xref: /openbmc/qemu/include/block/block.h (revision 91654e61)
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 #include "qemu/transactions.h"
13 
14 /*
15  * generated_co_wrapper
16  *
17  * Function specifier, which does nothing but mark functions to be
18  * generated by scripts/block-coroutine-wrapper.py
19  *
20  * Read more in docs/devel/block-coroutine-wrapper.rst
21  */
22 #define generated_co_wrapper
23 
24 /* block.c */
25 typedef struct BlockDriver BlockDriver;
26 typedef struct BdrvChild BdrvChild;
27 typedef struct BdrvChildClass BdrvChildClass;
28 
29 typedef struct BlockDriverInfo {
30     /* in bytes, 0 if irrelevant */
31     int cluster_size;
32     /* offset at which the VM state can be saved (0 if not possible) */
33     int64_t vm_state_offset;
34     bool is_dirty;
35     /*
36      * True if this block driver only supports compressed writes
37      */
38     bool needs_compressed_writes;
39 } BlockDriverInfo;
40 
41 typedef struct BlockFragInfo {
42     uint64_t allocated_clusters;
43     uint64_t total_clusters;
44     uint64_t fragmented_clusters;
45     uint64_t compressed_clusters;
46 } BlockFragInfo;
47 
48 typedef enum {
49     BDRV_REQ_COPY_ON_READ       = 0x1,
50     BDRV_REQ_ZERO_WRITE         = 0x2,
51 
52     /*
53      * The BDRV_REQ_MAY_UNMAP flag is used in write_zeroes requests to indicate
54      * that the block driver should unmap (discard) blocks if it is guaranteed
55      * that the result will read back as zeroes. The flag is only passed to the
56      * driver if the block device is opened with BDRV_O_UNMAP.
57      */
58     BDRV_REQ_MAY_UNMAP          = 0x4,
59 
60     BDRV_REQ_FUA                = 0x10,
61     BDRV_REQ_WRITE_COMPRESSED   = 0x20,
62 
63     /* Signifies that this write request will not change the visible disk
64      * content. */
65     BDRV_REQ_WRITE_UNCHANGED    = 0x40,
66 
67     /* Forces request serialisation. Use only with write requests. */
68     BDRV_REQ_SERIALISING        = 0x80,
69 
70     /* Execute the request only if the operation can be offloaded or otherwise
71      * be executed efficiently, but return an error instead of using a slow
72      * fallback. */
73     BDRV_REQ_NO_FALLBACK        = 0x100,
74 
75     /*
76      * BDRV_REQ_PREFETCH makes sense only in the context of copy-on-read
77      * (i.e., together with the BDRV_REQ_COPY_ON_READ flag or when a COR
78      * filter is involved), in which case it signals that the COR operation
79      * need not read the data into memory (qiov) but only ensure they are
80      * copied to the top layer (i.e., that COR operation is done).
81      */
82     BDRV_REQ_PREFETCH  = 0x200,
83 
84     /*
85      * If we need to wait for other requests, just fail immediately. Used
86      * only together with BDRV_REQ_SERIALISING.
87      */
88     BDRV_REQ_NO_WAIT = 0x400,
89 
90     /* Mask of valid flags */
91     BDRV_REQ_MASK               = 0x7ff,
92 } BdrvRequestFlags;
93 
94 typedef struct BlockSizes {
95     uint32_t phys;
96     uint32_t log;
97 } BlockSizes;
98 
99 typedef struct HDGeometry {
100     uint32_t heads;
101     uint32_t sectors;
102     uint32_t cylinders;
103 } HDGeometry;
104 
105 #define BDRV_O_NO_SHARE    0x0001 /* don't share permissions */
106 #define BDRV_O_RDWR        0x0002
107 #define BDRV_O_RESIZE      0x0004 /* request permission for resizing the node */
108 #define BDRV_O_SNAPSHOT    0x0008 /* open the file read only and save writes in a snapshot */
109 #define BDRV_O_TEMPORARY   0x0010 /* delete the file after use */
110 #define BDRV_O_NOCACHE     0x0020 /* do not use the host page cache */
111 #define BDRV_O_NATIVE_AIO  0x0080 /* use native AIO instead of the thread pool */
