xref: /openbmc/qemu/include/block/block.h (revision 4c4465ff)
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     /* Forces request serialisation. Use only with write requests. */
67     BDRV_REQ_SERIALISING        = 0x80,
68 
69     /* Execute the request only if the operation can be offloaded or otherwise
70      * be executed efficiently, but return an error instead of using a slow
71      * fallback. */
72     BDRV_REQ_NO_FALLBACK        = 0x100,
73 
74     /*
75      * BDRV_REQ_PREFETCH may be used only together with BDRV_REQ_COPY_ON_READ
76      * on read request and means that caller doesn't really need data to be
77      * written to qiov parameter which may be NULL.
78      */
79     BDRV_REQ_PREFETCH  = 0x200,
80 
81     /*
82      * If we need to wait for other requests, just fail immediately. Used
83      * only together with BDRV_REQ_SERIALISING.
84      */
85     BDRV_REQ_NO_WAIT = 0x400,
86 
87     /* Mask of valid flags */
88     BDRV_REQ_MASK               = 0x7ff,
89 } BdrvRequestFlags;
90 
91 typedef struct BlockSizes {
92     uint32_t phys;
93     uint32_t log;
94 } BlockSizes;
95 
96 typedef struct HDGeometry {
97     uint32_t heads;
98     uint32_t sectors;
99     uint32_t cylinders;
100 } HDGeometry;
101 
102 #define BDRV_O_RDWR        0x0002
103 #define BDRV_O_RESIZE      0x0004 /* request permission for resizing the node */
104 #define BDRV_O_SNAPSHOT    0x0008 /* open the file read only and save writes in a snapshot */
105 #define BDRV_O_TEMPORARY   0x0010 /* delete the file after use */
106 #define BDRV_O_NOCACHE     0x0020 /* do not use the host page cache */
107 #define BDRV_O_NATIVE_AIO  0x0080 /* use native AIO instead of the thread pool */
108 #define BDRV_O_NO_BACKING  0x0100 /* don't open the backing file */
109 #define BDRV_O_NO_FLUSH    0x0200 /* disable flushing on this disk */
110 #define BDRV_O_COPY_ON_READ 0x0400 /* copy read backing sectors into image */
111 #define BDRV_O_INACTIVE    0x0800  /* consistency hint for migration handoff */
112 #define BDRV_O_CHECK       0x1000  /* open solely for consistency check */
113 #define BDRV_O_ALLOW_RDWR  0x2000  /* allow reopen to change from r/o to r/w */
114 #define BDRV_O_UNMAP       0x4000  /* execute guest UNMAP/TRIM operations */
115 #define BDRV_O_PROTOCOL    0x8000  /* if no block driver is explicitly given:
116                                       select an appropriate protocol driver,
117                                       ignoring the format layer */
118 #define BDRV_O_NO_IO       0x10000 /* don't initialize for I/O */
119 #define BDRV_O_AUTO_RDONLY 0x20000 /* degrade to read-only if opening read-write fails */
120 #define BDRV_O_IO_URING    0x40000 /* use io_uring instead of the thread pool */
121 
122 #define BDRV_O_CACHE_MASK  (BDRV_O_NOCACHE | BDRV_O_NO_FLUSH)
123 
124 
125 /* Option names of options parsed by the block layer */
126 
127 #define BDRV_OPT_CACHE_WB       "cache.writeback"
128 #define BDRV_OPT_CACHE_DIRECT   "cache.direct"
129 #define BDRV_OPT_CACHE_NO_FLUSH "cache.no-flush"
130 #define BDRV_OPT_READ_ONLY      "read-only"
131 #define BDRV_OPT_AUTO_READ_ONLY "auto-read-only"
132 #define BDRV_OPT_DISCARD        "discard"
133 #define BDRV_OPT_FORCE_SHARE    "force-share"
134 
135 
136 #define BDRV_SECTOR_BITS   9
137 #define BDRV_SECTOR_SIZE   (1ULL << BDRV_SECTOR_BITS)
138 
139 #define BDRV_REQUEST_MAX_SECTORS MIN_CONST(SIZE_MAX >> BDRV_SECTOR_BITS, \
140                                            INT_MAX >> BDRV_SECTOR_BITS)
141 #define BDRV_REQUEST_MAX_BYTES (BDRV_REQUEST_MAX_SECTORS << BDRV_SECTOR_BITS)
142 
143 /*
144  * We want allow aligning requests and disk length up to any 32bit alignment
145  * and don't afraid of overflow.
