xref: /openbmc/qemu/include/block/block.h (revision 84ca4fa99d7b6c83ce31e1e41300d55cb4e97dcb)
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     /* The BDRV_REQ_MAY_UNMAP flag is used to indicate that the block driver
47      * is allowed to optimize a write zeroes request by unmapping (discarding)
48      * blocks if it is guaranteed that the result will read back as
49      * zeroes. The flag is only passed to the driver if the block device is
50      * opened with BDRV_O_UNMAP.
51      */
52     BDRV_REQ_MAY_UNMAP          = 0x4,
53     BDRV_REQ_NO_SERIALISING     = 0x8,
54     BDRV_REQ_FUA                = 0x10,
55     BDRV_REQ_WRITE_COMPRESSED   = 0x20,
56 
57     /* Signifies that this write request will not change the visible disk
58      * content. */
59     BDRV_REQ_WRITE_UNCHANGED    = 0x40,
60 
61     /* Mask of valid flags */
62     BDRV_REQ_MASK               = 0x7f,
63 } BdrvRequestFlags;
64 
65 typedef struct BlockSizes {
66     uint32_t phys;
67     uint32_t log;
68 } BlockSizes;
69 
70 typedef struct HDGeometry {
71     uint32_t heads;
72     uint32_t sectors;
73     uint32_t cylinders;
74 } HDGeometry;
75 
76 #define BDRV_O_RDWR        0x0002
77 #define BDRV_O_RESIZE      0x0004 /* request permission for resizing the node */
78 #define BDRV_O_SNAPSHOT    0x0008 /* open the file read only and save writes in a snapshot */
79 #define BDRV_O_TEMPORARY   0x0010 /* delete the file after use */
80 #define BDRV_O_NOCACHE     0x0020 /* do not use the host page cache */
81 #define BDRV_O_NATIVE_AIO  0x0080 /* use native AIO instead of the thread pool */
82 #define BDRV_O_NO_BACKING  0x0100 /* don't open the backing file */
83 #define BDRV_O_NO_FLUSH    0x0200 /* disable flushing on this disk */
84 #define BDRV_O_COPY_ON_READ 0x0400 /* copy read backing sectors into image */
85 #define BDRV_O_INACTIVE    0x0800  /* consistency hint for migration handoff */
86 #define BDRV_O_CHECK       0x1000  /* open solely for consistency check */
87 #define BDRV_O_ALLOW_RDWR  0x2000  /* allow reopen to change from r/o to r/w */
88 #define BDRV_O_UNMAP       0x4000  /* execute guest UNMAP/TRIM operations */
89 #define BDRV_O_PROTOCOL    0x8000  /* if no block driver is explicitly given:
90                                       select an appropriate protocol driver,
91                                       ignoring the format layer */
92 #define BDRV_O_NO_IO       0x10000 /* don't initialize for I/O */
93 
94 #define BDRV_O_CACHE_MASK  (BDRV_O_NOCACHE | BDRV_O_NO_FLUSH)
95 
96 
97 /* Option names of options parsed by the block layer */
98 
99 #define BDRV_OPT_CACHE_WB       "cache.writeback"
100 #define BDRV_OPT_CACHE_DIRECT   "cache.direct"
101 #define BDRV_OPT_CACHE_NO_FLUSH "cache.no-flush"
102 #define BDRV_OPT_READ_ONLY      "read-only"
103 #define BDRV_OPT_DISCARD        "discard"
104 #define BDRV_OPT_FORCE_SHARE    "force-share"
105 
106 
107 #define BDRV_SECTOR_BITS   9
108 #define BDRV_SECTOR_SIZE   (1ULL << BDRV_SECTOR_BITS)
109 #define BDRV_SECTOR_MASK   ~(BDRV_SECTOR_SIZE - 1)
110 
111 #define BDRV_REQUEST_MAX_SECTORS MIN(SIZE_MAX >> BDRV_SECTOR_BITS, \
112                                      INT_MAX >> BDRV_SECTOR_BITS)
113 #define BDRV_REQUEST_MAX_BYTES (BDRV_REQUEST_MAX_SECTORS << BDRV_SECTOR_BITS)
114 
115 /*
116  * Allocation status flags for bdrv_block_status() and friends.