112 #define BDRV_O_NO_BACKING  0x0100 /* don't open the backing file */
113 #define BDRV_O_NO_FLUSH    0x0200 /* disable flushing on this disk */
114 #define BDRV_O_COPY_ON_READ 0x0400 /* copy read backing sectors into image */
115 #define BDRV_O_INACTIVE    0x0800  /* consistency hint for migration handoff */
116 #define BDRV_O_CHECK       0x1000  /* open solely for consistency check */
117 #define BDRV_O_ALLOW_RDWR  0x2000  /* allow reopen to change from r/o to r/w */
118 #define BDRV_O_UNMAP       0x4000  /* execute guest UNMAP/TRIM operations */
119 #define BDRV_O_PROTOCOL    0x8000  /* if no block driver is explicitly given:
120                                       select an appropriate protocol driver,
121                                       ignoring the format layer */
122 #define BDRV_O_NO_IO       0x10000 /* don't initialize for I/O */
123 #define BDRV_O_AUTO_RDONLY 0x20000 /* degrade to read-only if opening read-write fails */
124 #define BDRV_O_IO_URING    0x40000 /* use io_uring instead of the thread pool */
125 
126 #define BDRV_O_CACHE_MASK  (BDRV_O_NOCACHE | BDRV_O_NO_FLUSH)
127 
128 
129 /* Option names of options parsed by the block layer */
130 
131 #define BDRV_OPT_CACHE_WB       "cache.writeback"
132 #define BDRV_OPT_CACHE_DIRECT   "cache.direct"
133 #define BDRV_OPT_CACHE_NO_FLUSH "cache.no-flush"
134 #define BDRV_OPT_READ_ONLY      "read-only"
135 #define BDRV_OPT_AUTO_READ_ONLY "auto-read-only"
136 #define BDRV_OPT_DISCARD        "discard"
137 #define BDRV_OPT_FORCE_SHARE    "force-share"
138 
139 
140 #define BDRV_SECTOR_BITS   9
141 #define BDRV_SECTOR_SIZE   (1ULL << BDRV_SECTOR_BITS)
142 
143 #define BDRV_REQUEST_MAX_SECTORS MIN_CONST(SIZE_MAX >> BDRV_SECTOR_BITS, \
144                                            INT_MAX >> BDRV_SECTOR_BITS)
145 #define BDRV_REQUEST_MAX_BYTES (BDRV_REQUEST_MAX_SECTORS << BDRV_SECTOR_BITS)
146 
147 /*
148  * We want allow aligning requests and disk length up to any 32bit alignment
149  * and don't afraid of overflow.
150  * To achieve it, and in the same time use some pretty number as maximum disk
151  * size, let's define maximum "length" (a limit for any offset/bytes request and
152  * for disk size) to be the greatest power of 2 less than INT64_MAX.
153  */
154 #define BDRV_MAX_ALIGNMENT (1L << 30)
155 #define BDRV_MAX_LENGTH (QEMU_ALIGN_DOWN(INT64_MAX, BDRV_MAX_ALIGNMENT))
156 
157 /*
158  * Allocation status flags for bdrv_block_status() and friends.
159  *
160  * Public flags:
161  * BDRV_BLOCK_DATA: allocation for data at offset is tied to this layer
162  * BDRV_BLOCK_ZERO: offset reads as zero
163  * BDRV_BLOCK_OFFSET_VALID: an associated offset exists for accessing raw data
164  * BDRV_BLOCK_ALLOCATED: the content of the block is determined by this
165  *                       layer rather than any backing, set by block layer
166  * BDRV_BLOCK_EOF: the returned pnum covers through end of file for this
167  *                 layer, set by block layer
168  *
169  * Internal flags:
170  * BDRV_BLOCK_RAW: for use by passthrough drivers, such as raw, to request
171  *                 that the block layer recompute the answer from the returned
172  *                 BDS; must be accompanied by just BDRV_BLOCK_OFFSET_VALID.
173  * BDRV_BLOCK_RECURSE: request that the block layer will recursively search for
174  *                     zeroes in file child of current block node inside
175  *                     returned region. Only valid together with both
176  *                     BDRV_BLOCK_DATA and BDRV_BLOCK_OFFSET_VALID. Should not
177  *                     appear with BDRV_BLOCK_ZERO.