146  * To achieve it, and in the same time use some pretty number as maximum disk
147  * size, let's define maximum "length" (a limit for any offset/bytes request and
148  * for disk size) to be the greatest power of 2 less than INT64_MAX.
149  */
150 #define BDRV_MAX_ALIGNMENT (1L << 30)
151 #define BDRV_MAX_LENGTH (QEMU_ALIGN_DOWN(INT64_MAX, BDRV_MAX_ALIGNMENT))
152 
153 /*
154  * Allocation status flags for bdrv_block_status() and friends.
155  *
156  * Public flags:
157  * BDRV_BLOCK_DATA: allocation for data at offset is tied to this layer
158  * BDRV_BLOCK_ZERO: offset reads as zero
159  * BDRV_BLOCK_OFFSET_VALID: an associated offset exists for accessing raw data
160  * BDRV_BLOCK_ALLOCATED: the content of the block is determined by this
161  *                       layer rather than any backing, set by block layer
162  * BDRV_BLOCK_EOF: the returned pnum covers through end of file for this
163  *                 layer, set by block layer
164  *
165  * Internal flags:
166  * BDRV_BLOCK_RAW: for use by passthrough drivers, such as raw, to request
167  *                 that the block layer recompute the answer from the returned
168  *                 BDS; must be accompanied by just BDRV_BLOCK_OFFSET_VALID.
169  * BDRV_BLOCK_RECURSE: request that the block layer will recursively search for
170  *                     zeroes in file child of current block node inside
171  *                     returned region. Only valid together with both
172  *                     BDRV_BLOCK_DATA and BDRV_BLOCK_OFFSET_VALID. Should not
173  *                     appear with BDRV_BLOCK_ZERO.
174  *
175  * If BDRV_BLOCK_OFFSET_VALID is set, the map parameter represents the
176  * host offset within the returned BDS that is allocated for the
177  * corresponding raw guest data.  However, whether that offset
178  * actually contains data also depends on BDRV_BLOCK_DATA, as follows:
179  *
180  * DATA ZERO OFFSET_VALID
181  *  t    t        t       sectors read as zero, returned file is zero at offset
182  *  t    f        t       sectors read as valid from file at offset
183  *  f    t        t       sectors preallocated, read as zero, returned file not
184  *                        necessarily zero at offset
185  *  f    f        t       sectors preallocated but read from backing_hd,
186  *                        returned file contains garbage at offset
187  *  t    t        f       sectors preallocated, read as zero, unknown offset
188  *  t    f        f       sectors read from unknown file or offset
189  *  f    t        f       not allocated or unknown offset, read as zero
190  *  f    f        f       not allocated or unknown offset, read from backing_hd
191  */
192 #define BDRV_BLOCK_DATA         0x01
193 #define BDRV_BLOCK_ZERO         0x02
194 #define BDRV_BLOCK_OFFSET_VALID 0x04
195 #define BDRV_BLOCK_RAW          0x08
196 #define BDRV_BLOCK_ALLOCATED    0x10
197 #define BDRV_BLOCK_EOF          0x20
198 #define BDRV_BLOCK_RECURSE      0x40
199 
200 typedef QTAILQ_HEAD(BlockReopenQueue, BlockReopenQueueEntry) BlockReopenQueue;
201 
202 typedef struct BDRVReopenState {
203     BlockDriverState *bs;
204     int flags;
205     BlockdevDetectZeroesOptions detect_zeroes;
206     bool backing_missing;
207     bool replace_backing_bs;  /* new_backing_bs is ignored if this is false */
208     BlockDriverState *new_backing_bs; /* If NULL then detach the current bs */
209     uint64_t perm, shared_perm;
210     QDict *options;
211     QDict *explicit_options;
212     void *opaque;
213 } BDRVReopenState;
214 
215 /*
216  * Block operation types
217  */
218 typedef enum BlockOpType {
219     BLOCK_OP_TYPE_BACKUP_SOURCE,
220     BLOCK_OP_TYPE_BACKUP_TARGET,
221     BLOCK_OP_TYPE_CHANGE,
222     BLOCK_OP_TYPE_COMMIT_SOURCE,
223     BLOCK_OP_TYPE_COMMIT_TARGET,
224     BLOCK_OP_TYPE_DATAPLANE,
225     BLOCK_OP_TYPE_DRIVE_DEL,
226     BLOCK_OP_TYPE_EJECT,
227     BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
228     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT,
229     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE,
230     BLOCK_OP_TYPE_MIRROR_SOURCE,
231     BLOCK_OP_TYPE_MIRROR_TARGET,
232     BLOCK_OP_TYPE_RESIZE,
233     BLOCK_OP_TYPE_STREAM,
234     BLOCK_OP_TYPE_REPLACE,
235     BLOCK_OP_TYPE_MAX,
236 } BlockOpType;
237 
238 /* Block node permission constants */
239 enum {
240     /**
241      * A user that has the "permission" of consistent reads is guaranteed that
242      * their view of the contents of the block device is complete and
243      * self-consistent, representing the contents of a disk at a specific
244      * point.