117  *
118  * Public flags:
119  * BDRV_BLOCK_DATA: allocation for data at offset is tied to this layer
120  * BDRV_BLOCK_ZERO: offset reads as zero
121  * BDRV_BLOCK_OFFSET_VALID: an associated offset exists for accessing raw data
122  * BDRV_BLOCK_ALLOCATED: the content of the block is determined by this
123  *                       layer rather than any backing, set by block layer
124  * BDRV_BLOCK_EOF: the returned pnum covers through end of file for this
125  *                 layer, set by block layer
126  *
127  * Internal flag:
128  * BDRV_BLOCK_RAW: for use by passthrough drivers, such as raw, to request
129  *                 that the block layer recompute the answer from the returned
130  *                 BDS; must be accompanied by just BDRV_BLOCK_OFFSET_VALID.
131  *
132  * If BDRV_BLOCK_OFFSET_VALID is set, the map parameter represents the
133  * host offset within the returned BDS that is allocated for the
134  * corresponding raw guest data.  However, whether that offset
135  * actually contains data also depends on BDRV_BLOCK_DATA, as follows:
136  *
137  * DATA ZERO OFFSET_VALID
138  *  t    t        t       sectors read as zero, returned file is zero at offset
139  *  t    f        t       sectors read as valid from file at offset
140  *  f    t        t       sectors preallocated, read as zero, returned file not
141  *                        necessarily zero at offset
142  *  f    f        t       sectors preallocated but read from backing_hd,
143  *                        returned file contains garbage at offset
144  *  t    t        f       sectors preallocated, read as zero, unknown offset
145  *  t    f        f       sectors read from unknown file or offset
146  *  f    t        f       not allocated or unknown offset, read as zero
147  *  f    f        f       not allocated or unknown offset, read from backing_hd
148  */
149 #define BDRV_BLOCK_DATA         0x01
150 #define BDRV_BLOCK_ZERO         0x02
151 #define BDRV_BLOCK_OFFSET_VALID 0x04
152 #define BDRV_BLOCK_RAW          0x08
153 #define BDRV_BLOCK_ALLOCATED    0x10
154 #define BDRV_BLOCK_EOF          0x20
155 #define BDRV_BLOCK_OFFSET_MASK  BDRV_SECTOR_MASK
156 
157 typedef QSIMPLEQ_HEAD(BlockReopenQueue, BlockReopenQueueEntry) BlockReopenQueue;
158 
159 typedef struct BDRVReopenState {
160     BlockDriverState *bs;
161     int flags;
162     uint64_t perm, shared_perm;
163     QDict *options;
164     QDict *explicit_options;
165     void *opaque;
166 } BDRVReopenState;
167 
168 /*
169  * Block operation types
170  */
171 typedef enum BlockOpType {
172     BLOCK_OP_TYPE_BACKUP_SOURCE,
173     BLOCK_OP_TYPE_BACKUP_TARGET,
174     BLOCK_OP_TYPE_CHANGE,
175     BLOCK_OP_TYPE_COMMIT_SOURCE,
176     BLOCK_OP_TYPE_COMMIT_TARGET,
177     BLOCK_OP_TYPE_DATAPLANE,
178     BLOCK_OP_TYPE_DRIVE_DEL,
179     BLOCK_OP_TYPE_EJECT,
180     BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
181     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT,
182     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE,
183     BLOCK_OP_TYPE_MIRROR_SOURCE,
184     BLOCK_OP_TYPE_MIRROR_TARGET,
185     BLOCK_OP_TYPE_RESIZE,
186     BLOCK_OP_TYPE_STREAM,
187     BLOCK_OP_TYPE_REPLACE,
188     BLOCK_OP_TYPE_MAX,
189 } BlockOpType;
190 
191 /* Block node permission constants */
192 enum {
193     /**
194      * A user that has the "permission" of consistent reads is guaranteed that
195      * their view of the contents of the block device is complete and
196      * self-consistent, representing the contents of a disk at a specific
197      * point.
