xref: /openbmc/qemu/include/block/block.h (revision 4d9c7c84)
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  * Allocation status flags for bdrv_block_status() and friends.
147  *
148  * Public flags:
149  * BDRV_BLOCK_DATA: allocation for data at offset is tied to this layer
150  * BDRV_BLOCK_ZERO: offset reads as zero
151  * BDRV_BLOCK_OFFSET_VALID: an associated offset exists for accessing raw data
152  * BDRV_BLOCK_ALLOCATED: the content of the block is determined by this
153  *                       layer rather than any backing, set by block layer
154  * BDRV_BLOCK_EOF: the returned pnum covers through end of file for this
155  *                 layer, set by block layer
156  *
157  * Internal flags:
158  * BDRV_BLOCK_RAW: for use by passthrough drivers, such as raw, to request
159  *                 that the block layer recompute the answer from the returned
160  *                 BDS; must be accompanied by just BDRV_BLOCK_OFFSET_VALID.
161  * BDRV_BLOCK_RECURSE: request that the block layer will recursively search for
162  *                     zeroes in file child of current block node inside
163  *                     returned region. Only valid together with both
164  *                     BDRV_BLOCK_DATA and BDRV_BLOCK_OFFSET_VALID. Should not
165  *                     appear with BDRV_BLOCK_ZERO.
166  *
167  * If BDRV_BLOCK_OFFSET_VALID is set, the map parameter represents the
168  * host offset within the returned BDS that is allocated for the
169  * corresponding raw guest data.  However, whether that offset
170  * actually contains data also depends on BDRV_BLOCK_DATA, as follows:
171  *
172  * DATA ZERO OFFSET_VALID
173  *  t    t        t       sectors read as zero, returned file is zero at offset
174  *  t    f        t       sectors read as valid from file at offset
175  *  f    t        t       sectors preallocated, read as zero, returned file not
176  *                        necessarily zero at offset
177  *  f    f        t       sectors preallocated but read from backing_hd,
178  *                        returned file contains garbage at offset
179  *  t    t        f       sectors preallocated, read as zero, unknown offset
180  *  t    f        f       sectors read from unknown file or offset
181  *  f    t        f       not allocated or unknown offset, read as zero
182  *  f    f        f       not allocated or unknown offset, read from backing_hd
183  */
184 #define BDRV_BLOCK_DATA         0x01
185 #define BDRV_BLOCK_ZERO         0x02
186 #define BDRV_BLOCK_OFFSET_VALID 0x04
187 #define BDRV_BLOCK_RAW          0x08
188 #define BDRV_BLOCK_ALLOCATED    0x10
189 #define BDRV_BLOCK_EOF          0x20
190 #define BDRV_BLOCK_RECURSE      0x40
191 
192 typedef QTAILQ_HEAD(BlockReopenQueue, BlockReopenQueueEntry) BlockReopenQueue;
193 
194 typedef struct BDRVReopenState {
195     BlockDriverState *bs;
196     int flags;
197     BlockdevDetectZeroesOptions detect_zeroes;
198     bool backing_missing;
199     bool replace_backing_bs;  /* new_backing_bs is ignored if this is false */
200     BlockDriverState *new_backing_bs; /* If NULL then detach the current bs */
201     uint64_t perm, shared_perm;
202     QDict *options;
203     QDict *explicit_options;
204     void *opaque;
205 } BDRVReopenState;
206 
207 /*
208  * Block operation types
209  */
210 typedef enum BlockOpType {
211     BLOCK_OP_TYPE_BACKUP_SOURCE,
212     BLOCK_OP_TYPE_BACKUP_TARGET,
213     BLOCK_OP_TYPE_CHANGE,
214     BLOCK_OP_TYPE_COMMIT_SOURCE,
215     BLOCK_OP_TYPE_COMMIT_TARGET,
216     BLOCK_OP_TYPE_DATAPLANE,
217     BLOCK_OP_TYPE_DRIVE_DEL,
218     BLOCK_OP_TYPE_EJECT,
219     BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
220     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT,
221     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE,
222     BLOCK_OP_TYPE_MIRROR_SOURCE,
223     BLOCK_OP_TYPE_MIRROR_TARGET,
224     BLOCK_OP_TYPE_RESIZE,
225     BLOCK_OP_TYPE_STREAM,
226     BLOCK_OP_TYPE_REPLACE,
227     BLOCK_OP_TYPE_MAX,
228 } BlockOpType;
229 
230 /* Block node permission constants */
231 enum {
232     /**
233      * A user that has the "permission" of consistent reads is guaranteed that
234      * their view of the contents of the block device is complete and
235      * self-consistent, representing the contents of a disk at a specific
236      * point.