245      *
246      * For most block devices (including their backing files) this is true, but
247      * the property cannot be maintained in a few situations like for
248      * intermediate nodes of a commit block job.
249      */
250     BLK_PERM_CONSISTENT_READ    = 0x01,
251 
252     /** This permission is required to change the visible disk contents. */
253     BLK_PERM_WRITE              = 0x02,
254 
255     /**
256      * This permission (which is weaker than BLK_PERM_WRITE) is both enough and
257      * required for writes to the block node when the caller promises that
258      * the visible disk content doesn't change.
259      *
260      * As the BLK_PERM_WRITE permission is strictly stronger, either is
261      * sufficient to perform an unchanging write.
262      */
263     BLK_PERM_WRITE_UNCHANGED    = 0x04,
264 
265     /** This permission is required to change the size of a block node. */
266     BLK_PERM_RESIZE             = 0x08,
267 
268     /**
269      * This permission is required to change the node that this BdrvChild
270      * points to.
271      */
272     BLK_PERM_GRAPH_MOD          = 0x10,
273 
274     BLK_PERM_ALL                = 0x1f,
275 
276     DEFAULT_PERM_PASSTHROUGH    = BLK_PERM_CONSISTENT_READ
277                                  | BLK_PERM_WRITE
278                                  | BLK_PERM_WRITE_UNCHANGED
279                                  | BLK_PERM_RESIZE,
280 
281     DEFAULT_PERM_UNCHANGED      = BLK_PERM_ALL & ~DEFAULT_PERM_PASSTHROUGH,
282 };
283 
284 /*
285  * Flags that parent nodes assign to child nodes to specify what kind of
286  * role(s) they take.
287  *
288  * At least one of DATA, METADATA, FILTERED, or COW must be set for
289  * every child.
290  */
291 enum BdrvChildRoleBits {
292     /*
293      * This child stores data.
294      * Any node may have an arbitrary number of such children.
295      */
296     BDRV_CHILD_DATA         = (1 << 0),
297 
298     /*
299      * This child stores metadata.
300      * Any node may have an arbitrary number of metadata-storing
301      * children.
302      */
303     BDRV_CHILD_METADATA     = (1 << 1),
304 
305     /*
306      * A child that always presents exactly the same visible data as
307      * the parent, e.g. by virtue of the parent forwarding all reads
308      * and writes.
309      * This flag is mutually exclusive with DATA, METADATA, and COW.
310      * Any node may have at most one filtered child at a time.
311      */
312     BDRV_CHILD_FILTERED     = (1 << 2),
313 
314     /*
315      * Child from which to read all data that isn't allocated in the
316      * parent (i.e., the backing child); such data is copied to the
317      * parent through COW (and optionally COR).
318      * This field is mutually exclusive with DATA, METADATA, and
319      * FILTERED.
320      * Any node may have at most one such backing child at a time.
321      */
322     BDRV_CHILD_COW          = (1 << 3),
323 
324     /*
325      * The primary child.  For most drivers, this is the child whose
326      * filename applies best to the parent node.
327      * Any node may have at most one primary child at a time.