198      *
199      * For most block devices (including their backing files) this is true, but
200      * the property cannot be maintained in a few situations like for
201      * intermediate nodes of a commit block job.
202      */
203     BLK_PERM_CONSISTENT_READ    = 0x01,
204 
205     /** This permission is required to change the visible disk contents. */
206     BLK_PERM_WRITE              = 0x02,
207 
208     /**
209      * This permission (which is weaker than BLK_PERM_WRITE) is both enough and
210      * required for writes to the block node when the caller promises that
211      * the visible disk content doesn't change.
212      *
213      * As the BLK_PERM_WRITE permission is strictly stronger, either is
214      * sufficient to perform an unchanging write.
215      */
216     BLK_PERM_WRITE_UNCHANGED    = 0x04,
217 
218     /** This permission is required to change the size of a block node. */
219     BLK_PERM_RESIZE             = 0x08,
220 
221     /**
222      * This permission is required to change the node that this BdrvChild
223      * points to.
224      */
225     BLK_PERM_GRAPH_MOD          = 0x10,
226 
227     BLK_PERM_ALL                = 0x1f,
228 
229     DEFAULT_PERM_PASSTHROUGH    = BLK_PERM_CONSISTENT_READ
230                                  | BLK_PERM_WRITE
231                                  | BLK_PERM_WRITE_UNCHANGED
232                                  | BLK_PERM_RESIZE,
233 
234     DEFAULT_PERM_UNCHANGED      = BLK_PERM_ALL & ~DEFAULT_PERM_PASSTHROUGH,
235 };
236 
237 char *bdrv_perm_names(uint64_t perm);
238 
239 /* disk I/O throttling */
240 void bdrv_init(void);
241 void bdrv_init_with_whitelist(void);
242 bool bdrv_uses_whitelist(void);
243 int bdrv_is_whitelisted(BlockDriver *drv, bool read_only);
244 BlockDriver *bdrv_find_protocol(const char *filename,
245                                 bool allow_protocol_prefix,
246                                 Error **errp);
247 BlockDriver *bdrv_find_format(const char *format_name);
248 int bdrv_create(BlockDriver *drv, const char* filename,
249                 QemuOpts *opts, Error **errp);
250 int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp);
251 BlockDriverState *bdrv_new(void);
252 void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top,
253                  Error **errp);
254 void bdrv_replace_node(BlockDriverState *from, BlockDriverState *to,
255                        Error **errp);
256 
257 int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough);
258 int bdrv_parse_discard_flags(const char *mode, int *flags);
259 BdrvChild *bdrv_open_child(const char *filename,
260                            QDict *options, const char *bdref_key,
261                            BlockDriverState* parent,
262                            const BdrvChildRole *child_role,
263                            bool allow_none, Error **errp);
264 BlockDriverState *bdrv_open_blockdev_ref(BlockdevRef *ref, Error **errp);
265 void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd,
266                          Error **errp);
267 int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options,
268                            const char *bdref_key, Error **errp);
269 BlockDriverState *bdrv_open(const char *filename, const char *reference,
270                             QDict *options, int flags, Error **errp);
271 BlockDriverState *bdrv_new_open_driver(BlockDriver *drv, const char *node_name,
272                                        int flags, Error **errp);
273 BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue,
274                                     BlockDriverState *bs,
275                                     QDict *options, int flags);
276 int bdrv_reopen_multiple(AioContext *ctx, BlockReopenQueue *bs_queue, Error **errp);
277 int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp);
278 int bdrv_reopen_prepare(BDRVReopenState *reopen_state,
279                         BlockReopenQueue *queue, Error **errp);
280 void bdrv_reopen_commit(BDRVReopenState *reopen_state);
281 void bdrv_reopen_abort(BDRVReopenState *reopen_state);
282 int bdrv_read(BdrvChild *child, int64_t sector_num,
283               uint8_t *buf, int nb_sectors);
284 int bdrv_write(BdrvChild *child, int64_t sector_num,
285                const uint8_t *buf, int nb_sectors);
286 int bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset,
287                        int bytes, BdrvRequestFlags flags);
288 int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags);
289 int bdrv_pread(BdrvChild *child, int64_t offset, void *buf, int bytes);
290 int bdrv_preadv(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
291 int bdrv_pwrite(BdrvChild *child, int64_t offset, const void *buf, int bytes);
292 int bdrv_pwritev(BdrvChild *child, int64_t offset, QEMUIOVector *qiov);
293 int bdrv_pwrite_sync(BdrvChild *child, int64_t offset,
294                      const void *buf, int count);
295 /*
296  * Efficiently zero a region of the disk image.  Note that this is a regular
297  * I/O request like read or write and should have a reasonable size.  This
298  * function is not suitable for zeroing the entire image in a single request
299  * because it may allocate memory for the entire region.