237      *
238      * For most block devices (including their backing files) this is true, but
239      * the property cannot be maintained in a few situations like for
240      * intermediate nodes of a commit block job.
241      */
242     BLK_PERM_CONSISTENT_READ    = 0x01,
243 
244     /** This permission is required to change the visible disk contents. */
245     BLK_PERM_WRITE              = 0x02,
246 
247     /**
248      * This permission (which is weaker than BLK_PERM_WRITE) is both enough and
249      * required for writes to the block node when the caller promises that
250      * the visible disk content doesn't change.
251      *
252      * As the BLK_PERM_WRITE permission is strictly stronger, either is
253      * sufficient to perform an unchanging write.
254      */
255     BLK_PERM_WRITE_UNCHANGED    = 0x04,
256 
257     /** This permission is required to change the size of a block node. */
258     BLK_PERM_RESIZE             = 0x08,
259 
260     /**
261      * This permission is required to change the node that this BdrvChild
262      * points to.
263      */
264     BLK_PERM_GRAPH_MOD          = 0x10,
265 
266     BLK_PERM_ALL                = 0x1f,
267 
268     DEFAULT_PERM_PASSTHROUGH    = BLK_PERM_CONSISTENT_READ
269                                  | BLK_PERM_WRITE
270                                  | BLK_PERM_WRITE_UNCHANGED
271                                  | BLK_PERM_RESIZE,
272 
273     DEFAULT_PERM_UNCHANGED      = BLK_PERM_ALL & ~DEFAULT_PERM_PASSTHROUGH,
274 };
275 
276 /*
277  * Flags that parent nodes assign to child nodes to specify what kind of
278  * role(s) they take.
279  *
280  * At least one of DATA, METADATA, FILTERED, or COW must be set for
281  * every child.
282  */
283 enum BdrvChildRoleBits {
284     /*
285      * This child stores data.
286      * Any node may have an arbitrary number of such children.
287      */
288     BDRV_CHILD_DATA         = (1 << 0),
289 
290     /*
291      * This child stores metadata.
292      * Any node may have an arbitrary number of metadata-storing
293      * children.
294      */
295     BDRV_CHILD_METADATA     = (1 << 1),
296 
297     /*
298      * A child that always presents exactly the same visible data as
299      * the parent, e.g. by virtue of the parent forwarding all reads
300      * and writes.
301      * This flag is mutually exclusive with DATA, METADATA, and COW.
302      * Any node may have at most one filtered child at a time.
303      */
304     BDRV_CHILD_FILTERED     = (1 << 2),
305 
306     /*
307      * Child from which to read all data that isn't allocated in the
308      * parent (i.e., the backing child); such data is copied to the
309      * parent through COW (and optionally COR).
310      * This field is mutually exclusive with DATA, METADATA, and
311      * FILTERED.
312      * Any node may have at most one such backing child at a time.
313      */
314     BDRV_CHILD_COW          = (1 << 3),
315 
316     /*
317      * The primary child.  For most drivers, this is the child whose
318      * filename applies best to the parent node.
319      * Any node may have at most one primary child at a time.