328      */
329     BDRV_CHILD_PRIMARY      = (1 << 4),
330 
331     /* Useful combination of flags */
332     BDRV_CHILD_IMAGE        = BDRV_CHILD_DATA
333                               | BDRV_CHILD_METADATA
334                               | BDRV_CHILD_PRIMARY,
335 };
336 
337 /* Mask of BdrvChildRoleBits values */
338 typedef unsigned int BdrvChildRole;
339 
340 char *bdrv_perm_names(uint64_t perm);
341 uint64_t bdrv_qapi_perm_to_blk_perm(BlockPermission qapi_perm);
342 
343 /* disk I/O throttling */
344 void bdrv_init(void);
345 void bdrv_init_with_whitelist(void);
346 bool bdrv_uses_whitelist(void);
347 int bdrv_is_whitelisted(BlockDriver *drv, bool read_only);
348 BlockDriver *bdrv_find_protocol(const char *filename,
349                                 bool allow_protocol_prefix,
350                                 Error **errp);
351 BlockDriver *bdrv_find_format(const char *format_name);
352 int bdrv_create(BlockDriver *drv, const char* filename,
353                 QemuOpts *opts, Error **errp);
354 int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp);
355 
356 BlockDriverState *bdrv_new(void);
357 void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top,
358                  Error **errp);
359 void bdrv_replace_node(BlockDriverState *from, BlockDriverState *to,
360                        Error **errp);
361 
362 int bdrv_parse_aio(const char *mode, int *flags);
363 int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough);
364 int bdrv_parse_discard_flags(const char *mode, int *flags);
365 BdrvChild *bdrv_open_child(const char *filename,
366                            QDict *options, const char *bdref_key,
367                            BlockDriverState* parent,
368                            const BdrvChildClass *child_class,
369                            BdrvChildRole child_role,
370                            bool allow_none, Error **errp);
371 BlockDriverState *bdrv_open_blockdev_ref(BlockdevRef *ref, Error **errp);
372 void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd,
373                          Error **errp);
374 int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options,
375                            const char *bdref_key, Error **errp);
376 BlockDriverState *bdrv_open(const char *filename, const char *reference,
377                             QDict *options, int flags, Error **errp);
378 BlockDriverState *bdrv_new_open_driver(BlockDriver *drv, const char *node_name,
379                                        int flags, Error **errp);
380 BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue,
381                                     BlockDriverState *bs, QDict *options,
382                                     bool keep_old_opts);
383 int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp);
384 int bdrv_reopen_set_read_only(BlockDriverState *bs, bool read_only,
385                               Error **errp);
386 int bdrv_reopen_prepare(BDRVReopenState *reopen_state,
387                         BlockReopenQueue *queue, Error **errp);
388 void bdrv_reopen_commit(BDRVReopenState *reopen_state);
389 void bdrv_reopen_abort(BDRVReopenState *reopen_state);
390 int bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset,
391                        int bytes, BdrvRequestFlags flags);
392 int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags);
393 int bdrv_pread(BdrvChild *child, int64_t offset, void *buf, int bytes);
394 int bdrv_pwrite(BdrvChild *child, int64_t offset, const void *buf, int bytes);
395 int bdrv_pwrite_sync(BdrvChild *child, int64_t offset,
396                      const void *buf, int count);
397 /*
398  * Efficiently zero a region of the disk image.  Note that this is a regular
399  * I/O request like read or write and should have a reasonable size.  This
400  * function is not suitable for zeroing the entire image in a single request
401  * because it may allocate memory for the entire region.