300  */
301 int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset,
302                                        int bytes, BdrvRequestFlags flags);
303 BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,
304     const char *backing_file);
305 void bdrv_refresh_filename(BlockDriverState *bs);
306 
307 int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset,
308                                   PreallocMode prealloc, Error **errp);
309 int bdrv_truncate(BdrvChild *child, int64_t offset, PreallocMode prealloc,
310                   Error **errp);
311 
312 int64_t bdrv_nb_sectors(BlockDriverState *bs);
313 int64_t bdrv_getlength(BlockDriverState *bs);
314 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs);
315 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
316                                BlockDriverState *in_bs, Error **errp);
317 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr);
318 void bdrv_refresh_limits(BlockDriverState *bs, Error **errp);
319 int bdrv_commit(BlockDriverState *bs);
320 int bdrv_change_backing_file(BlockDriverState *bs,
321     const char *backing_file, const char *backing_fmt);
322 void bdrv_register(BlockDriver *bdrv);
323 int bdrv_drop_intermediate(BlockDriverState *top, BlockDriverState *base,
324                            const char *backing_file_str);
325 BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
326                                     BlockDriverState *bs);
327 BlockDriverState *bdrv_find_base(BlockDriverState *bs);
328 
329 
330 typedef struct BdrvCheckResult {
331     int corruptions;
332     int leaks;
333     int check_errors;
334     int corruptions_fixed;
335     int leaks_fixed;
336     int64_t image_end_offset;
337     BlockFragInfo bfi;
338 } BdrvCheckResult;
339 
340 typedef enum {
341     BDRV_FIX_LEAKS    = 1,
342     BDRV_FIX_ERRORS   = 2,
343 } BdrvCheckMode;
344 
345 int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix);
346 
347 /* The units of offset and total_work_size may be chosen arbitrarily by the
348  * block driver; total_work_size may change during the course of the amendment
349  * operation */
350 typedef void BlockDriverAmendStatusCB(BlockDriverState *bs, int64_t offset,
351                                       int64_t total_work_size, void *opaque);
352 int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts,
353                        BlockDriverAmendStatusCB *status_cb, void *cb_opaque,
354                        Error **errp);
355 
356 /* external snapshots */
357 bool bdrv_recurse_is_first_non_filter(BlockDriverState *bs,
358                                       BlockDriverState *candidate);
359 bool bdrv_is_first_non_filter(BlockDriverState *candidate);
360 
361 /* check if a named node can be replaced when doing drive-mirror */
362 BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs,
363                                         const char *node_name, Error **errp);
364 
365 /* async block I/O */
366 void bdrv_aio_cancel(BlockAIOCB *acb);
367 void bdrv_aio_cancel_async(BlockAIOCB *acb);
368 
369 /* sg packet commands */
370 int bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf);
371 
372 /* Invalidate any cached metadata used by image formats */
373 void bdrv_invalidate_cache(BlockDriverState *bs, Error **errp);
374 void bdrv_invalidate_cache_all(Error **errp);
375 int bdrv_inactivate_all(void);
376 
377 /* Ensure contents are flushed to disk.  */
378 int bdrv_flush(BlockDriverState *bs);
379 int coroutine_fn bdrv_co_flush(BlockDriverState *bs);
380 int bdrv_flush_all(void);
381 void bdrv_close_all(void);
382 void bdrv_drain(BlockDriverState *bs);
383 void coroutine_fn bdrv_co_drain(BlockDriverState *bs);
384 void bdrv_drain_all_begin(void);
385 void bdrv_drain_all_end(void);
386 void bdrv_drain_all(void);
387 
388 /* Returns NULL when bs == NULL */
389 AioWait *bdrv_get_aio_wait(BlockDriverState *bs);
390 
391 #define BDRV_POLL_WHILE(bs, cond) ({                       \
392     BlockDriverState *bs_ = (bs);                          \
393     AIO_WAIT_WHILE(bdrv_get_aio_wait(bs_),                 \