320      */
321     BDRV_CHILD_PRIMARY      = (1 << 4),
322 
323     /* Useful combination of flags */
324     BDRV_CHILD_IMAGE        = BDRV_CHILD_DATA
325                               | BDRV_CHILD_METADATA
326                               | BDRV_CHILD_PRIMARY,
327 };
328 
329 /* Mask of BdrvChildRoleBits values */
330 typedef unsigned int BdrvChildRole;
331 
332 char *bdrv_perm_names(uint64_t perm);
333 uint64_t bdrv_qapi_perm_to_blk_perm(BlockPermission qapi_perm);
334 
335 /* disk I/O throttling */
336 void bdrv_init(void);
337 void bdrv_init_with_whitelist(void);
338 bool bdrv_uses_whitelist(void);
339 int bdrv_is_whitelisted(BlockDriver *drv, bool read_only);
340 BlockDriver *bdrv_find_protocol(const char *filename,
341                                 bool allow_protocol_prefix,
342                                 Error **errp);
343 BlockDriver *bdrv_find_format(const char *format_name);
344 int bdrv_create(BlockDriver *drv, const char* filename,
345                 QemuOpts *opts, Error **errp);
346 int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp);
347 
348 BlockDriverState *bdrv_new(void);
349 void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top,
350                  Error **errp);
351 void bdrv_replace_node(BlockDriverState *from, BlockDriverState *to,
352                        Error **errp);
353 
354 int bdrv_parse_aio(const char *mode, int *flags);
355 int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough);
356 int bdrv_parse_discard_flags(const char *mode, int *flags);
357 BdrvChild *bdrv_open_child(const char *filename,
358                            QDict *options, const char *bdref_key,
359                            BlockDriverState* parent,
360                            const BdrvChildClass *child_class,
361                            BdrvChildRole child_role,
362                            bool allow_none, Error **errp);
363 BlockDriverState *bdrv_open_blockdev_ref(BlockdevRef *ref, Error **errp);
364 void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd,
365                          Error **errp);
366 int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options,
367                            const char *bdref_key, Error **errp);
368 BlockDriverState *bdrv_open(const char *filename, const char *reference,
369                             QDict *options, int flags, Error **errp);
370 BlockDriverState *bdrv_new_open_driver(BlockDriver *drv, const char *node_name,
371                                        int flags, Error **errp);
372 BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue,
373                                     BlockDriverState *bs, QDict *options,
374                                     bool keep_old_opts);
375 int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp);
376 int bdrv_reopen_set_read_only(BlockDriverState *bs, bool read_only,
377                               Error **errp);
378 int bdrv_reopen_prepare(BDRVReopenState *reopen_state,
379                         BlockReopenQueue *queue, Error **errp);
380 void bdrv_reopen_commit(BDRVReopenState *reopen_state);
381 void bdrv_reopen_abort(BDRVReopenState *reopen_state);
382 int bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset,
383                        int bytes, BdrvRequestFlags flags);
384 int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags);
385 int bdrv_pread(BdrvChild *child, int64_t offset, void *buf, int bytes);
386 int bdrv_pwrite(BdrvChild *child, int64_t offset, const void *buf, int bytes);
387 int bdrv_pwrite_sync(BdrvChild *child, int64_t offset,
388                      const void *buf, int count);
389 /*
390  * Efficiently zero a region of the disk image.  Note that this is a regular
391  * I/O request like read or write and should have a reasonable size.  This
392  * function is not suitable for zeroing the entire image in a single request
393  * because it may allocate memory for the entire region.