402  */
403 int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset,
404                                        int bytes, BdrvRequestFlags flags);
405 BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,
406     const char *backing_file);
407 void bdrv_refresh_filename(BlockDriverState *bs);
408 
409 int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset, bool exact,
410                                   PreallocMode prealloc, BdrvRequestFlags flags,
411                                   Error **errp);
412 int generated_co_wrapper
413 bdrv_truncate(BdrvChild *child, int64_t offset, bool exact,
414               PreallocMode prealloc, BdrvRequestFlags flags, Error **errp);
415 
416 int64_t bdrv_nb_sectors(BlockDriverState *bs);
417 int64_t bdrv_getlength(BlockDriverState *bs);
418 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs);
419 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
420                                BlockDriverState *in_bs, Error **errp);
421 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr);
422 void bdrv_refresh_limits(BlockDriverState *bs, Error **errp);
423 int bdrv_commit(BlockDriverState *bs);
424 int bdrv_make_empty(BdrvChild *c, Error **errp);
425 int bdrv_change_backing_file(BlockDriverState *bs, const char *backing_file,
426                              const char *backing_fmt, bool warn);
427 void bdrv_register(BlockDriver *bdrv);
428 int bdrv_drop_intermediate(BlockDriverState *top, BlockDriverState *base,
429                            const char *backing_file_str);
430 BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
431                                     BlockDriverState *bs);
432 BlockDriverState *bdrv_find_base(BlockDriverState *bs);
433 bool bdrv_is_backing_chain_frozen(BlockDriverState *bs, BlockDriverState *base,
434                                   Error **errp);
435 int bdrv_freeze_backing_chain(BlockDriverState *bs, BlockDriverState *base,
436                               Error **errp);
437 void bdrv_unfreeze_backing_chain(BlockDriverState *bs, BlockDriverState *base);
438 int coroutine_fn bdrv_co_delete_file(BlockDriverState *bs, Error **errp);
439 
440 
441 typedef struct BdrvCheckResult {
442     int corruptions;
443     int leaks;
444     int check_errors;
445     int corruptions_fixed;
446     int leaks_fixed;
447     int64_t image_end_offset;
448     BlockFragInfo bfi;
449 } BdrvCheckResult;
450 
451 typedef enum {
452     BDRV_FIX_LEAKS    = 1,
453     BDRV_FIX_ERRORS   = 2,
454 } BdrvCheckMode;
455 
456 int generated_co_wrapper bdrv_check(BlockDriverState *bs, BdrvCheckResult *res,
457                                     BdrvCheckMode fix);
458 
459 /* The units of offset and total_work_size may be chosen arbitrarily by the
460  * block driver; total_work_size may change during the course of the amendment
461  * operation */
462 typedef void BlockDriverAmendStatusCB(BlockDriverState *bs, int64_t offset,
463                                       int64_t total_work_size, void *opaque);
464 int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts,
465                        BlockDriverAmendStatusCB *status_cb, void *cb_opaque,
466                        bool force,
467                        Error **errp);
468 
469 /* check if a named node can be replaced when doing drive-mirror */
470 BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs,
471                                         const char *node_name, Error **errp);
472 
473 /* async block I/O */
474 void bdrv_aio_cancel(BlockAIOCB *acb);
475 void bdrv_aio_cancel_async(BlockAIOCB *acb);
476 
477 /* sg packet commands */
478 int bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf);
479 
480 /* Invalidate any cached metadata used by image formats */
481 int generated_co_wrapper bdrv_invalidate_cache(BlockDriverState *bs,
482                                                Error **errp);
483 void bdrv_invalidate_cache_all(Error **errp);
484 int bdrv_inactivate_all(void);
485 
486 /* Ensure contents are flushed to disk.  */
487 int generated_co_wrapper bdrv_flush(BlockDriverState *bs);
488 int coroutine_fn bdrv_co_flush(BlockDriverState *bs);
489 int bdrv_flush_all(void);
490 void bdrv_close_all(void);
491 void bdrv_drain(BlockDriverState *bs);
492 void coroutine_fn bdrv_co_drain(BlockDriverState *bs);
493 void bdrv_drain_all_begin(void);
494 void bdrv_drain_all_end(void);
495 void bdrv_drain_all(void);
496 
497 #define BDRV_POLL_WHILE(bs, cond) ({                       \
498     BlockDriverState *bs_ = (bs);                          \
499     AIO_WAIT_WHILE(bdrv_get_aio_context(bs_),              \
500                    cond); })
501 
502 int generated_co_wrapper bdrv_pdiscard(BdrvChild *child, int64_t offset,
503                                        int64_t bytes);
504 int bdrv_co_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes);
505 int bdrv_has_zero_init_1(BlockDriverState *bs);
506 int bdrv_has_zero_init(BlockDriverState *bs);
507 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs);
508 int bdrv_block_status(BlockDriverState *bs, int64_t offset,
509                       int64_t bytes, int64_t *pnum, int64_t *map,
510                       BlockDriverState **file);
511 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base,
512                             int64_t offset, int64_t bytes, int64_t *pnum,
513                             int64_t *map, BlockDriverState **file);
514 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
515                       int64_t *pnum);
516 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
517                             bool include_base, int64_t offset, int64_t bytes,
518                             int64_t *pnum);
519 int coroutine_fn bdrv_co_is_zero_fast(BlockDriverState *bs, int64_t offset,
520                                       int64_t bytes);
521 
522 bool bdrv_is_read_only(BlockDriverState *bs);
523 int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only,
524                            bool ignore_allow_rdw, Error **errp);
525 int bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg,
526                               Error **errp);