178  *
179  * If BDRV_BLOCK_OFFSET_VALID is set, the map parameter represents the
180  * host offset within the returned BDS that is allocated for the
181  * corresponding raw guest data.  However, whether that offset
182  * actually contains data also depends on BDRV_BLOCK_DATA, as follows:
183  *
184  * DATA ZERO OFFSET_VALID
185  *  t    t        t       sectors read as zero, returned file is zero at offset
186  *  t    f        t       sectors read as valid from file at offset
187  *  f    t        t       sectors preallocated, read as zero, returned file not
188  *                        necessarily zero at offset
189  *  f    f        t       sectors preallocated but read from backing_hd,
190  *                        returned file contains garbage at offset
191  *  t    t        f       sectors preallocated, read as zero, unknown offset
192  *  t    f        f       sectors read from unknown file or offset
193  *  f    t        f       not allocated or unknown offset, read as zero
194  *  f    f        f       not allocated or unknown offset, read from backing_hd
195  */
196 #define BDRV_BLOCK_DATA         0x01
197 #define BDRV_BLOCK_ZERO         0x02
198 #define BDRV_BLOCK_OFFSET_VALID 0x04
199 #define BDRV_BLOCK_RAW          0x08
200 #define BDRV_BLOCK_ALLOCATED    0x10
201 #define BDRV_BLOCK_EOF          0x20
202 #define BDRV_BLOCK_RECURSE      0x40
203 
204 typedef QTAILQ_HEAD(BlockReopenQueue, BlockReopenQueueEntry) BlockReopenQueue;
205 
206 typedef struct BDRVReopenState {
207     BlockDriverState *bs;
208     int flags;
209     BlockdevDetectZeroesOptions detect_zeroes;
210     bool backing_missing;
211     BlockDriverState *old_backing_bs; /* keep pointer for permissions update */
212     BlockDriverState *old_file_bs; /* keep pointer for permissions update */
213     QDict *options;
214     QDict *explicit_options;
215     void *opaque;
216 } BDRVReopenState;
217 
218 /*
219  * Block operation types
220  */
221 typedef enum BlockOpType {
222     BLOCK_OP_TYPE_BACKUP_SOURCE,
223     BLOCK_OP_TYPE_BACKUP_TARGET,
224     BLOCK_OP_TYPE_CHANGE,
225     BLOCK_OP_TYPE_COMMIT_SOURCE,
226     BLOCK_OP_TYPE_COMMIT_TARGET,
227     BLOCK_OP_TYPE_DATAPLANE,
228     BLOCK_OP_TYPE_DRIVE_DEL,
229     BLOCK_OP_TYPE_EJECT,
230     BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
231     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT,
232     BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE,
233     BLOCK_OP_TYPE_MIRROR_SOURCE,
234     BLOCK_OP_TYPE_MIRROR_TARGET,
235     BLOCK_OP_TYPE_RESIZE,
236     BLOCK_OP_TYPE_STREAM,
237     BLOCK_OP_TYPE_REPLACE,
238     BLOCK_OP_TYPE_MAX,
239 } BlockOpType;
240 
241 /* Block node permission constants */
242 enum {
243     /**
244      * A user that has the "permission" of consistent reads is guaranteed that
245      * their view of the contents of the block device is complete and
246      * self-consistent, representing the contents of a disk at a specific
247      * point.
248      *
249      * For most block devices (including their backing files) this is true, but
250      * the property cannot be maintained in a few situations like for
251      * intermediate nodes of a commit block job.
252      */
253     BLK_PERM_CONSISTENT_READ    = 0x01,
254 
255     /** This permission is required to change the visible disk contents. */
256     BLK_PERM_WRITE              = 0x02,
257 
258     /**
259      * This permission (which is weaker than BLK_PERM_WRITE) is both enough and
260      * required for writes to the block node when the caller promises that
261      * the visible disk content doesn't change.
262      *
263      * As the BLK_PERM_WRITE permission is strictly stronger, either is
264      * sufficient to perform an unchanging write.
265      */
266     BLK_PERM_WRITE_UNCHANGED    = 0x04,
267 
268     /** This permission is required to change the size of a block node. */
269     BLK_PERM_RESIZE             = 0x08,
270 
271     /**
272      * There was a now-removed bit BLK_PERM_GRAPH_MOD, with value of 0x10. QEMU
273      * 6.1 and earlier may still lock the corresponding byte in block/file-posix
274      * locking.  So, implementing some new permission should be very careful to
275      * not interfere with this old unused thing.
276      */
277 
278     BLK_PERM_ALL                = 0x0f,
279 
280     DEFAULT_PERM_PASSTHROUGH    = BLK_PERM_CONSISTENT_READ
281                                  | BLK_PERM_WRITE
282                                  | BLK_PERM_WRITE_UNCHANGED
283                                  | BLK_PERM_RESIZE,
284 
285     DEFAULT_PERM_UNCHANGED      = BLK_PERM_ALL & ~DEFAULT_PERM_PASSTHROUGH,
286 };
287 
288 /*
289  * Flags that parent nodes assign to child nodes to specify what kind of
290  * role(s) they take.
291  *
292  * At least one of DATA, METADATA, FILTERED, or COW must be set for
293  * every child.
294  */
295 enum BdrvChildRoleBits {
296     /*
297      * This child stores data.
298      * Any node may have an arbitrary number of such children.
299      */
300     BDRV_CHILD_DATA         = (1 << 0),
301 
302     /*
303      * This child stores metadata.
304      * Any node may have an arbitrary number of metadata-storing
305      * children.
306      */
307     BDRV_CHILD_METADATA     = (1 << 1),
308 
309     /*
310      * A child that always presents exactly the same visible data as
311      * the parent, e.g. by virtue of the parent forwarding all reads
312      * and writes.
313      * This flag is mutually exclusive with DATA, METADATA, and COW.
314      * Any node may have at most one filtered child at a time.
315      */
316     BDRV_CHILD_FILTERED     = (1 << 2),
317 
318     /*
319      * Child from which to read all data that isn't allocated in the
320      * parent (i.e., the backing child); such data is copied to the
321      * parent through COW (and optionally COR).