394  */
395 int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset,
396                                        int bytes, BdrvRequestFlags flags);
397 BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,
398     const char *backing_file);
399 void bdrv_refresh_filename(BlockDriverState *bs);
400 
401 int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset, bool exact,
402                                   PreallocMode prealloc, BdrvRequestFlags flags,
403                                   Error **errp);
404 int generated_co_wrapper
405 bdrv_truncate(BdrvChild *child, int64_t offset, bool exact,
406               PreallocMode prealloc, BdrvRequestFlags flags, Error **errp);
407 
408 int64_t bdrv_nb_sectors(BlockDriverState *bs);
409 int64_t bdrv_getlength(BlockDriverState *bs);
410 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs);
411 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
412                                BlockDriverState *in_bs, Error **errp);
413 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr);
414 void bdrv_refresh_limits(BlockDriverState *bs, Error **errp);
415 int bdrv_commit(BlockDriverState *bs);
416 int bdrv_make_empty(BdrvChild *c, Error **errp);
417 int bdrv_change_backing_file(BlockDriverState *bs, const char *backing_file,
418                              const char *backing_fmt, bool warn);
419 void bdrv_register(BlockDriver *bdrv);
420 int bdrv_drop_intermediate(BlockDriverState *top, BlockDriverState *base,
421                            const char *backing_file_str);
422 BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
423                                     BlockDriverState *bs);
424 BlockDriverState *bdrv_find_base(BlockDriverState *bs);
425 bool bdrv_is_backing_chain_frozen(BlockDriverState *bs, BlockDriverState *base,
426                                   Error **errp);
427 int bdrv_freeze_backing_chain(BlockDriverState *bs, BlockDriverState *base,
428                               Error **errp);
429 void bdrv_unfreeze_backing_chain(BlockDriverState *bs, BlockDriverState *base);
430 int coroutine_fn bdrv_co_delete_file(BlockDriverState *bs, Error **errp);
431 
432 
433 typedef struct BdrvCheckResult {
434     int corruptions;
435     int leaks;
436     int check_errors;
437     int corruptions_fixed;
438     int leaks_fixed;
439     int64_t image_end_offset;
440     BlockFragInfo bfi;
441 } BdrvCheckResult;
442 
443 typedef enum {
444     BDRV_FIX_LEAKS    = 1,
445     BDRV_FIX_ERRORS   = 2,
446 } BdrvCheckMode;
447 
448 int generated_co_wrapper bdrv_check(BlockDriverState *bs, BdrvCheckResult *res,
449                                     BdrvCheckMode fix);
450 
451 /* The units of offset and total_work_size may be chosen arbitrarily by the
452  * block driver; total_work_size may change during the course of the amendment
453  * operation */
454 typedef void BlockDriverAmendStatusCB(BlockDriverState *bs, int64_t offset,
455                                       int64_t total_work_size, void *opaque);
456 int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts,
457                        BlockDriverAmendStatusCB *status_cb, void *cb_opaque,
458                        bool force,
459                        Error **errp);
460 
461 /* check if a named node can be replaced when doing drive-mirror */
462 BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs,
463                                         const char *node_name, Error **errp);
464 
465 /* async block I/O */
466 void bdrv_aio_cancel(BlockAIOCB *acb);
467 void bdrv_aio_cancel_async(BlockAIOCB *acb);
468 
469 /* sg packet commands */
470 int bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf);
471 
472 /* Invalidate any cached metadata used by image formats */
473 int generated_co_wrapper bdrv_invalidate_cache(BlockDriverState *bs,
474                                                Error **errp);
475 void bdrv_invalidate_cache_all(Error **errp);
476 int bdrv_inactivate_all(void);
477 
478 /* Ensure contents are flushed to disk.  */
479 int generated_co_wrapper bdrv_flush(BlockDriverState *bs);
480 int coroutine_fn bdrv_co_flush(BlockDriverState *bs);
481 int bdrv_flush_all(void);
482 void bdrv_close_all(void);
483 void bdrv_drain(BlockDriverState *bs);
484 void coroutine_fn bdrv_co_drain(BlockDriverState *bs);
485 void bdrv_drain_all_begin(void);
486 void bdrv_drain_all_end(void);
487 void bdrv_drain_all(void);
488 
489 #define BDRV_POLL_WHILE(bs, cond) ({                       \
490     BlockDriverState *bs_ = (bs);                          \
491     AIO_WAIT_WHILE(bdrv_get_aio_context(bs_),              \
492                    cond); })
493 
494 int generated_co_wrapper bdrv_pdiscard(BdrvChild *child, int64_t offset,
495                                        int64_t bytes);
496 int bdrv_co_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes);