527 bool bdrv_is_writable(BlockDriverState *bs);
528 bool bdrv_is_sg(BlockDriverState *bs);
529 bool bdrv_is_inserted(BlockDriverState *bs);
530 void bdrv_lock_medium(BlockDriverState *bs, bool locked);
531 void bdrv_eject(BlockDriverState *bs, bool eject_flag);
532 const char *bdrv_get_format_name(BlockDriverState *bs);
533 BlockDriverState *bdrv_find_node(const char *node_name);
534 BlockDeviceInfoList *bdrv_named_nodes_list(bool flat, Error **errp);
535 XDbgBlockGraph *bdrv_get_xdbg_block_graph(Error **errp);
536 BlockDriverState *bdrv_lookup_bs(const char *device,
537                                  const char *node_name,
538                                  Error **errp);
539 bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base);
540 BlockDriverState *bdrv_next_node(BlockDriverState *bs);
541 BlockDriverState *bdrv_next_all_states(BlockDriverState *bs);
542 
543 typedef struct BdrvNextIterator {
544     enum {
545         BDRV_NEXT_BACKEND_ROOTS,
546         BDRV_NEXT_MONITOR_OWNED,
547     } phase;
548     BlockBackend *blk;
549     BlockDriverState *bs;
550 } BdrvNextIterator;
551 
552 BlockDriverState *bdrv_first(BdrvNextIterator *it);
553 BlockDriverState *bdrv_next(BdrvNextIterator *it);
554 void bdrv_next_cleanup(BdrvNextIterator *it);
555 
556 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs);
557 bool bdrv_supports_compressed_writes(BlockDriverState *bs);
558 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
559                          void *opaque, bool read_only);
560 const char *bdrv_get_node_name(const BlockDriverState *bs);
561 const char *bdrv_get_device_name(const BlockDriverState *bs);
562 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
563 int bdrv_get_flags(BlockDriverState *bs);
564 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
565 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs,
566                                           Error **errp);
567 BlockStatsSpecific *bdrv_get_specific_stats(BlockDriverState *bs);
568 void bdrv_round_to_clusters(BlockDriverState *bs,
569                             int64_t offset, int64_t bytes,
570                             int64_t *cluster_offset,
571                             int64_t *cluster_bytes);
572 
573 void bdrv_get_backing_filename(BlockDriverState *bs,
574                                char *filename, int filename_size);
575 char *bdrv_get_full_backing_filename(BlockDriverState *bs, Error **errp);
576 char *bdrv_get_full_backing_filename_from_filename(const char *backed,
577                                                    const char *backing,
578                                                    Error **errp);
579 char *bdrv_dirname(BlockDriverState *bs, Error **errp);
580 
581 int path_has_protocol(const char *path);
582 int path_is_absolute(const char *path);
583 char *path_combine(const char *base_path, const char *filename);
584 
585 int generated_co_wrapper
586 bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
587 int generated_co_wrapper
588 bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
589 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
590                       int64_t pos, int size);
591 
592 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
593                       int64_t pos, int size);
594 
595 void bdrv_img_create(const char *filename, const char *fmt,
596                      const char *base_filename, const char *base_fmt,
597                      char *options, uint64_t img_size, int flags,
598                      bool quiet, Error **errp);
599 
600 /* Returns the alignment in bytes that is required so that no bounce buffer
601  * is required throughout the stack */
602 size_t bdrv_min_mem_align(BlockDriverState *bs);
603 /* Returns optimal alignment in bytes for bounce buffer */
604 size_t bdrv_opt_mem_align(BlockDriverState *bs);
605 void *qemu_blockalign(BlockDriverState *bs, size_t size);
606 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
607 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
608 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
609 bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov);
610 
611 void bdrv_enable_copy_on_read(BlockDriverState *bs);
612 void bdrv_disable_copy_on_read(BlockDriverState *bs);
613 
614 void bdrv_ref(BlockDriverState *bs);
615 void bdrv_unref(BlockDriverState *bs);
616 void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child);
617 BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
618                              BlockDriverState *child_bs,
619                              const char *child_name,
620                              const BdrvChildClass *child_class,
621                              BdrvChildRole child_role,
622                              Error **errp);
623 
624 bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp);
625 void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason);
626 void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason);
627 void bdrv_op_block_all(BlockDriverState *bs, Error *reason);
628 void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason);
629 bool bdrv_op_blocker_is_empty(BlockDriverState *bs);
630 
631 #define BLKDBG_EVENT(child, evt) \
632     do { \
633         if (child) { \
634             bdrv_debug_event(child->bs, evt); \
635         } \
636     } while (0)
637 
638 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
639 
640 int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event,
641                            const char *tag);
642 int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag);
643 int bdrv_debug_resume(BlockDriverState *bs, const char *tag);
644 bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag);
645 
646 /**
647  * bdrv_get_aio_context:
648  *
649  * Returns: the currently bound #AioContext
650  */
651 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
652 
653 /**
654  * Move the current coroutine to the AioContext of @bs and return the old
655  * AioContext of the coroutine. Increase bs->in_flight so that draining @bs
656  * will wait for the operation to proceed until the corresponding
657  * bdrv_co_leave().
