bool read_only);
272 BlockDriver *bdrv_find_protocol(const char *filename,
273                                 bool allow_protocol_prefix,
274                                 Error **errp);
275 BlockDriver *bdrv_find_format(const char *format_name);
276 int bdrv_create(BlockDriver *drv, const char* filename,
277                 QemuOpts *opts, Error **errp);
278 int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp);
279 BlockDriverState *bdrv_new(void);
280 void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top,
281                  Error **errp);
282 void bdrv_replace_node(BlockDriverState *from, BlockDriverState *to,
283                        Error **errp);
284 
285 int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough);
286 int bdrv_parse_discard_flags(const char *mode, int *flags);
287 BdrvChild *bdrv_open_child(const char *filename,
288                            QDict *options, const char *bdref_key,
289                            BlockDriverState* parent,
290                            const BdrvChildRole *child_role,
291                            bool allow_none, Error **errp);
292 BlockDriverState *bdrv_open_blockdev_ref(BlockdevRef *ref, Error **errp);
293 void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd,
294                          Error **errp);
295 int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options,
296                            const char *bdref_key, Error **errp);
297 BlockDriverState *bdrv_open(const char *filename, const char *reference,
298                             QDict *options, int flags, Error **errp);
299 BlockDriverState *bdrv_new_open_driver(BlockDriver *drv, const char *node_name,
300                                        int flags, Error **errp);
301 BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue,
302                                     BlockDriverState *bs, QDict *options);
303 int bdrv_reopen_multiple(AioContext *ctx, BlockReopenQueue *bs_queue, Error **errp);
304 int bdrv_reopen_set_read_only(BlockDriverState *bs, bool read_only,
305                               Error **errp);
306 int bdrv_reopen_prepare(BDRVReopenState *reopen_state,
307                         BlockReopenQueue *queue, Error **errp);
308 void bdrv_reopen_commit(BDRVReopenState *reopen_state);
309 void bdrv_reopen_abort(BDRVReopenState *reopen_state);
310 int bdrv_read(BdrvChild *child, int64_t sector_num,
311               uint8_t *buf, int nb_sectors);
312 int bdrv_write(BdrvChild *child, int64_t sector_num,
313                const uint8_t *buf, int nb_sectors);
314 int bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset,
315                        int bytes, BdrvRequestFlags flags);
316 int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags);
317 int bdrv_pread(BdrvChild *child, int64_t offset, void *buf, int bytes);
318 int bdrv_preadv(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
319 int bdrv_pwrite(BdrvChild *child, int64_t offset, const void *buf, int bytes);
320 int bdrv_pwritev(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
321 int bdrv_pwrite_sync(BdrvChild *child, int64_t offset,
322                      const void *buf, int count);
323 /*
324  * Efficiently zero a region of the disk image.  Note that this is a regular
325  * I/O request like read or write and should have a reasonable size.  This
326  * function is not suitable for zeroing the entire image in a single request
327  * because it may allocate memory for the entire region.
328  */
329 int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset,
330                                        int bytes, BdrvRequestFlags flags);
331 BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,
332     const char *backing_file);
333 void bdrv_refresh_filename(BlockDriverState *bs);
334 
335 int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset,
336                                   PreallocMode prealloc, Error **errp);
337 int bdrv_truncate(BdrvChild *child, int64_t offset, PreallocMode prealloc,
338                   Error **errp);
339 
340 int64_t bdrv_nb_sectors(BlockDriverState *bs);
341 int64_t bdrv_getlength(BlockDriverState *bs);
342 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs);
343 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
344                                BlockDriverState *in_bs, Error **errp);
345 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr);
346 void bdrv_refresh_limits(BlockDriverState *bs, Error **errp);
347 int bdrv_commit(BlockDriverState *bs);
348 int bdrv_change_backing_file(BlockDriverState *bs,
349     const char *backing_file, const char *backing_fmt);
350 void bdrv_register(BlockDriver *bdrv);
351 int bdrv_drop_intermediate(BlockDriverState *top, BlockDriverState *base,
352                            const char *backing_file_str);
353 BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
354                                     BlockDriverState *bs);
355 BlockDriverState *bdrv_find_base(BlockDriverState *bs);
356 
357 
358 typedef struct BdrvCheckResult {
359     int corruptions;
360     int leaks;
361     int check_errors;
362     int corruptions_fixed;
363     int leaks_fixed;
364     int64_t image_end_offset;
365     BlockFragInfo bfi;