497 int bdrv_has_zero_init_1(BlockDriverState *bs);
498 int bdrv_has_zero_init(BlockDriverState *bs);
499 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs);
500 int bdrv_block_status(BlockDriverState *bs, int64_t offset,
501                       int64_t bytes, int64_t *pnum, int64_t *map,
502                       BlockDriverState **file);
503 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base,
504                             int64_t offset, int64_t bytes, int64_t *pnum,
505                             int64_t *map, BlockDriverState **file);
506 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
507                       int64_t *pnum);
508 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
509                             bool include_base, int64_t offset, int64_t bytes,
510                             int64_t *pnum);
511 
512 bool bdrv_is_read_only(BlockDriverState *bs);
513 int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only,
514                            bool ignore_allow_rdw, Error **errp);
515 int bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg,
516                               Error **errp);
517 bool bdrv_is_writable(BlockDriverState *bs);
518 bool bdrv_is_sg(BlockDriverState *bs);
519 bool bdrv_is_inserted(BlockDriverState *bs);
520 void bdrv_lock_medium(BlockDriverState *bs, bool locked);
521 void bdrv_eject(BlockDriverState *bs, bool eject_flag);
522 const char *bdrv_get_format_name(BlockDriverState *bs);
523 BlockDriverState *bdrv_find_node(const char *node_name);
524 BlockDeviceInfoList *bdrv_named_nodes_list(bool flat, Error **errp);
525 XDbgBlockGraph *bdrv_get_xdbg_block_graph(Error **errp);
526 BlockDriverState *bdrv_lookup_bs(const char *device,
527                                  const char *node_name,
528                                  Error **errp);
529 bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base);
530 BlockDriverState *bdrv_next_node(BlockDriverState *bs);
531 BlockDriverState *bdrv_next_all_states(BlockDriverState *bs);
532 
533 typedef struct BdrvNextIterator {
534     enum {
535         BDRV_NEXT_BACKEND_ROOTS,
536         BDRV_NEXT_MONITOR_OWNED,
537     } phase;
538     BlockBackend *blk;
539     BlockDriverState *bs;
540 } BdrvNextIterator;
541 
542 BlockDriverState *bdrv_first(BdrvNextIterator *it);
543 BlockDriverState *bdrv_next(BdrvNextIterator *it);
544 void bdrv_next_cleanup(BdrvNextIterator *it);
545 
546 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs);
547 bool bdrv_supports_compressed_writes(BlockDriverState *bs);
548 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
549                          void *opaque, bool read_only);
550 const char *bdrv_get_node_name(const BlockDriverState *bs);
551 const char *bdrv_get_device_name(const BlockDriverState *bs);
552 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
553 int bdrv_get_flags(BlockDriverState *bs);
554 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
555 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs,
556                                           Error **errp);
557 BlockStatsSpecific *bdrv_get_specific_stats(BlockDriverState *bs);
558 void bdrv_round_to_clusters(BlockDriverState *bs,
559                             int64_t offset, int64_t bytes,
560                             int64_t *cluster_offset,
561                             int64_t *cluster_bytes);
562 
563 void bdrv_get_backing_filename(BlockDriverState *bs,
564                                char *filename, int filename_size);
565 char *bdrv_get_full_backing_filename(BlockDriverState *bs, Error **errp);
566 char *bdrv_get_full_backing_filename_from_filename(const char *backed,
567                                                    const char *backing,
568                                                    Error **errp);
569 char *bdrv_dirname(BlockDriverState *bs, Error **errp);
570 
571 int path_has_protocol(const char *path);
572 int path_is_absolute(const char *path);
573 char *path_combine(const char *base_path, const char *filename);
574 
575 int generated_co_wrapper
576 bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
577 int generated_co_wrapper
578 bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
579 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
580                       int64_t pos, int size);
581 
582 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
583                       int64_t pos, int size);
584 
585 void bdrv_img_create(const char *filename, const char *fmt,
586                      const char *base_filename, const char *base_fmt,
587                      char *options, uint64_t img_size, int flags,
588                      bool quiet, Error **errp);
589 
590 /* Returns the alignment in bytes that is required so that no bounce buffer