658  *
659  * Consequently, you can't call drain inside a bdrv_co_enter/leave() section as
660  * this will deadlock.
661  */
662 AioContext *coroutine_fn bdrv_co_enter(BlockDriverState *bs);
663 
664 /**
665  * Ends a section started by bdrv_co_enter(). Move the current coroutine back
666  * to old_ctx and decrease bs->in_flight again.
667  */
668 void coroutine_fn bdrv_co_leave(BlockDriverState *bs, AioContext *old_ctx);
669 
670 /**
671  * Locks the AioContext of @bs if it's not the current AioContext. This avoids
672  * double locking which could lead to deadlocks: This is a coroutine_fn, so we
673  * know we already own the lock of the current AioContext.
674  *
675  * May only be called in the main thread.
676  */
677 void coroutine_fn bdrv_co_lock(BlockDriverState *bs);
678 
679 /**
680  * Unlocks the AioContext of @bs if it's not the current AioContext.
681  */
682 void coroutine_fn bdrv_co_unlock(BlockDriverState *bs);
683 
684 /**
685  * Transfer control to @co in the aio context of @bs
686  */
687 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co);
688 
689 void bdrv_set_aio_context_ignore(BlockDriverState *bs,
690                                  AioContext *new_context, GSList **ignore);
691 int bdrv_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
692                              Error **errp);
693 int bdrv_child_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
694                                    BdrvChild *ignore_child, Error **errp);
695 bool bdrv_child_can_set_aio_context(BdrvChild *c, AioContext *ctx,
696                                     GSList **ignore, Error **errp);
697 bool bdrv_can_set_aio_context(BlockDriverState *bs, AioContext *ctx,
698                               GSList **ignore, Error **errp);
699 int bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz);
700 int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo);
701 
702 void bdrv_io_plug(BlockDriverState *bs);
703 void bdrv_io_unplug(BlockDriverState *bs);
704 
705 /**
706  * bdrv_parent_drained_begin_single:
707  *
708  * Begin a quiesced section for the parent of @c. If @poll is true, wait for
709  * any pending activity to cease.
710  */
711 void bdrv_parent_drained_begin_single(BdrvChild *c, bool poll);
712 
713 /**
714  * bdrv_parent_drained_end_single:
715  *
716  * End a quiesced section for the parent of @c.
717  *
718  * This polls @bs's AioContext until all scheduled sub-drained_ends
719  * have settled, which may result in graph changes.
720  */
721 void bdrv_parent_drained_end_single(BdrvChild *c);
722 
723 /**
724  * bdrv_drain_poll:
725  *
726  * Poll for pending requests in @bs, its parents (except for @ignore_parent),
727  * and if @recursive is true its children as well (used for subtree drain).
728  *
729  * If @ignore_bds_parents is true, parents that are BlockDriverStates must
730  * ignore the drain request because they will be drained separately (used for
731  * drain_all).
732  *
733  * This is part of bdrv_drained_begin.