322      * This field is mutually exclusive with DATA, METADATA, and
323      * FILTERED.
324      * Any node may have at most one such backing child at a time.
325      */
326     BDRV_CHILD_COW          = (1 << 3),
327 
328     /*
329      * The primary child.  For most drivers, this is the child whose
330      * filename applies best to the parent node.
331      * Any node may have at most one primary child at a time.
332      */
333     BDRV_CHILD_PRIMARY      = (1 << 4),
334 
335     /* Useful combination of flags */
336     BDRV_CHILD_IMAGE        = BDRV_CHILD_DATA
337                               | BDRV_CHILD_METADATA
338                               | BDRV_CHILD_PRIMARY,
339 };
340 
341 /* Mask of BdrvChildRoleBits values */
342 typedef unsigned int BdrvChildRole;
343 
344 char *bdrv_perm_names(uint64_t perm);
345 uint64_t bdrv_qapi_perm_to_blk_perm(BlockPermission qapi_perm);
346 
347 /* disk I/O throttling */
348 void bdrv_init(void);
349 void bdrv_init_with_whitelist(void);
350 bool bdrv_uses_whitelist(void);
351 int bdrv_is_whitelisted(BlockDriver *drv, bool read_only);
352 BlockDriver *bdrv_find_protocol(const char *filename,
353                                 bool allow_protocol_prefix,
354                                 Error **errp);
355 BlockDriver *bdrv_find_format(const char *format_name);
356 int bdrv_create(BlockDriver *drv, const char* filename,
357                 QemuOpts *opts, Error **errp);
358 int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp);
359 
360 BlockDriverState *bdrv_new(void);
361 int bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top,
362                 Error **errp);
363 int bdrv_replace_node(BlockDriverState *from, BlockDriverState *to,
364                       Error **errp);
365 int bdrv_replace_child_bs(BdrvChild *child, BlockDriverState *new_bs,
366                           Error **errp);
367 BlockDriverState *bdrv_insert_node(BlockDriverState *bs, QDict *node_options,
368                                    int flags, Error **errp);
369 int bdrv_drop_filter(BlockDriverState *bs, Error **errp);
370 
371 int bdrv_parse_aio(const char *mode, int *flags);
372 int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough);
373 int bdrv_parse_discard_flags(const char *mode, int *flags);
374 BdrvChild *bdrv_open_child(const char *filename,
375                            QDict *options, const char *bdref_key,
376                            BlockDriverState* parent,
377                            const BdrvChildClass *child_class,
378                            BdrvChildRole child_role,
379                            bool allow_none, Error **errp);
380 BlockDriverState *bdrv_open_blockdev_ref(BlockdevRef *ref, Error **errp);
381 int bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd,
382                         Error **errp);
383 int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options,
384                            const char *bdref_key, Error **errp);
385 BlockDriverState *bdrv_open(const char *filename, const char *reference,
386                             QDict *options, int flags, Error **errp);
387 BlockDriverState *bdrv_new_open_driver_opts(BlockDriver *drv,
388                                             const char *node_name,
389                                             QDict *options, int flags,
390                                             Error **errp);
391 BlockDriverState *bdrv_new_open_driver(BlockDriver *drv, const char *node_name,
392                                        int flags, Error **errp);
393 BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue,
394                                     BlockDriverState *bs, QDict *options,
395                                     bool keep_old_opts);
396 void bdrv_reopen_queue_free(BlockReopenQueue *bs_queue);
397 int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp);
398 int bdrv_reopen(BlockDriverState *bs, QDict *opts, bool keep_old_opts,
399                 Error **errp);
400 int bdrv_reopen_set_read_only(BlockDriverState *bs, bool read_only,
401                               Error **errp);
402 int bdrv_pwrite_zeroes(BdrvChild *child, int64_t offset,
403                        int64_t bytes, BdrvRequestFlags flags);
404 int bdrv_make_zero(BdrvChild *child, BdrvRequestFlags flags);
405 int bdrv_pread(BdrvChild *child, int64_t offset, void *buf, int64_t bytes);
406 int bdrv_pwrite(BdrvChild *child, int64_t offset, const void *buf,
407                 int64_t bytes);
408 int bdrv_pwrite_sync(BdrvChild *child, int64_t offset,
409                      const void *buf, int64_t bytes);
410 /*
411  * Efficiently zero a region of the disk image.  Note that this is a regular
412  * I/O request like read or write and should have a reasonable size.  This
413  * function is not suitable for zeroing the entire image in a single request
414  * because it may allocate memory for the entire region.