394                    bdrv_get_aio_context(bs_),              \
395                    cond); })
396 
397 int bdrv_pdiscard(BlockDriverState *bs, int64_t offset, int bytes);
398 int bdrv_co_pdiscard(BlockDriverState *bs, int64_t offset, int bytes);
399 int bdrv_has_zero_init_1(BlockDriverState *bs);
400 int bdrv_has_zero_init(BlockDriverState *bs);
401 bool bdrv_unallocated_blocks_are_zero(BlockDriverState *bs);
402 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs);
403 int bdrv_block_status(BlockDriverState *bs, int64_t offset,
404                       int64_t bytes, int64_t *pnum, int64_t *map,
405                       BlockDriverState **file);
406 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base,
407                             int64_t offset, int64_t bytes, int64_t *pnum,
408                             int64_t *map, BlockDriverState **file);
409 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
410                       int64_t *pnum);
411 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
412                             int64_t offset, int64_t bytes, int64_t *pnum);
413 
414 bool bdrv_is_read_only(BlockDriverState *bs);
415 int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only,
416                            bool ignore_allow_rdw, Error **errp);
417 int bdrv_set_read_only(BlockDriverState *bs, bool read_only, Error **errp);
418 bool bdrv_is_writable(BlockDriverState *bs);
419 bool bdrv_is_sg(BlockDriverState *bs);
420 bool bdrv_is_inserted(BlockDriverState *bs);
421 void bdrv_lock_medium(BlockDriverState *bs, bool locked);
422 void bdrv_eject(BlockDriverState *bs, bool eject_flag);
423 const char *bdrv_get_format_name(BlockDriverState *bs);
424 BlockDriverState *bdrv_find_node(const char *node_name);
425 BlockDeviceInfoList *bdrv_named_nodes_list(Error **errp);
426 BlockDriverState *bdrv_lookup_bs(const char *device,
427                                  const char *node_name,
428                                  Error **errp);
429 bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base);
430 BlockDriverState *bdrv_next_node(BlockDriverState *bs);
431 BlockDriverState *bdrv_next_all_states(BlockDriverState *bs);
432 
433 typedef struct BdrvNextIterator {
434     enum {
435         BDRV_NEXT_BACKEND_ROOTS,
436         BDRV_NEXT_MONITOR_OWNED,
437     } phase;
438     BlockBackend *blk;
439     BlockDriverState *bs;
440 } BdrvNextIterator;
441 
442 BlockDriverState *bdrv_first(BdrvNextIterator *it);
443 BlockDriverState *bdrv_next(BdrvNextIterator *it);
444 void bdrv_next_cleanup(BdrvNextIterator *it);
445 
446 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs);
447 bool bdrv_is_encrypted(BlockDriverState *bs);
448 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
449                          void *opaque);
450 const char *bdrv_get_node_name(const BlockDriverState *bs);
451 const char *bdrv_get_device_name(const BlockDriverState *bs);
452 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
453 int bdrv_get_flags(BlockDriverState *bs);
454 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
455 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs);
456 void bdrv_round_to_clusters(BlockDriverState *bs,
457                             int64_t offset, int64_t bytes,
458                             int64_t *cluster_offset,
459                             int64_t *cluster_bytes);
460 
461 const char *bdrv_get_encrypted_filename(BlockDriverState *bs);
462 void bdrv_get_backing_filename(BlockDriverState *bs,
463                                char *filename, int filename_size);
464 void bdrv_get_full_backing_filename(BlockDriverState *bs,
465                                     char *dest, size_t sz, Error **errp);
466 void bdrv_get_full_backing_filename_from_filename(const char *backed,
467                                                   const char *backing,
468                                                   char *dest, size_t sz,
469                                                   Error **errp);
470 
471 int path_has_protocol(const char *path);
472 int path_is_absolute(const char *path);
473 void path_combine(char *dest, int dest_size,