366 } BdrvCheckResult;
367 
368 typedef enum {
369     BDRV_FIX_LEAKS    = 1,
370     BDRV_FIX_ERRORS   = 2,
371 } BdrvCheckMode;
372 
373 int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix);
374 
375 /* The units of offset and total_work_size may be chosen arbitrarily by the
376  * block driver; total_work_size may change during the course of the amendment
377  * operation */
378 typedef void BlockDriverAmendStatusCB(BlockDriverState *bs, int64_t offset,
379                                       int64_t total_work_size, void *opaque);
380 int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts,
381                        BlockDriverAmendStatusCB *status_cb, void *cb_opaque,
382                        Error **errp);
383 
384 /* external snapshots */
385 bool bdrv_recurse_is_first_non_filter(BlockDriverState *bs,
386                                       BlockDriverState *candidate);
387 bool bdrv_is_first_non_filter(BlockDriverState *candidate);
388 
389 /* check if a named node can be replaced when doing drive-mirror */
390 BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs,
391                                         const char *node_name, Error **errp);
392 
393 /* async block I/O */
394 void bdrv_aio_cancel(BlockAIOCB *acb);
395 void bdrv_aio_cancel_async(BlockAIOCB *acb);
396 
397 /* sg packet commands */
398 int bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf);
399 
400 /* Invalidate any cached metadata used by image formats */
401 void bdrv_invalidate_cache(BlockDriverState *bs, Error **errp);
402 void bdrv_invalidate_cache_all(Error **errp);
403 int bdrv_inactivate_all(void);
404 
405 /* Ensure contents are flushed to disk.  */
406 int bdrv_flush(BlockDriverState *bs);
407 int coroutine_fn bdrv_co_flush(BlockDriverState *bs);
408 int bdrv_flush_all(void);
409 void bdrv_close_all(void);
410 void bdrv_drain(BlockDriverState *bs);
411 void coroutine_fn bdrv_co_drain(BlockDriverState *bs);
412 void bdrv_drain_all_begin(void);
413 void bdrv_drain_all_end(void);
414 void bdrv_drain_all(void);
415 
416 #define BDRV_POLL_WHILE(bs, cond) ({                       \
417     BlockDriverState *bs_ = (bs);                          \
418     AIO_WAIT_WHILE(bdrv_get_aio_context(bs_),              \
419                    cond); })
420 
421 int bdrv_pdiscard(BdrvChild *child, int64_t offset, int bytes);
422 int bdrv_co_pdiscard(BdrvChild *child, int64_t offset, int bytes);
423 int bdrv_has_zero_init_1(BlockDriverState *bs);
424 int bdrv_has_zero_init(BlockDriverState *bs);
425 bool bdrv_unallocated_blocks_are_zero(BlockDriverState *bs);
426 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs);
427 int bdrv_block_status(BlockDriverState *bs, int64_t offset,
428                       int64_t bytes, int64_t *pnum, int64_t *map,
429                       BlockDriverState **file);
430 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base,
431                             int64_t offset, int64_t bytes, int64_t *pnum,
432                             int64_t *map, BlockDriverState **file);
433 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
434                       int64_t *pnum);
435 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
436                             int64_t offset, int64_t bytes, int64_t *pnum);
437 
438 bool bdrv_is_read_only(BlockDriverState *bs);
439 int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only,
440                            bool ignore_allow_rdw, Error **errp);
441 int bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg,
442                               Error **errp);
443 bool bdrv_is_writable(BlockDriverState *bs);
444 bool bdrv_is_sg(BlockDriverState *bs);
445 bool bdrv_is_inserted(BlockDriverState *bs);
446 void bdrv_lock_medium(BlockDriverState *bs, bool locked);
447 void bdrv_eject(BlockDriverState *bs, bool eject_flag);
448 const char *bdrv_get_format_name(BlockDriverState *bs);
449 BlockDriverState *bdrv_find_node(const char *node_name);
450 BlockDeviceInfoList *bdrv_named_nodes_list(Error **errp);
451 XDbgBlockGraph *bdrv_get_xdbg_block_graph(Error **errp);
452 BlockDriverState *bdrv_lookup_bs(const char *device,
453                                  const char *node_name,
454                                  Error **errp);
455 bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base);
456 BlockDriverState *bdrv_next_node(BlockDriverState *bs);
457 BlockDriverState *bdrv_next_all_states(BlockDriverState *bs);
458 
459 typedef struct BdrvNextIterator {
460     enum {
461         BDRV_NEXT_BACKEND_ROOTS,
462         BDRV_NEXT_MONITOR_OWNED,
463     } phase;
464     BlockBackend *blk;
465     BlockDriverState *bs;
466 } BdrvNextIterator;
467 
468 BlockDriverState *bdrv_first(BdrvNextIterator *it);
469 BlockDriverState *bdrv_next(BdrvNextIterator *it);
470 void bdrv_next_cleanup(BdrvNextIterator *it);
471 
472 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs);
473 bool bdrv_is_encrypted(BlockDriverState *bs);
474 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
475                          void *opaque, bool read_only);
476 const char *bdrv_get_node_name(const BlockDriverState *bs);
477 const char *bdrv_get_device_name(const BlockDriverState *bs);