415  */
416 int coroutine_fn bdrv_co_pwrite_zeroes(BdrvChild *child, int64_t offset,
417                                        int64_t bytes, BdrvRequestFlags flags);
418 BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,
419     const char *backing_file);
420 void bdrv_refresh_filename(BlockDriverState *bs);
421 
422 int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset, bool exact,
423                                   PreallocMode prealloc, BdrvRequestFlags flags,
424                                   Error **errp);
425 int generated_co_wrapper
426 bdrv_truncate(BdrvChild *child, int64_t offset, bool exact,
427               PreallocMode prealloc, BdrvRequestFlags flags, Error **errp);
428 
429 int64_t bdrv_nb_sectors(BlockDriverState *bs);
430 int64_t bdrv_getlength(BlockDriverState *bs);
431 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs);
432 BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
433                                BlockDriverState *in_bs, Error **errp);
434 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr);
435 void bdrv_refresh_limits(BlockDriverState *bs, Transaction *tran, Error **errp);
436 int bdrv_commit(BlockDriverState *bs);
437 int bdrv_make_empty(BdrvChild *c, Error **errp);
438 int bdrv_change_backing_file(BlockDriverState *bs, const char *backing_file,
439                              const char *backing_fmt, bool warn);
440 void bdrv_register(BlockDriver *bdrv);
441 int bdrv_drop_intermediate(BlockDriverState *top, BlockDriverState *base,
442                            const char *backing_file_str);
443 BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
444                                     BlockDriverState *bs);
445 BlockDriverState *bdrv_find_base(BlockDriverState *bs);
446 bool bdrv_is_backing_chain_frozen(BlockDriverState *bs, BlockDriverState *base,
447                                   Error **errp);
448 int bdrv_freeze_backing_chain(BlockDriverState *bs, BlockDriverState *base,
449                               Error **errp);
450 void bdrv_unfreeze_backing_chain(BlockDriverState *bs, BlockDriverState *base);
451 int coroutine_fn bdrv_co_delete_file(BlockDriverState *bs, Error **errp);
452 void coroutine_fn bdrv_co_delete_file_noerr(BlockDriverState *bs);
453 
454 
455 typedef struct BdrvCheckResult {
456     int corruptions;
457     int leaks;
458     int check_errors;
459     int corruptions_fixed;
460     int leaks_fixed;
461     int64_t image_end_offset;
462     BlockFragInfo bfi;
463 } BdrvCheckResult;
464 
465 typedef enum {
466     BDRV_FIX_LEAKS    = 1,
467     BDRV_FIX_ERRORS   = 2,
468 } BdrvCheckMode;
469 
470 int generated_co_wrapper bdrv_check(BlockDriverState *bs, BdrvCheckResult *res,
471                                     BdrvCheckMode fix);
472 
473 /* The units of offset and total_work_size may be chosen arbitrarily by the
474  * block driver; total_work_size may change during the course of the amendment
475  * operation */
476 typedef void BlockDriverAmendStatusCB(BlockDriverState *bs, int64_t offset,
477                                       int64_t total_work_size, void *opaque);
478 int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts,
479                        BlockDriverAmendStatusCB *status_cb, void *cb_opaque,
480                        bool force,
481                        Error **errp);
482 
483 /* check if a named node can be replaced when doing drive-mirror */
484 BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs,
485                                         const char *node_name, Error **errp);
486 
487 /* async block I/O */
488 void bdrv_aio_cancel(BlockAIOCB *acb);
489 void bdrv_aio_cancel_async(BlockAIOCB *acb);
490 
491 /* sg packet commands */
492 int bdrv_co_ioctl(BlockDriverState *bs, int req, void *buf);
493 
494 /* Invalidate any cached metadata used by image formats */
495 int generated_co_wrapper bdrv_invalidate_cache(BlockDriverState *bs,
496                                                Error **errp);
497 void bdrv_invalidate_cache_all(Error **errp);
498 int bdrv_inactivate_all(void);
499 
500 /* Ensure contents are flushed to disk.  */
501 int generated_co_wrapper bdrv_flush(BlockDriverState *bs);
502 int coroutine_fn bdrv_co_flush(BlockDriverState *bs);
503 int bdrv_flush_all(void);
504 void bdrv_close_all(void);
505 void bdrv_drain(BlockDriverState *bs);
506 void coroutine_fn bdrv_co_drain(BlockDriverState *bs);
507 void bdrv_drain_all_begin(void);
508 void bdrv_drain_all_end(void);
509 void bdrv_drain_all(void);
510 
511 #define BDRV_POLL_WHILE(bs, cond) ({                       \
512     BlockDriverState *bs_ = (bs);                          \
513     AIO_WAIT_WHILE(bdrv_get_aio_context(bs_),              \
514                    cond); })
515 
516 int generated_co_wrapper bdrv_pdiscard(BdrvChild *child, int64_t offset,
517                                        int64_t bytes);
518 int bdrv_co_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes);
519 int bdrv_has_zero_init_1(BlockDriverState *bs);
520 int bdrv_has_zero_init(BlockDriverState *bs);
521 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs);
522 int bdrv_block_status(BlockDriverState *bs, int64_t offset,