474                   const char *base_path,
475                   const char *filename);
476 
477 int bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
478 int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
479 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
480                       int64_t pos, int size);
481 
482 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
483                       int64_t pos, int size);
484 
485 void bdrv_img_create(const char *filename, const char *fmt,
486                      const char *base_filename, const char *base_fmt,
487                      char *options, uint64_t img_size, int flags,
488                      bool quiet, Error **errp);
489 
490 /* Returns the alignment in bytes that is required so that no bounce buffer
491  * is required throughout the stack */
492 size_t bdrv_min_mem_align(BlockDriverState *bs);
493 /* Returns optimal alignment in bytes for bounce buffer */
494 size_t bdrv_opt_mem_align(BlockDriverState *bs);
495 void *qemu_blockalign(BlockDriverState *bs, size_t size);
496 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
497 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
498 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
499 bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov);
500 
501 void bdrv_enable_copy_on_read(BlockDriverState *bs);
502 void bdrv_disable_copy_on_read(BlockDriverState *bs);
503 
504 void bdrv_ref(BlockDriverState *bs);
505 void bdrv_unref(BlockDriverState *bs);
506 void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child);
507 BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
508                              BlockDriverState *child_bs,
509                              const char *child_name,
510                              const BdrvChildRole *child_role,
511                              Error **errp);
512 
513 bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp);
514 void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason);
515 void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason);
516 void bdrv_op_block_all(BlockDriverState *bs, Error *reason);
517 void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason);
518 bool bdrv_op_blocker_is_empty(BlockDriverState *bs);
519 
520 #define BLKDBG_EVENT(child, evt) \
521     do { \
522         if (child) { \
523             bdrv_debug_event(child->bs, evt); \
524         } \
525     } while (0)
526 
527 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
528 
529 int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event,
530                            const char *tag);
531 int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag);
532 int bdrv_debug_resume(BlockDriverState *bs, const char *tag);
533 bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag);
534 
535 /**
536  * bdrv_get_aio_context:
537  *
538  * Returns: the currently bound #AioContext
539  */
540 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
541 
542 /**
543  * Transfer control to @co in the aio context of @bs
544  */
545 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co);
546 
547 /**
548  * bdrv_set_aio_context:
549  *
550  * Changes the #AioContext used for fd handlers, timers, and BHs by this
551  * BlockDriverState and all its children.
552  *
553  * This function must be called with iothread lock held.
554  */
555 void bdrv_set_aio_context(BlockDriverState *bs, AioContext *new_context);
556 int bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz);
557 int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo);
558 
559 void bdrv_io_plug(BlockDriverState *bs);
560 void bdrv_io_unplug(BlockDriverState *bs);
561 
562 /**
563  * bdrv_parent_drained_begin:
564  *
565  * Begin a quiesced section of all users of @bs. This is part of
566  * bdrv_drained_begin.
567  */
568 void bdrv_parent_drained_begin(BlockDriverState *bs, BdrvChild *ignore,
569                                bool ignore_bds_parents);
570 
571 /**
572  * bdrv_parent_drained_end:
573  *
574  * End a quiesced section of all users of @bs. This is part of
575  * bdrv_drained_end.