478 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
479 int bdrv_get_flags(BlockDriverState *bs);
480 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
481 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs,
482                                           Error **errp);
483 void bdrv_round_to_clusters(BlockDriverState *bs,
484                             int64_t offset, int64_t bytes,
485                             int64_t *cluster_offset,
486                             int64_t *cluster_bytes);
487 
488 void bdrv_get_backing_filename(BlockDriverState *bs,
489                                char *filename, int filename_size);
490 char *bdrv_get_full_backing_filename(BlockDriverState *bs, Error **errp);
491 char *bdrv_get_full_backing_filename_from_filename(const char *backed,
492                                                    const char *backing,
493                                                    Error **errp);
494 char *bdrv_dirname(BlockDriverState *bs, Error **errp);
495 
496 int path_has_protocol(const char *path);
497 int path_is_absolute(const char *path);
498 char *path_combine(const char *base_path, const char *filename);
499 
500 int bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
501 int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
502 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
503                       int64_t pos, int size);
504 
505 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
506                       int64_t pos, int size);
507 
508 void bdrv_img_create(const char *filename, const char *fmt,
509                      const char *base_filename, const char *base_fmt,
510                      char *options, uint64_t img_size, int flags,
511                      bool quiet, Error **errp);
512 
513 /* Returns the alignment in bytes that is required so that no bounce buffer
514  * is required throughout the stack */
515 size_t bdrv_min_mem_align(BlockDriverState *bs);
516 /* Returns optimal alignment in bytes for bounce buffer */
517 size_t bdrv_opt_mem_align(BlockDriverState *bs);
518 void *qemu_blockalign(BlockDriverState *bs, size_t size);
519 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
520 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
521 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
522 bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov);
523 
524 void bdrv_enable_copy_on_read(BlockDriverState *bs);
525 void bdrv_disable_copy_on_read(BlockDriverState *bs);
526 
527 void bdrv_ref(BlockDriverState *bs);
528 void bdrv_unref(BlockDriverState *bs);
529 void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child);
530 BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
531                              BlockDriverState *child_bs,
532                              const char *child_name,
533                              const BdrvChildRole *child_role,
534                              Error **errp);
535 
536 bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp);
537 void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason);
538 void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason);
539 void bdrv_op_block_all(BlockDriverState *bs, Error *reason);
540 void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason);
541 bool bdrv_op_blocker_is_empty(BlockDriverState *bs);
542 
543 #define BLKDBG_EVENT(child, evt) \
544     do { \
545         if (child) { \
546             bdrv_debug_event(child->bs, evt); \
547         } \
548     } while (0)
549 
550 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
551 
552 int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event,
553                            const char *tag);
554 int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag);
555 int bdrv_debug_resume(BlockDriverState *bs, const char *tag);
556 bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag);
557 
558 /**
559  * bdrv_get_aio_context:
560  *
561  * Returns: the currently bound #AioContext
562  */
563 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
564 
565 /**
566  * Transfer control to @co in the aio context of @bs
567  */
568 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co);
569 
570 /**
571  * bdrv_set_aio_context:
572  *
573  * Changes the #AioContext used for fd handlers, timers, and BHs by this
574  * BlockDriverState and all its children.
575  *
576  * This function must be called with iothread lock held.
577  */
578 void bdrv_set_aio_context(BlockDriverState *bs, AioContext *new_context);
579 int bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz);
580 int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo);
581 
582 void bdrv_io_plug(BlockDriverState *bs);
583 void bdrv_io_unplug(BlockDriverState *bs);
584 
585 /**
586  * bdrv_parent_drained_begin:
587  *
588  * Begin a quiesced section of all users of @bs. This is part of
589  * bdrv_drained_begin.
590  */
591 void bdrv_parent_drained_begin(BlockDriverState *bs, BdrvChild *ignore,
592                                bool ignore_bds_parents);
593 
594 /**
595  * bdrv_parent_drained_begin_single:
596  *
597  * Begin a quiesced section for the parent of @c. If @poll is true, wait for
598  * any pending activity to cease.