523                       int64_t bytes, int64_t *pnum, int64_t *map,
524                       BlockDriverState **file);
525 int bdrv_block_status_above(BlockDriverState *bs, BlockDriverState *base,
526                             int64_t offset, int64_t bytes, int64_t *pnum,
527                             int64_t *map, BlockDriverState **file);
528 int bdrv_is_allocated(BlockDriverState *bs, int64_t offset, int64_t bytes,
529                       int64_t *pnum);
530 int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
531                             bool include_base, int64_t offset, int64_t bytes,
532                             int64_t *pnum);
533 int coroutine_fn bdrv_co_is_zero_fast(BlockDriverState *bs, int64_t offset,
534                                       int64_t bytes);
535 
536 bool bdrv_is_read_only(BlockDriverState *bs);
537 int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only,
538                            bool ignore_allow_rdw, Error **errp);
539 int bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg,
540                               Error **errp);
541 bool bdrv_is_writable(BlockDriverState *bs);
542 bool bdrv_is_sg(BlockDriverState *bs);
543 bool bdrv_is_inserted(BlockDriverState *bs);
544 void bdrv_lock_medium(BlockDriverState *bs, bool locked);
545 void bdrv_eject(BlockDriverState *bs, bool eject_flag);
546 const char *bdrv_get_format_name(BlockDriverState *bs);
547 BlockDriverState *bdrv_find_node(const char *node_name);
548 BlockDeviceInfoList *bdrv_named_nodes_list(bool flat, Error **errp);
549 XDbgBlockGraph *bdrv_get_xdbg_block_graph(Error **errp);
550 BlockDriverState *bdrv_lookup_bs(const char *device,
551                                  const char *node_name,
552                                  Error **errp);
553 bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base);
554 BlockDriverState *bdrv_next_node(BlockDriverState *bs);
555 BlockDriverState *bdrv_next_all_states(BlockDriverState *bs);
556 
557 typedef struct BdrvNextIterator {
558     enum {
559         BDRV_NEXT_BACKEND_ROOTS,
560         BDRV_NEXT_MONITOR_OWNED,
561     } phase;
562     BlockBackend *blk;
563     BlockDriverState *bs;
564 } BdrvNextIterator;
565 
566 BlockDriverState *bdrv_first(BdrvNextIterator *it);
567 BlockDriverState *bdrv_next(BdrvNextIterator *it);
568 void bdrv_next_cleanup(BdrvNextIterator *it);
569 
570 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs);
571 bool bdrv_supports_compressed_writes(BlockDriverState *bs);
572 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
573                          void *opaque, bool read_only);
574 const char *bdrv_get_node_name(const BlockDriverState *bs);
575 const char *bdrv_get_device_name(const BlockDriverState *bs);
576 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs);
577 int bdrv_get_flags(BlockDriverState *bs);
578 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi);
579 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs,
580                                           Error **errp);
581 BlockStatsSpecific *bdrv_get_specific_stats(BlockDriverState *bs);
582 void bdrv_round_to_clusters(BlockDriverState *bs,
583                             int64_t offset, int64_t bytes,
584                             int64_t *cluster_offset,
585                             int64_t *cluster_bytes);
586 
587 void bdrv_get_backing_filename(BlockDriverState *bs,
588                                char *filename, int filename_size);
589 char *bdrv_get_full_backing_filename(BlockDriverState *bs, Error **errp);
590 char *bdrv_get_full_backing_filename_from_filename(const char *backed,
591                                                    const char *backing,
592                                                    Error **errp);
593 char *bdrv_dirname(BlockDriverState *bs, Error **errp);
594 
595 int path_has_protocol(const char *path);
596 int path_is_absolute(const char *path);
597 char *path_combine(const char *base_path, const char *filename);
598 
599 int generated_co_wrapper
600 bdrv_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
601 int generated_co_wrapper
602 bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos);
603 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
604                       int64_t pos, int size);
605 
606 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
607                       int64_t pos, int size);
608 
609 void bdrv_img_create(const char *filename, const char *fmt,
610                      const char *base_filename, const char *base_fmt,
611                      char *options, uint64_t img_size, int flags,
612                      bool quiet, Error **errp);
613 
614 /* Returns the alignment in bytes that is required so that no bounce buffer
615  * is required throughout the stack */
616 size_t bdrv_min_mem_align(BlockDriverState *bs);
617 /* Returns optimal alignment in bytes for bounce buffer */
618 size_t bdrv_opt_mem_align(BlockDriverState *bs);
619 void *qemu_blockalign(BlockDriverState *bs, size_t size);
620 void *qemu_blockalign0(BlockDriverState *bs, size_t size);
621 void *qemu_try_blockalign(BlockDriverState *bs, size_t size);
622 void *qemu_try_blockalign0(BlockDriverState *bs, size_t size);