576  */
577 void bdrv_parent_drained_end(BlockDriverState *bs, BdrvChild *ignore,
578                              bool ignore_bds_parents);
579 
580 /**
581  * bdrv_drain_poll:
582  *
583  * Poll for pending requests in @bs, its parents (except for @ignore_parent),
584  * and if @recursive is true its children as well (used for subtree drain).
585  *
586  * If @ignore_bds_parents is true, parents that are BlockDriverStates must
587  * ignore the drain request because they will be drained separately (used for
588  * drain_all).
589  *
590  * This is part of bdrv_drained_begin.
591  */
592 bool bdrv_drain_poll(BlockDriverState *bs, bool recursive,
593                      BdrvChild *ignore_parent, bool ignore_bds_parents);
594 
595 /**
596  * bdrv_drained_begin:
597  *
598  * Begin a quiesced section for exclusive access to the BDS, by disabling
599  * external request sources including NBD server and device model. Note that
600  * this doesn't block timers or coroutines from submitting more requests, which
601  * means block_job_pause is still necessary.
602  *
603  * This function can be recursive.
604  */
605 void bdrv_drained_begin(BlockDriverState *bs);
606 
607 /**
608  * bdrv_do_drained_begin_quiesce:
609  *
610  * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already
611  * running requests to complete.
612  */
613 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs,
614                                    BdrvChild *parent, bool ignore_bds_parents);
615 
616 /**
617  * Like bdrv_drained_begin, but recursively begins a quiesced section for
618  * exclusive access to all child nodes as well.
619  */
620 void bdrv_subtree_drained_begin(BlockDriverState *bs);
621 
622 /**
623  * bdrv_drained_end:
624  *
625  * End a quiescent section started by bdrv_drained_begin().
626  */
627 void bdrv_drained_end(BlockDriverState *bs);
628 
629 /**
630  * End a quiescent section started by bdrv_subtree_drained_begin().
631  */
632 void bdrv_subtree_drained_end(BlockDriverState *bs);
633 
634 void bdrv_add_child(BlockDriverState *parent, BlockDriverState *child,
635                     Error **errp);
636 void bdrv_del_child(BlockDriverState *parent, BdrvChild *child, Error **errp);
637 
638 bool bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
639                                      uint32_t granularity, Error **errp);
640 /**
641  *
642  * bdrv_register_buf/bdrv_unregister_buf:
643  *
644  * Register/unregister a buffer for I/O. For example, VFIO drivers are
645  * interested to know the memory areas that would later be used for I/O, so
646  * that they can prepare IOMMU mapping etc., to get better performance.
647  */
648 void bdrv_register_buf(BlockDriverState *bs, void *host, size_t size);
649 void bdrv_unregister_buf(BlockDriverState *bs, void *host);
650 
651 /**
652  *
653  * bdrv_co_copy_range:
654  *
655  * Do offloaded copy between two children. If the operation is not implemented
656  * by the driver, or if the backend storage doesn't support it, a negative
657  * error code will be returned.
658  *
659  * Note: block layer doesn't emulate or fallback to a bounce buffer approach
660  * because usually the caller shouldn't attempt offloaded copy any more (e.g.
661  * calling copy_file_range(2)) after the first error, thus it should fall back
662  * to a read+write path in the caller level.
663  *
664  * @src: Source child to copy data from
665  * @src_offset: offset in @src image to read data
666  * @dst: Destination child to copy data to
667  * @dst_offset: offset in @dst image to write data
668  * @bytes: number of bytes to copy
669  * @flags: request flags. Supported flags:
670  *         BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero
671  *                               write on @dst as if bdrv_co_pwrite_zeroes is
672  *                               called. Used to simplify caller code, or
673  *                               during BlockDriver.bdrv_co_copy_range_from()
674  *                               recursion.
675  *         BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping
676  *                                   requests currently in flight.
677  *
678  * Returns: 0 if succeeded; negative error code if failed.
679  **/
680 int coroutine_fn bdrv_co_copy_range(BdrvChild *src, uint64_t src_offset,
681                                     BdrvChild *dst, uint64_t dst_offset,
682                                     uint64_t bytes, BdrvRequestFlags flags);
683 #endif
684