599  */
600 void bdrv_parent_drained_begin_single(BdrvChild *c, bool poll);
601 
602 /**
603  * bdrv_parent_drained_end:
604  *
605  * End a quiesced section of all users of @bs. This is part of
606  * bdrv_drained_end.
607  */
608 void bdrv_parent_drained_end(BlockDriverState *bs, BdrvChild *ignore,
609                              bool ignore_bds_parents);
610 
611 /**
612  * bdrv_drain_poll:
613  *
614  * Poll for pending requests in @bs, its parents (except for @ignore_parent),
615  * and if @recursive is true its children as well (used for subtree drain).
616  *
617  * If @ignore_bds_parents is true, parents that are BlockDriverStates must
618  * ignore the drain request because they will be drained separately (used for
619  * drain_all).
620  *
621  * This is part of bdrv_drained_begin.
622  */
623 bool bdrv_drain_poll(BlockDriverState *bs, bool recursive,
624                      BdrvChild *ignore_parent, bool ignore_bds_parents);
625 
626 /**
627  * bdrv_drained_begin:
628  *
629  * Begin a quiesced section for exclusive access to the BDS, by disabling
630  * external request sources including NBD server and device model. Note that
631  * this doesn't block timers or coroutines from submitting more requests, which
632  * means block_job_pause is still necessary.
633  *
634  * This function can be recursive.
635  */
636 void bdrv_drained_begin(BlockDriverState *bs);
637 
638 /**
639  * bdrv_do_drained_begin_quiesce:
640  *
641  * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already
642  * running requests to complete.
643  */
644 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs,
645                                    BdrvChild *parent, bool ignore_bds_parents);
646 
647 /**
648  * Like bdrv_drained_begin, but recursively begins a quiesced section for
649  * exclusive access to all child nodes as well.
650  */
651 void bdrv_subtree_drained_begin(BlockDriverState *bs);
652 
653 /**
654  * bdrv_drained_end:
655  *
656  * End a quiescent section started by bdrv_drained_begin().
657  */
658 void bdrv_drained_end(BlockDriverState *bs);
659 
660 /**
661  * End a quiescent section started by bdrv_subtree_drained_begin().
662  */
663 void bdrv_subtree_drained_end(BlockDriverState *bs);
664 
665 void bdrv_add_child(BlockDriverState *parent, BlockDriverState *child,
666                     Error **errp);
667 void bdrv_del_child(BlockDriverState *parent, BdrvChild *child, Error **errp);
668 
669 bool bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
670                                      uint32_t granularity, Error **errp);
671 /**
672  *
673  * bdrv_register_buf/bdrv_unregister_buf:
674  *
675  * Register/unregister a buffer for I/O. For example, VFIO drivers are
676  * interested to know the memory areas that would later be used for I/O, so
677  * that they can prepare IOMMU mapping etc., to get better performance.
678  */
679 void bdrv_register_buf(BlockDriverState *bs, void *host, size_t size);
680 void bdrv_unregister_buf(BlockDriverState *bs, void *host);
681 
682 /**
683  *
684  * bdrv_co_copy_range:
685  *
686  * Do offloaded copy between two children. If the operation is not implemented
687  * by the driver, or if the backend storage doesn't support it, a negative
688  * error code will be returned.
689  *
690  * Note: block layer doesn't emulate or fallback to a bounce buffer approach
691  * because usually the caller shouldn't attempt offloaded copy any more (e.g.
692  * calling copy_file_range(2)) after the first error, thus it should fall back
693  * to a read+write path in the caller level.
694  *
695  * @src: Source child to copy data from
696  * @src_offset: offset in @src image to read data
697  * @dst: Destination child to copy data to
698  * @dst_offset: offset in @dst image to write data
699  * @bytes: number of bytes to copy
700  * @flags: request flags. Supported flags:
701  *         BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero
702  *                               write on @dst as if bdrv_co_pwrite_zeroes is
703  *                               called. Used to simplify caller code, or
704  *                               during BlockDriver.bdrv_co_copy_range_from()
705  *                               recursion.
706  *         BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping
707  *                                   requests currently in flight.
708  *
709  * Returns: 0 if succeeded; negative error code if failed.
710  **/
711 int coroutine_fn bdrv_co_copy_range(BdrvChild *src, uint64_t src_offset,
712                                     BdrvChild *dst, uint64_t dst_offset,
713                                     uint64_t bytes, BdrvRequestFlags read_flags,
714                                     BdrvRequestFlags write_flags);
715 #endif
716