591  * is required throughout the stack */
592 size_t bdrv_min_mem_align(BlockDriverState *bs);
593 /* Returns optimal alignment in bytes for bounce buffer */
594 size_t bdrv_opt_mem_align(BlockDriverState *bs);
595 void *qemu_blockalign(BlockDriverState *bs, size_t size);
596 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
597 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
598 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
599 bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov);
600 
601 void bdrv_enable_copy_on_read(BlockDriverState *bs);
602 void bdrv_disable_copy_on_read(BlockDriverState *bs);
603 
604 void bdrv_ref(BlockDriverState *bs);
605 void bdrv_unref(BlockDriverState *bs);
606 void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child);
607 BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
608                              BlockDriverState *child_bs,
609                              const char *child_name,
610                              const BdrvChildClass *child_class,
611                              BdrvChildRole child_role,
612                              Error **errp);
613 
614 bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp);
615 void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason);
616 void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason);
617 void bdrv_op_block_all(BlockDriverState *bs, Error *reason);
618 void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason);
619 bool bdrv_op_blocker_is_empty(BlockDriverState *bs);
620 
621 #define BLKDBG_EVENT(child, evt) \
622     do { \
623         if (child) { \
624             bdrv_debug_event(child->bs, evt); \
625         } \
626     } while (0)
627 
628 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
629 
630 int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event,
631                            const char *tag);
632 int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag);
633 int bdrv_debug_resume(BlockDriverState *bs, const char *tag);
634 bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag);
635 
636 /**
637  * bdrv_get_aio_context:
638  *
639  * Returns: the currently bound #AioContext
640  */
641 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
642 
643 /**
644  * Transfer control to @co in the aio context of @bs
645  */
646 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co);
647 
648 void bdrv_set_aio_context_ignore(BlockDriverState *bs,
649                                  AioContext *new_context, GSList **ignore);
650 int bdrv_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
651                              Error **errp);
652 int bdrv_child_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
653                                    BdrvChild *ignore_child, Error **errp);
654 bool bdrv_child_can_set_aio_context(BdrvChild *c, AioContext *ctx,
655                                     GSList **ignore, Error **errp);
656 bool bdrv_can_set_aio_context(BlockDriverState *bs, AioContext *ctx,
657                               GSList **ignore, Error **errp);
658 int bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz);
659 int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo);
660 
661 void bdrv_io_plug(BlockDriverState *bs);
662 void bdrv_io_unplug(BlockDriverState *bs);
663 
664 /**
665  * bdrv_parent_drained_begin_single:
666  *
667  * Begin a quiesced section for the parent of @c. If @poll is true, wait for
668  * any pending activity to cease.
669  */
670 void bdrv_parent_drained_begin_single(BdrvChild *c, bool poll);
671 
672 /**
673  * bdrv_parent_drained_end_single:
674  *
675  * End a quiesced section for the parent of @c.
676  *
677  * This polls @bs's AioContext until all scheduled sub-drained_ends
678  * have settled, which may result in graph changes.
679  */
680 void bdrv_parent_drained_end_single(BdrvChild *c);
681 
682 /**
683  * bdrv_drain_poll:
684  *
685  * Poll for pending requests in @bs, its parents (except for @ignore_parent),
686  * and if @recursive is true its children as well (used for subtree drain).
687  *
688  * If @ignore_bds_parents is true, parents that are BlockDriverStates must
689  * ignore the drain request because they will be drained separately (used for
690  * drain_all).
691  *
692  * This is part of bdrv_drained_begin.
693  */
694 bool bdrv_drain_poll(BlockDriverState *bs, bool recursive,
695                      BdrvChild *ignore_parent, bool ignore_bds_parents);
696 
697 /**
698  * bdrv_drained_begin:
699  *
700  * Begin a quiesced section for exclusive access to the BDS, by disabling
701  * external request sources including NBD server and device model. Note that
702  * this doesn't block timers or coroutines from submitting more requests, which
703  * means block_job_pause is still necessary.
704  *
705  * This function can be recursive.