623 bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov);
624 
625 void bdrv_enable_copy_on_read(BlockDriverState *bs);
626 void bdrv_disable_copy_on_read(BlockDriverState *bs);
627 
628 void bdrv_ref(BlockDriverState *bs);
629 void bdrv_unref(BlockDriverState *bs);
630 void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child);
631 BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
632                              BlockDriverState *child_bs,
633                              const char *child_name,
634                              const BdrvChildClass *child_class,
635                              BdrvChildRole child_role,
636                              Error **errp);
637 
638 bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp);
639 void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason);
640 void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason);
641 void bdrv_op_block_all(BlockDriverState *bs, Error *reason);
642 void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason);
643 bool bdrv_op_blocker_is_empty(BlockDriverState *bs);
644 
645 #define BLKDBG_EVENT(child, evt) \
646     do { \
647         if (child) { \
648             bdrv_debug_event(child->bs, evt); \
649         } \
650     } while (0)
651 
652 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event);
653 
654 int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event,
655                            const char *tag);
656 int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag);
657 int bdrv_debug_resume(BlockDriverState *bs, const char *tag);
658 bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag);
659 
660 /**
661  * bdrv_get_aio_context:
662  *
663  * Returns: the currently bound #AioContext
664  */
665 AioContext *bdrv_get_aio_context(BlockDriverState *bs);
666 
667 /**
668  * Move the current coroutine to the AioContext of @bs and return the old
669  * AioContext of the coroutine. Increase bs->in_flight so that draining @bs
670  * will wait for the operation to proceed until the corresponding
671  * bdrv_co_leave().
672  *
673  * Consequently, you can't call drain inside a bdrv_co_enter/leave() section as
674  * this will deadlock.
675  */
676 AioContext *coroutine_fn bdrv_co_enter(BlockDriverState *bs);
677 
678 /**
679  * Ends a section started by bdrv_co_enter(). Move the current coroutine back
680  * to old_ctx and decrease bs->in_flight again.
681  */
682 void coroutine_fn bdrv_co_leave(BlockDriverState *bs, AioContext *old_ctx);
683 
684 /**
685  * Locks the AioContext of @bs if it's not the current AioContext. This avoids
686  * double locking which could lead to deadlocks: This is a coroutine_fn, so we
687  * know we already own the lock of the current AioContext.
688  *
689  * May only be called in the main thread.
690  */
691 void coroutine_fn bdrv_co_lock(BlockDriverState *bs);
692 
693 /**
694  * Unlocks the AioContext of @bs if it's not the current AioContext.
695  */
696 void coroutine_fn bdrv_co_unlock(BlockDriverState *bs);
697 
698 /**
699  * Transfer control to @co in the aio context of @bs
700  */
701 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co);
702 
703 void bdrv_set_aio_context_ignore(BlockDriverState *bs,
704                                  AioContext *new_context, GSList **ignore);
705 int bdrv_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
706                              Error **errp);
707 int bdrv_child_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
708                                    BdrvChild *ignore_child, Error **errp);
709 bool bdrv_child_can_set_aio_context(BdrvChild *c, AioContext *ctx,
710                                     GSList **ignore, Error **errp);
711 bool bdrv_can_set_aio_context(BlockDriverState *bs, AioContext *ctx,
712                               GSList **ignore, Error **errp);
713 AioContext *bdrv_child_get_parent_aio_context(BdrvChild *c);
714 AioContext *child_of_bds_get_parent_aio_context(BdrvChild *c);
715 
716 int bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz);
717 int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo);
718 
719 void bdrv_io_plug(BlockDriverState *bs);
720 void bdrv_io_unplug(BlockDriverState *bs);
721 
722 /**
723  * bdrv_parent_drained_begin_single:
724  *
725  * Begin a quiesced section for the parent of @c. If @poll is true, wait for
726  * any pending activity to cease.
727  */
728 void bdrv_parent_drained_begin_single(BdrvChild *c, bool poll);
729 
730 /**
731  * bdrv_parent_drained_end_single:
732  *
733  * End a quiesced section for the parent of @c.
734  *
735  * This polls @bs's AioContext until all scheduled sub-drained_ends
736  * have settled, which may result in graph changes.
737  */
738 void bdrv_parent_drained_end_single(BdrvChild *c);
739 
740 /**
741  * bdrv_drain_poll:
742  *
743  * Poll for pending requests in @bs, its parents (except for @ignore_parent),
744  * and if @recursive is true its children as well (used for subtree drain).
745  *
746  * If @ignore_bds_parents is true, parents that are BlockDriverStates must
747  * ignore the drain request because they will be drained separately (used for
748  * drain_all).
749  *
750  * This is part of bdrv_drained_begin.