734  */
735 bool bdrv_drain_poll(BlockDriverState *bs, bool recursive,
736                      BdrvChild *ignore_parent, bool ignore_bds_parents);
737 
738 /**
739  * bdrv_drained_begin:
740  *
741  * Begin a quiesced section for exclusive access to the BDS, by disabling
742  * external request sources including NBD server and device model. Note that
743  * this doesn't block timers or coroutines from submitting more requests, which
744  * means block_job_pause is still necessary.
745  *
746  * This function can be recursive.
747  */
748 void bdrv_drained_begin(BlockDriverState *bs);
749 
750 /**
751  * bdrv_do_drained_begin_quiesce:
752  *
753  * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already
754  * running requests to complete.
755  */
756 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs,
757                                    BdrvChild *parent, bool ignore_bds_parents);
758 
759 /**
760  * Like bdrv_drained_begin, but recursively begins a quiesced section for
761  * exclusive access to all child nodes as well.
762  */
763 void bdrv_subtree_drained_begin(BlockDriverState *bs);
764 
765 /**
766  * bdrv_drained_end:
767  *
768  * End a quiescent section started by bdrv_drained_begin().
769  *
770  * This polls @bs's AioContext until all scheduled sub-drained_ends
771  * have settled.  On one hand, that may result in graph changes.  On
772  * the other, this requires that the caller either runs in the main
773  * loop; or that all involved nodes (@bs and all of its parents) are
774  * in the caller's AioContext.
775  */
776 void bdrv_drained_end(BlockDriverState *bs);
777 
778 /**
779  * bdrv_drained_end_no_poll:
780  *
781  * Same as bdrv_drained_end(), but do not poll for the subgraph to
782  * actually become unquiesced.  Therefore, no graph changes will occur
783  * with this function.
784  *
785  * *drained_end_counter is incremented for every background operation
786  * that is scheduled, and will be decremented for every operation once
787  * it settles.  The caller must poll until it reaches 0.  The counter
788  * should be accessed using atomic operations only.
789  */
790 void bdrv_drained_end_no_poll(BlockDriverState *bs, int *drained_end_counter);
791 
792 /**
793  * End a quiescent section started by bdrv_subtree_drained_begin().
794  */
795 void bdrv_subtree_drained_end(BlockDriverState *bs);
796 
797 void bdrv_add_child(BlockDriverState *parent, BlockDriverState *child,
798                     Error **errp);
799 void bdrv_del_child(BlockDriverState *parent, BdrvChild *child, Error **errp);
800 
801 bool bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
802                                      uint32_t granularity, Error **errp);
803 /**
804  *
805  * bdrv_register_buf/bdrv_unregister_buf:
806  *
807  * Register/unregister a buffer for I/O. For example, VFIO drivers are
808  * interested to know the memory areas that would later be used for I/O, so
809  * that they can prepare IOMMU mapping etc., to get better performance.
810  */
811 void bdrv_register_buf(BlockDriverState *bs, void *host, size_t size);
812 void bdrv_unregister_buf(BlockDriverState *bs, void *host);
813 
814 /**
815  *
816  * bdrv_co_copy_range:
817  *
818  * Do offloaded copy between two children. If the operation is not implemented
819  * by the driver, or if the backend storage doesn't support it, a negative
820  * error code will be returned.
821  *
822  * Note: block layer doesn't emulate or fallback to a bounce buffer approach
823  * because usually the caller shouldn't attempt offloaded copy any more (e.g.
824  * calling copy_file_range(2)) after the first error, thus it should fall back
825  * to a read+write path in the caller level.
826  *
827  * @src: Source child to copy data from
828  * @src_offset: offset in @src image to read data
829  * @dst: Destination child to copy data to
830  * @dst_offset: offset in @dst image to write data
831  * @bytes: number of bytes to copy
832  * @flags: request flags. Supported flags:
833  *         BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero
834  *                               write on @dst as if bdrv_co_pwrite_zeroes is
835  *                               called. Used to simplify caller code, or
836  *                               during BlockDriver.bdrv_co_copy_range_from()
837  *                               recursion.
838  *         BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping
839  *                                   requests currently in flight.
840  *
841  * Returns: 0 if succeeded; negative error code if failed.
842  **/
843 int coroutine_fn bdrv_co_copy_range(BdrvChild *src, uint64_t src_offset,
844                                     BdrvChild *dst, uint64_t dst_offset,
845                                     uint64_t bytes, BdrvRequestFlags read_flags,
846                                     BdrvRequestFlags write_flags);
847 #endif
848