706  */
707 void bdrv_drained_begin(BlockDriverState *bs);
708 
709 /**
710  * bdrv_do_drained_begin_quiesce:
711  *
712  * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already
713  * running requests to complete.
714  */
715 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs,
716                                    BdrvChild *parent, bool ignore_bds_parents);
717 
718 /**
719  * Like bdrv_drained_begin, but recursively begins a quiesced section for
720  * exclusive access to all child nodes as well.
721  */
722 void bdrv_subtree_drained_begin(BlockDriverState *bs);
723 
724 /**
725  * bdrv_drained_end:
726  *
727  * End a quiescent section started by bdrv_drained_begin().
728  *
729  * This polls @bs's AioContext until all scheduled sub-drained_ends
730  * have settled.  On one hand, that may result in graph changes.  On
731  * the other, this requires that the caller either runs in the main
732  * loop; or that all involved nodes (@bs and all of its parents) are
733  * in the caller's AioContext.
734  */
735 void bdrv_drained_end(BlockDriverState *bs);
736 
737 /**
738  * bdrv_drained_end_no_poll:
739  *
740  * Same as bdrv_drained_end(), but do not poll for the subgraph to
741  * actually become unquiesced.  Therefore, no graph changes will occur
742  * with this function.
743  *
744  * *drained_end_counter is incremented for every background operation
745  * that is scheduled, and will be decremented for every operation once
746  * it settles.  The caller must poll until it reaches 0.  The counter
747  * should be accessed using atomic operations only.
748  */
749 void bdrv_drained_end_no_poll(BlockDriverState *bs, int *drained_end_counter);
750 
751 /**
752  * End a quiescent section started by bdrv_subtree_drained_begin().
753  */
754 void bdrv_subtree_drained_end(BlockDriverState *bs);
755 
756 void bdrv_add_child(BlockDriverState *parent, BlockDriverState *child,
757                     Error **errp);
758 void bdrv_del_child(BlockDriverState *parent, BdrvChild *child, Error **errp);
759 
760 bool bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
761                                      uint32_t granularity, Error **errp);
762 /**
763  *
764  * bdrv_register_buf/bdrv_unregister_buf:
765  *
766  * Register/unregister a buffer for I/O. For example, VFIO drivers are
767  * interested to know the memory areas that would later be used for I/O, so
768  * that they can prepare IOMMU mapping etc., to get better performance.
769  */
770 void bdrv_register_buf(BlockDriverState *bs, void *host, size_t size);
771 void bdrv_unregister_buf(BlockDriverState *bs, void *host);
772 
773 /**
774  *
775  * bdrv_co_copy_range:
776  *
777  * Do offloaded copy between two children. If the operation is not implemented
778  * by the driver, or if the backend storage doesn't support it, a negative
779  * error code will be returned.
780  *
781  * Note: block layer doesn't emulate or fallback to a bounce buffer approach
782  * because usually the caller shouldn't attempt offloaded copy any more (e.g.
783  * calling copy_file_range(2)) after the first error, thus it should fall back
784  * to a read+write path in the caller level.
785  *
786  * @src: Source child to copy data from
787  * @src_offset: offset in @src image to read data
788  * @dst: Destination child to copy data to
789  * @dst_offset: offset in @dst image to write data
790  * @bytes: number of bytes to copy
791  * @flags: request flags. Supported flags:
792  *         BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero
793  *                               write on @dst as if bdrv_co_pwrite_zeroes is
794  *                               called. Used to simplify caller code, or
795  *                               during BlockDriver.bdrv_co_copy_range_from()
796  *                               recursion.
797  *         BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping
798  *                                   requests currently in flight.
799  *
800  * Returns: 0 if succeeded; negative error code if failed.
801  **/
802 int coroutine_fn bdrv_co_copy_range(BdrvChild *src, uint64_t src_offset,
803                                     BdrvChild *dst, uint64_t dst_offset,
804                                     uint64_t bytes, BdrvRequestFlags read_flags,
805                                     BdrvRequestFlags write_flags);
806 #endif
807