751  */
752 bool bdrv_drain_poll(BlockDriverState *bs, bool recursive,
753                      BdrvChild *ignore_parent, bool ignore_bds_parents);
754 
755 /**
756  * bdrv_drained_begin:
757  *
758  * Begin a quiesced section for exclusive access to the BDS, by disabling
759  * external request sources including NBD server, block jobs, and device model.
760  *
761  * This function can be recursive.
762  */
763 void bdrv_drained_begin(BlockDriverState *bs);
764 
765 /**
766  * bdrv_do_drained_begin_quiesce:
767  *
768  * Quiesces a BDS like bdrv_drained_begin(), but does not wait for already
769  * running requests to complete.
770  */
771 void bdrv_do_drained_begin_quiesce(BlockDriverState *bs,
772                                    BdrvChild *parent, bool ignore_bds_parents);
773 
774 /**
775  * Like bdrv_drained_begin, but recursively begins a quiesced section for
776  * exclusive access to all child nodes as well.
777  */
778 void bdrv_subtree_drained_begin(BlockDriverState *bs);
779 
780 /**
781  * bdrv_drained_end:
782  *
783  * End a quiescent section started by bdrv_drained_begin().
784  *
785  * This polls @bs's AioContext until all scheduled sub-drained_ends
786  * have settled.  On one hand, that may result in graph changes.  On
787  * the other, this requires that the caller either runs in the main
788  * loop; or that all involved nodes (@bs and all of its parents) are
789  * in the caller's AioContext.
790  */
791 void bdrv_drained_end(BlockDriverState *bs);
792 
793 /**
794  * bdrv_drained_end_no_poll:
795  *
796  * Same as bdrv_drained_end(), but do not poll for the subgraph to
797  * actually become unquiesced.  Therefore, no graph changes will occur
798  * with this function.
799  *
800  * *drained_end_counter is incremented for every background operation
801  * that is scheduled, and will be decremented for every operation once
802  * it settles.  The caller must poll until it reaches 0.  The counter
803  * should be accessed using atomic operations only.
804  */
805 void bdrv_drained_end_no_poll(BlockDriverState *bs, int *drained_end_counter);
806 
807 /**
808  * End a quiescent section started by bdrv_subtree_drained_begin().
809  */
810 void bdrv_subtree_drained_end(BlockDriverState *bs);
811 
812 void bdrv_add_child(BlockDriverState *parent, BlockDriverState *child,
813                     Error **errp);
814 void bdrv_del_child(BlockDriverState *parent, BdrvChild *child, Error **errp);
815 
816 bool bdrv_can_store_new_dirty_bitmap(BlockDriverState *bs, const char *name,
817                                      uint32_t granularity, Error **errp);
818 /**
819  *
820  * bdrv_register_buf/bdrv_unregister_buf:
821  *
822  * Register/unregister a buffer for I/O. For example, VFIO drivers are
823  * interested to know the memory areas that would later be used for I/O, so
824  * that they can prepare IOMMU mapping etc., to get better performance.
825  */
826 void bdrv_register_buf(BlockDriverState *bs, void *host, size_t size);
827 void bdrv_unregister_buf(BlockDriverState *bs, void *host);
828 
829 /**
830  *
831  * bdrv_co_copy_range:
832  *
833  * Do offloaded copy between two children. If the operation is not implemented
834  * by the driver, or if the backend storage doesn't support it, a negative
835  * error code will be returned.
836  *
837  * Note: block layer doesn't emulate or fallback to a bounce buffer approach
838  * because usually the caller shouldn't attempt offloaded copy any more (e.g.
839  * calling copy_file_range(2)) after the first error, thus it should fall back
840  * to a read+write path in the caller level.
841  *
842  * @src: Source child to copy data from
843  * @src_offset: offset in @src image to read data
844  * @dst: Destination child to copy data to
845  * @dst_offset: offset in @dst image to write data
846  * @bytes: number of bytes to copy
847  * @flags: request flags. Supported flags:
848  *         BDRV_REQ_ZERO_WRITE - treat the @src range as zero data and do zero
849  *                               write on @dst as if bdrv_co_pwrite_zeroes is
850  *                               called. Used to simplify caller code, or
851  *                               during BlockDriver.bdrv_co_copy_range_from()
852  *                               recursion.
853  *         BDRV_REQ_NO_SERIALISING - do not serialize with other overlapping
854  *                                   requests currently in flight.
855  *
856  * Returns: 0 if succeeded; negative error code if failed.
857  **/
858 int coroutine_fn bdrv_co_copy_range(BdrvChild *src, int64_t src_offset,
859                                     BdrvChild *dst, int64_t dst_offset,
860                                     int64_t bytes, BdrvRequestFlags read_flags,
861                                     BdrvRequestFlags write_flags);
862 
863 void bdrv_cancel_in_flight(BlockDriverState *bs);
864 
865 #endif
866